[llvm] 46cad21 - [RISCV] Update RISCVISD::VFWMADD_VLL isel patterns to allow a mask that isn't all ones.
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Mon May 15 22:47:39 PDT 2023
Author: Craig Topper
Date: 2023-05-15T22:47:30-07:00
New Revision: 46cad213ee71a77eefe94fa852576b247fbdafb9
URL: https://github.com/llvm/llvm-project/commit/46cad213ee71a77eefe94fa852576b247fbdafb9
DIFF: https://github.com/llvm/llvm-project/commit/46cad213ee71a77eefe94fa852576b247fbdafb9.diff
LOG: [RISCV] Update RISCVISD::VFWMADD_VLL isel patterns to allow a mask that isn't all ones.
Fixes an isel crash after 6e6bed575777be7be55482090414e153ed6f7557.
Add more tests so we have coverage for this.
Added:
llvm/test/CodeGen/RISCV/rvv/vfwmacc-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfwmsac-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfwnmacc-vp.ll
llvm/test/CodeGen/RISCV/rvv/vfwnmsac-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 ea5084620bb5a..23b971583e96b 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -1530,18 +1530,18 @@ multiclass VPatWidenFPMulAccVL_VV_VF<SDNode vop, string instruction_name> {
defvar wti = vtiToWti.Wti;
def : Pat<(vop (vti.Vector vti.RegClass:$rs1),
(vti.Vector vti.RegClass:$rs2),
- (wti.Vector wti.RegClass:$rd), (vti.Mask true_mask),
+ (wti.Vector wti.RegClass:$rd), (vti.Mask V0),
VLOpFrag),
- (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX)
+ (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX #"_MASK")
wti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
- GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
+ (vti.Mask V0), GPR:$vl, vti.Log2SEW, TA_MA)>;
def : Pat<(vop (vti.Vector (SplatFPOp vti.ScalarRegClass:$rs1)),
(vti.Vector vti.RegClass:$rs2),
- (wti.Vector wti.RegClass:$rd), (vti.Mask true_mask),
+ (wti.Vector wti.RegClass:$rd), (vti.Mask V0),
VLOpFrag),
- (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+ (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX #"_MASK")
wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
- GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
+ (vti.Mask V0), GPR:$vl, vti.Log2SEW, TA_MA)>;
}
}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfwmacc-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfwmacc-vp.ll
new file mode 100644
index 0000000000000..5622046db7694
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfwmacc-vp.ll
@@ -0,0 +1,756 @@
+; 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 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.fpext.nxv1f32.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x float> @vfmacc_vv_nxv1f32(<vscale x 1 x half> %a, <vscale x 1 x half> %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x float> %bext, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfmacc_vv_nxv1f32_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x float> %bext, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32(<vscale x 1 x half> %va, half %b, <vscale x 1 x float> %vc, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %va, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vaext, <vscale x 1 x float> %vbext, <vscale x 1 x float> %vc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32_commute(<vscale x 1 x half> %va, half %b, <vscale x 1 x float> %vc, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32_commute:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %va, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vbext, <vscale x 1 x float> %vaext, <vscale x 1 x float> %vc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfmacc_vf_nxv1f32_unmasked(<vscale x 1 x half> %va, half %b, <vscale x 1 x float> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %va, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vaext, <vscale x 1 x float> %vbext, <vscale x 1 x float> %vc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+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)
+declare <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vfmacc_vv_nxv2f32(<vscale x 2 x half> %a, <vscale x 2 x half> %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x float> %bext, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfmacc_vv_nxv2f32_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x float> %bext, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32(<vscale x 2 x half> %va, half %b, <vscale x 2 x float> %vc, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vaext, <vscale x 2 x float> %vbext, <vscale x 2 x float> %vc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfmacc_vf_nxv2f32_unmasked(<vscale x 2 x half> %va, half %b, <vscale x 2 x float> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vaext, <vscale x 2 x float> %vbext, <vscale x 2 x float> %vc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+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)
+declare <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x float> @vfmacc_vv_nxv4f32(<vscale x 4 x half> %a, <vscale x 4 x half> %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x float> %bext, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfmacc_vv_nxv4f32_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x float> %bext, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32(<vscale x 4 x half> %va, half %b, <vscale x 4 x float> %vc, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %vaext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %va, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vaext, <vscale x 4 x float> %vbext, <vscale x 4 x float> %vc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfmacc_vf_nxv4f32_unmasked(<vscale x 4 x half> %va, half %b, <vscale x 4 x float> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %vaext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %va, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vaext, <vscale x 4 x float> %vbext, <vscale x 4 x float> %vc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+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)
+declare <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x float> @vfmacc_vv_nxv8f32(<vscale x 8 x half> %a, <vscale x 8 x half> %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x float> %bext, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfmacc_vv_nxv8f32_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x float> %bext, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32(<vscale x 8 x half> %va, half %b, <vscale x 8 x float> %vc, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %vaext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %va, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vaext, <vscale x 8 x float> %vbext, <vscale x 8 x float> %vc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfmacc_vf_nxv8f32_unmasked(<vscale x 8 x half> %va, half %b, <vscale x 8 x float> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %vaext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %va, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vaext, <vscale x 8 x float> %vbext, <vscale x 8 x float> %vc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+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)
+declare <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+
