[llvm] r345473 - [VectorLegalizer] Enable TargetLowering::expandFP_TO_UINT support.
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sun Oct 28 06:07:25 PDT 2018
Author: rksimon
Date: Sun Oct 28 06:07:25 2018
New Revision: 345473
URL: http://llvm.org/viewvc/llvm-project?rev=345473&view=rev
Log:
[VectorLegalizer] Enable TargetLowering::expandFP_TO_UINT support.
Add vector support to TargetLowering::expandFP_TO_UINT.
This exposes an issue in X86TargetLowering::LowerVSELECT which was assuming that the select mask was the same width as the LHS/RHS ops - as long as the result is a sign splat we can easily sext/trunk this.
Modified:
llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
llvm/trunk/test/CodeGen/X86/ftrunc.ll
llvm/trunk/test/CodeGen/X86/vec_cast3.ll
llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll
Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp?rev=345473&r1=345472&r2=345473&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp Sun Oct 28 06:07:25 2018
@@ -86,9 +86,10 @@ class VectorLegalizer {
/// operations to legalize them.
SDValue Expand(SDValue Op);
- /// Implements expansion for FNEG; falls back to UnrollVectorOp if
- /// FSUB isn't legal.
- ///
+ /// Implements expansion for FP_TO_UINT; falls back to UnrollVectorOp if
+ /// FP_TO_SINT isn't legal.
+ SDValue ExpandFP_TO_UINT(SDValue Op);
+
/// Implements expansion for UINT_TO_FLOAT; falls back to UnrollVectorOp if
/// SINT_TO_FLOAT and SHR on vectors isn't legal.
SDValue ExpandUINT_TO_FLOAT(SDValue Op);
@@ -709,6 +710,8 @@ SDValue VectorLegalizer::Expand(SDValue
return ExpandVSELECT(Op);
case ISD::SELECT:
return ExpandSELECT(Op);
+ case ISD::FP_TO_UINT:
+ return ExpandFP_TO_UINT(Op);
case ISD::UINT_TO_FP:
return ExpandUINT_TO_FLOAT(Op);
case ISD::FNEG:
@@ -1018,6 +1021,16 @@ SDValue VectorLegalizer::ExpandVSELECT(S
return DAG.getNode(ISD::BITCAST, DL, Op.getValueType(), Val);
}
+SDValue VectorLegalizer::ExpandFP_TO_UINT(SDValue Op) {
+ // Attempt to expand using TargetLowering.
+ SDValue Result;
+ if (TLI.expandFP_TO_UINT(Op.getNode(), Result, DAG))
+ return Result;
+
+ // Otherwise go ahead and unroll.
+ return DAG.UnrollVectorOp(Op.getNode());
+}
+
SDValue VectorLegalizer::ExpandUINT_TO_FLOAT(SDValue Op) {
EVT VT = Op.getOperand(0).getValueType();
SDLoc DL(Op);
Modified: llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=345473&r1=345472&r2=345473&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp Sun Oct 28 06:07:25 2018
@@ -4147,6 +4147,11 @@ bool TargetLowering::expandFP_TO_UINT(SD
EVT SetCCVT =
getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), SrcVT);
+ // Only expand vector types if we have the appropriate vector bit operations.
+ if (DstVT.isVector() && (!isOperationLegalOrCustom(ISD::FP_TO_SINT, DstVT) ||
+ !isOperationLegalOrCustomOrPromote(ISD::XOR, SrcVT)))
+ return false;
+
// Expand based on maximum range of FP_TO_SINT:
// True = fp_to_sint(Src)
// False = 0x8000000000000000 + fp_to_sint(Src - 0x8000000000000000)
Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=345473&r1=345472&r2=345473&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Sun Oct 28 06:07:25 2018
@@ -15698,7 +15698,9 @@ SDValue X86TargetLowering::LowerVSELECT(
// If this VSELECT has a vector if i1 as a mask, it will be directly matched
// with patterns on the mask registers on AVX-512.
- if (Cond.getScalarValueSizeInBits() == 1)
+ MVT CondVT = Cond.getSimpleValueType();
+ unsigned CondEltSize = Cond.getScalarValueSizeInBits();
+ if (CondEltSize == 1)
return Op;
// Variable blends are only legal from SSE4.1 onward.
@@ -15707,24 +15709,34 @@ SDValue X86TargetLowering::LowerVSELECT(
SDLoc dl(Op);
MVT VT = Op.getSimpleValueType();
+ unsigned EltSize = VT.getScalarSizeInBits();
+ unsigned NumElts = VT.getVectorNumElements();
// If the VSELECT is on a 512-bit type, we have to convert a non-i1 condition
// into an i1 condition so that we can use the mask-based 512-bit blend
// instructions.
if (VT.getSizeInBits() == 512) {
- // The vNi1 condition case should be handled above as it can be trivially
- // lowered.
- assert(Cond.getScalarValueSizeInBits() == VT.getScalarSizeInBits() &&
- "Should have a size-matched integer condition!");
// Build a mask by testing the condition against zero.
- MVT MaskVT = MVT::getVectorVT(MVT::i1, VT.getVectorNumElements());
+ MVT MaskVT = MVT::getVectorVT(MVT::i1, NumElts);
SDValue Mask = DAG.getSetCC(dl, MaskVT, Cond,
- getZeroVector(VT, Subtarget, DAG, dl),
+ getZeroVector(CondVT, Subtarget, DAG, dl),
ISD::SETNE);
// Now return a new VSELECT using the mask.
return DAG.getSelect(dl, VT, Mask, LHS, RHS);
}
+ // SEXT/TRUNC cases where the mask doesn't match the destination size.
+ if (CondEltSize != EltSize) {
+ // If we don't have a sign splat, rely on the expansion.
+ if (CondEltSize != DAG.ComputeNumSignBits(Cond))
+ return SDValue();
+
+ MVT NewCondSVT = MVT::getIntegerVT(EltSize);
+ MVT NewCondVT = MVT::getVectorVT(NewCondSVT, NumElts);
+ Cond = DAG.getSExtOrTrunc(Cond, dl, NewCondVT);
+ return DAG.getNode(ISD::VSELECT, dl, VT, Cond, LHS, RHS);
+ }
+
// Only some types will be legal on some subtargets. If we can emit a legal
// VSELECT-matching blend, return Op, and but if we need to expand, return
// a null value.
@@ -15743,7 +15755,7 @@ SDValue X86TargetLowering::LowerVSELECT(
case MVT::v8i16:
case MVT::v16i16: {
// Bitcast everything to the vXi8 type and use a vXi8 vselect.
- MVT CastVT = MVT::getVectorVT(MVT::i8, VT.getVectorNumElements() * 2);
+ MVT CastVT = MVT::getVectorVT(MVT::i8, NumElts * 2);
Cond = DAG.getBitcast(CastVT, Cond);
LHS = DAG.getBitcast(CastVT, LHS);
RHS = DAG.getBitcast(CastVT, RHS);
Modified: llvm/trunk/test/CodeGen/X86/ftrunc.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/ftrunc.ll?rev=345473&r1=345472&r2=345473&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/ftrunc.ll (original)
+++ llvm/trunk/test/CodeGen/X86/ftrunc.ll Sun Oct 28 06:07:25 2018
@@ -63,25 +63,19 @@ define double @trunc_unsigned_f64(double
define <4 x float> @trunc_unsigned_v4f32(<4 x float> %x) #0 {
; SSE2-LABEL: trunc_unsigned_v4f32:
; SSE2: # %bb.0:
+; SSE2-NEXT: movaps {{.*#+}} xmm2 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
; SSE2-NEXT: movaps %xmm0, %xmm1
-; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE2-NEXT: cvttss2si %xmm1, %rax
-; SSE2-NEXT: movd %eax, %xmm1
-; SSE2-NEXT: movaps %xmm0, %xmm2
-; SSE2-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; SSE2-NEXT: cvttss2si %xmm2, %rax
-; SSE2-NEXT: movd %eax, %xmm2
-; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE2-NEXT: cvttss2si %xmm0, %rax
-; SSE2-NEXT: movd %eax, %xmm1
-; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-NEXT: cvttss2si %xmm0, %rax
-; SSE2-NEXT: movd %eax, %xmm0
-; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; SSE2-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535]
-; SSE2-NEXT: pand %xmm1, %xmm0
-; SSE2-NEXT: por {{.*}}(%rip), %xmm0
+; SSE2-NEXT: cmpltps %xmm2, %xmm1
+; SSE2-NEXT: cvttps2dq %xmm0, %xmm3
+; SSE2-NEXT: subps %xmm2, %xmm0
+; SSE2-NEXT: cvttps2dq %xmm0, %xmm0
+; SSE2-NEXT: xorps {{.*}}(%rip), %xmm0
+; SSE2-NEXT: andps %xmm1, %xmm3
+; SSE2-NEXT: andnps %xmm0, %xmm1
+; SSE2-NEXT: orps %xmm3, %xmm1
+; SSE2-NEXT: movaps {{.*#+}} xmm0 = [65535,65535,65535,65535]
+; SSE2-NEXT: andps %xmm1, %xmm0
+; SSE2-NEXT: orps {{.*}}(%rip), %xmm0
; SSE2-NEXT: psrld $16, %xmm1
; SSE2-NEXT: por {{.*}}(%rip), %xmm1
; SSE2-NEXT: addps {{.*}}(%rip), %xmm1
Modified: llvm/trunk/test/CodeGen/X86/vec_cast3.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_cast3.ll?rev=345473&r1=345472&r2=345473&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vec_cast3.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vec_cast3.ll Sun Oct 28 06:07:25 2018
@@ -317,25 +317,13 @@ define <2 x i32> @cvt_v2f32_v2u32(<2 x f
;
; CHECK-WIDE-LABEL: cvt_v2f32_v2u32:
; CHECK-WIDE: ## %bb.0:
-; CHECK-WIDE-NEXT: subl $68, %esp
-; CHECK-WIDE-NEXT: .cfi_def_cfa_offset 72
-; CHECK-WIDE-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: flds {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: fisttpll {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: flds {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: fisttpll {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: flds {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: fisttpll {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: flds {{[0-9]+}}(%esp)
-; CHECK-WIDE-NEXT: fisttpll (%esp)
-; CHECK-WIDE-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; CHECK-WIDE-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; CHECK-WIDE-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; CHECK-WIDE-NEXT: vpinsrd $3, (%esp), %xmm0, %xmm0
-; CHECK-WIDE-NEXT: addl $68, %esp
+; CHECK-WIDE-NEXT: vmovaps {{.*#+}} xmm1 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
+; CHECK-WIDE-NEXT: vcmpltps %xmm1, %xmm0, %xmm2
+; CHECK-WIDE-NEXT: vsubps %xmm1, %xmm0, %xmm1
+; CHECK-WIDE-NEXT: vcvttps2dq %xmm1, %xmm1
+; CHECK-WIDE-NEXT: vxorps LCPI11_1, %xmm1, %xmm1
+; CHECK-WIDE-NEXT: vcvttps2dq %xmm0, %xmm0
+; CHECK-WIDE-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; CHECK-WIDE-NEXT: retl
%res = fptoui <2 x float> %src to <2 x i32>
ret <2 x i32> %res
Modified: llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll?rev=345473&r1=345472&r2=345473&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll Sun Oct 28 06:07:25 2018
@@ -627,16 +627,36 @@ define <4 x i32> @fptoui_4f64_to_2i32(<2
; SSE-NEXT: movaps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; VEX-LABEL: fptoui_4f64_to_2i32:
-; VEX: # %bb.0:
-; VEX-NEXT: vcvttsd2si %xmm0, %rax
-; VEX-NEXT: vmovd %eax, %xmm1
-; VEX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; VEX-NEXT: vcvttsd2si %xmm0, %rax
-; VEX-NEXT: vmovd %eax, %xmm0
-; VEX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; VEX-NEXT: vmovq {{.*#+}} xmm0 = xmm0[0],zero
-; VEX-NEXT: retq
+; AVX1-LABEL: fptoui_4f64_to_2i32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovapd %xmm0, %xmm0
+; AVX1-NEXT: vmovapd {{.*#+}} ymm1 = [2147483648,2147483648,2147483648,2147483648]
+; AVX1-NEXT: vcmpltpd %ymm1, %ymm0, %ymm2
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vpackssdw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vsubpd %ymm1, %ymm0, %ymm1
+; AVX1-NEXT: vcvttpd2dq %ymm1, %xmm1
+; AVX1-NEXT: vxorpd {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vcvttpd2dq %ymm0, %xmm0
+; AVX1-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: fptoui_4f64_to_2i32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovapd %xmm0, %xmm0
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm1 = [2147483648,2147483648,2147483648,2147483648]
+; AVX2-NEXT: vcmpltpd %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX2-NEXT: vpackssdw %xmm3, %xmm2, %xmm2
+; AVX2-NEXT: vsubpd %ymm1, %ymm0, %ymm1
+; AVX2-NEXT: vcvttpd2dq %ymm1, %xmm1
+; AVX2-NEXT: vbroadcastss {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
+; AVX2-NEXT: vxorpd %xmm3, %xmm1, %xmm1
+; AVX2-NEXT: vcvttpd2dq %ymm0, %xmm0
+; AVX2-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
;
; AVX512F-LABEL: fptoui_4f64_to_2i32:
; AVX512F: # %bb.0:
@@ -930,21 +950,34 @@ define <4 x i32> @fptoui_4f64_to_4i32(<4
; SSE-NEXT: movaps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; VEX-LABEL: fptoui_4f64_to_4i32:
-; VEX: # %bb.0:
-; VEX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; VEX-NEXT: vcvttsd2si %xmm1, %rax
-; VEX-NEXT: vcvttsd2si %xmm0, %rcx
-; VEX-NEXT: vmovd %ecx, %xmm1
-; VEX-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
-; VEX-NEXT: vextractf128 $1, %ymm0, %xmm0
-; VEX-NEXT: vcvttsd2si %xmm0, %rax
-; VEX-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
-; VEX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; VEX-NEXT: vcvttsd2si %xmm0, %rax
-; VEX-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
-; VEX-NEXT: vzeroupper
-; VEX-NEXT: retq
+; AVX1-LABEL: fptoui_4f64_to_4i32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovapd {{.*#+}} ymm1 = [2147483648,2147483648,2147483648,2147483648]
+; AVX1-NEXT: vcmpltpd %ymm1, %ymm0, %ymm2
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vpackssdw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vsubpd %ymm1, %ymm0, %ymm1
+; AVX1-NEXT: vcvttpd2dq %ymm1, %xmm1
+; AVX1-NEXT: vxorpd {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vcvttpd2dq %ymm0, %xmm0
+; AVX1-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: fptoui_4f64_to_4i32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm1 = [2147483648,2147483648,2147483648,2147483648]
+; AVX2-NEXT: vcmpltpd %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX2-NEXT: vpackssdw %xmm3, %xmm2, %xmm2
+; AVX2-NEXT: vsubpd %ymm1, %ymm0, %ymm1
+; AVX2-NEXT: vcvttpd2dq %ymm1, %xmm1
+; AVX2-NEXT: vbroadcastss {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
+; AVX2-NEXT: vxorpd %xmm3, %xmm1, %xmm1
+; AVX2-NEXT: vcvttpd2dq %ymm0, %xmm0
+; AVX2-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
;
; AVX512F-LABEL: fptoui_4f64_to_4i32:
; AVX512F: # %bb.0:
@@ -1570,39 +1603,41 @@ define <2 x i32> @fptoui_2f32_to_2i32(<2
define <4 x i32> @fptoui_4f32_to_4i32(<4 x float> %a) {
; SSE-LABEL: fptoui_4f32_to_4i32:
; SSE: # %bb.0:
+; SSE-NEXT: movaps {{.*#+}} xmm2 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
; SSE-NEXT: movaps %xmm0, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: cvttss2si %xmm1, %rax
-; SSE-NEXT: movd %eax, %xmm1
-; SSE-NEXT: movaps %xmm0, %xmm2
-; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; SSE-NEXT: cvttss2si %xmm2, %rax
-; SSE-NEXT: movd %eax, %xmm2
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: cvttss2si %xmm0, %rax
-; SSE-NEXT: movd %eax, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE-NEXT: cvttss2si %xmm0, %rax
-; SSE-NEXT: movd %eax, %xmm0
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: cmpltps %xmm2, %xmm1
+; SSE-NEXT: cvttps2dq %xmm0, %xmm3
+; SSE-NEXT: subps %xmm2, %xmm0
+; SSE-NEXT: cvttps2dq %xmm0, %xmm0
+; SSE-NEXT: xorps {{.*}}(%rip), %xmm0
+; SSE-NEXT: andps %xmm1, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm1
+; SSE-NEXT: orps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; VEX-LABEL: fptoui_4f32_to_4i32:
-; VEX: # %bb.0:
-; VEX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; VEX-NEXT: vcvttss2si %xmm1, %rax
-; VEX-NEXT: vcvttss2si %xmm0, %rcx
-; VEX-NEXT: vmovd %ecx, %xmm1
-; VEX-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
-; VEX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; VEX-NEXT: vcvttss2si %xmm2, %rax
-; VEX-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
-; VEX-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; VEX-NEXT: vcvttss2si %xmm0, %rax
-; VEX-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
-; VEX-NEXT: retq
+; AVX1-LABEL: fptoui_4f32_to_4i32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps {{.*#+}} xmm1 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
+; AVX1-NEXT: vcmpltps %xmm1, %xmm0, %xmm2
+; AVX1-NEXT: vsubps %xmm1, %xmm0, %xmm1
+; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
+; AVX1-NEXT: vxorps {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vcvttps2dq %xmm0, %xmm0
+; AVX1-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: fptoui_4f32_to_4i32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
+; AVX2-NEXT: vcmpltps %xmm1, %xmm0, %xmm2
+; AVX2-NEXT: vsubps %xmm1, %xmm0, %xmm1
+; AVX2-NEXT: vcvttps2dq %xmm1, %xmm1
+; AVX2-NEXT: vbroadcastss {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
+; AVX2-NEXT: vxorps %xmm3, %xmm1, %xmm1
+; AVX2-NEXT: vcvttps2dq %xmm0, %xmm0
+; AVX2-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: retq
;
; AVX512F-LABEL: fptoui_4f32_to_4i32:
; AVX512F: # %bb.0:
@@ -1853,95 +1888,51 @@ define <2 x i64> @fptoui_4f32_to_2i64(<4
define <8 x i32> @fptoui_8f32_to_8i32(<8 x float> %a) {
; SSE-LABEL: fptoui_8f32_to_8i32:
; SSE: # %bb.0:
+; SSE-NEXT: movaps {{.*#+}} xmm4 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
; SSE-NEXT: movaps %xmm0, %xmm2
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; SSE-NEXT: cvttss2si %xmm0, %rax
-; SSE-NEXT: movd %eax, %xmm0
-; SSE-NEXT: movaps %xmm2, %xmm3
-; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm2[1]
-; SSE-NEXT: cvttss2si %xmm3, %rax
-; SSE-NEXT: movd %eax, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: cvttss2si %xmm2, %rax
-; SSE-NEXT: movd %eax, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,1,2,3]
-; SSE-NEXT: cvttss2si %xmm2, %rax
-; SSE-NEXT: movd %eax, %xmm2
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
-; SSE-NEXT: movaps %xmm1, %xmm2
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
-; SSE-NEXT: cvttss2si %xmm2, %rax
-; SSE-NEXT: movd %eax, %xmm2
+; SSE-NEXT: cmpltps %xmm4, %xmm2
+; SSE-NEXT: cvttps2dq %xmm0, %xmm3
+; SSE-NEXT: subps %xmm4, %xmm0
+; SSE-NEXT: cvttps2dq %xmm0, %xmm0
+; SSE-NEXT: movaps {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648]
+; SSE-NEXT: xorps %xmm5, %xmm0
+; SSE-NEXT: andps %xmm2, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm2
+; SSE-NEXT: orps %xmm3, %xmm2
; SSE-NEXT: movaps %xmm1, %xmm3
-; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
-; SSE-NEXT: cvttss2si %xmm3, %rax
-; SSE-NEXT: movd %eax, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: cvttss2si %xmm1, %rax
-; SSE-NEXT: movd %eax, %xmm2
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
-; SSE-NEXT: cvttss2si %xmm1, %rax
-; SSE-NEXT: movd %eax, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: cmpltps %xmm4, %xmm3
+; SSE-NEXT: cvttps2dq %xmm1, %xmm0
+; SSE-NEXT: subps %xmm4, %xmm1
+; SSE-NEXT: cvttps2dq %xmm1, %xmm1
+; SSE-NEXT: xorps %xmm5, %xmm1
+; SSE-NEXT: andps %xmm3, %xmm0
+; SSE-NEXT: andnps %xmm1, %xmm3
+; SSE-NEXT: orps %xmm0, %xmm3
+; SSE-NEXT: movaps %xmm2, %xmm0
+; SSE-NEXT: movaps %xmm3, %xmm1
; SSE-NEXT: retq
;
; AVX1-LABEL: fptoui_8f32_to_8i32:
; AVX1: # %bb.0:
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; AVX1-NEXT: vcvttss2si %xmm2, %rax
-; AVX1-NEXT: vcvttss2si %xmm1, %rcx
-; AVX1-NEXT: vmovd %ecx, %xmm2
-; AVX1-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
-; AVX1-NEXT: vcvttss2si %xmm3, %rax
-; AVX1-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
-; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; AVX1-NEXT: vcvttss2si %xmm1, %rax
-; AVX1-NEXT: vpinsrd $3, %eax, %xmm2, %xmm1
-; AVX1-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; AVX1-NEXT: vcvttss2si %xmm2, %rax
-; AVX1-NEXT: vcvttss2si %xmm0, %rcx
-; AVX1-NEXT: vmovd %ecx, %xmm2
-; AVX1-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
-; AVX1-NEXT: vcvttss2si %xmm3, %rax
-; AVX1-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
-; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; AVX1-NEXT: vcvttss2si %xmm0, %rax
-; AVX1-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
-; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{.*#+}} ymm1 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
+; AVX1-NEXT: vcmpltps %ymm1, %ymm0, %ymm2
+; AVX1-NEXT: vsubps %ymm1, %ymm0, %ymm1
+; AVX1-NEXT: vcvttps2dq %ymm1, %ymm1
+; AVX1-NEXT: vxorps {{.*}}(%rip), %ymm1, %ymm1
+; AVX1-NEXT: vcvttps2dq %ymm0, %ymm0
+; AVX1-NEXT: vblendvps %ymm2, %ymm0, %ymm1, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: fptoui_8f32_to_8i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; AVX2-NEXT: vcvttss2si %xmm2, %rax
-; AVX2-NEXT: vcvttss2si %xmm1, %rcx
-; AVX2-NEXT: vmovd %ecx, %xmm2
-; AVX2-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
-; AVX2-NEXT: vcvttss2si %xmm3, %rax
-; AVX2-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
-; AVX2-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; AVX2-NEXT: vcvttss2si %xmm1, %rax
-; AVX2-NEXT: vpinsrd $3, %eax, %xmm2, %xmm1
-; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; AVX2-NEXT: vcvttss2si %xmm2, %rax
-; AVX2-NEXT: vcvttss2si %xmm0, %rcx
-; AVX2-NEXT: vmovd %ecx, %xmm2
-; AVX2-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
-; AVX2-NEXT: vcvttss2si %xmm3, %rax
-; AVX2-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
-; AVX2-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; AVX2-NEXT: vcvttss2si %xmm0, %rax
-; AVX2-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
-; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: vbroadcastss {{.*#+}} ymm1 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9]
+; AVX2-NEXT: vcmpltps %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vsubps %ymm1, %ymm0, %ymm1
+; AVX2-NEXT: vcvttps2dq %ymm1, %ymm1
+; AVX2-NEXT: vbroadcastss {{.*#+}} ymm3 = [2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648]
+; AVX2-NEXT: vxorps %ymm3, %ymm1, %ymm1
+; AVX2-NEXT: vcvttps2dq %ymm0, %ymm0
+; AVX2-NEXT: vblendvps %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: fptoui_8f32_to_8i32:
More information about the llvm-commits
mailing list