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

llvm-svn: 332206

6 files changed, 16 insertions, 22 deletions

diff --git a/llvm/include/llvm/IR/IntrinsicsX86.td b/llvm/include/llvm/IR/IntrinsicsX86.td
index aa8df70e7a2..8be8aed0820 100644
--- a/llvm/include/llvm/IR/IntrinsicsX86.td
+++ b/llvm/include/llvm/IR/IntrinsicsX86.td
@@ -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]>;
diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index 0a3ecaed6d7..43ed98f673e 100644
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -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());
diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index 050c2e4d00a..48c1913e57c 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/lib/Target/X86/X86InstrAVX512.td
@@ -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 {
diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index 5c1e263608f..e18ea717ee3 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -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:
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
index 3d7cfac82f4..eed3078da0b 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
@@ -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
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
index 1e9ba6cb117..e91f7322452 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
@@ -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: