[X86] replace vextractf128 intrinsics with generic shuffles (LLVM)

[spatel]

Summary

This is very much like D8086 (checked in here: http://llvm.org/viewvc/llvm-project?rev=231794&view=rev ).
This is also the LLVM sibling to the cfe D8275 patch.

Now that we've replaced the vinsertf128 intrinsics, do the same for their extract twins.

There are several existing regression tests of the memory folding versions of these intrinsics in 'avx-vextractf128.ll'.
I'm not sure if it's better to leave those as-is or consolidate them into the 'avx-intrinsics-x86-upgrade.ll' file.
If anyone has a suggestion about that, please let me know.

[craig.topper]

LGTM

Are you going to do the AVX2 version of these too?

[spatel]

In D8276#139537, @craig.topper wrote:

LGTM

Are you going to do the AVX2 version of these too?

Yes, I'm planning to fix up as many of these as possible. Simon may have some similar patches to commit as well.

[spatel]

In D8276#139537, @craig.topper wrote:

Are you going to do the AVX2 version of these too?

Those appear to be pure copy / paste / delete, so I went ahead and checked them in here:

http://reviews.llvm.org/rL232109
http://reviews.llvm.org/rL232120

include/llvm/IR/IntrinsicsX86.td

let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx_cmp_pd_256 : GCCBuiltin<"__builtin_ia32_cmppd256">,
         Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty,
                   llvm_v4f64_ty, llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx_cmp_ps_256 : GCCBuiltin<"__builtin_ia32_cmpps256">,
         Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty,
                   llvm_v8f32_ty, llvm_i8_ty], [IntrNoMem]>;
 }
 
-// Vector extract and insert
-let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
-  def int_x86_avx_vextractf128_pd_256 :
-        GCCBuiltin<"__builtin_ia32_vextractf128_pd256">,
-        Intrinsic<[llvm_v2f64_ty], [llvm_v4f64_ty, llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx_vextractf128_ps_256 :
-        GCCBuiltin<"__builtin_ia32_vextractf128_ps256">,
-        Intrinsic<[llvm_v4f32_ty], [llvm_v8f32_ty, llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx_vextractf128_si_256 :
-        GCCBuiltin<"__builtin_ia32_vextractf128_si256">,
-        Intrinsic<[llvm_v4i32_ty], [llvm_v8i32_ty, llvm_i8_ty], [IntrNoMem]>;
-}
-
 // Vector convert
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx_cvtdq2_pd_256 : GCCBuiltin<"__builtin_ia32_cvtdq2pd256">,
         Intrinsic<[llvm_v4f64_ty], [llvm_v4i32_ty], [IntrNoMem]>;
   def int_x86_avx_cvtdq2_ps_256 : GCCBuiltin<"__builtin_ia32_cvtdq2ps256">,
         Intrinsic<[llvm_v8f32_ty], [llvm_v8i32_ty], [IntrNoMem]>;
   def int_x86_avx_cvt_pd2_ps_256 : GCCBuiltin<"__builtin_ia32_cvtpd2ps256">,
         Intrinsic<[llvm_v4f32_ty], [llvm_v4f64_ty], [IntrNoMem]>;

lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

uint64_t Idx = (cast<ConstantInt>(I.getArgOperand(2))->getZExtValue() & 1) *
                    ElVT.getVectorNumElements();
     Res =
         DAG.getNode(ISD::INSERT_SUBVECTOR, sdl, DestVT,
                     getValue(I.getArgOperand(0)), getValue(I.getArgOperand(1)),
                     DAG.getConstant(Idx, TLI.getVectorIdxTy()));
     setValue(&I, Res);
     return nullptr;
   }
-  case Intrinsic::x86_avx_vextractf128_pd_256:
-  case Intrinsic::x86_avx_vextractf128_ps_256:
-  case Intrinsic::x86_avx_vextractf128_si_256:
   case Intrinsic::x86_avx2_vextracti128: {
     EVT DestVT = TLI.getValueType(I.getType());
     uint64_t Idx = (cast<ConstantInt>(I.getArgOperand(1))->getZExtValue() & 1) *
                    DestVT.getVectorNumElements();
     Res = DAG.getNode(ISD::EXTRACT_SUBVECTOR, sdl, DestVT,
                       getValue(I.getArgOperand(0)),
                       DAG.getConstant(Idx, TLI.getVectorIdxTy()));
     setValue(&I, Res);

lib/IR/AutoUpgrade.cpp

case 'x': {
     if (Name.startswith("x86.sse2.pcmpeq.") ||
         Name.startswith("x86.sse2.pcmpgt.") ||
         Name.startswith("x86.avx2.pcmpeq.") ||
         Name.startswith("x86.avx2.pcmpgt.") ||
         Name.startswith("x86.avx.vpermil.") ||
         Name == "x86.avx.vinsertf128.pd.256" ||
         Name == "x86.avx.vinsertf128.ps.256" ||
         Name == "x86.avx.vinsertf128.si.256" ||
+        Name == "x86.avx.vextractf128.pd.256" ||
+        Name == "x86.avx.vextractf128.ps.256" ||
+        Name == "x86.avx.vextractf128.si.256" ||
         Name == "x86.avx.movnt.dq.256" ||
         Name == "x86.avx.movnt.pd.256" ||
         Name == "x86.avx.movnt.ps.256" ||
         Name == "x86.sse42.crc32.64.8" ||
         Name == "x86.avx.vbroadcast.ss" ||
         Name == "x86.avx.vbroadcast.ss.256" ||
         Name == "x86.avx.vbroadcast.sd.256" ||
         Name == "x86.sse2.psll.dq" ||
if (Name.startswith("llvm.x86.sse2.pcmpeq.") ||
       }
       // The high half of the result is either the low half of the 2nd operand
       // (the inserted vector) or the high half of the 1st operand.
       for (unsigned i = NumElts / 2; i != NumElts; ++i) {
         unsigned Idx = Imm ? (i + NumElts / 2) : i;
         Idxs2.push_back(Builder.getInt32(Idx));
       }
       Rep = Builder.CreateShuffleVector(Op0, Rep, ConstantVector::get(Idxs2));
+    } else if (Name == "llvm.x86.avx.vextractf128.pd.256" ||
+               Name == "llvm.x86.avx.vextractf128.ps.256" ||
+               Name == "llvm.x86.avx.vextractf128.si.256") {
+      Value *Op0 = CI->getArgOperand(0);
+      unsigned Imm = cast<ConstantInt>(CI->getArgOperand(1))->getZExtValue();
+      VectorType *VecTy = cast<VectorType>(CI->getType());
+      unsigned NumElts = VecTy->getNumElements();
+      
+      // Mask off the high bits of the immediate value; hardware ignores those.
+      Imm = Imm & 1;
+
+      // Get indexes for either the high half or low half of the input vector.
+      SmallVector<Constant*, 4> Idxs(NumElts);
+      for (unsigned i = 0; i != NumElts; ++i) {
+        unsigned Idx = Imm ? (i + NumElts) : i;
+        Idxs[i] = Builder.getInt32(Idx);
+      }
+
+      Value *UndefV = UndefValue::get(Op0->getType());
+      Rep = Builder.CreateShuffleVector(Op0, UndefV, ConstantVector::get(Idxs));
     } else {
       bool PD128 = false, PD256 = false, PS128 = false, PS256 = false;
       if (Name == "llvm.x86.avx.vpermil.pd.256")
         PD256 = true;
       else if (Name == "llvm.x86.avx.vpermil.pd")
         PD128 = true;
       else if (Name == "llvm.x86.avx.vpermil.ps.256")
         PS256 = true;

test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll

 define <8 x i32> @test_x86_avx_vinsertf128_si_256_2(<8 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL:    test_x86_avx_vinsertf128_si_256_2: 
 ; CHECK-NOT:      vinsertf128 $1, %xmm1, %ymm0, %ymm0
   %res = call <8 x i32> @llvm.x86.avx.vinsertf128.si.256(<8 x i32> %a0, <4 x i32> %a1, i8 2)
   ret <8 x i32> %res
 }
 declare <8 x i32> @llvm.x86.avx.vinsertf128.si.256(<8 x i32>, <4 x i32>, i8) nounwind readnone
 
+; We don't check any vextractf128 variant with immediate 0 because that's just a move. 
+
+define <2 x double> @test_x86_avx_vextractf128_pd_256_1(<4 x double> %a0) {
+; CHECK-LABEL:       test_x86_avx_vextractf128_pd_256_1: 
+; CHECK:             vextractf128 $1, %ymm0, %xmm0
+  %res = call <2 x double> @llvm.x86.avx.vextractf128.pd.256(<4 x double> %a0, i8 1)
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.avx.vextractf128.pd.256(<4 x double>, i8) nounwind readnone
+
+define <4 x float> @test_x86_avx_vextractf128_ps_256_1(<8 x float> %a0) {
+; CHECK-LABEL:       test_x86_avx_vextractf128_ps_256_1: 
+; CHECK:             vextractf128 $1, %ymm0, %xmm0
+  %res = call <4 x float> @llvm.x86.avx.vextractf128.ps.256(<8 x float> %a0, i8 1)
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.avx.vextractf128.ps.256(<8 x float>, i8) nounwind readnone
+
+define <4 x i32> @test_x86_avx_vextractf128_si_256_1(<8 x i32> %a0) {
+; CHECK-LABEL:    test_x86_avx_vextractf128_si_256_1: 
+; CHECK:          vextractf128 $1, %ymm0, %xmm0
+  %res = call <4 x i32> @llvm.x86.avx.vextractf128.si.256(<8 x i32> %a0, i8 1)
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.avx.vextractf128.si.256(<8 x i32>, i8) nounwind readnone
+
+; Verify that high bits of the immediate are masked off. This should be the equivalent
+; of a vextractf128 $0 which should be optimized away, so just check that it's
+; not a vextractf128 of any kind.
+define <2 x double> @test_x86_avx_extractf128_pd_256_2(<4 x double> %a0) {
+; CHECK-LABEL:       test_x86_avx_extractf128_pd_256_2: 
+; CHECK-NOT:         vextractf128
+  %res = call <2 x double> @llvm.x86.avx.vextractf128.pd.256(<4 x double> %a0, i8 2)
+  ret <2 x double> %res
+}
+
+
 define <4 x double> @test_x86_avx_blend_pd_256(<4 x double> %a0, <4 x double> %a1) {
 ; CHECK-LABEL:       test_x86_avx_blend_pd_256: 
 ; CHECK:             vblendpd
   %res = call <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double> %a0, <4 x double> %a1, i32 7) ; <<4 x double>> [#uses=1]
   ret <4 x double> %res
 }
 declare <4 x double> @llvm.x86.avx.blend.pd.256(<4 x double>, <4 x double>, i32) nounwind readnone
 

test/CodeGen/X86/avx-intrinsics-x86.ll

 define <8 x float> @test_x86_avx_vbroadcastf128_ps_256(i8* %a0) {
   ; CHECK: vbroadcastf128
   %res = call <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8* %a0) ; <<8 x float>> [#uses=1]
   ret <8 x float> %res
 }
 declare <8 x float> @llvm.x86.avx.vbroadcastf128.ps.256(i8*) nounwind readonly
 
 
-define <2 x double> @test_x86_avx_vextractf128_pd_256(<4 x double> %a0) {
-  ; CHECK: vextractf128
-  %res = call <2 x double> @llvm.x86.avx.vextractf128.pd.256(<4 x double> %a0, i8 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.avx.vextractf128.pd.256(<4 x double>, i8) nounwind readnone
-
-
-define <4 x float> @test_x86_avx_vextractf128_ps_256(<8 x float> %a0) {
-  ; CHECK: vextractf128
-  %res = call <4 x float> @llvm.x86.avx.vextractf128.ps.256(<8 x float> %a0, i8 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.avx.vextractf128.ps.256(<8 x float>, i8) nounwind readnone
-
-
-define <4 x i32> @test_x86_avx_vextractf128_si_256(<8 x i32> %a0) {
-  ; CHECK: vextractf128
-  %res = call <4 x i32> @llvm.x86.avx.vextractf128.si.256(<8 x i32> %a0, i8 7) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.avx.vextractf128.si.256(<8 x i32>, i8) nounwind readnone
-
-
 define <4 x double> @test_x86_avx_vperm2f128_pd_256(<4 x double> %a0, <4 x double> %a1) {
   ; CHECK: vperm2f128
   %res = call <4 x double> @llvm.x86.avx.vperm2f128.pd.256(<4 x double> %a0, <4 x double> %a1, i8 7) ; <<4 x double>> [#uses=1]
   ret <4 x double> %res
 }
 declare <4 x double> @llvm.x86.avx.vperm2f128.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone