[llvm] r263303 - [X86][SSE] Simplify vector LOAD + EXTEND on pre-SSE41 hardware

Simon Pilgrim via llvm-commits llvm-commits at lists.llvm.org
Mon Mar 14 16:03:18 PDT 2016


Yes revert if there is a problem but please raise a bugzilla with a 
repro, even if it isn't greatly reduced.
Simon.

On 14/03/2016 15:44, Alina Sbirlea wrote:
> As far as I can tell the problematic change is the check in 
> llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> Working on a testcase in the meantime.
>
>
> Thank you,
> Alina
>
> On Mon, Mar 14, 2016 at 3:31 PM, Eric Christopher <echristo at gmail.com 
> <mailto:echristo at gmail.com>> wrote:
>
>     Hi Simon,
>
>     I'm seeing asserts with this and halide on x86. I'm working with
>     Alina to get a testcase for you, but would you mind terribly if we
>     reverted this in the meantime?
>
>     -eric
>
>     On Fri, Mar 11, 2016 at 2:22 PM Simon Pilgrim via llvm-commits
>     <llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>>
>     wrote:
>
>         Author: rksimon
>         Date: Fri Mar 11 16:18:05 2016
>         New Revision: 263303
>
>         URL: http://llvm.org/viewvc/llvm-project?rev=263303&view=rev
>         Log:
>         [X86][SSE] Simplify vector LOAD + EXTEND on pre-SSE41 hardware
>
>         Improve vector extension of vectors on hardware without
>         dedicated VSEXT/VZEXT instructions.
>
>         We already convert these to
>         SIGN_EXTEND_VECTOR_INREG/ZERO_EXTEND_VECTOR_INREG but can
>         further improve this by using the legalizer instead of
>         prematurely splitting into legal vectors in the combine as
>         this only properly helps for lowering to VSEXT/VZEXT.
>
>         Removes a lot of unnecessary any_extend + mask pattern - (Fix
>         for PR25718).
>
>         Differential Revision: http://reviews.llvm.org/D17932
>
>         Modified:
>         llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeTypes.h
>         llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
>             llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
>             llvm/trunk/test/CodeGen/X86/vector-zext.ll
>
>         Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeTypes.h
>         URL:
>         http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeTypes.h?rev=263303&r1=263302&r2=263303&view=diff
>         ==============================================================================
>         --- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeTypes.h (original)
>         +++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeTypes.h Fri
>         Mar 11 16:18:05 2016
>         @@ -653,6 +653,7 @@ private:
>            void SplitVecRes_UnaryOp(SDNode *N, SDValue &Lo, SDValue &Hi);
>            void SplitVecRes_ExtendOp(SDNode *N, SDValue &Lo, SDValue &Hi);
>            void SplitVecRes_InregOp(SDNode *N, SDValue &Lo, SDValue &Hi);
>         +  void SplitVecRes_ExtVecInRegOp(SDNode *N, SDValue &Lo,
>         SDValue &Hi);
>
>            void SplitVecRes_BITCAST(SDNode *N, SDValue &Lo, SDValue &Hi);
>            void SplitVecRes_BUILD_VECTOR(SDNode *N, SDValue &Lo,
>         SDValue &Hi);
>
>         Modified:
>         llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
>         URL:
>         http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp?rev=263303&r1=263302&r2=263303&view=diff
>         ==============================================================================
>         ---
>         llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
>         (original)
>         +++
>         llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
>         Fri Mar 11 16:18:05 2016
>         @@ -621,6 +621,12 @@ void DAGTypeLegalizer::SplitVectorResult
>          SplitVecRes_VECTOR_SHUFFLE(cast<ShuffleVectorSDNode>(N), Lo, Hi);
>              break;
>
>         +  case ISD::ANY_EXTEND_VECTOR_INREG:
>         +  case ISD::SIGN_EXTEND_VECTOR_INREG:
>         +  case ISD::ZERO_EXTEND_VECTOR_INREG:
>         +    SplitVecRes_ExtVecInRegOp(N, Lo, Hi);
>         +    break;
>         +
>            case ISD::BITREVERSE:
>            case ISD::BSWAP:
>            case ISD::CONVERT_RNDSAT:
>         @@ -917,6 +923,39 @@ void DAGTypeLegalizer::SplitVecRes_Inreg
>                             DAG.getValueType(HiVT));
>          }
>
>         +void DAGTypeLegalizer::SplitVecRes_ExtVecInRegOp(SDNode *N,
>         SDValue &Lo,
>         +  SDValue &Hi) {
>         +  unsigned Opcode = N->getOpcode();
>         +  SDValue N0 = N->getOperand(0);
>         +
>         +  SDLoc dl(N);
>         +  SDValue InLo, InHi;
>         +  GetSplitVector(N0, InLo, InHi);
>         +  EVT InLoVT = InLo.getValueType();
>         +  unsigned InNumElements = InLoVT.getVectorNumElements();
>         +
>         +  EVT OutLoVT, OutHiVT;
>         +  std::tie(OutLoVT, OutHiVT) =
>         DAG.GetSplitDestVTs(N->getValueType(0));
>         +  unsigned OutNumElements = OutLoVT.getVectorNumElements();
>         +  assert((2 * OutNumElements) <= InNumElements &&
>         +         "Illegal extend vector in reg split");
>         +
>         +  // *_EXTEND_VECTOR_INREG instructions extend the lowest
>         elements of the
>         +  // input vector (i.e. we only use InLo):
>         +  // OutLo will extend the first OutNumElements from InLo.
>         +  // OutHi will extend the next OutNumElements from InLo.
>         +
>         +  // Shuffle the elements from InLo for OutHi into the bottom
>         elements to
>         +  // create a 'fake' InHi.
>         +  SmallVector<int, 8> SplitHi(InNumElements, -1);
>         +  for (unsigned i = 0; i != OutNumElements; ++i)
>         +    SplitHi[i] = i + OutNumElements;
>         +  InHi = DAG.getVectorShuffle(InLoVT, dl, InLo,
>         DAG.getUNDEF(InLoVT), SplitHi);
>         +
>         +  Lo = DAG.getNode(Opcode, dl, OutLoVT, InLo);
>         +  Hi = DAG.getNode(Opcode, dl, OutHiVT, InHi);
>         +}
>         +
>          void DAGTypeLegalizer::SplitVecRes_INSERT_VECTOR_ELT(SDNode
>         *N, SDValue &Lo,
>         SDValue &Hi) {
>            SDValue Vec = N->getOperand(0);
>
>         Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
>         URL:
>         http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=263303&r1=263302&r2=263303&view=diff
>         ==============================================================================
>         --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
>         +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Mar 11
>         16:18:05 2016
>         @@ -28534,7 +28534,9 @@ static SDValue combineToExtendVectorInRe
>
>            // If target-size is 128-bits (or 256-bits on AVX2 target),
>         then convert to
>            // ISD::*_EXTEND_VECTOR_INREG which ensures lowering to
>         X86ISD::V*EXT.
>         -  if (VT.is128BitVector() || (VT.is256BitVector() &&
>         Subtarget.hasInt256())) {
>         +  // Also use this if we don't have SSE41 to allow the
>         legalizer do its job.
>         +  if (!Subtarget.hasSSE41() || VT.is128BitVector() ||
>         +      (VT.is256BitVector() && Subtarget.hasInt256())) {
>              SDValue ExOp = ExtendVecSize(DL, N0, VT.getSizeInBits());
>              return Opcode == ISD::SIGN_EXTEND
>                         ? DAG.getSignExtendVectorInReg(ExOp, DL, VT)
>
>         Modified: llvm/trunk/test/CodeGen/X86/vector-zext.ll
>         URL:
>         http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-zext.ll?rev=263303&r1=263302&r2=263303&view=diff
>         ==============================================================================
>         --- llvm/trunk/test/CodeGen/X86/vector-zext.ll (original)
>         +++ llvm/trunk/test/CodeGen/X86/vector-zext.ll Fri Mar 11
>         16:18:05 2016
>         @@ -544,23 +544,20 @@ define <4 x i64> @load_zext_4i8_to_4i64(
>          ; SSE2-LABEL: load_zext_4i8_to_4i64:
>          ; SSE2:       # BB#0: # %entry
>          ; SSE2-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
>         -; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>         -; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,1,1,3]
>         -; SSE2-NEXT:    movdqa {{.*#+}} xmm2 =
>         [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
>         -; SSE2-NEXT:    pand %xmm2, %xmm0
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
>         -; SSE2-NEXT:    pand %xmm2, %xmm1
>         +; SSE2-NEXT:    pxor %xmm2, %xmm2
>         +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
>         +; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>         +; SSE2-NEXT:    movdqa %xmm1, %xmm0
>         +; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 =
>         xmm0[0],xmm2[0],xmm0[1],xmm2[1]
>         +; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 =
>         xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>          ; SSE2-NEXT:    retq
>          ;
>          ; SSSE3-LABEL: load_zext_4i8_to_4i64:
>          ; SSSE3:       # BB#0: # %entry
>          ; SSSE3-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
>         -; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>         -; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
>          ; SSSE3-NEXT:    movdqa %xmm1, %xmm0
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 =
>         xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 =
>         xmm1[8],zero,zero,zero,zero,zero,zero,zero,xmm1[12],zero,zero,zero,zero,zero,zero,zero
>         +; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 =
>         xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
>         +; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 =
>         xmm1[2],zero,zero,zero,zero,zero,zero,zero,xmm1[3],zero,zero,zero,zero,zero,zero,zero
>          ; SSSE3-NEXT:    retq
>          ;
>          ; SSE41-LABEL: load_zext_4i8_to_4i64:
>         @@ -625,22 +622,21 @@ define <8 x i32> @load_zext_8i8_to_8i32(
>          ; SSE2-LABEL: load_zext_8i8_to_8i32:
>          ; SSE2:       # BB#0: # %entry
>          ; SSE2-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
>         -; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>         +; SSE2-NEXT:    pxor %xmm2, %xmm2
>         +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
>          ; SSE2-NEXT:    movdqa %xmm1, %xmm0
>         -; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
>         -; SSE2-NEXT:    movdqa {{.*#+}} xmm2 =
>         [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
>         -; SSE2-NEXT:    pand %xmm2, %xmm0
>         -; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 =
>         xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>         -; SSE2-NEXT:    pand %xmm2, %xmm1
>         +; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 =
>         xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
>         +; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 =
>         xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
>          ; SSE2-NEXT:    retq
>          ;
>          ; SSSE3-LABEL: load_zext_8i8_to_8i32:
>          ; SSSE3:       # BB#0: # %entry
>          ; SSSE3-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
>         -; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>         +; SSSE3-NEXT:    pxor %xmm2, %xmm2
>         +; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
>          ; SSSE3-NEXT:    movdqa %xmm1, %xmm0
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 =
>         xmm0[0],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[6],zero,zero,zero
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 =
>         xmm1[8],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[14],zero,zero,zero
>         +; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 =
>         xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
>         +; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm1 =
>         xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
>          ; SSSE3-NEXT:    retq
>          ;
>          ; SSE41-LABEL: load_zext_8i8_to_8i32:
>         @@ -674,34 +670,33 @@ entry:
>          define <8 x i64> @load_zext_8i8_to_8i64(<8 x i8> *%ptr) {
>          ; SSE2-LABEL: load_zext_8i8_to_8i64:
>          ; SSE2:       # BB#0: # %entry
>         -; SSE2-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
>         -; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 =
>         xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,0,3]
>         -; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,6,7]
>         -; SSE2-NEXT:    movdqa {{.*#+}} xmm4 =
>         [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
>         -; SSE2-NEXT:    pand %xmm4, %xmm0
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,3]
>         -; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,6,7]