+define <vscale x 16 x float> @vfmacc_vv_nxv16f32(<vscale x 16 x half> %a, <vscale x 16 x half> %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: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %bext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %b, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x float> %bext, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfmacc_vv_nxv16f32_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %b, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x float> %bext, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32(<vscale x 16 x half> %va, half %b, <vscale x 16 x float> %vc, <vscale x 16 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %vaext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %va, <vscale x 16 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vaext, <vscale x 16 x float> %vbext, <vscale x 16 x float> %vc, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfmacc_vf_nxv16f32_unmasked(<vscale x 16 x half> %va, half %b, <vscale x 16 x float> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %vaext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %va, <vscale x 16 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %allones, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vaext, <vscale x 16 x float> %vbext, <vscale x 16 x float> %vc, <vscale x 16 x i1> %allones, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+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)
+declare <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64(<vscale x 1 x float> %a, <vscale x 1 x float> %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x double> %bext, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x double> %bext, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64(<vscale x 1 x float> %va, float %b, <vscale x 1 x double> %vc, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %va, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vaext, <vscale x 1 x double> %vbext, <vscale x 1 x double> %vc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfmacc_vf_nxv1f64_unmasked(<vscale x 1 x float> %va, float %b, <vscale x 1 x double> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.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
+ %vaext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %va, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vaext, <vscale x 1 x double> %vbext, <vscale x 1 x double> %vc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+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)
+declare <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64(<vscale x 2 x float> %a, <vscale x 2 x float> %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x double> %bext, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x double> %bext, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64(<vscale x 2 x float> %va, float %b, <vscale x 2 x double> %vc, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %vaext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vaext, <vscale x 2 x double> %vbext, <vscale x 2 x double> %vc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfmacc_vf_nxv2f64_unmasked(<vscale x 2 x float> %va, float %b, <vscale x 2 x double> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %vaext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vaext, <vscale x 2 x double> %vbext, <vscale x 2 x double> %vc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+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)
+declare <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64(<vscale x 4 x float> %a, <vscale x 4 x float> %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x double> %bext, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x double> %bext, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64(<vscale x 4 x float> %va, float %b, <vscale x 4 x double> %vc, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %vaext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %va, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vaext, <vscale x 4 x double> %vbext, <vscale x 4 x double> %vc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfmacc_vf_nxv4f64_unmasked(<vscale x 4 x float> %va, float %b, <vscale x 4 x double> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %vaext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %va, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vaext, <vscale x 4 x double> %vbext, <vscale x 4 x double> %vc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+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)
+declare <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64(<vscale x 8 x float> %a, <vscale x 8 x float> %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: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x double> %bext, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x double> %bext, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64(<vscale x 8 x float> %va, float %b, <vscale x 8 x double> %vc, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %vaext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %va, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vaext, <vscale x 8 x double> %vbext, <vscale x 8 x double> %vc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfmacc_vf_nxv8f64_unmasked(<vscale x 8 x float> %va, float %b, <vscale x 8 x double> %vc, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %vaext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %va, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vaext, <vscale x 8 x double> %vbext, <vscale x 8 x double> %vc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+declare <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f64_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v11, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v11, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x double> %bext, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfmacc_vv_nxv1f64_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv1f64_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v11, v8
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9
+; CHECK-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v11, v8
+; 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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x double> %bext, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f64_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v12, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x double> %bext, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfmacc_vv_nxv2f64_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv2f64_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9
+; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT: vfwmacc.vv v10, v12, v8
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x double> %bext, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+declare <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f64_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v10, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v16, v9, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v12, v10, v16, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x double> %bext, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfmacc_vv_nxv4f64_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv4f64_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v10, v8
+; CHECK-NEXT: vfwcvt.f.f.v v16, v9
+; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vfwmacc.vv v12, v10, v16
