[llvm] 671cbc1 - [AArch64][SVE] Add mul/mla/mls lane & dup intrinsics
Kerry McLaughlin via llvm-commits
llvm-commits at lists.llvm.org
Thu Feb 13 02:41:25 PST 2020
Author: Kerry McLaughlin
Date: 2020-02-13T10:32:59Z
New Revision: 671cbc1fbba049ec2343cdcff069ce59c0160e31
URL: https://github.com/llvm/llvm-project/commit/671cbc1fbba049ec2343cdcff069ce59c0160e31
DIFF: https://github.com/llvm/llvm-project/commit/671cbc1fbba049ec2343cdcff069ce59c0160e31.diff
LOG: [AArch64][SVE] Add mul/mla/mls lane & dup intrinsics
Summary:
Implements the following intrinsics:
- @llvm.aarch64.sve.dup
- @llvm.aarch64.sve.mul.lane
- @llvm.aarch64.sve.mla.lane
- @llvm.aarch64.sve.mls.lane
Reviewers: c-rhodes, sdesmalen, dancgr, efriedma, rengolin
Reviewed By: sdesmalen
Subscribers: tschuett, kristof.beyls, hiraditya, rkruppe, psnobl, cfe-commits, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74222
Added:
llvm/test/CodeGen/AArch64/sve-intrinsics-scalar-to-vec.ll
llvm/test/CodeGen/AArch64/sve2-intrinsics-int-mul-lane.ll
Modified:
llvm/include/llvm/IR/IntrinsicsAArch64.td
llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
llvm/lib/Target/AArch64/AArch64ISelLowering.h
llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
llvm/lib/Target/AArch64/SVEInstrFormats.td
Removed:
################################################################################
diff --git a/llvm/include/llvm/IR/IntrinsicsAArch64.td b/llvm/include/llvm/IR/IntrinsicsAArch64.td
index fbc808ca642c..245eb1358f10 100644
--- a/llvm/include/llvm/IR/IntrinsicsAArch64.td
+++ b/llvm/include/llvm/IR/IntrinsicsAArch64.td
@@ -948,6 +948,13 @@ let TargetPrefix = "aarch64" in { // All intrinsics start with "llvm.aarch64.".
llvm_i32_ty],
[IntrNoMem, ImmArg<3>, ImmArg<4>]>;
+ class AdvSIMD_SVE_DUP_Intrinsic
+ : Intrinsic<[llvm_anyvector_ty],
+ [LLVMMatchType<0>,
+ LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
+ LLVMVectorElementType<0>],
+ [IntrNoMem]>;
+
class AdvSIMD_SVE_EXPA_Intrinsic
: Intrinsic<[llvm_anyvector_ty],
[LLVMVectorOfBitcastsToInt<0>],
@@ -1224,6 +1231,12 @@ def int_aarch64_sve_ldff1 : AdvSIMD_1Vec_PredFaultingLoad_Intrinsic;
def int_aarch64_sve_stnt1 : AdvSIMD_1Vec_PredStore_Intrinsic;
+//
+// Scalar to vector operations
+//
+
+def int_aarch64_sve_dup : AdvSIMD_SVE_DUP_Intrinsic;
+
//
// Integer arithmetic
//
@@ -1235,6 +1248,7 @@ def int_aarch64_sve_subr : AdvSIMD_Pred2VectorArg_Intrinsic;
def int_aarch64_sve_pmul : AdvSIMD_2VectorArg_Intrinsic;
def int_aarch64_sve_mul : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_mul_lane : AdvSIMD_2VectorArgIndexed_Intrinsic;
def int_aarch64_sve_smulh : AdvSIMD_Pred2VectorArg_Intrinsic;
def int_aarch64_sve_umulh : AdvSIMD_Pred2VectorArg_Intrinsic;
@@ -1253,7 +1267,9 @@ def int_aarch64_sve_uabd : AdvSIMD_Pred2VectorArg_Intrinsic;
def int_aarch64_sve_mad : AdvSIMD_Pred3VectorArg_Intrinsic;
def int_aarch64_sve_msb : AdvSIMD_Pred3VectorArg_Intrinsic;
def int_aarch64_sve_mla : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_mla_lane : AdvSIMD_3VectorArgIndexed_Intrinsic;
def int_aarch64_sve_mls : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_mls_lane : AdvSIMD_3VectorArgIndexed_Intrinsic;
def int_aarch64_sve_saddv : AdvSIMD_SVE_SADDV_Reduce_Intrinsic;
def int_aarch64_sve_uaddv : AdvSIMD_SVE_SADDV_Reduce_Intrinsic;
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index e01762e2ca9b..9736a18832c0 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -1425,6 +1425,7 @@ const char *AArch64TargetLowering::getTargetNodeName(unsigned Opcode) const {
