[llvm] 74a130a - [RISCV] Add isel patterns for vfmacc, vfnmacc, vfmsac and vfnmsac.
Yeting Kuo via llvm-commits
llvm-commits at lists.llvm.org
Tue Oct 4 18:56:56 PDT 2022
Author: Yeting Kuo
Date: 2022-10-05T09:56:43+08:00
New Revision: 74a130af9735d37100408604828707db2922a971
URL: https://github.com/llvm/llvm-project/commit/74a130af9735d37100408604828707db2922a971
DIFF: https://github.com/llvm/llvm-project/commit/74a130af9735d37100408604828707db2922a971.diff
LOG: [RISCV] Add isel patterns for vfmacc, vfnmacc, vfmsac and vfnmsac.
The patch selects VSELECT_VL/VP_MERGE_VL that uses VF(N)M(ACC|SAC) as its
true operand and the adden of the true operand as its false operand.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D135080
Added:
llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmacc-vp.ll
llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmsac-vp.ll
llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmacc-vp.ll
llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmsac-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfmacc-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfmsac-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfnmacc-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfnmsac-vp.ll
Modified:
llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
Removed:
################################################################################
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index d31361554874a..a11ef668494e3 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -327,6 +327,34 @@ def riscv_fpextend_vl_oneuse : PatFrag<(ops node:$A, node:$B, node:$C),
return N->hasOneUse();
}]>;
+def riscv_vfmadd_vl_oneuse : PatFrag<(ops node:$A, node:$B, node:$C, node:$D,
+ node:$E),
+ (riscv_vfmadd_vl node:$A, node:$B,
+ node:$C, node:$D, node:$E), [{
+ return N->hasOneUse();
+}]>;
+
+def riscv_vfnmadd_vl_oneuse : PatFrag<(ops node:$A, node:$B, node:$C, node:$D,
+ node:$E),
+ (riscv_vfnmadd_vl node:$A, node:$B,
+ node:$C, node:$D, node:$E), [{
+ return N->hasOneUse();
+}]>;
+
+def riscv_vfmsub_vl_oneuse : PatFrag<(ops node:$A, node:$B, node:$C, node:$D,
+ node:$E),
+ (riscv_vfmsub_vl node:$A, node:$B,
+ node:$C, node:$D, node:$E), [{
+ return N->hasOneUse();
+}]>;
+
+def riscv_vfnmsub_vl_oneuse : PatFrag<(ops node:$A, node:$B, node:$C, node:$D,
+ node:$E),
+ (riscv_vfnmsub_vl node:$A, node:$B,
+ node:$C, node:$D, node:$E), [{
+ return N->hasOneUse();
+}]>;
+
foreach kind = ["ADD", "UMAX", "SMAX", "UMIN", "SMIN", "AND", "OR", "XOR",
"FADD", "SEQ_FADD", "FMIN", "FMAX"] in
def rvv_vecreduce_#kind#_vl : SDNode<"RISCVISD::VECREDUCE_"#kind#"_VL", SDTRVVVecReduce>;
@@ -1057,6 +1085,54 @@ multiclass VPatFPMulAddVL_VV_VF<SDNode vop, string instruction_name> {
}
}
+multiclass VPatFPMulAccVL_VV_VF<PatFrag vop, string instruction_name> {
+ foreach vti = AllFloatVectors in {
+ defvar suffix = vti.LMul.MX;
+ def : Pat<(riscv_vp_merge_vl (vti.Mask true_mask),
+ (vti.Vector (vop vti.RegClass:$rs1, vti.RegClass:$rs2,
+ vti.RegClass:$rd, (vti.Mask true_mask), VLOpFrag)),
+ vti.RegClass:$rd, VLOpFrag),
+ (!cast<Instruction>(instruction_name#"_VV_"# suffix)
+ vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+ GPR:$vl, vti.Log2SEW, TAIL_UNDISTURBED_MASK_UNDISTURBED)>;
+ def : Pat<(riscv_vp_merge_vl (vti.Mask V0),
+ (vti.Vector (vop vti.RegClass:$rs1, vti.RegClass:$rs2,
+ vti.RegClass:$rd, (vti.Mask true_mask), VLOpFrag)),
+ vti.RegClass:$rd, VLOpFrag),
+ (!cast<Instruction>(instruction_name#"_VV_"# suffix #"_MASK")
+ vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+ (vti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_UNDISTURBED_MASK_UNDISTURBED)>;
+ def : Pat<(riscv_vp_merge_vl (vti.Mask true_mask),
+ (vti.Vector (vop (SplatFPOp vti.ScalarRegClass:$rs1), vti.RegClass:$rs2,
+ vti.RegClass:$rd, (vti.Mask true_mask), VLOpFrag)),
+ vti.RegClass:$rd, VLOpFrag),
+ (!cast<Instruction>(instruction_name#"_V" # vti.ScalarSuffix # "_" # suffix)
+ vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+ GPR:$vl, vti.Log2SEW, TAIL_UNDISTURBED_MASK_UNDISTURBED)>;
+ def : Pat<(riscv_vp_merge_vl (vti.Mask V0),
+ (vti.Vector (vop (SplatFPOp vti.ScalarRegClass:$rs1), vti.RegClass:$rs2,
+ vti.RegClass:$rd, (vti.Mask true_mask), VLOpFrag)),
+ vti.RegClass:$rd, VLOpFrag),
+ (!cast<Instruction>(instruction_name#"_V" # vti.ScalarSuffix # "_" # suffix # "_MASK")
+ vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+ (vti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_UNDISTURBED_MASK_UNDISTURBED)>;
+ def : Pat<(riscv_vselect_vl (vti.Mask V0),
+ (vti.Vector (vop vti.RegClass:$rs1, vti.RegClass:$rs2,
+ vti.RegClass:$rd, (vti.Mask true_mask), VLOpFrag)),
+ vti.RegClass:$rd, VLOpFrag),
+ (!cast<Instruction>(instruction_name#"_VV_"# suffix #"_MASK")
+ vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+ (vti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
+ def : Pat<(riscv_vselect_vl (vti.Mask V0),
+ (vti.Vector (vop (SplatFPOp vti.ScalarRegClass:$rs1), vti.RegClass:$rs2,
+ vti.RegClass:$rd, (vti.Mask true_mask), VLOpFrag)),
+ vti.RegClass:$rd, VLOpFrag),
+ (!cast<Instruction>(instruction_name#"_V" # vti.ScalarSuffix # "_" # suffix # "_MASK")
+ vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+ (vti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
+ }
+}
+
multiclass VPatWidenFPMulAccVL_VV_VF<SDNode vop, string instruction_name> {
foreach vtiToWti = AllWidenableFloatVectors in {
defvar vti = vtiToWti.Vti;
@@ -1393,6 +1469,10 @@ defm : VPatFPMulAddVL_VV_VF<riscv_vfmadd_vl, "PseudoVFMADD">;
defm : VPatFPMulAddVL_VV_VF<riscv_vfmsub_vl, "PseudoVFMSUB">;
defm : VPatFPMulAddVL_VV_VF<riscv_vfnmadd_vl, "PseudoVFNMADD">;
defm : VPatFPMulAddVL_VV_VF<riscv_vfnmsub_vl, "PseudoVFNMSUB">;
+defm : VPatFPMulAccVL_VV_VF<riscv_vfmadd_vl_oneuse, "PseudoVFMACC">;
+defm : VPatFPMulAccVL_VV_VF<riscv_vfmsub_vl_oneuse, "PseudoVFMSAC">;
+defm : VPatFPMulAccVL_VV_VF<riscv_vfnmadd_vl_oneuse, "PseudoVFNMACC">;
+defm : VPatFPMulAccVL_VV_VF<riscv_vfnmsub_vl_oneuse, "PseudoVFNMSAC">;
// 14.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
defm : VPatWidenFPMulAccVL_VV_VF<riscv_vfmadd_vl, "PseudoVFWMACC">;
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmacc-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmacc-vp.ll
new file mode 100644
index 0000000000000..7fa78ee4b6cf9
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmacc-vp.ll
@@ -0,0 +1,1529 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <2 x half> @llvm.vp.fma.v2f16(<2 x half>, <2 x half>, <2 x half>, <2 x i1>, i32)
+declare <2 x half> @llvm.vp.fneg.v2f16(<2 x half>, <2 x i1>, i32)
+declare <2 x half> @llvm.vp.merge.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+declare <2 x half> @llvm.vp.select.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+
+define <2 x half> @vfmacc_vv_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vv_v2f16_unmasked(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %allones, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vf_v2f16(<2 x half> %va, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %va, <2 x half> %vb, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vf_v2f16_commute(<2 x half> %va, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %vb, <2 x half> %va, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vf_v2f16_unmasked(<2 x half> %va, half %b, <2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %va, <2 x half> %vb, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %allones, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vv_v2f16_ta(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vf_v2f16_ta(<2 x half> %va, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %va, <2 x half> %vb, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmacc_vf_v2f16_commute_ta(<2 x half> %va, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %vb, <2 x half> %va, <2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+declare <4 x half> @llvm.vp.fma.v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x i1>, i32)
+declare <4 x half> @llvm.vp.fneg.v4f16(<4 x half>, <4 x i1>, i32)
+declare <4 x half> @llvm.vp.merge.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+declare <4 x half> @llvm.vp.select.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+
+define <4 x half> @vfmacc_vv_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vv_v4f16_unmasked(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %allones, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vf_v4f16(<4 x half> %va, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %va, <4 x half> %vb, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vf_v4f16_commute(<4 x half> %va, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %vb, <4 x half> %va, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vf_v4f16_unmasked(<4 x half> %va, half %b, <4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %va, <4 x half> %vb, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %allones, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vv_v4f16_ta(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vf_v4f16_ta(<4 x half> %va, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %va, <4 x half> %vb, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmacc_vf_v4f16_commute_ta(<4 x half> %va, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %vb, <4 x half> %va, <4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+declare <8 x half> @llvm.vp.fma.v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x i1>, i32)
+declare <8 x half> @llvm.vp.fneg.v8f16(<8 x half>, <8 x i1>, i32)
+declare <8 x half> @llvm.vp.merge.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+declare <8 x half> @llvm.vp.select.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+
+define <8 x half> @vfmacc_vv_v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vv_v8f16_unmasked(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %allones, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vf_v8f16(<8 x half> %va, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %va, <8 x half> %vb, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vf_v8f16_commute(<8 x half> %va, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %vb, <8 x half> %va, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vf_v8f16_unmasked(<8 x half> %va, half %b, <8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %va, <8 x half> %vb, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %allones, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vv_v8f16_ta(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vf_v8f16_ta(<8 x half> %va, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %va, <8 x half> %vb, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmacc_vf_v8f16_commute_ta(<8 x half> %va, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %vb, <8 x half> %va, <8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+declare <16 x half> @llvm.vp.fma.v16f16(<16 x half>, <16 x half>, <16 x half>, <16 x i1>, i32)
+declare <16 x half> @llvm.vp.fneg.v16f16(<16 x half>, <16 x i1>, i32)
+declare <16 x half> @llvm.vp.merge.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+declare <16 x half> @llvm.vp.select.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+
+define <16 x half> @vfmacc_vv_v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vv_v16f16_unmasked(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %allones, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vf_v16f16(<16 x half> %va, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %va, <16 x half> %vb, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vf_v16f16_commute(<16 x half> %va, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %vb, <16 x half> %va, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vf_v16f16_unmasked(<16 x half> %va, half %b, <16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %va, <16 x half> %vb, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %allones, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vv_v16f16_ta(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vf_v16f16_ta(<16 x half> %va, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %va, <16 x half> %vb, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmacc_vf_v16f16_commute_ta(<16 x half> %va, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %vb, <16 x half> %va, <16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+declare <32 x half> @llvm.vp.fma.v32f16(<32 x half>, <32 x half>, <32 x half>, <32 x i1>, i32)
+declare <32 x half> @llvm.vp.fneg.v32f16(<32 x half>, <32 x i1>, i32)
+declare <32 x half> @llvm.vp.merge.v32f16(<32 x i1>, <32 x half>, <32 x half>, i32)
+declare <32 x half> @llvm.vp.select.v32f16(<32 x i1>, <32 x half>, <32 x half>, i32)
+
+define <32 x half> @vfmacc_vv_v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vv_v32f16_unmasked(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %allones, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vf_v32f16(<32 x half> %va, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %va, <32 x half> %vb, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vf_v32f16_commute(<32 x half> %va, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %vb, <32 x half> %va, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vf_v32f16_unmasked(<32 x half> %va, half %b, <32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %va, <32 x half> %vb, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %allones, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vv_v32f16_ta(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vf_v32f16_ta(<32 x half> %va, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %va, <32 x half> %vb, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmacc_vf_v32f16_commute_ta(<32 x half> %va, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %vb, <32 x half> %va, <32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+declare <2 x float> @llvm.vp.fma.v2f32(<2 x float>, <2 x float>, <2 x float>, <2 x i1>, i32)
+declare <2 x float> @llvm.vp.fneg.v2f32(<2 x float>, <2 x i1>, i32)
+declare <2 x float> @llvm.vp.merge.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+declare <2 x float> @llvm.vp.select.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+
+define <2 x float> @vfmacc_vv_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vv_v2f32_unmasked(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %allones, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vf_v2f32(<2 x float> %va, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %va, <2 x float> %vb, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vf_v2f32_commute(<2 x float> %va, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %vb, <2 x float> %va, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vf_v2f32_unmasked(<2 x float> %va, float %b, <2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %va, <2 x float> %vb, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %allones, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vv_v2f32_ta(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vf_v2f32_ta(<2 x float> %va, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %va, <2 x float> %vb, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmacc_vf_v2f32_commute_ta(<2 x float> %va, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %vb, <2 x float> %va, <2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+declare <4 x float> @llvm.vp.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, <4 x i1>, i32)
+declare <4 x float> @llvm.vp.fneg.v4f32(<4 x float>, <4 x i1>, i32)
+declare <4 x float> @llvm.vp.merge.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+declare <4 x float> @llvm.vp.select.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+
+define <4 x float> @vfmacc_vv_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vv_v4f32_unmasked(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %allones, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vf_v4f32(<4 x float> %va, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %va, <4 x float> %vb, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vf_v4f32_commute(<4 x float> %va, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %vb, <4 x float> %va, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vf_v4f32_unmasked(<4 x float> %va, float %b, <4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %va, <4 x float> %vb, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %allones, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vv_v4f32_ta(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vf_v4f32_ta(<4 x float> %va, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %va, <4 x float> %vb, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmacc_vf_v4f32_commute_ta(<4 x float> %va, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %vb, <4 x float> %va, <4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+declare <8 x float> @llvm.vp.fma.v8f32(<8 x float>, <8 x float>, <8 x float>, <8 x i1>, i32)
+declare <8 x float> @llvm.vp.fneg.v8f32(<8 x float>, <8 x i1>, i32)
+declare <8 x float> @llvm.vp.merge.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+declare <8 x float> @llvm.vp.select.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+
+define <8 x float> @vfmacc_vv_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vv_v8f32_unmasked(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %allones, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vf_v8f32(<8 x float> %va, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %va, <8 x float> %vb, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vf_v8f32_commute(<8 x float> %va, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %vb, <8 x float> %va, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vf_v8f32_unmasked(<8 x float> %va, float %b, <8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %va, <8 x float> %vb, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %allones, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vv_v8f32_ta(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vf_v8f32_ta(<8 x float> %va, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %va, <8 x float> %vb, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmacc_vf_v8f32_commute_ta(<8 x float> %va, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %vb, <8 x float> %va, <8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+declare <16 x float> @llvm.vp.fma.v16f32(<16 x float>, <16 x float>, <16 x float>, <16 x i1>, i32)
+declare <16 x float> @llvm.vp.fneg.v16f32(<16 x float>, <16 x i1>, i32)
+declare <16 x float> @llvm.vp.merge.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+declare <16 x float> @llvm.vp.select.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+
+define <16 x float> @vfmacc_vv_v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vv_v16f32_unmasked(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %allones, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vf_v16f32(<16 x float> %va, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vf_v16f32_commute(<16 x float> %va, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %vb, <16 x float> %va, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vf_v16f32_unmasked(<16 x float> %va, float %b, <16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %allones, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vv_v16f32_ta(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vf_v16f32_ta(<16 x float> %va, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmacc_vf_v16f32_commute_ta(<16 x float> %va, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %vb, <16 x float> %va, <16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+declare <2 x double> @llvm.vp.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, <2 x i1>, i32)
+declare <2 x double> @llvm.vp.fneg.v2f64(<2 x double>, <2 x i1>, i32)
+declare <2 x double> @llvm.vp.merge.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+declare <2 x double> @llvm.vp.select.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+
+define <2 x double> @vfmacc_vv_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vv_v2f64_unmasked(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %allones, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vf_v2f64(<2 x double> %va, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %va, <2 x double> %vb, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vf_v2f64_commute(<2 x double> %va, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %vb, <2 x double> %va, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vf_v2f64_unmasked(<2 x double> %va, double %b, <2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %va, <2 x double> %vb, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %allones, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vv_v2f64_ta(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vf_v2f64_ta(<2 x double> %va, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %va, <2 x double> %vb, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmacc_vf_v2f64_commute_ta(<2 x double> %va, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %vb, <2 x double> %va, <2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+declare <4 x double> @llvm.vp.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, <4 x i1>, i32)
+declare <4 x double> @llvm.vp.fneg.v4f64(<4 x double>, <4 x i1>, i32)
+declare <4 x double> @llvm.vp.merge.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+declare <4 x double> @llvm.vp.select.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+
+define <4 x double> @vfmacc_vv_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vv_v4f64_unmasked(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %allones, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vf_v4f64(<4 x double> %va, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %va, <4 x double> %vb, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vf_v4f64_commute(<4 x double> %va, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %vb, <4 x double> %va, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vf_v4f64_unmasked(<4 x double> %va, double %b, <4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %va, <4 x double> %vb, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %allones, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vv_v4f64_ta(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vf_v4f64_ta(<4 x double> %va, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %va, <4 x double> %vb, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmacc_vf_v4f64_commute_ta(<4 x double> %va, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %vb, <4 x double> %va, <4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+declare <8 x double> @llvm.vp.fma.v8f64(<8 x double>, <8 x double>, <8 x double>, <8 x i1>, i32)
+declare <8 x double> @llvm.vp.fneg.v8f64(<8 x double>, <8 x i1>, i32)
+declare <8 x double> @llvm.vp.merge.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+declare <8 x double> @llvm.vp.select.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+
+define <8 x double> @vfmacc_vv_v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vv_v8f64_unmasked(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %allones, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vf_v8f64(<8 x double> %va, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %va, <8 x double> %vb, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vf_v8f64_commute(<8 x double> %va, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %vb, <8 x double> %va, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vf_v8f64_unmasked(<8 x double> %va, double %b, <8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %va, <8 x double> %vb, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %allones, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vv_v8f64_ta(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_v8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vf_v8f64_ta(<8 x double> %va, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %va, <8 x double> %vb, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmacc_vf_v8f64_commute_ta(<8 x double> %va, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_v8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %vb, <8 x double> %va, <8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmsac-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmsac-vp.ll
new file mode 100644
index 0000000000000..3e1f3e720360e
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfmsac-vp.ll
@@ -0,0 +1,1625 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <2 x half> @llvm.vp.fma.v2f16(<2 x half>, <2 x half>, <2 x half>, <2 x i1>, i32)
+declare <2 x half> @llvm.vp.fneg.v2f16(<2 x half>, <2 x i1>, i32)
+declare <2 x half> @llvm.vp.merge.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+declare <2 x half> @llvm.vp.select.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+
+define <2 x half> @vfmsac_vv_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vv_v2f16_unmasked(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %allones, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vf_v2f16(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %vb, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vf_v2f16_commute(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %vb, <2 x half> %a, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vf_v2f16_unmasked(<2 x half> %a, half %b, <2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %vb, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %allones, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vv_v2f16_ta(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vf_v2f16_ta(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %a, <2 x half> %vb, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfmsac_vf_v2f16_commute_ta(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %vb, <2 x half> %a, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+declare <4 x half> @llvm.vp.fma.v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x i1>, i32)
+declare <4 x half> @llvm.vp.fneg.v4f16(<4 x half>, <4 x i1>, i32)
+declare <4 x half> @llvm.vp.merge.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+declare <4 x half> @llvm.vp.select.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+
+define <4 x half> @vfmsac_vv_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vv_v4f16_unmasked(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %allones, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vf_v4f16(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %vb, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vf_v4f16_commute(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %vb, <4 x half> %a, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vf_v4f16_unmasked(<4 x half> %a, half %b, <4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %vb, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %allones, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vv_v4f16_ta(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vf_v4f16_ta(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %a, <4 x half> %vb, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfmsac_vf_v4f16_commute_ta(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %vb, <4 x half> %a, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+declare <8 x half> @llvm.vp.fma.v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x i1>, i32)
+declare <8 x half> @llvm.vp.fneg.v8f16(<8 x half>, <8 x i1>, i32)
+declare <8 x half> @llvm.vp.merge.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+declare <8 x half> @llvm.vp.select.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+
+define <8 x half> @vfmsac_vv_v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vv_v8f16_unmasked(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %allones, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vf_v8f16(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %vb, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vf_v8f16_commute(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %vb, <8 x half> %a, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vf_v8f16_unmasked(<8 x half> %a, half %b, <8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %vb, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %allones, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vv_v8f16_ta(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vf_v8f16_ta(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %a, <8 x half> %vb, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfmsac_vf_v8f16_commute_ta(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %vb, <8 x half> %a, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+declare <16 x half> @llvm.vp.fma.v16f16(<16 x half>, <16 x half>, <16 x half>, <16 x i1>, i32)
+declare <16 x half> @llvm.vp.fneg.v16f16(<16 x half>, <16 x i1>, i32)
+declare <16 x half> @llvm.vp.merge.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+declare <16 x half> @llvm.vp.select.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+
+define <16 x half> @vfmsac_vv_v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vv_v16f16_unmasked(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %allones, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vf_v16f16(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %vb, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vf_v16f16_commute(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %vb, <16 x half> %a, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vf_v16f16_unmasked(<16 x half> %a, half %b, <16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %vb, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %allones, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vv_v16f16_ta(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vf_v16f16_ta(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %a, <16 x half> %vb, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfmsac_vf_v16f16_commute_ta(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %vb, <16 x half> %a, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+declare <32 x half> @llvm.vp.fma.v32f16(<32 x half>, <32 x half>, <32 x half>, <32 x i1>, i32)
+declare <32 x half> @llvm.vp.fneg.v32f16(<32 x half>, <32 x i1>, i32)
+declare <32 x half> @llvm.vp.merge.v32f16(<32 x i1>, <32 x half>, <32 x half>, i32)
+declare <32 x half> @llvm.vp.select.v32f16(<32 x i1>, <32 x half>, <32 x half>, i32)
+
+define <32 x half> @vfmsac_vv_v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vv_v32f16_unmasked(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %allones, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vf_v32f16(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %vb, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vf_v32f16_commute(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %vb, <32 x half> %a, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vf_v32f16_unmasked(<32 x half> %a, half %b, <32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %vb, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %allones, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vv_v32f16_ta(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vf_v32f16_ta(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %a, <32 x half> %vb, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfmsac_vf_v32f16_commute_ta(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %vb, <32 x half> %a, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+declare <2 x float> @llvm.vp.fma.v2f32(<2 x float>, <2 x float>, <2 x float>, <2 x i1>, i32)
+declare <2 x float> @llvm.vp.fneg.v2f32(<2 x float>, <2 x i1>, i32)
+declare <2 x float> @llvm.vp.merge.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+declare <2 x float> @llvm.vp.select.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+
+define <2 x float> @vfmsac_vv_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vv_v2f32_unmasked(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %allones, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vf_v2f32(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %vb, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vf_v2f32_commute(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %vb, <2 x float> %a, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vf_v2f32_unmasked(<2 x float> %a, float %b, <2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %vb, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %allones, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vv_v2f32_ta(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vf_v2f32_ta(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %a, <2 x float> %vb, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfmsac_vf_v2f32_commute_ta(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %vb, <2 x float> %a, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+declare <4 x float> @llvm.vp.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, <4 x i1>, i32)
+declare <4 x float> @llvm.vp.fneg.v4f32(<4 x float>, <4 x i1>, i32)
+declare <4 x float> @llvm.vp.merge.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+declare <4 x float> @llvm.vp.select.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+
+define <4 x float> @vfmsac_vv_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vv_v4f32_unmasked(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %allones, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vf_v4f32(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %vb, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vf_v4f32_commute(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %vb, <4 x float> %a, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vf_v4f32_unmasked(<4 x float> %a, float %b, <4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %vb, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %allones, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vv_v4f32_ta(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vf_v4f32_ta(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a, <4 x float> %vb, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfmsac_vf_v4f32_commute_ta(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %vb, <4 x float> %a, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+declare <8 x float> @llvm.vp.fma.v8f32(<8 x float>, <8 x float>, <8 x float>, <8 x i1>, i32)
