RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-01 05:01:59 +01:00
Code Issues Projects Releases Wiki Activity

[X86] Remove and autoupgrade the cvtusi2sd intrinsic. Use uitofp+insertelement instead.

Browse Source 
llvm-svn: 332206
This commit is contained in:
Craig Topper 2018-05-14 00:06:49 +00:00
parent db32709343
commit 2e9c58a690
6 changed files with 16 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
include/llvm/IR/IntrinsicsX86.td
View File

@ -3611,9 +3611,6 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
Intrinsic<[llvm_i32_ty], [llvm_v2f64_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_cvttsd2usi64 : GCCBuiltin<"__builtin_ia32_vcvttsd2usi64">,
Intrinsic<[llvm_i64_ty], [llvm_v2f64_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_cvtusi2sd : GCCBuiltin<"__builtin_ia32_cvtusi2sd32">,
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty,
llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_cvtusi642sd : GCCBuiltin<"__builtin_ia32_cvtusi2sd64">,
Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty,
llvm_i64_ty, llvm_i32_ty], [IntrNoMem]>;

5
lib/IR/AutoUpgrade.cpp
View File

@ -181,6 +181,7 @@ static bool ShouldUpgradeX86Intrinsic(Function *F, StringRef Name) {
Name == "avx512.mask.cvttps2dq.256" || // Added in 7.0
Name == "avx512.mask.cvtps2pd.128" || // Added in 7.0
Name == "avx512.mask.cvtps2pd.256" || // Added in 7.0
Name == "avx512.cvtusi2sd" || // Added in 7.0
Name == "avx512.mask.permvar.sf.256" || // Added in 7.0
Name == "avx512.mask.permvar.si.256" || // Added in 7.0
Name == "sse2.pmulu.dq" || // Added in 7.0
@ -1566,6 +1567,10 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
Rep = Builder.CreateSIToFP(CI->getArgOperand(1),
CI->getType()->getVectorElementType());
Rep = Builder.CreateInsertElement(CI->getArgOperand(0), Rep, (uint64_t)0);
} else if (IsX86 && Name == "avx512.cvtusi2sd") {
Rep = Builder.CreateUIToFP(CI->getArgOperand(1),
CI->getType()->getVectorElementType());
Rep = Builder.CreateInsertElement(CI->getArgOperand(0), Rep, (uint64_t)0);
} else if (IsX86 && Name == "sse2.cvtss2sd") {
Rep = Builder.CreateExtractElement(CI->getArgOperand(1), (uint64_t)0);
Rep = Builder.CreateFPExt(Rep, CI->getType()->getVectorElementType());

7
lib/Target/X86/X86InstrAVX512.td
View File

@ -6921,13 +6921,6 @@ let Predicates = [HasAVX512] in {
(VCVTSD2SI64Zrm_Int sse_load_f64:$src)>;
} // HasAVX512
let Predicates = [HasAVX512] in {
def : Pat<(int_x86_avx512_cvtusi2sd VR128X:$src1, GR32:$src2),
(VCVTUSI2SDZrr_Int VR128X:$src1, GR32:$src2)>;
def : Pat<(int_x86_avx512_cvtusi2sd VR128X:$src1, (loadi32 addr:$src2)),
(VCVTUSI2SDZrm_Int VR128X:$src1, addr:$src2)>;
} // Predicates = [HasAVX512]
// Patterns used for matching vcvtsi2s{s,d} intrinsic sequences from clang
// which produce unnecessary vmovs{s,d} instructions
let Predicates = [HasAVX512] in {

1
lib/Transforms/Instrumentation/MemorySanitizer.cpp
View File

@ -2556,7 +2556,6 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
case Intrinsic::x86_avx512_cvttss2usi:
case Intrinsic::x86_avx512_cvttsd2usi64:
case Intrinsic::x86_avx512_cvttsd2usi:
case Intrinsic::x86_avx512_cvtusi2sd:
case Intrinsic::x86_avx512_cvtusi2ss:
case Intrinsic::x86_avx512_cvtusi642sd:
case Intrinsic::x86_avx512_cvtusi642ss:

11
test/CodeGen/X86/avx512-intrinsics-upgrade.ll
View File

@ -4120,3 +4120,14 @@ define <8 x i64> @test_mul_epu32_rmbkz(<16 x i32> %a, i64* %ptr_b, i8 %mask) {
}
declare <8 x i64> @llvm.x86.avx512.pmulu.dq.512(<16 x i32>, <16 x i32>)
define <2 x double> @test_x86_avx512_mm_cvtu32_sd(<2 x double> %a, i32 %b)
; CHECK-LABEL: test_x86_avx512_mm_cvtu32_sd:
; CHECK: ## %bb.0:
; CHECK-NEXT: vcvtusi2sdl %edi, %xmm0, %xmm0
; CHECK-NEXT: retq
{
%res = call <2 x double> @llvm.x86.avx512.cvtusi2sd(<2 x double> %a, i32 %b) ; <<<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.avx512.cvtusi2sd(<2 x double>, i32) nounwind readnone

11
test/CodeGen/X86/avx512-intrinsics.ll
View File

@ -1954,17 +1954,6 @@ define <4 x float> @_mm_cvtu64_ss(<4 x float> %a, i64 %b)
}
declare <4 x float> @llvm.x86.avx512.cvtusi642ss(<4 x float>, i64, i32) nounwind readnone
define <2 x double> @test_x86_avx512_mm_cvtu32_sd(<2 x double> %a, i32 %b)
; CHECK-LABEL: test_x86_avx512_mm_cvtu32_sd:
; CHECK: ## %bb.0:
; CHECK-NEXT: vcvtusi2sdl %edi, %xmm0, %xmm0
; CHECK-NEXT: retq
{
%res = call <2 x double> @llvm.x86.avx512.cvtusi2sd(<2 x double> %a, i32 %b) ; <<<2 x double>> [#uses=1]
ret <2 x double> %res
}
declare <2 x double> @llvm.x86.avx512.cvtusi2sd(<2 x double>, i32) nounwind readnone
define <2 x double> @test_x86_avx512_mm_cvtu64_sd(<2 x double> %a, i64 %b)
; CHECK-LABEL: test_x86_avx512_mm_cvtu64_sd:
; CHECK: ## %bb.0: