[X86][SSE] Reimplement SSE fp2si conversion intrinsics instead of using generic IR

D20859 and D20860 attempted to replace the SSE (V)CVTTPS2DQ and VCVTTPD2DQ truncating conversions with generic IR instead.

It turns out that the behaviour of these intrinsics is different enough from generic IR that this will cause problems, INF/NAN/out of range values are guaranteed to result in a 0x80000000 value - which plays havoc with constant folding which converts them to either zero or UNDEF. This is also an issue with the scalar implementations (which were already generic IR and what I was trying to match).

This patch changes both scalar and packed versions back to using x86-specific builtins.

It also deals with the other scalar conversion cases that are runtime rounding mode dependent and can have similar issues with constant folding.

A companion clang patch is at D22105

Differential Revision: https://reviews.llvm.org/D22106

llvm-svn: 275981

diff --git a/llvm/include/llvm/IR/IntrinsicsX86.td b/llvm/include/llvm/IR/IntrinsicsX86.td
index 74c9715..b965f08 100644 (file)
--- a/llvm/include/llvm/IR/IntrinsicsX86.td
+++ b/llvm/include/llvm/IR/IntrinsicsX86.td
@@ -479,6 +479,8 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
               Intrinsic<[llvm_v4f32_ty], [llvm_v2f64_ty], [IntrNoMem]>;
   def int_x86_sse2_cvtps2dq : GCCBuiltin<"__builtin_ia32_cvtps2dq">,
               Intrinsic<[llvm_v4i32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
+  def int_x86_sse2_cvttps2dq : GCCBuiltin<"__builtin_ia32_cvttps2dq">,
+              Intrinsic<[llvm_v4i32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
   def int_x86_sse2_cvtsd2si : GCCBuiltin<"__builtin_ia32_cvtsd2si">,
               Intrinsic<[llvm_i32_ty], [llvm_v2f64_ty], [IntrNoMem]>;
   def int_x86_sse2_cvtsd2si64 : GCCBuiltin<"__builtin_ia32_cvtsd2si64">,
@@ -1512,8 +1514,12 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
         Intrinsic<[llvm_v4f32_ty], [llvm_v4f64_ty], [IntrNoMem]>;
   def int_x86_avx_cvt_ps2dq_256 : GCCBuiltin<"__builtin_ia32_cvtps2dq256">,
         Intrinsic<[llvm_v8i32_ty], [llvm_v8f32_ty], [IntrNoMem]>;
+  def int_x86_avx_cvtt_pd2dq_256 : GCCBuiltin<"__builtin_ia32_cvttpd2dq256">,
+        Intrinsic<[llvm_v4i32_ty], [llvm_v4f64_ty], [IntrNoMem]>;
   def int_x86_avx_cvt_pd2dq_256 : GCCBuiltin<"__builtin_ia32_cvtpd2dq256">,
         Intrinsic<[llvm_v4i32_ty], [llvm_v4f64_ty], [IntrNoMem]>;
+  def int_x86_avx_cvtt_ps2dq_256 : GCCBuiltin<"__builtin_ia32_cvttps2dq256">,
+        Intrinsic<[llvm_v8i32_ty], [llvm_v8f32_ty], [IntrNoMem]>;
 }
 
 // Vector bit test

diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index 6c471ab..c9adaa7 100644 (file)
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -1424,8 +1424,8 @@ Constant *ConstantFoldBinaryFP(double (*NativeFP)(double, double), double V,
 /// integer type Ty is used to select how many bits are available for the
 /// result. Returns null if the conversion cannot be performed, otherwise
 /// returns the Constant value resulting from the conversion.
-Constant *ConstantFoldConvertToInt(const APFloat &Val, bool roundTowardZero,
-                                   Type *Ty) {
+Constant *ConstantFoldSSEConvertToInt(const APFloat &Val, bool roundTowardZero,
+                                      Type *Ty) {
   // All of these conversion intrinsics form an integer of at most 64bits.
   unsigned ResultWidth = Ty->getIntegerBitWidth();
   assert(ResultWidth <= 64 &&
@@ -1438,7 +1438,8 @@ Constant *ConstantFoldConvertToInt(const APFloat &Val, bool roundTowardZero,
   APFloat::opStatus status = Val.convertToInteger(&UIntVal, ResultWidth,
                                                   /*isSigned=*/true, mode,
                                                   &isExact);
-  if (status != APFloat::opOK && status != APFloat::opInexact)
+  if (status != APFloat::opOK &&
+      (!roundTowardZero || status != APFloat::opInexact))
     return nullptr;
   return ConstantInt::get(Ty, UIntVal, /*isSigned=*/true);
 }
@@ -1676,17 +1677,17 @@ Constant *ConstantFoldScalarCall(StringRef Name, unsigned IntrinsicID, Type *Ty,
       case Intrinsic::x86_sse2_cvtsd2si:
       case Intrinsic::x86_sse2_cvtsd2si64:
         if (ConstantFP *FPOp =
-              dyn_cast_or_null<ConstantFP>(Op->getAggregateElement(0U)))
-          return ConstantFoldConvertToInt(FPOp->getValueAPF(),
-                                          /*roundTowardZero=*/false, Ty);
+                dyn_cast_or_null<ConstantFP>(Op->getAggregateElement(0U)))
+          return ConstantFoldSSEConvertToInt(FPOp->getValueAPF(),
+                                             /*roundTowardZero=*/false, Ty);
       case Intrinsic::x86_sse_cvttss2si:
       case Intrinsic::x86_sse_cvttss2si64:
       case Intrinsic::x86_sse2_cvttsd2si:
       case Intrinsic::x86_sse2_cvttsd2si64:
         if (ConstantFP *FPOp =
-              dyn_cast_or_null<ConstantFP>(Op->getAggregateElement(0U)))
-          return ConstantFoldConvertToInt(FPOp->getValueAPF(),
-                                          /*roundTowardZero=*/true, Ty);
+                dyn_cast_or_null<ConstantFP>(Op->getAggregateElement(0U)))
+          return ConstantFoldSSEConvertToInt(FPOp->getValueAPF(),
+                                             /*roundTowardZero=*/true, Ty);
       }
     }
 

diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index 431e51b..2e4a2f8 100644 (file)
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -251,8 +251,6 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
          Name == "sse2.cvtps2pd" ||
          Name == "avx.cvtdq2.pd.256" ||
          Name == "avx.cvt.ps2.pd.256" ||
-         Name == "sse2.cvttps2dq" ||
-         Name.startswith("avx.cvtt.") ||
          Name.startswith("avx.vinsertf128.") ||
          Name == "avx2.vinserti128" ||
          Name.startswith("avx.vextractf128.") ||
@@ -712,12 +710,6 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
         Rep = Builder.CreateSIToFP(Rep, DstTy, "cvtdq2pd");
       else
         Rep = Builder.CreateFPExt(Rep, DstTy, "cvtps2pd");
-    } else if (IsX86 && (Name == "sse2.cvttps2dq" ||
-                         Name.startswith("avx.cvtt."))) {
-      // Truncation (round to zero) float/double to i32 vector conversion.
-      Value *Src = CI->getArgOperand(0);
-      VectorType *DstTy = cast<VectorType>(CI->getType());
-      Rep = Builder.CreateFPToSI(Src, DstTy, "cvtt");
     } else if (IsX86 && Name.startswith("sse4a.movnt.")) {
       Module *M = F->getParent();
       SmallVector<Metadata *, 1> Elts;

diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 3972fa9..34a5aaf 100644 (file)
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -2009,24 +2009,35 @@ def CVTPD2DQrr  : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
 // SSE2 packed instructions with XS prefix
 def VCVTTPS2DQrr : VS2SI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                          "cvttps2dq\t{$src, $dst|$dst, $src}",
-                         [], IIC_SSE_CVT_PS_RR>, VEX, Sched<[WriteCvtF2I]>;
+                         [(set VR128:$dst,
+                           (int_x86_sse2_cvttps2dq VR128:$src))],
+                         IIC_SSE_CVT_PS_RR>, VEX, Sched<[WriteCvtF2I]>;
 def VCVTTPS2DQrm : VS2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                          "cvttps2dq\t{$src, $dst|$dst, $src}",
-                         [], IIC_SSE_CVT_PS_RM>, VEX, Sched<[WriteCvtF2ILd]>;
+                         [(set VR128:$dst, (int_x86_sse2_cvttps2dq
+                                            (loadv4f32 addr:$src)))],
+                         IIC_SSE_CVT_PS_RM>, VEX, Sched<[WriteCvtF2ILd]>;
 def VCVTTPS2DQYrr : VS2SI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
                           "cvttps2dq\t{$src, $dst|$dst, $src}",
-                          [], IIC_SSE_CVT_PS_RR>, VEX, VEX_L, Sched<[WriteCvtF2I]>;
+                          [(set VR256:$dst,
+                            (int_x86_avx_cvtt_ps2dq_256 VR256:$src))],
+                          IIC_SSE_CVT_PS_RR>, VEX, VEX_L, Sched<[WriteCvtF2I]>;
 def VCVTTPS2DQYrm : VS2SI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
                           "cvttps2dq\t{$src, $dst|$dst, $src}",
-                          [], IIC_SSE_CVT_PS_RM>, VEX, VEX_L,
+                          [(set VR256:$dst, (int_x86_avx_cvtt_ps2dq_256
+                                             (loadv8f32 addr:$src)))],
+                          IIC_SSE_CVT_PS_RM>, VEX, VEX_L,
                           Sched<[WriteCvtF2ILd]>;
 
 def CVTTPS2DQrr : S2SI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        "cvttps2dq\t{$src, $dst|$dst, $src}",
-                       [], IIC_SSE_CVT_PS_RR>, Sched<[WriteCvtF2I]>;
+                       [(set VR128:$dst, (int_x86_sse2_cvttps2dq VR128:$src))],
+                       IIC_SSE_CVT_PS_RR>, Sched<[WriteCvtF2I]>;
 def CVTTPS2DQrm : S2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                        "cvttps2dq\t{$src, $dst|$dst, $src}",
-                       [], IIC_SSE_CVT_PS_RM>, Sched<[WriteCvtF2ILd]>;
+                       [(set VR128:$dst,
+                         (int_x86_sse2_cvttps2dq (memopv4f32 addr:$src)))],
+                       IIC_SSE_CVT_PS_RM>, Sched<[WriteCvtF2ILd]>;
 
 let Predicates = [HasAVX] in {
   def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
@@ -2096,10 +2107,14 @@ def VCVTTPD2DQXrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
 // YMM only
 def VCVTTPD2DQYrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
                          "cvttpd2dq{y}\t{$src, $dst|$dst, $src}",
-                         [], IIC_SSE_CVT_PD_RR>, VEX, VEX_L, Sched<[WriteCvtF2I]>;
+                         [(set VR128:$dst,
+                           (int_x86_avx_cvtt_pd2dq_256 VR256:$src))],
+                         IIC_SSE_CVT_PD_RR>, VEX, VEX_L, Sched<[WriteCvtF2I]>;
 def VCVTTPD2DQYrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
                          "cvttpd2dq{y}\t{$src, $dst|$dst, $src}",
-                         [], IIC_SSE_CVT_PD_RM>, VEX, VEX_L, Sched<[WriteCvtF2ILd]>;
+                         [(set VR128:$dst,
+                          (int_x86_avx_cvtt_pd2dq_256 (loadv4f64 addr:$src)))],
+                         IIC_SSE_CVT_PD_RM>, VEX, VEX_L, Sched<[WriteCvtF2ILd]>;
 def : InstAlias<"vcvttpd2dq\t{$src, $dst|$dst, $src}",
                 (VCVTTPD2DQYrr VR128:$dst, VR256:$src), 0>;
 

diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
index c7cf857..f886e1f 100644 (file)
--- a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
@@ -681,10 +681,11 @@ define <2 x i64> @test_mm256_cvttpd_epi32(<4 x double> %a0) nounwind {
 ; X64-NEXT:    vcvttpd2dqy %ymm0, %xmm0
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
-  %cvt = fptosi <4 x double> %a0 to <4 x i32>
+  %cvt = call <4 x i32> @llvm.x86.avx.cvtt.pd2dq.256(<4 x double> %a0)
   %res = bitcast <4 x i32> %cvt to <2 x i64>
   ret <2 x i64> %res
 }
+declare <4 x i32> @llvm.x86.avx.cvtt.pd2dq.256(<4 x double>) nounwind readnone
 
 define <4 x i64> @test_mm256_cvttps_epi32(<8 x float> %a0) nounwind {
 ; X32-LABEL: test_mm256_cvttps_epi32:
@@ -696,10 +697,11 @@ define <4 x i64> @test_mm256_cvttps_epi32(<8 x float> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    vcvttps2dq %ymm0, %ymm0
 ; X64-NEXT:    retq
-  %cvt = fptosi <8 x float> %a0 to <8 x i32>
+  %cvt = call <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float> %a0)
   %res = bitcast <8 x i32> %cvt to <4 x i64>
   ret <4 x i64> %res
 }
+declare <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float>) nounwind readnone
 
 define <4 x double> @test_mm256_div_pd(<4 x double> %a0, <4 x double> %a1) nounwind {
 ; X32-LABEL: test_mm256_div_pd:

diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll b/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll
index a7b4c6b..0630fd8 100644 (file)
--- a/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll
@@ -359,35 +359,12 @@ define <4 x double> @test_x86_avx_cvt_ps2_pd_256(<4 x float> %a0) {
 declare <4 x double> @llvm.x86.avx.cvt.ps2.pd.256(<4 x float>) nounwind readnone
 
 
-define <4 x i32> @test_x86_avx_cvtt_pd2dq_256(<4 x double> %a0) {
-; CHECK-LABEL: test_x86_avx_cvtt_pd2dq_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vcvttpd2dqy %ymm0, %xmm0
-; CHECK-NEXT:    vzeroupper
-; CHECK-NEXT:    retl
-  %res = call <4 x i32> @llvm.x86.avx.cvtt.pd2dq.256(<4 x double> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.avx.cvtt.pd2dq.256(<4 x double>) nounwind readnone
-
-
-define <8 x i32> @test_x86_avx_cvtt_ps2dq_256(<8 x float> %a0) {
-; CHECK-LABEL: test_x86_avx_cvtt_ps2dq_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vcvttps2dq %ymm0, %ymm0
-; CHECK-NEXT:    retl
-  %res = call <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float> %a0) ; <<8 x i32>> [#uses=1]
-  ret <8 x i32> %res
-}
-declare <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float>) nounwind readnone
-
-
 define void @test_x86_sse2_storeu_dq(i8* %a0, <16 x i8> %a1) {
   ; add operation forces the execution domain.
 ; CHECK-LABEL: test_x86_sse2_storeu_dq:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT:    vpaddb LCPI34_0, %xmm0, %xmm0
+; CHECK-NEXT:    vpaddb LCPI32_0, %xmm0, %xmm0
 ; CHECK-NEXT:    vmovdqu %xmm0, (%eax)
 ; CHECK-NEXT:    retl
   %a2 = add <16 x i8> %a1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>

diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll b/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll
index 3576329..c5d60da 100644 (file)
--- a/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll
@@ -3431,6 +3431,39 @@ define <8 x float> @test_x86_avx_cvtdq2_ps_256(<8 x i32> %a0) {
 declare <8 x float> @llvm.x86.avx.cvtdq2.ps.256(<8 x i32>) nounwind readnone
 
 
+define <4 x i32> @test_x86_avx_cvtt_pd2dq_256(<4 x double> %a0) {
+; AVX-LABEL: test_x86_avx_cvtt_pd2dq_256:
+; AVX:       ## BB#0:
+; AVX-NEXT:    vcvttpd2dqy %ymm0, %xmm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retl
+;
+; AVX512VL-LABEL: test_x86_avx_cvtt_pd2dq_256:
+; AVX512VL:       ## BB#0:
+; AVX512VL-NEXT:    vcvttpd2dqy %ymm0, %xmm0
+; AVX512VL-NEXT:    retl
+  %res = call <4 x i32> @llvm.x86.avx.cvtt.pd2dq.256(<4 x double> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx.cvtt.pd2dq.256(<4 x double>) nounwind readnone
+
+
+define <8 x i32> @test_x86_avx_cvtt_ps2dq_256(<8 x float> %a0) {
+; AVX-LABEL: test_x86_avx_cvtt_ps2dq_256:
+; AVX:       ## BB#0:
+; AVX-NEXT:    vcvttps2dq %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; AVX512VL-LABEL: test_x86_avx_cvtt_ps2dq_256:
+; AVX512VL:       ## BB#0:
+; AVX512VL-NEXT:    vcvttps2dq %ymm0, %ymm0
+; AVX512VL-NEXT:    retl
+  %res = call <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float> %a0) ; <<8 x i32>> [#uses=1]
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx.cvtt.ps2dq.256(<8 x float>) nounwind readnone
+
+
 define <8 x float> @test_x86_avx_dp_ps_256(<8 x float> %a0, <8 x float> %a1) {
 ; AVX-LABEL: test_x86_avx_dp_ps_256:
 ; AVX:       ## BB#0:
@@ -4552,7 +4585,7 @@ define void @movnt_dq(i8* %p, <2 x i64> %a1) nounwind {
 ; AVX-LABEL: movnt_dq:
 ; AVX:       ## BB#0:
 ; AVX-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX-NEXT:    vpaddq LCPI254_0, %xmm0, %xmm0
+; AVX-NEXT:    vpaddq LCPI256_0, %xmm0, %xmm0
 ; AVX-NEXT:    vmovntdq %ymm0, (%eax)
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retl
@@ -4560,7 +4593,7 @@ define void @movnt_dq(i8* %p, <2 x i64> %a1) nounwind {
 ; AVX512VL-LABEL: movnt_dq:
 ; AVX512VL:       ## BB#0:
 ; AVX512VL-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512VL-NEXT:    vpaddq LCPI254_0, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpaddq LCPI256_0, %xmm0, %xmm0
 ; AVX512VL-NEXT:    vmovntdq %ymm0, (%eax)
 ; AVX512VL-NEXT:    retl
   %a2 = add <2 x i64> %a1, <i64 1, i64 1>

diff --git a/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel-x86_64.ll b/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel-x86_64.ll
index 2102b42..aad00e7 100644 (file)
--- a/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel-x86_64.ll
+++ b/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel-x86_64.ll
@@ -6,13 +6,12 @@
 define <4 x float> @test_mm_cvtsi64_ss(<4 x float> %a0, i64 %a1) nounwind {
 ; X64-LABEL: test_mm_cvtsi64_ss:
 ; X64:       # BB#0:
-; X64-NEXT:    cvtsi2ssq %rdi, %xmm1
-; X64-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; X64-NEXT:    cvtsi2ssq %rdi, %xmm0
 ; X64-NEXT:    retq
-  %cvt = sitofp i64 %a1 to float
-  %res = insertelement <4 x float> %a0, float %cvt, i32 0
+  %res = call <4 x float> @llvm.x86.sse.cvtsi642ss(<4 x float> %a0, i64 %a1)
   ret <4 x float> %res
 }
+declare <4 x float> @llvm.x86.sse.cvtsi642ss(<4 x float>, i64) nounwind readnone
 
 define i64 @test_mm_cvtss_si64(<4 x float> %a0) nounwind {
 ; X64-LABEL: test_mm_cvtss_si64:
@@ -29,7 +28,7 @@ define i64 @test_mm_cvttss_si64(<4 x float> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    cvttss2si %xmm0, %rax
 ; X64-NEXT:    retq
-  %cvt = extractelement <4 x float> %a0, i32 0
-  %res = fptosi float %cvt to i64
+  %res = call i64 @llvm.x86.sse.cvttss2si64(<4 x float> %a0)
   ret i64 %res
 }
+declare i64 @llvm.x86.sse.cvttss2si64(<4 x float>) nounwind readnone

diff --git a/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll
index 090ddfd..4715b7f 100644 (file)
--- a/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll
@@ -707,20 +707,17 @@ declare i32 @llvm.x86.sse.cvtss2si(<4 x float>) nounwind readnone
 define <4 x float> @test_mm_cvtsi32_ss(<4 x float> %a0, i32 %a1) nounwind {
 ; X32-LABEL: test_mm_cvtsi32_ss:
 ; X32:       # BB#0:
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    cvtsi2ssl %eax, %xmm1
-; X32-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; X32-NEXT:    cvtsi2ssl {{[0-9]+}}(%esp), %xmm0
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_mm_cvtsi32_ss:
 ; X64:       # BB#0:
-; X64-NEXT:    cvtsi2ssl %edi, %xmm1
-; X64-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; X64-NEXT:    cvtsi2ssl %edi, %xmm0
 ; X64-NEXT:    retq
-  %cvt = sitofp i32 %a1 to float
-  %res = insertelement <4 x float> %a0, float %cvt, i32 0
+  %res = call <4 x float> @llvm.x86.sse.cvtsi2ss(<4 x float> %a0, i32 %a1)
   ret <4 x float> %res
 }
+declare <4 x float> @llvm.x86.sse.cvtsi2ss(<4 x float>, i32) nounwind readnone
 
 define float @test_mm_cvtss_f32(<4 x float> %a0) nounwind {
 ; X32-LABEL: test_mm_cvtss_f32:
@@ -762,10 +759,10 @@ define i32 @test_mm_cvttss_si(<4 x float> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    cvttss2si %xmm0, %eax
 ; X64-NEXT:    retq
-  %cvt = extractelement <4 x float> %a0, i32 0
-  %res = fptosi float %cvt to i32
+  %res = call i32 @llvm.x86.sse.cvttss2si(<4 x float> %a0)
   ret i32 %res
 }
+declare i32 @llvm.x86.sse.cvttss2si(<4 x float>) nounwind readnone
 
 define i32 @test_mm_cvttss_si32(<4 x float> %a0) nounwind {
 ; X32-LABEL: test_mm_cvttss_si32:
@@ -777,8 +774,7 @@ define i32 @test_mm_cvttss_si32(<4 x float> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    cvttss2si %xmm0, %eax
 ; X64-NEXT:    retq
-  %cvt = extractelement <4 x float> %a0, i32 0
-  %res = fptosi float %cvt to i32
+  %res = call i32 @llvm.x86.sse.cvttss2si(<4 x float> %a0)
   ret i32 %res
 }
 

diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel-x86_64.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel-x86_64.ll
index f5ecfa4..6b9dc40 100644 (file)
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel-x86_64.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel-x86_64.ll
@@ -25,13 +25,12 @@ define i64 @test_mm_cvtsi128_si64(<2 x i64> %a0) nounwind {
 define <2 x double> @test_mm_cvtsi64_sd(<2 x double> %a0, i64 %a1) nounwind {
 ; X64-LABEL: test_mm_cvtsi64_sd:
 ; X64:       # BB#0:
-; X64-NEXT:    cvtsi2sdq %rdi, %xmm1
-; X64-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; X64-NEXT:    cvtsi2sdq %rdi, %xmm0
 ; X64-NEXT:    retq
-  %cvt = sitofp i64 %a1 to double
-  %res = insertelement <2 x double> %a0, double %cvt, i32 0
+  %res = call <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double> %a0, i64 %a1)
   ret <2 x double> %res
 }
+declare <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double>, i64) nounwind readnone
 
 define <2 x i64> @test_mm_cvtsi64_si128(i64 %a0) nounwind {
 ; X64-LABEL: test_mm_cvtsi64_si128:
@@ -48,10 +47,10 @@ define i64 @test_mm_cvttsd_si64(<2 x double> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    cvttsd2si %xmm0, %rax
 ; X64-NEXT:    retq
-  %ext = extractelement <2 x double> %a0, i32 0
-  %res = fptosi double %ext to i64
+  %res = call i64 @llvm.x86.sse2.cvttsd2si64(<2 x double> %a0)
   ret i64 %res
 }
+declare i64 @llvm.x86.sse2.cvttsd2si64(<2 x double>) nounwind readnone
 
 define <2 x i64> @test_mm_loadu_si64(i64* %a0) nounwind {
 ; X64-LABEL: test_mm_loadu_si64:

diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
index fa71325..803f364 100644 (file)
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
@@ -1208,6 +1208,21 @@ define i32 @test_mm_cvtsd_si32(<2 x double> %a0) nounwind {
 }
 declare i32 @llvm.x86.sse2.cvtsd2si(<2 x double>) nounwind readnone
 
+define <4 x float> @test_mm_cvtsd_ss(<4 x float> %a0, <2 x double> %a1) {
+; X32-LABEL: test_mm_cvtsd_ss:
+; X32:       # BB#0:
+; X32-NEXT:    cvtsd2ss %xmm1, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: test_mm_cvtsd_ss:
+; X64:       # BB#0:
+; X64-NEXT:    cvtsd2ss %xmm1, %xmm0
+; X64-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float> %a0, <2 x double> %a1)
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float>, <2 x double>) nounwind readnone
+
 define i32 @test_mm_cvtsi128_si32(<2 x i64> %a0) nounwind {
 ; X32-LABEL: test_mm_cvtsi128_si32:
 ; X32:       # BB#0:
@@ -1303,10 +1318,11 @@ define <2 x i64> @test_mm_cvttps_epi32(<4 x float> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    cvttps2dq %xmm0, %xmm0
 ; X64-NEXT:    retq
-  %res = fptosi <4 x float> %a0 to <4 x i32>
+  %res = call <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float> %a0)
   %bc = bitcast <4 x i32> %res to <2 x i64>
   ret <2 x i64> %bc
 }
+declare <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float>) nounwind readnone
 
 define i32 @test_mm_cvttsd_si32(<2 x double> %a0) nounwind {
 ; X32-LABEL: test_mm_cvttsd_si32:
@@ -1318,10 +1334,10 @@ define i32 @test_mm_cvttsd_si32(<2 x double> %a0) nounwind {
 ; X64:       # BB#0:
 ; X64-NEXT:    cvttsd2si %xmm0, %eax
 ; X64-NEXT:    retq
-  %ext = extractelement <2 x double> %a0, i32 0
-  %res = fptosi double %ext to i32
+  %res = call i32 @llvm.x86.sse2.cvttsd2si(<2 x double> %a0)
   ret i32 %res
 }
+declare i32 @llvm.x86.sse2.cvttsd2si(<2 x double>) nounwind readnone
 
 define <2 x double> @test_mm_div_pd(<2 x double> %a0, <2 x double> %a1) nounwind {
 ; X32-LABEL: test_mm_div_pd:

diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll
index ae6626b..27a3fce 100644 (file)
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll
@@ -66,17 +66,6 @@ define <2 x double> @test_x86_sse2_cvtps2pd(<4 x float> %a0) {
 declare <2 x double> @llvm.x86.sse2.cvtps2pd(<4 x float>) nounwind readnone
 
 
-define <4 x i32> @test_x86_sse2_cvttps2dq(<4 x float> %a0) {
-; CHECK-LABEL: test_x86_sse2_cvttps2dq:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
-; CHECK-NEXT:    retl
-  %res = call <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float>) nounwind readnone
-
-
 define void @test_x86_sse2_storel_dq(i8* %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: test_x86_sse2_storel_dq:
 ; CHECK:       ## BB#0:
@@ -94,7 +83,7 @@ define void @test_x86_sse2_storeu_dq(i8* %a0, <16 x i8> %a1) {
 ; CHECK-LABEL: test_x86_sse2_storeu_dq:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT:    paddb LCPI8_0, %xmm0
+; CHECK-NEXT:    paddb LCPI7_0, %xmm0
 ; CHECK-NEXT:    movdqu %xmm0, (%eax)
 ; CHECK-NEXT:    retl
   %a2 = add <16 x i8> %a1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>

diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
index 617e30e..e682e8d 100644 (file)
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
@@ -1,4 +1,4 @@
-; NOTE: Assertions have been autogenerated by update_llc_test_checks.py
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=-avx,+sse2 | FileCheck %s --check-prefix=SSE
 ; RUN: llc < %s -mtriple=i386-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=KNL
 
@@ -322,6 +322,22 @@ define <4 x i32> @test_x86_sse2_cvttpd2dq(<2 x double> %a0) {
 declare <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double>) nounwind readnone
 
 
+define <4 x i32> @test_x86_sse2_cvttps2dq(<4 x float> %a0) {
+; SSE-LABEL: test_x86_sse2_cvttps2dq:
+; SSE:       ## BB#0:
+; SSE-NEXT:    cvttps2dq %xmm0, %xmm0
+; SSE-NEXT:    retl
+;
+; KNL-LABEL: test_x86_sse2_cvttps2dq:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vcvttps2dq %xmm0, %xmm0
+; KNL-NEXT:    retl
+  %res = call <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float>) nounwind readnone
+
+
 define i32 @test_x86_sse2_cvttsd2si(<2 x double> %a0) {
 ; SSE-LABEL: test_x86_sse2_cvttsd2si:
 ; SSE:       ## BB#0:

diff --git a/llvm/test/Transforms/ConstProp/calls.ll b/llvm/test/Transforms/ConstProp/calls.ll
index a445ac8..736d8c1 100644 (file)
--- a/llvm/test/Transforms/ConstProp/calls.ll
+++ b/llvm/test/Transforms/ConstProp/calls.ll
@@ -193,11 +193,13 @@ entry:
   ret i1 %b
 }
 
-; TODO: Inexact values should not fold as they are dependent on rounding mode
+; Inexact values should not fold as they are dependent on rounding mode
 define i1 @test_sse_cvts_inexact() nounwind readnone {
 ; CHECK-LABEL: @test_sse_cvts_inexact(
-; CHECK-NOT: call
-; CHECK: ret i1 true
+; CHECK: call
+; CHECK: call
+; CHECK: call
+; CHECK: call
 entry:
   %i0 = tail call i32 @llvm.x86.sse.cvtss2si(<4 x float> <float 1.75, float undef, float undef, float undef>) nounwind
   %i1 = tail call i64 @llvm.x86.sse.cvtss2si64(<4 x float> <float 1.75, float undef, float undef, float undef>) nounwind