+declare <8 x float> @llvm.vp.fneg.v8f32(<8 x float>, <8 x i1>, i32)
+declare <8 x float> @llvm.vp.merge.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+declare <8 x float> @llvm.vp.select.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+
+define <8 x float> @vfmsac_vv_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vv_v8f32_unmasked(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %allones, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vf_v8f32(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %vb, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vf_v8f32_commute(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %vb, <8 x float> %a, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vf_v8f32_unmasked(<8 x float> %a, float %b, <8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %vb, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %allones, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vv_v8f32_ta(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vf_v8f32_ta(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %a, <8 x float> %vb, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfmsac_vf_v8f32_commute_ta(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %vb, <8 x float> %a, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+declare <16 x float> @llvm.vp.fma.v16f32(<16 x float>, <16 x float>, <16 x float>, <16 x i1>, i32)
+declare <16 x float> @llvm.vp.fneg.v16f32(<16 x float>, <16 x i1>, i32)
+declare <16 x float> @llvm.vp.merge.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+declare <16 x float> @llvm.vp.select.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+
+define <16 x float> @vfmsac_vv_v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vv_v16f32_unmasked(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %allones, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vf_v16f32(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %vb, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vf_v16f32_commute(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %vb, <16 x float> %a, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vf_v16f32_unmasked(<16 x float> %a, float %b, <16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %vb, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %allones, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vv_v16f32_ta(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vf_v16f32_ta(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %a, <16 x float> %vb, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfmsac_vf_v16f32_commute_ta(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %vb, <16 x float> %a, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+declare <2 x double> @llvm.vp.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, <2 x i1>, i32)
+declare <2 x double> @llvm.vp.fneg.v2f64(<2 x double>, <2 x i1>, i32)
+declare <2 x double> @llvm.vp.merge.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+declare <2 x double> @llvm.vp.select.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+
+define <2 x double> @vfmsac_vv_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vv_v2f64_unmasked(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %allones, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vf_v2f64(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %vb, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vf_v2f64_commute(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %vb, <2 x double> %a, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vf_v2f64_unmasked(<2 x double> %a, double %b, <2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %vb, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %allones, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vv_v2f64_ta(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vf_v2f64_ta(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %a, <2 x double> %vb, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfmsac_vf_v2f64_commute_ta(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %vb, <2 x double> %a, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+declare <4 x double> @llvm.vp.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, <4 x i1>, i32)
+declare <4 x double> @llvm.vp.fneg.v4f64(<4 x double>, <4 x i1>, i32)
+declare <4 x double> @llvm.vp.merge.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+declare <4 x double> @llvm.vp.select.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+
+define <4 x double> @vfmsac_vv_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vv_v4f64_unmasked(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %allones, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vf_v4f64(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %vb, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vf_v4f64_commute(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %vb, <4 x double> %a, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vf_v4f64_unmasked(<4 x double> %a, double %b, <4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %vb, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %allones, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vv_v4f64_ta(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vf_v4f64_ta(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %a, <4 x double> %vb, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfmsac_vf_v4f64_commute_ta(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %vb, <4 x double> %a, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+declare <8 x double> @llvm.vp.fma.v8f64(<8 x double>, <8 x double>, <8 x double>, <8 x i1>, i32)
+declare <8 x double> @llvm.vp.fneg.v8f64(<8 x double>, <8 x i1>, i32)
+declare <8 x double> @llvm.vp.merge.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+declare <8 x double> @llvm.vp.select.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+
+define <8 x double> @vfmsac_vv_v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vv_v8f64_unmasked(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %allones, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vf_v8f64(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %vb, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vf_v8f64_commute(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %vb, <8 x double> %a, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vf_v8f64_unmasked(<8 x double> %a, double %b, <8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %vb, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %allones, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vv_v8f64_ta(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vv_v8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vf_v8f64_ta(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %a, <8 x double> %vb, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfmsac_vf_v8f64_commute_ta(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmsac_vf_v8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %vb, <8 x double> %a, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmacc-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmacc-vp.ll
new file mode 100644
index 0000000000000..fecb828c87833
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmacc-vp.ll
@@ -0,0 +1,1721 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <2 x half> @llvm.vp.fma.v2f16(<2 x half>, <2 x half>, <2 x half>, <2 x i1>, i32)
+declare <2 x half> @llvm.vp.fneg.v2f16(<2 x half>, <2 x i1>, i32)
+declare <2 x half> @llvm.vp.merge.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+declare <2 x half> @llvm.vp.select.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+
+define <2 x half> @vfnmacc_vv_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %nega, <2 x half> %b, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vv_v2f16_unmasked(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %nega, <2 x half> %b, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %allones, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vf_v2f16(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %nega, <2 x half> %vb, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vf_v2f16_commute(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %vb, <2 x half> %nega, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vf_v2f16_unmasked(<2 x half> %a, half %b, <2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %nega, <2 x half> %vb, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.merge.v2f16(<2 x i1> %allones, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vv_v2f16_ta(<2 x half> %a, <2 x half> %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %nega, <2 x half> %b, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vf_v2f16_ta(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %nega, <2 x half> %vb, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+define <2 x half> @vfnmacc_vf_v2f16_commute_ta(<2 x half> %a, half %b, <2 x half> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x half> poison, half %b, i32 0
+ %vb = shufflevector <2 x half> %elt.head, <2 x half> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x half> @llvm.vp.fma.v2f16(<2 x half> %vb, <2 x half> %nega, <2 x half> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %m, <2 x half> %v, <2 x half> %c, i32 %evl)
+ ret <2 x half> %u
+}
+
+declare <4 x half> @llvm.vp.fma.v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x i1>, i32)
+declare <4 x half> @llvm.vp.fneg.v4f16(<4 x half>, <4 x i1>, i32)
+declare <4 x half> @llvm.vp.merge.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+declare <4 x half> @llvm.vp.select.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+
+define <4 x half> @vfnmacc_vv_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %nega, <4 x half> %b, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vv_v4f16_unmasked(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %nega, <4 x half> %b, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %allones, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vf_v4f16(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %nega, <4 x half> %vb, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vf_v4f16_commute(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %vb, <4 x half> %nega, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vf_v4f16_unmasked(<4 x half> %a, half %b, <4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %nega, <4 x half> %vb, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.merge.v4f16(<4 x i1> %allones, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vv_v4f16_ta(<4 x half> %a, <4 x half> %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %nega, <4 x half> %b, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vf_v4f16_ta(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %nega, <4 x half> %vb, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+define <4 x half> @vfnmacc_vf_v4f16_commute_ta(<4 x half> %a, half %b, <4 x half> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x half> poison, half %b, i32 0
+ %vb = shufflevector <4 x half> %elt.head, <4 x half> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x half> @llvm.vp.fma.v4f16(<4 x half> %vb, <4 x half> %nega, <4 x half> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %m, <4 x half> %v, <4 x half> %c, i32 %evl)
+ ret <4 x half> %u
+}
+
+declare <8 x half> @llvm.vp.fma.v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x i1>, i32)
+declare <8 x half> @llvm.vp.fneg.v8f16(<8 x half>, <8 x i1>, i32)
+declare <8 x half> @llvm.vp.merge.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+declare <8 x half> @llvm.vp.select.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+
+define <8 x half> @vfnmacc_vv_v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %nega, <8 x half> %b, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vv_v8f16_unmasked(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %nega, <8 x half> %b, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %allones, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vf_v8f16(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %nega, <8 x half> %vb, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vf_v8f16_commute(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %vb, <8 x half> %nega, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vf_v8f16_unmasked(<8 x half> %a, half %b, <8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %nega, <8 x half> %vb, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.merge.v8f16(<8 x i1> %allones, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vv_v8f16_ta(<8 x half> %a, <8 x half> %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %nega, <8 x half> %b, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vf_v8f16_ta(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %nega, <8 x half> %vb, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+define <8 x half> @vfnmacc_vf_v8f16_commute_ta(<8 x half> %a, half %b, <8 x half> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x half> poison, half %b, i32 0
+ %vb = shufflevector <8 x half> %elt.head, <8 x half> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x half> @llvm.vp.fma.v8f16(<8 x half> %vb, <8 x half> %nega, <8 x half> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %m, <8 x half> %v, <8 x half> %c, i32 %evl)
+ ret <8 x half> %u
+}
+
+declare <16 x half> @llvm.vp.fma.v16f16(<16 x half>, <16 x half>, <16 x half>, <16 x i1>, i32)
+declare <16 x half> @llvm.vp.fneg.v16f16(<16 x half>, <16 x i1>, i32)
+declare <16 x half> @llvm.vp.merge.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+declare <16 x half> @llvm.vp.select.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+
+define <16 x half> @vfnmacc_vv_v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %nega, <16 x half> %b, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vv_v16f16_unmasked(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %nega, <16 x half> %b, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %allones, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vf_v16f16(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %nega, <16 x half> %vb, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vf_v16f16_commute(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %vb, <16 x half> %nega, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vf_v16f16_unmasked(<16 x half> %a, half %b, <16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %nega, <16 x half> %vb, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.merge.v16f16(<16 x i1> %allones, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vv_v16f16_ta(<16 x half> %a, <16 x half> %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %nega, <16 x half> %b, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vf_v16f16_ta(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %nega, <16 x half> %vb, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+define <16 x half> @vfnmacc_vf_v16f16_commute_ta(<16 x half> %a, half %b, <16 x half> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x half> poison, half %b, i32 0
+ %vb = shufflevector <16 x half> %elt.head, <16 x half> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x half> @llvm.vp.fma.v16f16(<16 x half> %vb, <16 x half> %nega, <16 x half> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %m, <16 x half> %v, <16 x half> %c, i32 %evl)
+ ret <16 x half> %u
+}
+
+declare <32 x half> @llvm.vp.fma.v32f16(<32 x half>, <32 x half>, <32 x half>, <32 x i1>, i32)
+declare <32 x half> @llvm.vp.fneg.v32f16(<32 x half>, <32 x i1>, i32)
+declare <32 x half> @llvm.vp.merge.v32f16(<32 x i1>, <32 x half>, <32 x half>, i32)
+declare <32 x half> @llvm.vp.select.v32f16(<32 x i1>, <32 x half>, <32 x half>, i32)
+
+define <32 x half> @vfnmacc_vv_v32f16(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %nega, <32 x half> %b, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vv_v32f16_unmasked(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %nega, <32 x half> %b, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %allones, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vf_v32f16(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %nega, <32 x half> %vb, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vf_v32f16_commute(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %vb, <32 x half> %nega, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vf_v32f16_unmasked(<32 x half> %a, half %b, <32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %nega, <32 x half> %vb, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.merge.v32f16(<32 x i1> %allones, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vv_v32f16_ta(<32 x half> %a, <32 x half> %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %nega, <32 x half> %b, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vf_v32f16_ta(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %nega, <32 x half> %vb, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+define <32 x half> @vfnmacc_vf_v32f16_commute_ta(<32 x half> %a, half %b, <32 x half> %c, <32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <32 x half> poison, half %b, i32 0
