summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--llvm/lib/Target/X86/InstPrinter/X86InstComments.cpp6
-rw-r--r--llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp13
-rw-r--r--llvm/lib/Target/X86/Utils/X86ShuffleDecode.h4
-rw-r--r--llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll8
-rw-r--r--llvm/test/CodeGen/X86/avx-intrinsics-x86.ll8
5 files changed, 31 insertions, 8 deletions
diff --git a/llvm/lib/Target/X86/InstPrinter/X86InstComments.cpp b/llvm/lib/Target/X86/InstPrinter/X86InstComments.cpp
index 97e9ee3519b..f5379566b61 100644
--- a/llvm/lib/Target/X86/InstPrinter/X86InstComments.cpp
+++ b/llvm/lib/Target/X86/InstPrinter/X86InstComments.cpp
@@ -944,6 +944,12 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
Src2Name = getRegName(MI->getOperand(2).getReg());
break;
+ case X86::VBROADCASTF128:
+ case X86::VBROADCASTI128:
+ DecodeSubVectorBroadcast(MVT::v4f64, MVT::v2f64, ShuffleMask);
+ DestName = getRegName(MI->getOperand(0).getReg());
+ break;
+
CASE_PMOVZX(PMOVZXBW, r)
CASE_PMOVZX(PMOVZXBD, r)
CASE_PMOVZX(PMOVZXBQ, r)
diff --git a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index 1a51aa5858e..4fe9a9e949d 100644
--- a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -275,6 +275,19 @@ void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {
}
}
+/// Decodes a broadcast of a subvector to a larger vector type.
+void DecodeSubVectorBroadcast(MVT DstVT, MVT SrcVT,
+ SmallVectorImpl<int> &ShuffleMask) {
+ assert(SrcVT.getScalarType() == DstVT.getScalarType() &&
+ "Non matching vector element types");
+ unsigned NumElts = SrcVT.getVectorNumElements();
+ unsigned Scale = DstVT.getSizeInBits() / SrcVT.getSizeInBits();
+
+ for (unsigned i = 0; i != Scale; ++i)
+ for (unsigned j = 0; j != NumElts; ++j)
+ ShuffleMask.push_back(j);
+}
+
/// \brief Decode a shuffle packed values at 128-bit granularity
/// (SHUFF32x4/SHUFF64x2/SHUFI32x4/SHUFI64x2)
/// immediate mask into a shuffle mask.
diff --git a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
index 7caccefc0a8..30b17052b72 100644
--- a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -88,6 +88,10 @@ void DecodeUNPCKHMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
/// datatypes and vector widths.
void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
+/// Decodes a broadcast of a subvector to a larger vector type.
+void DecodeSubVectorBroadcast(MVT DstVT, MVT SrcVT,
+ SmallVectorImpl<int> &ShuffleMask);
+
/// Decode a PSHUFB mask from a raw array of constants such as from
/// BUILD_VECTOR.
void DecodePSHUFBMask(ArrayRef<uint64_t> RawMask,
diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
index cca8f99ec01..c7cf857e1d4 100644
--- a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
@@ -200,12 +200,12 @@ define <4 x double> @test_mm256_broadcast_pd(<2 x double>* %a0) nounwind {
; X32-LABEL: test_mm256_broadcast_pd:
; X32: # BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vbroadcastf128 (%eax), %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_mm256_broadcast_pd:
; X64: # BB#0:
-; X64-NEXT: vbroadcastf128 (%rdi), %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%arg0 = bitcast <2 x double>* %a0 to i8*
%res = call <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8* %arg0)
@@ -217,12 +217,12 @@ define <8 x float> @test_mm256_broadcast_ps(<4 x float>* %a0) nounwind {
; X32-LABEL: test_mm256_broadcast_ps:
; X32: # BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT: vbroadcastf128 (%eax), %ymm0
+; X32-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X32-NEXT: retl
;
; X64-LABEL: test_mm256_broadcast_ps:
; X64: # BB#0:
-; X64-NEXT: vbroadcastf128 (%rdi), %ymm0
+; X64-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; X64-NEXT: retq
%arg0 = bitcast <4 x float>* %a0 to i8*
%res = call <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8* %arg0)
diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll b/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll
index a7421be86e9..35763297d81 100644
--- a/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll
@@ -3941,13 +3941,13 @@ define <4 x double> @test_x86_avx_vbroadcastf128_pd_256(i8* %a0) {
; AVX-LABEL: test_x86_avx_vbroadcastf128_pd_256:
; AVX: ## BB#0:
; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX-NEXT: vbroadcastf128 (%eax), %ymm0
+; AVX-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX-NEXT: retl
;
; AVX512VL-LABEL: test_x86_avx_vbroadcastf128_pd_256:
; AVX512VL: ## BB#0:
; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX512VL-NEXT: vbroadcastf128 (%eax), %ymm0
+; AVX512VL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX512VL-NEXT: retl
%res = call <4 x double> @llvm.x86.avx.vbroadcastf128.pd.256(i8* %a0) ; <<4 x double>> [#uses=1]
ret <4 x double> %res
@@ -3959,13 +3959,13 @@ define <8 x float> @test_x86_avx_vbroadcastf128_ps_256(i8* %a0) {
; AVX-LABEL: test_x86_avx_vbroadcastf128_ps_256:
; AVX: ## BB#0:
; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX-NEXT: vbroadcastf128 (%eax), %ymm0
+; AVX-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX-NEXT: retl
;
; AVX512VL-LABEL: test_x86_avx_vbroadcastf128_ps_256:
; AVX512VL: ## BB#0:
; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
-; AVX512VL-NEXT: vbroadcastf128 (%eax), %ymm0
+; AVX512VL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
; AVX512VL-NEXT: retl
%res = call <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8* %a0) ; <<8 x float>> [#uses=1]
ret <8 x float> %res
OpenPOWER on IntegriCloud