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x double> %bext, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+declare <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64_nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f64_nxv8f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v24, v10, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v16, v12, v24, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x double> %bext, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfmacc_vv_nxv8f64_nxv8f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfmacc_vv_nxv8f64_nxv8f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8
+; CHECK-NEXT: vfwcvt.f.f.v v24, v10
+; CHECK-NEXT: vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT: vfwmacc.vv v16, v12, v24
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x double> %bext, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfwmsac-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfwmsac-vp.ll
new file mode 100644
index 0000000000000..578caa36855da
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfwmsac-vp.ll
@@ -0,0 +1,349 @@
+; 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 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.fpext.nxv1f32.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x float> @vmfsac_vv_nxv1f32(<vscale x 1 x half> %a, <vscale x 1 x half> %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x float> %bext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vmfsac_vv_nxv1f32_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <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> %aext, <vscale x 1 x float> %bext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32(<vscale x 1 x half> %a, half %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x float> %vbext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32_commute(<vscale x 1 x half> %a, half %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vbext, <vscale x 1 x float> %aext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vmfsac_vf_nxv1f32_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <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> %aext, <vscale x 1 x float> %vbext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+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.fpext.nxv2f32.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vmfsac_vv_nxv2f32(<vscale x 2 x half> %a, <vscale x 2 x half> %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x float> %bext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vmfsac_vv_nxv2f32_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <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> %aext, <vscale x 2 x float> %bext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32(<vscale x 2 x half> %a, half %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x float> %vbext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32_commute(<vscale x 2 x half> %a, half %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vbext, <vscale x 2 x float> %aext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vmfsac_vf_nxv2f32_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <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> %aext, <vscale x 2 x float> %vbext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+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.fpext.nxv4f32.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x float> @vmfsac_vv_nxv4f32(<vscale x 4 x half> %a, <vscale x 4 x half> %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x float> %bext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vmfsac_vv_nxv4f32_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <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> %aext, <vscale x 4 x float> %bext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32(<vscale x 4 x half> %a, half %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x float> %vbext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32_commute(<vscale x 4 x half> %a, half %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vbext, <vscale x 4 x float> %aext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vmfsac_vf_nxv4f32_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <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> %aext, <vscale x 4 x float> %vbext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+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.fpext.nxv8f32.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x float> @vmfsac_vv_nxv8f32(<vscale x 8 x half> %a, <vscale x 8 x half> %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x float> %bext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vmfsac_vv_nxv8f32_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <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> %aext, <vscale x 8 x float> %bext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32(<vscale x 8 x half> %a, half %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x float> %vbext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32_commute(<vscale x 8 x half> %a, half %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vbext, <vscale x 8 x float> %aext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vmfsac_vf_nxv8f32_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vmfsac_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <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> %aext, <vscale x 8 x float> %vbext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfwnmacc-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfwnmacc-vp.ll
new file mode 100644
index 0000000000000..a77501bd5552a
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfwnmacc-vp.ll
@@ -0,0 +1,975 @@
+; 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 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.fpext.nxv1f32.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x float> @vfnmacc_vv_nxv1f32(<vscale x 1 x half> %a, <vscale x 1 x half> %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %bext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmacc_vv_nxv1f32_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <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> %bext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32(<vscale x 1 x half> %a, half %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vbext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32_commute(<vscale x 1 x half> %a, half %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vbext, <vscale x 1 x float> %nega, <vscale x 1 x float> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmacc_vf_nxv1f32_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <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> %vbext, <vscale x 1 x float> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+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.fpext.nxv2f32.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vfnmacc_vv_nxv2f32(<vscale x 2 x half> %a, <vscale x 2 x half> %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %bext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmacc_vv_nxv2f32_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <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> %bext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32(<vscale x 2 x half> %a, half %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vbext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32_commute(<vscale x 2 x half> %a, half %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vbext, <vscale x 2 x float> %nega, <vscale x 2 x float> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmacc_vf_nxv2f32_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <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> %vbext, <vscale x 2 x float> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+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.fpext.nxv4f32.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x float> @vfnmacc_vv_nxv4f32(<vscale x 4 x half> %a, <vscale x 4 x half> %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %bext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmacc_vv_nxv4f32_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <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> %bext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32(<vscale x 4 x half> %a, half %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vbext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32_commute(<vscale x 4 x half> %a, half %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vbext, <vscale x 4 x float> %nega, <vscale x 4 x float> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmacc_vf_nxv4f32_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <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> %vbext, <vscale x 4 x float> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+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.fpext.nxv8f32.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x float> @vfnmacc_vv_nxv8f32(<vscale x 8 x half> %a, <vscale x 8 x half> %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %bext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmacc_vv_nxv8f32_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <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> %bext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32(<vscale x 8 x half> %a, half %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vbext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32_commute(<vscale x 8 x half> %a, half %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vbext, <vscale x 8 x float> %nega, <vscale x 8 x float> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmacc_vf_nxv8f32_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <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> %vbext, <vscale x 8 x float> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+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.fpext.nxv16f32.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+
+define <vscale x 16 x float> @vfnmacc_vv_nxv16f32(<vscale x 16 x half> %a, <vscale x 16 x half> %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: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %bext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %b, <vscale x 16 x i1> %m, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x i1> %m, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %bext, <vscale x 16 x float> %negc, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmacc_vv_nxv16f32_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %b, <vscale x 16 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <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> %bext, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32(<vscale x 16 x half> %a, half %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, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %m, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x i1> %m, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vbext, <vscale x 16 x float> %negc, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32_commute(<vscale x 16 x half> %a, half %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, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %m, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x i1> %m, i32 %evl)
+ %negc = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vbext, <vscale x 16 x float> %nega, <vscale x 16 x float> %negc, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmacc_vf_nxv16f32_unmasked(<vscale x 16 x half> %a, half %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <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> %vbext, <vscale x 16 x float> %negc, <vscale x 16 x i1> %allones, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+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.fpext.nxv1f64.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64(<vscale x 1 x float> %a, <vscale x 1 x float> %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %bext, <vscale x 1 x double> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <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> %bext, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64(<vscale x 1 x float> %a, float %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vbext, <vscale x 1 x double> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64_commute(<vscale x 1 x float> %a, float %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vbext, <vscale x 1 x double> %nega, <vscale x 1 x double> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmacc_vf_nxv1f64_unmasked(<vscale x 1 x float> %a, float %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <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> %vbext, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+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.fpext.nxv2f64.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64(<vscale x 2 x float> %a, <vscale x 2 x float> %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %bext, <vscale x 2 x double> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <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> %bext, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64(<vscale x 2 x float> %a, float %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vbext, <vscale x 2 x double> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64_commute(<vscale x 2 x float> %a, float %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vbext, <vscale x 2 x double> %nega, <vscale x 2 x double> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmacc_vf_nxv2f64_unmasked(<vscale x 2 x float> %a, float %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <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> %vbext, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+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.fpext.nxv4f64.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64(<vscale x 4 x float> %a, <vscale x 4 x float> %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %bext, <vscale x 4 x double> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <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> %bext, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64(<vscale x 4 x float> %a, float %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vbext, <vscale x 4 x double> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64_commute(<vscale x 4 x float> %a, float %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vbext, <vscale x 4 x double> %nega, <vscale x 4 x double> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmacc_vf_nxv4f64_unmasked(<vscale x 4 x float> %a, float %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <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> %vbext, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+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.fpext.nxv8f64.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64(<vscale x 8 x float> %a, <vscale x 8 x float> %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: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %bext, <vscale x 8 x double> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v16, v8, v12
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <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> %bext, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64(<vscale x 8 x float> %a, float %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, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vbext, <vscale x 8 x double> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64_commute(<vscale x 8 x float> %a, float %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, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vbext, <vscale x 8 x double> %nega, <vscale x 8 x double> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmacc_vf_nxv8f64_unmasked(<vscale x 8 x float> %a, float %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <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> %vbext, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+declare <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64_nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f64_nxv1f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v11, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v11, v8, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %negc = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %bext, <vscale x 1 x double> %negc, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmacc_vv_nxv1f64_nxv1f16_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv1f64_nxv1f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v11, v8
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9
+; CHECK-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v11, v8
+; 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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <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> %bext, <vscale x 1 x double> %negc, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64_nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f64_nxv2f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v12, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %negc = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %bext, <vscale x 2 x double> %negc, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmacc_vv_nxv2f64_nxv2f16_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv2f64_nxv2f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8
+; CHECK-NEXT: vfwcvt.f.f.v v8, v9
+; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v10, v12, v8
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <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> %bext, <vscale x 2 x double> %negc, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+declare <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64_nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f64_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v10, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v16, v9, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v12, v10, v16, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %negc = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %bext, <vscale x 4 x double> %negc, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmacc_vv_nxv4f64_nxv4f16_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv4f64_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v10, v8
+; CHECK-NEXT: vfwcvt.f.f.v v16, v9
+; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v12, v10, v16
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <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> %bext, <vscale x 4 x double> %negc, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+declare <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64_nxv4f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f64_nxv4f16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8, v0.t
+; CHECK-NEXT: vfwcvt.f.f.v v24, v10, v0.t
+; CHECK-NEXT: vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v16, v12, v24, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %negc = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %bext, <vscale x 8 x double> %negc, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmacc_vv_nxv8f64_nxv4f16_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmacc_vv_nxv8f64_nxv4f16_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwcvt.f.f.v v12, v8
+; CHECK-NEXT: vfwcvt.f.f.v v24, v10
+; CHECK-NEXT: vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmacc.vv v16, v12, v24
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <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> %bext, <vscale x 8 x double> %negc, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vfwnmsac-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vfwnmsac-vp.ll
new file mode 100644
index 0000000000000..2d2f64d9c8ca6
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vfwnmsac-vp.ll
@@ -0,0 +1,761 @@
+; 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 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.fpext.nxv1f32.nxv1f16(<vscale x 1 x half>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32(<vscale x 1 x half> %a, <vscale x 1 x half> %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %bext, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmsac_vv_nxv1f32_unmasked(<vscale x 1 x half> %a, <vscale x 1 x half> %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <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> %bext, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32(<vscale x 1 x half> %a, half %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %nega, <vscale x 1 x float> %vbext, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_commute(<vscale x 1 x half> %a, half %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, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x float> @llvm.vp.fma.nxv1f32(<vscale x 1 x float> %vbext, <vscale x 1 x float> %nega, <vscale x 1 x float> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+define <vscale x 1 x float> @vfnmsac_vf_nxv1f32_unmasked(<vscale x 1 x half> %a, half %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f32.nxv1f16(<vscale x 1 x half> %vb, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> %aext, <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> %vbext, <vscale x 1 x float> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x float> %v
+}
+
+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.fpext.nxv2f32.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32(<vscale x 2 x half> %a, <vscale x 2 x half> %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %bext, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmsac_vv_nxv2f32_unmasked(<vscale x 2 x half> %a, <vscale x 2 x half> %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <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> %bext, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32(<vscale x 2 x half> %a, half %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %nega, <vscale x 2 x float> %vbext, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_commute(<vscale x 2 x half> %a, half %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, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x float> @llvm.vp.fma.nxv2f32(<vscale x 2 x float> %vbext, <vscale x 2 x float> %nega, <vscale x 2 x float> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vfnmsac_vf_nxv2f32_unmasked(<vscale x 2 x half> %a, half %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f32.nxv2f16(<vscale x 2 x half> %vb, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> %aext, <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> %vbext, <vscale x 2 x float> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x float> %v
+}
+
+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.fpext.nxv4f32.nxv4f16(<vscale x 4 x half>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32(<vscale x 4 x half> %a, <vscale x 4 x half> %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %bext, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmsac_vv_nxv4f32_unmasked(<vscale x 4 x half> %a, <vscale x 4 x half> %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <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> %bext, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32(<vscale x 4 x half> %a, half %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %nega, <vscale x 4 x float> %vbext, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_commute(<vscale x 4 x half> %a, half %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, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x float> @llvm.vp.fma.nxv4f32(<vscale x 4 x float> %vbext, <vscale x 4 x float> %nega, <vscale x 4 x float> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+define <vscale x 4 x float> @vfnmsac_vf_nxv4f32_unmasked(<vscale x 4 x half> %a, half %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f32.nxv4f16(<vscale x 4 x half> %vb, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> %aext, <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> %vbext, <vscale x 4 x float> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x float> %v
+}
+
+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.fpext.nxv8f32.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32(<vscale x 8 x half> %a, <vscale x 8 x half> %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %bext, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmsac_vv_nxv8f32_unmasked(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <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> %bext, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32(<vscale x 8 x half> %a, half %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %nega, <vscale x 8 x float> %vbext, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_commute(<vscale x 8 x half> %a, half %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, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x float> @llvm.vp.fma.nxv8f32(<vscale x 8 x float> %vbext, <vscale x 8 x float> %nega, <vscale x 8 x float> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+define <vscale x 8 x float> @vfnmsac_vf_nxv8f32_unmasked(<vscale x 8 x half> %a, half %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f32.nxv8f16(<vscale x 8 x half> %vb, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> %aext, <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> %vbext, <vscale x 8 x float> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x float> %v
+}
+
+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.fpext.nxv16f32.nxv16f16(<vscale x 16 x half>, <vscale x 16 x i1>, i32)
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32(<vscale x 16 x half> %a, <vscale x 16 x half> %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: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %bext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %b, <vscale x 16 x i1> %m, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %bext, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmsac_vv_nxv16f32_unmasked(<vscale x 16 x half> %a, <vscale x 16 x half> %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %b, <vscale x 16 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <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> %bext, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32(<vscale x 16 x half> %a, half %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, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %m, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %nega, <vscale x 16 x float> %vbext, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_commute(<vscale x 16 x half> %a, half %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, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %m, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <vscale x 16 x i1> %m, i32 %evl)
+ %v = call <vscale x 16 x float> @llvm.vp.fma.nxv16f32(<vscale x 16 x float> %vbext, <vscale x 16 x float> %nega, <vscale x 16 x float> %c, <vscale x 16 x i1> %m, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+define <vscale x 16 x float> @vfnmsac_vf_nxv16f32_unmasked(<vscale x 16 x half> %a, half %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv16f32_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %a, <vscale x 16 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f32.nxv16f16(<vscale x 16 x half> %vb, <vscale x 16 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> %aext, <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> %vbext, <vscale x 16 x float> %c, <vscale x 16 x i1> %allones, i32 %evl)
+ ret <vscale x 16 x float> %v
+}
+
+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.fpext.nxv1f64.nxv1f32(<vscale x 1 x float>, <vscale x 1 x i1>, i32)
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64(<vscale x 1 x float> %a, <vscale x 1 x float> %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv1r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %b, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %bext, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmsac_vv_nxv1f64_unmasked(<vscale x 1 x float> %a, <vscale x 1 x float> %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %b, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <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> %bext, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64(<vscale x 1 x float> %a, float %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %nega, <vscale x 1 x double> %vbext, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_commute(<vscale x 1 x float> %a, float %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, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %m, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <vscale x 1 x i1> %m, i32 %evl)
+ %v = call <vscale x 1 x double> @llvm.vp.fma.nxv1f64(<vscale x 1 x double> %vbext, <vscale x 1 x double> %nega, <vscale x 1 x double> %c, <vscale x 1 x i1> %m, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+define <vscale x 1 x double> @vfnmsac_vf_nxv1f64_unmasked(<vscale x 1 x float> %a, float %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv1f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; CHECK-NEXT: vfwnmsac.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
+ %aext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %a, <vscale x 1 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f64.nxv1f32(<vscale x 1 x float> %vb, <vscale x 1 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> %aext, <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> %vbext, <vscale x 1 x double> %c, <vscale x 1 x i1> %allones, i32 %evl)
+ ret <vscale x 1 x double> %v
+}
+
+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.fpext.nxv2f64.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64(<vscale x 2 x float> %a, <vscale x 2 x float> %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; CHECK-NEXT: ret
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %b, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %bext, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmsac_vv_nxv2f64_unmasked(<vscale x 2 x float> %a, <vscale x 2 x float> %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v10, v8, v9
+; CHECK-NEXT: vmv2r.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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %b, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <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> %bext, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64(<vscale x 2 x float> %a, float %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %nega, <vscale x 2 x double> %vbext, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_commute(<vscale x 2 x float> %a, float %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, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v10, fa0, v8, v0.t
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %m, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <vscale x 2 x i1> %m, i32 %evl)
+ %v = call <vscale x 2 x double> @llvm.vp.fma.nxv2f64(<vscale x 2 x double> %vbext, <vscale x 2 x double> %nega, <vscale x 2 x double> %c, <vscale x 2 x i1> %m, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vfnmsac_vf_nxv2f64_unmasked(<vscale x 2 x float> %a, float %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv2f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v10, fa0, v8
+; CHECK-NEXT: vmv2r.v v8, v10
+; 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
+ %aext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %a, <vscale x 2 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %vb, <vscale x 2 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> %aext, <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> %vbext, <vscale x 2 x double> %c, <vscale x 2 x i1> %allones, i32 %evl)
+ ret <vscale x 2 x double> %v
+}
+
+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.fpext.nxv4f64.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, i32)
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64(<vscale x 4 x float> %a, <vscale x 4 x float> %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v12, v8, v10, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; CHECK-NEXT: ret
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %b, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %bext, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmsac_vv_nxv4f64_unmasked(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v12, v8, v10
+; CHECK-NEXT: vmv4r.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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %b, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <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> %bext, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64(<vscale x 4 x float> %a, float %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %nega, <vscale x 4 x double> %vbext, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_commute(<vscale x 4 x float> %a, float %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, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v12, fa0, v8, v0.t
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %m, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <vscale x 4 x i1> %m, i32 %evl)
+ %v = call <vscale x 4 x double> @llvm.vp.fma.nxv4f64(<vscale x 4 x double> %vbext, <vscale x 4 x double> %nega, <vscale x 4 x double> %c, <vscale x 4 x i1> %m, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+define <vscale x 4 x double> @vfnmsac_vf_nxv4f64_unmasked(<vscale x 4 x float> %a, float %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv4f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v12, fa0, v8
+; CHECK-NEXT: vmv4r.v v8, v12
+; 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
+ %aext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> %vb, <vscale x 4 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> %aext, <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> %vbext, <vscale x 4 x double> %c, <vscale x 4 x i1> %allones, i32 %evl)
+ ret <vscale x 4 x double> %v
+}
+
+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.fpext.nxv8f64.nxv8f32(<vscale x 8 x float>, <vscale x 8 x i1>, i32)
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64(<vscale x 8 x float> %a, <vscale x 8 x float> %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: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v16, v8, v12, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; CHECK-NEXT: ret
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %b, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %bext, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmsac_vv_nxv8f64_unmasked(<vscale x 8 x float> %a, <vscale x 8 x float> %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vv_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vv v16, v8, v12
+; CHECK-NEXT: vmv8r.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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %bext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %b, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <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> %bext, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64(<vscale x 8 x float> %a, float %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, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %nega, <vscale x 8 x double> %vbext, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_commute(<vscale x 8 x float> %a, float %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, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v16, fa0, v8, v0.t
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %m, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %m, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <vscale x 8 x i1> %m, i32 %evl)
+ %v = call <vscale x 8 x double> @llvm.vp.fma.nxv8f64(<vscale x 8 x double> %vbext, <vscale x 8 x double> %nega, <vscale x 8 x double> %c, <vscale x 8 x i1> %m, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
+
+define <vscale x 8 x double> @vfnmsac_vf_nxv8f64_unmasked(<vscale x 8 x float> %a, float %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: vfnmsac_vf_nxv8f64_unmasked:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; CHECK-NEXT: vfwnmsac.vf v16, fa0, v8
+; CHECK-NEXT: vmv8r.v v8, v16
+; 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
+ %aext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %a, <vscale x 8 x i1> %allones, i32 %evl)
+ %vbext = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f64.nxv8f32(<vscale x 8 x float> %vb, <vscale x 8 x i1> %allones, i32 %evl)
+ %nega = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> %aext, <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> %vbext, <vscale x 8 x double> %c, <vscale x 8 x i1> %allones, i32 %evl)
+ ret <vscale x 8 x double> %v
+}
More information about the llvm-commits
mailing list