+ %vb = shufflevector <32 x half> %elt.head, <32 x half> poison, <32 x i32> zeroinitializer
+ %splat = insertelement <32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <32 x i1> %splat, <32 x i1> poison, <32 x i32> zeroinitializer
+ %nega = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %a, <32 x i1> %allones, i32 %evl)
+ %negc = call <32 x half> @llvm.vp.fneg.v32f16(<32 x half> %c, <32 x i1> %allones, i32 %evl)
+ %v = call <32 x half> @llvm.vp.fma.v32f16(<32 x half> %vb, <32 x half> %nega, <32 x half> %negc, <32 x i1> %allones, i32 %evl)
+ %u = call <32 x half> @llvm.vp.select.v32f16(<32 x i1> %m, <32 x half> %v, <32 x half> %c, i32 %evl)
+ ret <32 x half> %u
+}
+
+declare <2 x float> @llvm.vp.fma.v2f32(<2 x float>, <2 x float>, <2 x float>, <2 x i1>, i32)
+declare <2 x float> @llvm.vp.fneg.v2f32(<2 x float>, <2 x i1>, i32)
+declare <2 x float> @llvm.vp.merge.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+declare <2 x float> @llvm.vp.select.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+
+define <2 x float> @vfnmacc_vv_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %nega, <2 x float> %b, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vv_v2f32_unmasked(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %nega, <2 x float> %b, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %allones, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vf_v2f32(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %nega, <2 x float> %vb, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vf_v2f32_commute(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %vb, <2 x float> %nega, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vf_v2f32_unmasked(<2 x float> %a, float %b, <2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %nega, <2 x float> %vb, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.merge.v2f32(<2 x i1> %allones, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vv_v2f32_ta(<2 x float> %a, <2 x float> %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %nega, <2 x float> %b, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vf_v2f32_ta(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %nega, <2 x float> %vb, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+define <2 x float> @vfnmacc_vf_v2f32_commute_ta(<2 x float> %a, float %b, <2 x float> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x float> poison, float %b, i32 0
+ %vb = shufflevector <2 x float> %elt.head, <2 x float> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x float> @llvm.vp.fma.v2f32(<2 x float> %vb, <2 x float> %nega, <2 x float> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %m, <2 x float> %v, <2 x float> %c, i32 %evl)
+ ret <2 x float> %u
+}
+
+declare <4 x float> @llvm.vp.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, <4 x i1>, i32)
+declare <4 x float> @llvm.vp.fneg.v4f32(<4 x float>, <4 x i1>, i32)
+declare <4 x float> @llvm.vp.merge.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+declare <4 x float> @llvm.vp.select.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+
+define <4 x float> @vfnmacc_vv_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %nega, <4 x float> %b, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vv_v4f32_unmasked(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %nega, <4 x float> %b, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %allones, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vf_v4f32(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %nega, <4 x float> %vb, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vf_v4f32_commute(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %vb, <4 x float> %nega, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vf_v4f32_unmasked(<4 x float> %a, float %b, <4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %nega, <4 x float> %vb, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.merge.v4f32(<4 x i1> %allones, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vv_v4f32_ta(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %nega, <4 x float> %b, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vf_v4f32_ta(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %nega, <4 x float> %vb, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+define <4 x float> @vfnmacc_vf_v4f32_commute_ta(<4 x float> %a, float %b, <4 x float> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x float> poison, float %b, i32 0
+ %vb = shufflevector <4 x float> %elt.head, <4 x float> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %vb, <4 x float> %nega, <4 x float> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %m, <4 x float> %v, <4 x float> %c, i32 %evl)
+ ret <4 x float> %u
+}
+
+declare <8 x float> @llvm.vp.fma.v8f32(<8 x float>, <8 x float>, <8 x float>, <8 x i1>, i32)
+declare <8 x float> @llvm.vp.fneg.v8f32(<8 x float>, <8 x i1>, i32)
+declare <8 x float> @llvm.vp.merge.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+declare <8 x float> @llvm.vp.select.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+
+define <8 x float> @vfnmacc_vv_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %nega, <8 x float> %b, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vv_v8f32_unmasked(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %nega, <8 x float> %b, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %allones, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vf_v8f32(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %nega, <8 x float> %vb, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vf_v8f32_commute(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %vb, <8 x float> %nega, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vf_v8f32_unmasked(<8 x float> %a, float %b, <8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %nega, <8 x float> %vb, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.merge.v8f32(<8 x i1> %allones, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vv_v8f32_ta(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %nega, <8 x float> %b, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vf_v8f32_ta(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %nega, <8 x float> %vb, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+define <8 x float> @vfnmacc_vf_v8f32_commute_ta(<8 x float> %a, float %b, <8 x float> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x float> poison, float %b, i32 0
+ %vb = shufflevector <8 x float> %elt.head, <8 x float> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x float> @llvm.vp.fma.v8f32(<8 x float> %vb, <8 x float> %nega, <8 x float> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %m, <8 x float> %v, <8 x float> %c, i32 %evl)
+ ret <8 x float> %u
+}
+
+declare <16 x float> @llvm.vp.fma.v16f32(<16 x float>, <16 x float>, <16 x float>, <16 x i1>, i32)
+declare <16 x float> @llvm.vp.fneg.v16f32(<16 x float>, <16 x i1>, i32)
+declare <16 x float> @llvm.vp.merge.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+declare <16 x float> @llvm.vp.select.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+
+define <16 x float> @vfnmacc_vv_v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %nega, <16 x float> %b, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vv_v16f32_unmasked(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %nega, <16 x float> %b, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %allones, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vf_v16f32(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %nega, <16 x float> %vb, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vf_v16f32_commute(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %vb, <16 x float> %nega, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vf_v16f32_unmasked(<16 x float> %a, float %b, <16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %nega, <16 x float> %vb, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.merge.v16f32(<16 x i1> %allones, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vv_v16f32_ta(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %nega, <16 x float> %b, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vf_v16f32_ta(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %nega, <16 x float> %vb, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+define <16 x float> @vfnmacc_vf_v16f32_commute_ta(<16 x float> %a, float %b, <16 x float> %c, <16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <16 x float> poison, float %b, i32 0
+ %vb = shufflevector <16 x float> %elt.head, <16 x float> poison, <16 x i32> zeroinitializer
+ %splat = insertelement <16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <16 x i1> %splat, <16 x i1> poison, <16 x i32> zeroinitializer
+ %nega = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %a, <16 x i1> %allones, i32 %evl)
+ %negc = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> %c, <16 x i1> %allones, i32 %evl)
+ %v = call <16 x float> @llvm.vp.fma.v16f32(<16 x float> %vb, <16 x float> %nega, <16 x float> %negc, <16 x i1> %allones, i32 %evl)
+ %u = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %m, <16 x float> %v, <16 x float> %c, i32 %evl)
+ ret <16 x float> %u
+}
+
+declare <2 x double> @llvm.vp.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, <2 x i1>, i32)
+declare <2 x double> @llvm.vp.fneg.v2f64(<2 x double>, <2 x i1>, i32)
+declare <2 x double> @llvm.vp.merge.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+declare <2 x double> @llvm.vp.select.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+
+define <2 x double> @vfnmacc_vv_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %nega, <2 x double> %b, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vv_v2f64_unmasked(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %nega, <2 x double> %b, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %allones, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vf_v2f64(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %nega, <2 x double> %vb, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vf_v2f64_commute(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %vb, <2 x double> %nega, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vf_v2f64_unmasked(<2 x double> %a, double %b, <2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %nega, <2 x double> %vb, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.merge.v2f64(<2 x i1> %allones, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vv_v2f64_ta(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %nega, <2 x double> %b, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vf_v2f64_ta(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %nega, <2 x double> %vb, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+define <2 x double> @vfnmacc_vf_v2f64_commute_ta(<2 x double> %a, double %b, <2 x double> %c, <2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <2 x double> poison, double %b, i32 0
+ %vb = shufflevector <2 x double> %elt.head, <2 x double> poison, <2 x i32> zeroinitializer
+ %splat = insertelement <2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <2 x i1> %splat, <2 x i1> poison, <2 x i32> zeroinitializer
+ %nega = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %a, <2 x i1> %allones, i32 %evl)
+ %negc = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> %c, <2 x i1> %allones, i32 %evl)
+ %v = call <2 x double> @llvm.vp.fma.v2f64(<2 x double> %vb, <2 x double> %nega, <2 x double> %negc, <2 x i1> %allones, i32 %evl)
+ %u = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %m, <2 x double> %v, <2 x double> %c, i32 %evl)
+ ret <2 x double> %u
+}
+
+declare <4 x double> @llvm.vp.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, <4 x i1>, i32)
+declare <4 x double> @llvm.vp.fneg.v4f64(<4 x double>, <4 x i1>, i32)
+declare <4 x double> @llvm.vp.merge.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+declare <4 x double> @llvm.vp.select.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+
+define <4 x double> @vfnmacc_vv_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %nega, <4 x double> %b, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vv_v4f64_unmasked(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %nega, <4 x double> %b, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %allones, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vf_v4f64(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %nega, <4 x double> %vb, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vf_v4f64_commute(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %vb, <4 x double> %nega, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vf_v4f64_unmasked(<4 x double> %a, double %b, <4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %nega, <4 x double> %vb, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.merge.v4f64(<4 x i1> %allones, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vv_v4f64_ta(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %nega, <4 x double> %b, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vf_v4f64_ta(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %nega, <4 x double> %vb, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+define <4 x double> @vfnmacc_vf_v4f64_commute_ta(<4 x double> %a, double %b, <4 x double> %c, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <4 x double> poison, double %b, i32 0
+ %vb = shufflevector <4 x double> %elt.head, <4 x double> poison, <4 x i32> zeroinitializer
+ %splat = insertelement <4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <4 x i1> %splat, <4 x i1> poison, <4 x i32> zeroinitializer
+ %nega = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %a, <4 x i1> %allones, i32 %evl)
+ %negc = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> %c, <4 x i1> %allones, i32 %evl)
+ %v = call <4 x double> @llvm.vp.fma.v4f64(<4 x double> %vb, <4 x double> %nega, <4 x double> %negc, <4 x i1> %allones, i32 %evl)
+ %u = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %m, <4 x double> %v, <4 x double> %c, i32 %evl)
+ ret <4 x double> %u
+}
+
+declare <8 x double> @llvm.vp.fma.v8f64(<8 x double>, <8 x double>, <8 x double>, <8 x i1>, i32)
+declare <8 x double> @llvm.vp.fneg.v8f64(<8 x double>, <8 x i1>, i32)
+declare <8 x double> @llvm.vp.merge.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+declare <8 x double> @llvm.vp.select.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+
+define <8 x double> @vfnmacc_vv_v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %nega, <8 x double> %b, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vv_v8f64_unmasked(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %nega, <8 x double> %b, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %allones, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vf_v8f64(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %nega, <8 x double> %vb, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vf_v8f64_commute(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %vb, <8 x double> %nega, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vf_v8f64_unmasked(<8 x double> %a, double %b, <8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %nega, <8 x double> %vb, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.merge.v8f64(<8 x i1> %allones, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vv_v8f64_ta(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_v8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %nega, <8 x double> %b, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vf_v8f64_ta(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %nega, <8 x double> %vb, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
+
+define <8 x double> @vfnmacc_vf_v8f64_commute_ta(<8 x double> %a, double %b, <8 x double> %c, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_v8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <8 x double> poison, double %b, i32 0
+ %vb = shufflevector <8 x double> %elt.head, <8 x double> poison, <8 x i32> zeroinitializer
+ %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+ %nega = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %a, <8 x i1> %allones, i32 %evl)
+ %negc = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> %c, <8 x i1> %allones, i32 %evl)
+ %v = call <8 x double> @llvm.vp.fma.v8f64(<8 x double> %vb, <8 x double> %nega, <8 x double> %negc, <8 x i1> %allones, i32 %evl)
+ %u = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %m, <8 x double> %v, <8 x double> %c, i32 %evl)
+ ret <8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmsac-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmsac-vp.ll
new file mode 100644
index 0000000000000..d3321e1d8c7c2
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfnmsac-vp.ll
@@ -0,0 +1,2039 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+declare <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+
+define <vscale x 1 x half> @vfnmsac_vv_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vv_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_commute(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %nega, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vv_nxv1f16_ta(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_commute_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %nega, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+declare <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+declare <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+
+define <vscale x 2 x half> @vfnmsac_vv_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vv_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_commute(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %nega, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vv_nxv2f16_ta(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_commute_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %nega, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+declare <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+declare <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+
+define <vscale x 4 x half> @vfnmsac_vv_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vv_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_commute(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %nega, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vv_nxv4f16_ta(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_commute_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %nega, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+declare <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+
+define <vscale x 8 x half> @vfnmsac_vv_nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vv_nxv8f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_commute(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %nega, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vv_nxv8f16_ta(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_commute_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %nega, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+declare <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+declare <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+
+define <vscale x 16 x half> @vfnmsac_vv_nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vv_nxv16f16_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_commute(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %nega, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_unmasked(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vv_nxv16f16_ta(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_commute_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %nega, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+declare <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+declare <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+
+define <vscale x 32 x half> @vfnmsac_vv_nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vv_nxv32f16_unmasked(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_commute(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %nega, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_unmasked(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vv_nxv32f16_ta(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_commute_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %nega, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+declare <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+declare <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_commute(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %nega, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_unmasked(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32_ta(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_commute_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %nega, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+declare <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+declare <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_commute(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %nega, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_unmasked(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32_ta(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_commute_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %nega, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+declare <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_commute(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %nega, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_unmasked(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32_ta(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_commute_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %nega, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+declare <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+declare <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_commute(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %nega, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_unmasked(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32_ta(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_commute_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %nega, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+declare <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+declare <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32_unmasked(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_commute(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %nega, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_unmasked(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32_ta(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_commute_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %nega, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+declare <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+declare <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64_unmasked(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_commute(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %nega, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_unmasked(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64_ta(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_commute_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %nega, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+declare <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+declare <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64_unmasked(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_commute(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %nega, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_unmasked(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64_ta(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_commute_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %nega, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+declare <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+declare <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64_unmasked(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_commute(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %nega, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_unmasked(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64_ta(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_commute_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %nega, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+declare <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+declare <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64_unmasked(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_commute(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %nega, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_unmasked(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64_ta(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_commute_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %nega, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfmacc-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfmacc-vp.ll
new file mode 100644
index 0000000000000..bfb53d8ee3162
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfmacc-vp.ll
@@ -0,0 +1,1919 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+declare <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+
+define <vscale x 1 x half> @vfmacc_vv_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vv_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vf_nxv1f16(<vscale x 1 x half> %va, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %va, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vf_nxv1f16_commute(<vscale x 1 x half> %va, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %va, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vf_nxv1f16_unmasked(<vscale x 1 x half> %va, half %b, <vscale x 1 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %va, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vv_nxv1f16_ta(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vf_nxv1f16_ta(<vscale x 1 x half> %va, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %va, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfmacc_vf_nxv1f16_commute_ta(<vscale x 1 x half> %va, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %va, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+declare <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+declare <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+
+define <vscale x 2 x half> @vfmacc_vv_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vv_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vf_nxv2f16(<vscale x 2 x half> %va, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vf_nxv2f16_commute(<vscale x 2 x half> %va, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %va, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vf_nxv2f16_unmasked(<vscale x 2 x half> %va, half %b, <vscale x 2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vv_nxv2f16_ta(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vf_nxv2f16_ta(<vscale x 2 x half> %va, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfmacc_vf_nxv2f16_commute_ta(<vscale x 2 x half> %va, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %va, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+declare <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+declare <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+
+define <vscale x 4 x half> @vfmacc_vv_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vv_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vf_nxv4f16(<vscale x 4 x half> %va, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %va, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vf_nxv4f16_commute(<vscale x 4 x half> %va, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %va, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vf_nxv4f16_unmasked(<vscale x 4 x half> %va, half %b, <vscale x 4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %va, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vv_nxv4f16_ta(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vf_nxv4f16_ta(<vscale x 4 x half> %va, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %va, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfmacc_vf_nxv4f16_commute_ta(<vscale x 4 x half> %va, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %va, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+declare <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+
+define <vscale x 8 x half> @vfmacc_vv_nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vv_nxv8f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vf_nxv8f16(<vscale x 8 x half> %va, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %va, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vf_nxv8f16_commute(<vscale x 8 x half> %va, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %va, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vf_nxv8f16_unmasked(<vscale x 8 x half> %va, half %b, <vscale x 8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %va, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vv_nxv8f16_ta(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vf_nxv8f16_ta(<vscale x 8 x half> %va, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %va, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfmacc_vf_nxv8f16_commute_ta(<vscale x 8 x half> %va, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %va, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+declare <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+declare <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+
+define <vscale x 16 x half> @vfmacc_vv_nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vv_nxv16f16_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vf_nxv16f16(<vscale x 16 x half> %va, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %va, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vf_nxv16f16_commute(<vscale x 16 x half> %va, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %va, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vf_nxv16f16_unmasked(<vscale x 16 x half> %va, half %b, <vscale x 16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %va, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vv_nxv16f16_ta(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vf_nxv16f16_ta(<vscale x 16 x half> %va, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %va, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfmacc_vf_nxv16f16_commute_ta(<vscale x 16 x half> %va, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %va, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+declare <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+declare <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+
+define <vscale x 32 x half> @vfmacc_vv_nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vv_nxv32f16_unmasked(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vf_nxv32f16(<vscale x 32 x half> %va, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %va, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vf_nxv32f16_commute(<vscale x 32 x half> %va, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %va, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vf_nxv32f16_unmasked(<vscale x 32 x half> %va, half %b, <vscale x 32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %va, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vv_nxv32f16_ta(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vf_nxv32f16_ta(<vscale x 32 x half> %va, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %va, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfmacc_vf_nxv32f16_commute_ta(<vscale x 32 x half> %va, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %va, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+declare <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+declare <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+
+define <vscale x 1 x float> @vfmacc_vv_nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vv_nxv1f32_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32(<vscale x 1 x float> %va, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %va, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32_commute(<vscale x 1 x float> %va, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %va, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32_unmasked(<vscale x 1 x float> %va, float %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %va, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vv_nxv1f32_ta(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32_ta(<vscale x 1 x float> %va, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %va, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32_commute_ta(<vscale x 1 x float> %va, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %va, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+declare <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+declare <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+
+define <vscale x 2 x float> @vfmacc_vv_nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vv_nxv2f32_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32(<vscale x 2 x float> %va, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32_commute(<vscale x 2 x float> %va, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %va, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32_unmasked(<vscale x 2 x float> %va, float %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vv_nxv2f32_ta(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32_ta(<vscale x 2 x float> %va, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32_commute_ta(<vscale x 2 x float> %va, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %va, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+declare <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+
+define <vscale x 4 x float> @vfmacc_vv_nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vv_nxv4f32_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32(<vscale x 4 x float> %va, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %va, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32_commute(<vscale x 4 x float> %va, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %va, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32_unmasked(<vscale x 4 x float> %va, float %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %va, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vv_nxv4f32_ta(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32_ta(<vscale x 4 x float> %va, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %va, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32_commute_ta(<vscale x 4 x float> %va, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %va, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+declare <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+declare <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+
+define <vscale x 8 x float> @vfmacc_vv_nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vv_nxv8f32_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32(<vscale x 8 x float> %va, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %va, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32_commute(<vscale x 8 x float> %va, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %va, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32_unmasked(<vscale x 8 x float> %va, float %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %va, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vv_nxv8f32_ta(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32_ta(<vscale x 8 x float> %va, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %va, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32_commute_ta(<vscale x 8 x float> %va, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %va, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+declare <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+declare <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+
+define <vscale x 16 x float> @vfmacc_vv_nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vv_nxv16f32_unmasked(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32(<vscale x 16 x float> %va, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %va, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32_commute(<vscale x 16 x float> %va, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %va, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32_unmasked(<vscale x 16 x float> %va, float %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %va, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vv_nxv16f32_ta(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32_ta(<vscale x 16 x float> %va, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %va, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32_commute_ta(<vscale x 16 x float> %va, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %va, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+declare <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+declare <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64_unmasked(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64(<vscale x 1 x double> %va, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %va, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64_commute(<vscale x 1 x double> %va, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %va, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64_unmasked(<vscale x 1 x double> %va, double %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %va, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64_ta(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64_ta(<vscale x 1 x double> %va, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %va, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64_commute_ta(<vscale x 1 x double> %va, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %va, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+declare <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+declare <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64_unmasked(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64(<vscale x 2 x double> %va, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64_commute(<vscale x 2 x double> %va, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %va, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64_unmasked(<vscale x 2 x double> %va, double %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64_ta(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64_ta(<vscale x 2 x double> %va, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64_commute_ta(<vscale x 2 x double> %va, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %va, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+declare <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+declare <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64_unmasked(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64(<vscale x 4 x double> %va, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %va, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64_commute(<vscale x 4 x double> %va, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %va, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64_unmasked(<vscale x 4 x double> %va, double %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %va, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64_ta(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64_ta(<vscale x 4 x double> %va, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %va, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64_commute_ta(<vscale x 4 x double> %va, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %va, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+declare <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+declare <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64_unmasked(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64(<vscale x 8 x double> %va, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %va, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64_commute(<vscale x 8 x double> %va, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %va, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64_unmasked(<vscale x 8 x double> %va, double %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %va, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64_ta(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vfmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64_ta(<vscale x 8 x double> %va, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %va, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64_commute_ta(<vscale x 8 x double> %va, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %va, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfmsac-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfmsac-vp.ll
new file mode 100644
index 0000000000000..42808658c14c7
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfmsac-vp.ll
@@ -0,0 +1,2039 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+declare <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+
+define <vscale x 1 x half> @vmfsac_vv_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vv_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vf_nxv1f16(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %vb, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vf_nxv1f16_commute(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %a, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vf_nxv1f16_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %vb, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vv_nxv1f16_ta(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vf_nxv1f16_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %vb, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vmfsac_vf_nxv1f16_commute_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %a, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+declare <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+declare <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+
+define <vscale x 2 x half> @vmfsac_vv_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vv_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vf_nxv2f16(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %vb, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vf_nxv2f16_commute(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %a, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vf_nxv2f16_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %vb, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vv_nxv2f16_ta(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vf_nxv2f16_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %vb, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vmfsac_vf_nxv2f16_commute_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %a, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+declare <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+declare <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+
+define <vscale x 4 x half> @vmfsac_vv_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vv_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vf_nxv4f16(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %vb, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vf_nxv4f16_commute(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %a, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vf_nxv4f16_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %vb, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vv_nxv4f16_ta(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vf_nxv4f16_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %vb, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vmfsac_vf_nxv4f16_commute_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %a, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+declare <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+
+define <vscale x 8 x half> @vmfsac_vv_nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vv_nxv8f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vf_nxv8f16(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %vb, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vf_nxv8f16_commute(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %a, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vf_nxv8f16_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %vb, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vv_nxv8f16_ta(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vf_nxv8f16_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %vb, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vmfsac_vf_nxv8f16_commute_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %a, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+declare <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+declare <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+
+define <vscale x 16 x half> @vmfsac_vv_nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vv_nxv16f16_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vf_nxv16f16(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %vb, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vf_nxv16f16_commute(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %a, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vf_nxv16f16_unmasked(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %vb, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vv_nxv16f16_ta(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vf_nxv16f16_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %vb, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vmfsac_vf_nxv16f16_commute_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %a, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+declare <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+declare <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+
+define <vscale x 32 x half> @vmfsac_vv_nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vv_nxv32f16_unmasked(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vf_nxv32f16(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %vb, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vf_nxv32f16_commute(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %a, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vf_nxv32f16_unmasked(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %vb, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vv_nxv32f16_ta(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vf_nxv32f16_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %vb, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vmfsac_vf_nxv32f16_commute_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %a, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+declare <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+declare <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+
+define <vscale x 1 x float> @vmfsac_vv_nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vv_nxv1f32_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %vb, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32_commute(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %a, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32_unmasked(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %vb, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vv_nxv1f32_ta(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %vb, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32_commute_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %a, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+declare <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+declare <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+
+define <vscale x 2 x float> @vmfsac_vv_nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vv_nxv2f32_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %vb, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32_commute(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %a, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32_unmasked(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %vb, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vv_nxv2f32_ta(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %vb, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32_commute_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %a, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+declare <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+
+define <vscale x 4 x float> @vmfsac_vv_nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vv_nxv4f32_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %vb, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32_commute(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %a, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32_unmasked(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %vb, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vv_nxv4f32_ta(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %vb, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32_commute_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %a, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+declare <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+declare <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+
+define <vscale x 8 x float> @vmfsac_vv_nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vv_nxv8f32_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %vb, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32_commute(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %a, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32_unmasked(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %vb, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vv_nxv8f32_ta(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %vb, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32_commute_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %a, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+declare <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+declare <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+
+define <vscale x 16 x float> @vmfsac_vv_nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vv_nxv16f32_unmasked(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vf_nxv16f32(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %vb, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vf_nxv16f32_commute(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %a, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vf_nxv16f32_unmasked(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %vb, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vv_nxv16f32_ta(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vf_nxv16f32_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %vb, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vmfsac_vf_nxv16f32_commute_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %a, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+declare <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+declare <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+
+define <vscale x 1 x double> @vmfsac_vv_nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vv_nxv1f64_unmasked(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vf_nxv1f64(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %vb, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vf_nxv1f64_commute(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %a, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vf_nxv1f64_unmasked(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %vb, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vv_nxv1f64_ta(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vf_nxv1f64_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %vb, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vmfsac_vf_nxv1f64_commute_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %a, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+declare <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+declare <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+define <vscale x 2 x double> @vmfsac_vv_nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vv_nxv2f64_unmasked(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vf_nxv2f64(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %vb, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vf_nxv2f64_commute(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %a, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vf_nxv2f64_unmasked(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %vb, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vv_nxv2f64_ta(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vf_nxv2f64_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %vb, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vmfsac_vf_nxv2f64_commute_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %a, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+declare <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+declare <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+
+define <vscale x 4 x double> @vmfsac_vv_nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vv_nxv4f64_unmasked(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vf_nxv4f64(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %vb, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vf_nxv4f64_commute(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %a, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vf_nxv4f64_unmasked(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %vb, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vv_nxv4f64_ta(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vf_nxv4f64_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %vb, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vmfsac_vf_nxv4f64_commute_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %a, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+declare <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+declare <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+
+define <vscale x 8 x double> @vmfsac_vv_nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vv_nxv8f64_unmasked(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vf_nxv8f64(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %vb, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vf_nxv8f64_commute(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %a, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vf_nxv8f64_unmasked(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %vb, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vv_nxv8f64_ta(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vfmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vf_nxv8f64_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %vb, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vmfsac_vf_nxv8f64_commute_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %a, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfnmacc-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfnmacc-vp.ll
new file mode 100644
index 0000000000000..f098982d308f8
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfnmacc-vp.ll
@@ -0,0 +1,2159 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+declare <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+
+define <vscale x 1 x half> @vfnmacc_vv_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vv_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vf_nxv1f16(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vf_nxv1f16_commute(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %nega, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vf_nxv1f16_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vv_nxv1f16_ta(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vf_nxv1f16_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmacc_vf_nxv1f16_commute_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %nega, <vscale x 1 x half> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+declare <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+declare <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+
+define <vscale x 2 x half> @vfnmacc_vv_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vv_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vf_nxv2f16(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vf_nxv2f16_commute(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %nega, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vf_nxv2f16_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vv_nxv2f16_ta(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vf_nxv2f16_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmacc_vf_nxv2f16_commute_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %nega, <vscale x 2 x half> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+declare <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+declare <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+
+define <vscale x 4 x half> @vfnmacc_vv_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vv_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vf_nxv4f16(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vf_nxv4f16_commute(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %nega, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vf_nxv4f16_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vv_nxv4f16_ta(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vf_nxv4f16_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmacc_vf_nxv4f16_commute_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %nega, <vscale x 4 x half> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+declare <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+
+define <vscale x 8 x half> @vfnmacc_vv_nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vv_nxv8f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vf_nxv8f16(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vf_nxv8f16_commute(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %nega, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vf_nxv8f16_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vv_nxv8f16_ta(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vf_nxv8f16_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmacc_vf_nxv8f16_commute_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %nega, <vscale x 8 x half> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+declare <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+declare <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+
+define <vscale x 16 x half> @vfnmacc_vv_nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vv_nxv16f16_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vf_nxv16f16(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vf_nxv16f16_commute(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %nega, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vf_nxv16f16_unmasked(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vv_nxv16f16_ta(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vf_nxv16f16_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmacc_vf_nxv16f16_commute_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %nega, <vscale x 16 x half> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+declare <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+declare <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+
+define <vscale x 32 x half> @vfnmacc_vv_nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vv_nxv32f16_unmasked(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vf_nxv32f16(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vf_nxv32f16_commute(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %nega, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vf_nxv32f16_unmasked(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vv_nxv32f16_ta(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vf_nxv32f16_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmacc_vf_nxv32f16_commute_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %nega, <vscale x 32 x half> %negc, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+declare <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+declare <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+
+define <vscale x 1 x float> @vfnmacc_vv_nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vv_nxv1f32_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32_commute(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %nega, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32_unmasked(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vv_nxv1f32_ta(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32_commute_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %nega, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+declare <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+declare <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+
+define <vscale x 2 x float> @vfnmacc_vv_nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vv_nxv2f32_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32_commute(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %nega, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32_unmasked(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vv_nxv2f32_ta(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32_commute_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %nega, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+declare <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+
+define <vscale x 4 x float> @vfnmacc_vv_nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vv_nxv4f32_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32_commute(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %nega, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32_unmasked(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vv_nxv4f32_ta(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32_commute_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %nega, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+declare <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+declare <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+
+define <vscale x 8 x float> @vfnmacc_vv_nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vv_nxv8f32_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32_commute(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %nega, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32_unmasked(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vv_nxv8f32_ta(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32_commute_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %nega, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+declare <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+declare <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+
+define <vscale x 16 x float> @vfnmacc_vv_nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vv_nxv16f32_unmasked(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32_commute(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %nega, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32_unmasked(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vv_nxv16f32_ta(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32_commute_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %nega, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+declare <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+declare <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64_unmasked(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64_commute(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %nega, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64_unmasked(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64_ta(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64_commute_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmacc.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %nega, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+declare <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+declare <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64_unmasked(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64_commute(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %nega, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64_unmasked(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64_ta(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64_commute_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %nega, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+declare <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+declare <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64_unmasked(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64_commute(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %nega, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64_unmasked(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64_ta(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64_commute_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %nega, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+declare <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+declare <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64_unmasked(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64_commute(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %nega, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64_unmasked(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64_ta(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64_commute_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %nega, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfnmsac-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfnmsac-vp.ll
new file mode 100644
index 0000000000000..d3321e1d8c7c2
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfnmsac-vp.ll
@@ -0,0 +1,2039 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=ilp32d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v,+m -target-abi=lp64d \
+; RUN: -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+declare <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+
+define <vscale x 1 x half> @vfnmsac_vv_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vv_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_commute(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %nega, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.merge.nxv1f16(<vscale x 1 x i1> %allones, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vv_nxv1f16_ta(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %nega, <vscale x 1 x half> %vb, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+define <vscale x 1 x half> @vfnmsac_vf_nxv1f16_commute_ta(<vscale x 1 x half> %a, half %b, <vscale x 1 x half> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 1 x half> %elt.head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x half> @llvm.vp.fma.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x half> %nega, <vscale x 1 x half> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %m, <vscale x 1 x half> %v, <vscale x 1 x half> %c, i32 %evl)
+ ret <vscale x 1 x half> %u
+}
+
+declare <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+declare <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+
+define <vscale x 2 x half> @vfnmsac_vv_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vv_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_commute(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %nega, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.merge.nxv2f16(<vscale x 2 x i1> %allones, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vv_nxv2f16_ta(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %nega, <vscale x 2 x half> %vb, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+define <vscale x 2 x half> @vfnmsac_vf_nxv2f16_commute_ta(<vscale x 2 x half> %a, half %b, <vscale x 2 x half> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 2 x half> %elt.head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x half> @llvm.vp.fma.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x half> %nega, <vscale x 2 x half> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %m, <vscale x 2 x half> %v, <vscale x 2 x half> %c, i32 %evl)
+ ret <vscale x 2 x half> %u
+}
+
+declare <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+declare <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+
+define <vscale x 4 x half> @vfnmsac_vv_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vv_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_commute(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %nega, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.merge.nxv4f16(<vscale x 4 x i1> %allones, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vv_nxv4f16_ta(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %nega, <vscale x 4 x half> %vb, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+define <vscale x 4 x half> @vfnmsac_vf_nxv4f16_commute_ta(<vscale x 4 x half> %a, half %b, <vscale x 4 x half> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 4 x half> %elt.head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x half> @llvm.vp.fma.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x half> %nega, <vscale x 4 x half> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %m, <vscale x 4 x half> %v, <vscale x 4 x half> %c, i32 %evl)
+ ret <vscale x 4 x half> %u
+}
+
+declare <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+
+define <vscale x 8 x half> @vfnmsac_vv_nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vv_nxv8f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_commute(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %nega, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.merge.nxv8f16(<vscale x 8 x i1> %allones, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vv_nxv8f16_ta(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %nega, <vscale x 8 x half> %vb, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+define <vscale x 8 x half> @vfnmsac_vf_nxv8f16_commute_ta(<vscale x 8 x half> %a, half %b, <vscale x 8 x half> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 8 x half> %elt.head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x half> @llvm.vp.fma.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x half> %nega, <vscale x 8 x half> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %m, <vscale x 8 x half> %v, <vscale x 8 x half> %c, i32 %evl)
+ ret <vscale x 8 x half> %u
+}
+
+declare <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+declare <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+
+define <vscale x 16 x half> @vfnmsac_vv_nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vv_nxv16f16_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_commute(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %nega, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_unmasked(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.merge.nxv16f16(<vscale x 16 x i1> %allones, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vv_nxv16f16_ta(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %nega, <vscale x 16 x half> %vb, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+define <vscale x 16 x half> @vfnmsac_vf_nxv16f16_commute_ta(<vscale x 16 x half> %a, half %b, <vscale x 16 x half> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 16 x half> %elt.head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x half> @llvm.vp.fma.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x half> %nega, <vscale x 16 x half> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %m, <vscale x 16 x half> %v, <vscale x 16 x half> %c, i32 %evl)
+ ret <vscale x 16 x half> %u
+}
+
+declare <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half>, <vscale x 32 x i1>, i32)
+declare <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+declare <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+
+define <vscale x 32 x half> @vfnmsac_vv_nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vv_nxv32f16_unmasked(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_commute(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %nega, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_unmasked(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.merge.nxv32f16(<vscale x 32 x i1> %allones, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vv_nxv32f16_ta(<vscale x 32 x half> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re16.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %nega, <vscale x 32 x half> %vb, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+define <vscale x 32 x half> @vfnmsac_vf_nxv32f16_commute_ta(<vscale x 32 x half> %a, half %b, <vscale x 32 x half> %c, <vscale x 32 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv32f16_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 32 x half> poison, half %b, i32 0
+ %vb = shufflevector <vscale x 32 x half> %elt.head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer
+ %splat = insertelement <vscale x 32 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 32 x i1> %splat, <vscale x 32 x i1> poison, <vscale x 32 x i32> zeroinitializer
+ %nega = call <vscale x 32 x half> @llvm.vp.fneg.nxv32f16(<vscale x 32 x half> %a, <vscale x 32 x i1> %allones, i32 %evl)
+ %v = call <vscale x 32 x half> @llvm.vp.fma.nxv32f16(<vscale x 32 x half> %vb, <vscale x 32 x half> %nega, <vscale x 32 x half> %c, <vscale x 32 x i1> %allones, i32 %evl)
+ %u = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %m, <vscale x 32 x half> %v, <vscale x 32 x half> %c, i32 %evl)
+ ret <vscale x 32 x half> %u
+}
+
+declare <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+declare <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_commute(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %nega, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_unmasked(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.merge.nxv1f32(<vscale x 1 x i1> %allones, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32_ta(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vb, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_commute_ta(<vscale x 1 x float> %a, float %b, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 1 x float> %elt.head, <vscale x 1 x float> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x float> %nega, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %m, <vscale x 1 x float> %v, <vscale x 1 x float> %c, i32 %evl)
+ ret <vscale x 1 x float> %u
+}
+
+declare <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+declare <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_commute(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %nega, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_unmasked(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %allones, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32_ta(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vb, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_commute_ta(<vscale x 2 x float> %a, float %b, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 2 x float> %elt.head, <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x float> %nega, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %v, <vscale x 2 x float> %c, i32 %evl)
+ ret <vscale x 2 x float> %u
+}
+
+declare <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_commute(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %nega, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_unmasked(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> %allones, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32_ta(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vb, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_commute_ta(<vscale x 4 x float> %a, float %b, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 4 x float> %elt.head, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x float> %nega, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %m, <vscale x 4 x float> %v, <vscale x 4 x float> %c, i32 %evl)
+ ret <vscale x 4 x float> %u
+}
+
+declare <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+declare <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_commute(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %nega, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_unmasked(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.merge.nxv8f32(<vscale x 8 x i1> %allones, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32_ta(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vb, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_commute_ta(<vscale x 8 x float> %a, float %b, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 8 x float> %elt.head, <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x float> %nega, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %m, <vscale x 8 x float> %v, <vscale x 8 x float> %c, i32 %evl)
+ ret <vscale x 8 x float> %u
+}
+
+declare <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float>, <vscale x 16 x i1>, i32)
+declare <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+declare <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32_unmasked(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_commute(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %nega, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_unmasked(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.merge.nxv16f32(<vscale x 16 x i1> %allones, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32_ta(<vscale x 16 x float> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re32.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vb, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_commute_ta(<vscale x 16 x float> %a, float %b, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 16 x float> poison, float %b, i32 0
+ %vb = shufflevector <vscale x 16 x float> %elt.head, <vscale x 16 x float> poison, <vscale x 16 x i32> zeroinitializer
+ %splat = insertelement <vscale x 16 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 16 x i1> %splat, <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vb, <vscale x 16 x float> %nega, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ %u = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %m, <vscale x 16 x float> %v, <vscale x 16 x float> %c, i32 %evl)
+ ret <vscale x 16 x float> %u
+}
+
+declare <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+declare <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64_unmasked(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_commute(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %nega, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_unmasked(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, tu, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8
+; CHECK-NEXT: vmv1r.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.merge.nxv1f64(<vscale x 1 x i1> %allones, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64_ta(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vb, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_commute_ta(<vscale x 1 x double> %a, double %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT: vfnmsac.vf v9, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v9
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 1 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 1 x double> %elt.head, <vscale x 1 x double> poison, <vscale x 1 x i32> zeroinitializer
+ %splat = insertelement <vscale x 1 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 1 x i1> %splat, <vscale x 1 x i1> poison, <vscale x 1 x i32> zeroinitializer
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vb, <vscale x 1 x double> %nega, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ %u = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %m, <vscale x 1 x double> %v, <vscale x 1 x double> %c, i32 %evl)
+ ret <vscale x 1 x double> %u
+}
+
+declare <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+declare <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+declare <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64_unmasked(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv2r.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_commute(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %nega, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_unmasked(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, tu, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %allones, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64_ta(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vb, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_commute_ta(<vscale x 2 x double> %a, double %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT: vfnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v10
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 2 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 2 x double> %elt.head, <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
+ %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vb, <vscale x 2 x double> %nega, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ %u = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %v, <vscale x 2 x double> %c, i32 %evl)
+ ret <vscale x 2 x double> %u
+}
+
+declare <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double>, <vscale x 4 x i1>, i32)
+declare <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+declare <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64_unmasked(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv4r.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_commute(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %nega, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_unmasked(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, tu, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.merge.nxv4f64(<vscale x 4 x i1> %allones, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64_ta(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vb, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_commute_ta(<vscale x 4 x double> %a, double %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT: vfnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v12
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 4 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 4 x double> %elt.head, <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
+ %splat = insertelement <vscale x 4 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 4 x i1> %splat, <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vb, <vscale x 4 x double> %nega, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ %u = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %m, <vscale x 4 x double> %v, <vscale x 4 x double> %c, i32 %evl)
+ ret <vscale x 4 x double> %u
+}
+
+declare <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double>, <vscale x 8 x i1>, i32)
+declare <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+declare <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64_unmasked(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16
+; CHECK-NEXT: vmv8r.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_commute(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %nega, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_unmasked(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, tu, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.merge.nxv8f64(<vscale x 8 x i1> %allones, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64_ta(<vscale x 8 x double> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vl8re64.v v24, (a0)
+; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vv v24, v8, v16, v0.t
+; CHECK-NEXT: vmv.v.v v8, v24
+; CHECK-NEXT: ret
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vb, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_commute_ta(<vscale x 8 x double> %a, double %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_commute_ta:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT: vfnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv.v.v v8, v16
+; CHECK-NEXT: ret
+ %elt.head = insertelement <vscale x 8 x double> poison, double %b, i32 0
+ %vb = shufflevector <vscale x 8 x double> %elt.head, <vscale x 8 x double> poison, <vscale x 8 x i32> zeroinitializer
+ %splat = insertelement <vscale x 8 x i1> poison, i1 -1, i32 0
+ %allones = shufflevector <vscale x 8 x i1> %splat, <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vb, <vscale x 8 x double> %nega, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ %u = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %m, <vscale x 8 x double> %v, <vscale x 8 x double> %c, i32 %evl)
+ ret <vscale x 8 x double> %u
+}
More information about the llvm-commits
mailing list