case AArch64ISD::LDP: return "AArch64ISD::LDP";
case AArch64ISD::STP: return "AArch64ISD::STP";
case AArch64ISD::STNP: return "AArch64ISD::STNP";
+ case AArch64ISD::DUP_PRED: return "AArch64ISD::DUP_PRED";
}
return nullptr;
}
@@ -10917,6 +10918,18 @@ static SDValue LowerSVEIntReduction(SDNode *N, unsigned Opc,
return SDValue();
}
+static SDValue LowerSVEIntrinsicDUP(SDNode *N, SelectionDAG &DAG) {
+ SDLoc dl(N);
+ SDValue Scalar = N->getOperand(3);
+ EVT ScalarTy = Scalar.getValueType();
+
+ if ((ScalarTy == MVT::i8) || (ScalarTy == MVT::i16))
+ Scalar = DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i32, Scalar);
+
+ return DAG.getNode(AArch64ISD::DUP_PRED, dl, N->getValueType(0),
+ N->getOperand(1), N->getOperand(2), Scalar);
+}
+
static SDValue LowerSVEIntrinsicEXT(SDNode *N, SelectionDAG &DAG) {
SDLoc dl(N);
LLVMContext &Ctx = *DAG.getContext();
@@ -11105,6 +11118,8 @@ static SDValue performIntrinsicCombine(SDNode *N,
return LowerSVEIntReduction(N, AArch64ISD::EORV_PRED, DAG);
case Intrinsic::aarch64_sve_andv:
return LowerSVEIntReduction(N, AArch64ISD::ANDV_PRED, DAG);
+ case Intrinsic::aarch64_sve_dup:
+ return LowerSVEIntrinsicDUP(N, DAG);
case Intrinsic::aarch64_sve_ext:
return LowerSVEIntrinsicEXT(N, DAG);
case Intrinsic::aarch64_sve_cmpeq_wide:
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
index 363302c359ff..f664484a8803 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -215,6 +215,8 @@ enum NodeType : unsigned {
PTEST,
PTRUE,
+ DUP_PRED,
+
LDNF1,
LDNF1S,
LDFF1,
diff --git a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
index a986348dff32..5bcc60fcb828 100644
--- a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
@@ -96,6 +96,9 @@ def AArch64rev : SDNode<"AArch64ISD::REV", SDT_AArch64Rev>;
def SDT_AArch64PTest : SDTypeProfile<0, 2, [SDTCisVec<0>, SDTCisSameAs<0,1>]>;
def AArch64ptest : SDNode<"AArch64ISD::PTEST", SDT_AArch64PTest>;
+def SDT_AArch64DUP_PRED : SDTypeProfile<1, 3, [SDTCisVec<0>, SDTCisSameAs<0,1>, SDTCisVec<2>, SDTCVecEltisVT<2,i1>]>;
+def AArch64dup_pred : SDNode<"AArch64ISD::DUP_PRED", SDT_AArch64DUP_PRED>;
+
let Predicates = [HasSVE] in {
defm RDFFR_PPz : sve_int_rdffr_pred<0b0, "rdffr", int_aarch64_sve_rdffr_z>;
@@ -287,8 +290,8 @@ let Predicates = [HasSVE] in {
defm DUP_ZZI : sve_int_perm_dup_i<"dup">;
// Splat scalar register (predicated)
- defm CPY_ZPmR : sve_int_perm_cpy_r<"cpy">;
- defm CPY_ZPmV : sve_int_perm_cpy_v<"cpy">;
+ defm CPY_ZPmR : sve_int_perm_cpy_r<"cpy", AArch64dup_pred>;
+ defm CPY_ZPmV : sve_int_perm_cpy_v<"cpy", AArch64dup_pred>;
// Select elements from either vector (predicated)
defm SEL_ZPZZ : sve_int_sel_vvv<"sel", vselect>;
@@ -1396,8 +1399,8 @@ let Predicates = [HasSVE] in {
let Predicates = [HasSVE2] in {
// SVE2 integer multiply-add (indexed)
- defm MLA_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b0, "mla", null_frag>;
- defm MLS_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b1, "mls", null_frag>;
+ defm MLA_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b0, "mla", int_aarch64_sve_mla_lane>;
+ defm MLS_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b1, "mls", int_aarch64_sve_mls_lane>;
// SVE2 saturating multiply-add high (indexed)
defm SQRDMLAH_ZZZI : sve2_int_mla_by_indexed_elem<0b10, 0b0, "sqrdmlah", int_aarch64_sve_sqrdmlah_lane>;
@@ -1408,7 +1411,7 @@ let Predicates = [HasSVE2] in {
defm SQRDMLSH_ZZZ : sve2_int_mla<0b1, "sqrdmlsh", int_aarch64_sve_sqrdmlsh>;
// SVE2 integer multiply (indexed)
- defm MUL_ZZZI : sve2_int_mul_by_indexed_elem<0b1110, "mul", null_frag>;
+ defm MUL_ZZZI : sve2_int_mul_by_indexed_elem<0b1110, "mul", int_aarch64_sve_mul_lane>;
// SVE2 saturating multiply high (indexed)
defm SQDMULH_ZZZI : sve2_int_mul_by_indexed_elem<0b1100, "sqdmulh", int_aarch64_sve_sqdmulh_lane>;
diff --git a/llvm/lib/Target/AArch64/SVEInstrFormats.td b/llvm/lib/Target/AArch64/SVEInstrFormats.td
index 2df9f890baa1..d0f80b869ab0 100644
--- a/llvm/lib/Target/AArch64/SVEInstrFormats.td
+++ b/llvm/lib/Target/AArch64/SVEInstrFormats.td
@@ -5603,7 +5603,7 @@ class sve_int_perm_cpy_r<bits<2> sz8_64, string asm, ZPRRegOp zprty,
let ElementSize = zprty.ElementSize;
}
-multiclass sve_int_perm_cpy_r<string asm> {
+multiclass sve_int_perm_cpy_r<string asm, SDPatternOperator op> {
def _B : sve_int_perm_cpy_r<0b00, asm, ZPR8, GPR32sp>;
def _H : sve_int_perm_cpy_r<0b01, asm, ZPR16, GPR32sp>;
def _S : sve_int_perm_cpy_r<0b10, asm, ZPR32, GPR32sp>;
@@ -5617,6 +5617,11 @@ multiclass sve_int_perm_cpy_r<string asm> {
(!cast<Instruction>(NAME # _S) ZPR32:$Zd, PPR3bAny:$Pg, GPR32sp:$Rn), 1>;
def : InstAlias<"mov $Zd, $Pg/m, $Rn",
(!cast<Instruction>(NAME # _D) ZPR64:$Zd, PPR3bAny:$Pg, GPR64sp:$Rn), 1>;
+
+ def : SVE_3_Op_Pat<nxv16i8, op, nxv16i8, nxv16i1, i32, !cast<Instruction>(NAME # _B)>;
+ def : SVE_3_Op_Pat<nxv8i16, op, nxv8i16, nxv8i1, i32, !cast<Instruction>(NAME # _H)>;
+ def : SVE_3_Op_Pat<nxv4i32, op, nxv4i32, nxv4i1, i32, !cast<Instruction>(NAME # _S)>;
+ def : SVE_3_Op_Pat<nxv2i64, op, nxv2i64, nxv2i1, i64, !cast<Instruction>(NAME # _D)>;
}
class sve_int_perm_cpy_v<bits<2> sz8_64, string asm, ZPRRegOp zprty,
@@ -5640,7 +5645,7 @@ class sve_int_perm_cpy_v<bits<2> sz8_64, string asm, ZPRRegOp zprty,
let ElementSize = zprty.ElementSize;
}
-multiclass sve_int_perm_cpy_v<string asm> {
+multiclass sve_int_perm_cpy_v<string asm, SDPatternOperator op> {
def _B : sve_int_perm_cpy_v<0b00, asm, ZPR8, FPR8>;
def _H : sve_int_perm_cpy_v<0b01, asm, ZPR16, FPR16>;
def _S : sve_int_perm_cpy_v<0b10, asm, ZPR32, FPR32>;
@@ -5654,6 +5659,11 @@ multiclass sve_int_perm_cpy_v<string asm> {
(!cast<Instruction>(NAME # _S) ZPR32:$Zd, PPR3bAny:$Pg, FPR32:$Vn), 1>;
def : InstAlias<"mov $Zd, $Pg/m, $Vn",
(!cast<Instruction>(NAME # _D) ZPR64:$Zd, PPR3bAny:$Pg, FPR64:$Vn), 1>;
+
+ def : SVE_3_Op_Pat<nxv8f16, op, nxv8f16, nxv8i1, f16, !cast<Instruction>(NAME # _H)>;
+ def : SVE_3_Op_Pat<nxv4f32, op, nxv4f32, nxv4i1, f32, !cast<Instruction>(NAME # _S)>;
+ def : SVE_3_Op_Pat<nxv2f32, op, nxv2f32, nxv2i1, f32, !cast<Instruction>(NAME # _S)>;
+ def : SVE_3_Op_Pat<nxv2f64, op, nxv2f64, nxv2i1, f64, !cast<Instruction>(NAME # _D)>;
}
class sve_int_perm_compact<bit sz, string asm, ZPRRegOp zprty>
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-scalar-to-vec.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-scalar-to-vec.ll
new file mode 100644
index 000000000000..5baafcd9e7da
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-scalar-to-vec.ll
@@ -0,0 +1,83 @@
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
+
+;
+; DUP
+;
+
+define <vscale x 16 x i8> @dup_i8(<vscale x 16 x i8> %a, <vscale x 16 x i1> %pg, i8 %b) {
+; CHECK-LABEL: dup_i8:
+; CHECK: mov z0.b, p0/m, w0
+; CHECK-NEXT: ret
+ %out = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.nxv16i8(<vscale x 16 x i8> %a,
+ <vscale x 16 x i1> %pg,
+ i8 %b)
+ ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @dup_i16(<vscale x 8 x i16> %a, <vscale x 8 x i1> %pg, i16 %b) {
+; CHECK-LABEL: dup_i16:
+; CHECK: mov z0.h, p0/m, w0
+; CHECK-NEXT: ret
+ %out = call <vscale x 8 x i16> @llvm.aarch64.sve.dup.nxv8i16(<vscale x 8 x i16> %a,
+ <vscale x 8 x i1> %pg,
+ i16 %b)
+ ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @dup_i32(<vscale x 4 x i32> %a, <vscale x 4 x i1> %pg, i32 %b) {
+; CHECK-LABEL: dup_i32:
+; CHECK: mov z0.s, p0/m, w0
+; CHECK-NEXT: ret
+ %out = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.nxv4i32(<vscale x 4 x i32> %a,
+ <vscale x 4 x i1> %pg,
+ i32 %b)
+ ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @dup_i64(<vscale x 2 x i64> %a, <vscale x 2 x i1> %pg, i64 %b) {
+; CHECK-LABEL: dup_i64:
+; CHECK: mov z0.d, p0/m, x0
+; CHECK-NEXT: ret
+ %out = call <vscale x 2 x i64> @llvm.aarch64.sve.dup.nxv2i64(<vscale x 2 x i64> %a,
+ <vscale x 2 x i1> %pg,
+ i64 %b)
+ ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 8 x half> @dup_f16(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, half %b) {
+; CHECK-LABEL: dup_f16:
+; CHECK: mov z0.h, p0/m, h1
+; CHECK-NEXT: ret
+ %out = call <vscale x 8 x half> @llvm.aarch64.sve.dup.nxv8f16(<vscale x 8 x half> %a,
+ <vscale x 8 x i1> %pg,
+ half %b)
+ ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @dup_f32(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, float %b) {
+; CHECK-LABEL: dup_f32:
+; CHECK: mov z0.s, p0/m, s1
+; CHECK-NEXT: ret
+ %out = call <vscale x 4 x float> @llvm.aarch64.sve.dup.nxv4f32(<vscale x 4 x float> %a,
+ <vscale x 4 x i1> %pg,
+ float %b)
+ ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @dup_f64(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, double %b) {
+; CHECK-LABEL: dup_f64:
+; CHECK: mov z0.d, p0/m, d1
+; CHECK-NEXT: ret
+ %out = call <vscale x 2 x double> @llvm.aarch64.sve.dup.nxv2f64(<vscale x 2 x double> %a,
+ <vscale x 2 x i1> %pg,
+ double %b)
+ ret <vscale x 2 x double> %out
+}
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.dup.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i1>, i8)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.dup.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i1>, i16)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.dup.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i1>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.dup.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, i64)
+declare <vscale x 8 x half> @llvm.aarch64.sve.dup.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, half)
+declare <vscale x 4 x float> @llvm.aarch64.sve.dup.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, float)
+declare <vscale x 2 x double> @llvm.aarch64.sve.dup.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, double)
diff --git a/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-mul-lane.ll b/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-mul-lane.ll
new file mode 100644
index 000000000000..6e7e2eabd2a0
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-mul-lane.ll
@@ -0,0 +1,119 @@
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s | FileCheck %s
+
+;
+; MUL
+;
+
+define <vscale x 2 x i64> @mul_lane_d(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: mul_lane_d:
+; CHECK: mul z0.d, z0.d, z1.d[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 2 x i64> @llvm.aarch64.sve.mul.lane.nxv2i64(<vscale x 2 x i64> %a,
+ <vscale x 2 x i64> %b,
+ i32 1)
+ ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 4 x i32> @mul_lane_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: mul_lane_s:
+; CHECK: mul z0.s, z0.s, z1.s[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 4 x i32> @llvm.aarch64.sve.mul.lane.nxv4i32(<vscale x 4 x i32> %a,
+ <vscale x 4 x i32> %b,
+ i32 1)
+ ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 8 x i16> @mul_lane_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: mul_lane_h:
+; CHECK: mul z0.h, z0.h, z1.h[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 8 x i16> @llvm.aarch64.sve.mul.lane.nxv8i16(<vscale x 8 x i16> %a,
+ <vscale x 8 x i16> %b,
+ i32 1)
+ ret <vscale x 8 x i16> %out
+}
+
+;
+; MLA
+;
+
+define <vscale x 2 x i64> @mla_lane_d(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
+; CHECK-LABEL: mla_lane_d:
+; CHECK: mla z0.d, z1.d, z2.d[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 2 x i64> @llvm.aarch64.sve.mla.lane.nxv2i64(<vscale x 2 x i64> %a,
+ <vscale x 2 x i64> %b,
+ <vscale x 2 x i64> %c,
+ i32 1)
+ ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 4 x i32> @mla_lane_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
+; CHECK-LABEL: mla_lane_s:
+; CHECK: mla z0.s, z1.s, z2.s[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 4 x i32> @llvm.aarch64.sve.mla.lane.nxv4i32(<vscale x 4 x i32> %a,
+ <vscale x 4 x i32> %b,
+ <vscale x 4 x i32> %c,
+ i32 1)
+ ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 8 x i16> @mla_lane_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
+; CHECK-LABEL: mla_lane_h:
+; CHECK: mla z0.h, z1.h, z2.h[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 8 x i16> @llvm.aarch64.sve.mla.lane.nxv8i16(<vscale x 8 x i16> %a,
+ <vscale x 8 x i16> %b,
+ <vscale x 8 x i16> %c,
+ i32 1)
+ ret <vscale x 8 x i16> %out
+}
+
+;
+; MLS
+;
+
+define <vscale x 2 x i64> @mls_lane_d(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
+; CHECK-LABEL: mls_lane_d:
+; CHECK: mls z0.d, z1.d, z2.d[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 2 x i64> @llvm.aarch64.sve.mls.lane.nxv2i64(<vscale x 2 x i64> %a,
+ <vscale x 2 x i64> %b,
+ <vscale x 2 x i64> %c,
+ i32 1)
+ ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 4 x i32> @mls_lane_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
+; CHECK-LABEL: mls_lane_s:
+; CHECK: mls z0.s, z1.s, z2.s[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 4 x i32> @llvm.aarch64.sve.mls.lane.nxv4i32(<vscale x 4 x i32> %a,
+ <vscale x 4 x i32> %b,
+ <vscale x 4 x i32> %c,
+ i32 1)
+ ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 8 x i16> @mls_lane_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
+; CHECK-LABEL: mls_lane_h:
+; CHECK: mls z0.h, z1.h, z2.h[1]
+; CHECK-NEXT: ret
+ %out = call <vscale x 8 x i16> @llvm.aarch64.sve.mls.lane.nxv8i16(<vscale x 8 x i16> %a,
+ <vscale x 8 x i16> %b,
+ <vscale x 8 x i16> %c,
+ i32 1)
+ ret <vscale x 8 x i16> %out
+}
+
+declare <vscale x 8 x i16> @llvm.aarch64.sve.mul.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.mul.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.mul.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.mla.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.mla.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.mla.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.mls.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.mls.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.mls.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, i32)
More information about the llvm-commits
mailing list