>         -; SSE2-NEXT:    pand %xmm4, %xmm1
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,1,2,3]
>         -; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,6,7]
>         -; SSE2-NEXT:    pand %xmm4, %xmm2
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,3,3]
>         -; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,6,7]
>         -; SSE2-NEXT:    pand %xmm4, %xmm3
>         +; SSE2-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
>         +; SSE2-NEXT:    pxor %xmm4, %xmm4
>         +; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,2,3]
>         +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 =
>         xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
>         +; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
>         +; SSE2-NEXT:    movdqa %xmm1, %xmm0
>         +; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 =
>         xmm0[0],xmm4[0],xmm0[1],xmm4[1]
>         +; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 =
>         xmm1[2],xmm4[2],xmm1[3],xmm4[3]
>         +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 =
>         xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
>         +; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 =
>         xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
>         +; SSE2-NEXT:    movdqa %xmm3, %xmm2
>         +; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 =
>         xmm2[0],xmm4[0],xmm2[1],xmm4[1]
>         +; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 =
>         xmm3[2],xmm4[2],xmm3[3],xmm4[3]
>          ; SSE2-NEXT:    retq
>          ;
>          ; SSSE3-LABEL: load_zext_8i8_to_8i64:
>          ; SSSE3:       # BB#0: # %entry
>         -; SSSE3-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
>         -; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 =
>         xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
>         -; SSSE3-NEXT:    movdqa %xmm3, %xmm0
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 =
>         xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero
>         -; SSSE3-NEXT:    movdqa %xmm3, %xmm1
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 =
>         xmm1[4],zero,zero,zero,zero,zero,zero,zero,xmm1[6],zero,zero,zero,zero,zero,zero,zero
>         +; SSSE3-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
>         +; SSSE3-NEXT:    movdqa {{.*#+}} xmm4 =
>         [0,128,128,128,128,128,128,128,1,128,128,128,128,128,128,128]
>         +; SSSE3-NEXT:    movdqa %xmm1, %xmm0
>         +; SSSE3-NEXT:    pshufb %xmm4, %xmm0
>         +; SSSE3-NEXT:    movdqa {{.*#+}} xmm5 =
>         [2,128,128,128,128,128,128,128,3,128,128,128,128,128,128,128]
>         +; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,2,3]
>         +; SSSE3-NEXT:    pshufb %xmm5, %xmm1
>          ; SSSE3-NEXT:    movdqa %xmm3, %xmm2
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm2 =
>         xmm2[8],zero,zero,zero,zero,zero,zero,zero,xmm2[10],zero,zero,zero,zero,zero,zero,zero
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm3 =
>         xmm3[12],zero,zero,zero,zero,zero,zero,zero,xmm3[14],zero,zero,zero,zero,zero,zero,zero
>         +; SSSE3-NEXT:    pshufb %xmm4, %xmm2
>         +; SSSE3-NEXT:    pshufb %xmm5, %xmm3
>          ; SSSE3-NEXT:    retq
>          ;
>          ; SSE41-LABEL: load_zext_8i8_to_8i64:
>         @@ -851,21 +846,21 @@ define <4 x i64> @load_zext_4i16_to_4i64
>          ; SSE2-LABEL: load_zext_4i16_to_4i64:
>          ; SSE2:       # BB#0: # %entry
>          ; SSE2-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
>         -; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,1,1,3]
>         -; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,0,0,65535,0,0,0]
>         -; SSE2-NEXT:    pand %xmm2, %xmm0
>         -; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
>         -; SSE2-NEXT:    pand %xmm2, %xmm1
>         +; SSE2-NEXT:    pxor %xmm2, %xmm2
>         +; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>         +; SSE2-NEXT:    movdqa %xmm1, %xmm0
>         +; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 =
>         xmm0[0],xmm2[0],xmm0[1],xmm2[1]
>         +; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 =
>         xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>          ; SSE2-NEXT:    retq
>          ;
>          ; SSSE3-LABEL: load_zext_4i16_to_4i64:
>          ; SSSE3:       # BB#0: # %entry
>          ; SSSE3-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
>         -; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
>         +; SSSE3-NEXT:    pxor %xmm2, %xmm2
>         +; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 =
>         xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>          ; SSSE3-NEXT:    movdqa %xmm1, %xmm0
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 =
>         xmm0[0,1],zero,zero,zero,zero,zero,zero,xmm0[4,5],zero,zero,zero,zero,zero,zero
>         -; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 =
>         xmm1[8,9],zero,zero,zero,zero,zero,zero,xmm1[12,13],zero,zero,zero,zero,zero,zero
>         +; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 =
>         xmm0[0],xmm2[0],xmm0[1],xmm2[1]
>         +; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 =
>         xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>          ; SSSE3-NEXT:    retq
>          ;
>          ; SSE41-LABEL: load_zext_4i16_to_4i64:
>
>
>         _______________________________________________
>         llvm-commits mailing list
>         llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>
>         http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160314/757cb3b0/attachment.html>


More information about the llvm-commits mailing list