Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -1679,6 +1679,8 @@
   setTargetDAGCombine(ISD::ZERO_EXTEND_VECTOR_INREG);
   setTargetDAGCombine(ISD::SINT_TO_FP);
   setTargetDAGCombine(ISD::UINT_TO_FP);
+  setTargetDAGCombine(ISD::FP_TO_SINT);
+  setTargetDAGCombine(ISD::FP_TO_UINT);
   setTargetDAGCombine(ISD::SETCC);
   setTargetDAGCombine(ISD::MUL);
   setTargetDAGCombine(ISD::XOR);
@@ -36905,6 +36907,32 @@
   return SDValue();
 }
 
+static SDValue combineIntToFP_AVX512(SDNode *N, SelectionDAG &DAG,
+                                     const X86Subtarget &Subtarget) {
+  SDValue Op0 = N->getOperand(0);
+  EVT VT = N->getValueType(0);
+  EVT InVT = Op0.getValueType();
+
+  // AVX512DQ 32-bit targets don't have i64 scalar float/double conversions,
+  // so perform it entirely as a vector.
+  if (!Subtarget.hasDQI() || InVT != MVT::i64 ||
+      (VT != MVT::f32 && VT != MVT::f64) ||
+      DAG.getTargetLoweringInfo().isTypeLegal(InVT))
+    return SDValue();
+
+  SDLoc DL(N);
+  MVT VecInVT = Subtarget.hasVLX() ? MVT::v4i64 : MVT::v8i64;
+  EVT VecVT =
+      EVT::getVectorVT(*DAG.getContext(), VT, VecInVT.getVectorNumElements());
+
+  SDValue ZeroIdx = DAG.getIntPtrConstant(0, DL);
+  SDValue InVec =
+      DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, VecInVT,
+                  getZeroVector(VecInVT, Subtarget, DAG, DL), Op0, ZeroIdx);
+  SDValue CvtVec = DAG.getNode(N->getOpcode(), SDLoc(N), VecVT, InVec);
+  return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, CvtVec, ZeroIdx);
+}
+
 static SDValue combineUIntToFP(SDNode *N, SelectionDAG &DAG,
                                const X86Subtarget &Subtarget) {
   SDValue Op0 = N->getOperand(0);
@@ -36930,6 +36958,11 @@
   if (DAG.SignBitIsZero(Op0))
     return DAG.getNode(ISD::SINT_TO_FP, SDLoc(N), VT, Op0);
 
+  // Attempt AVX512-only combines.
+  if (Subtarget.hasAVX512())
+    if (SDValue Cvt = combineIntToFP_AVX512(N, DAG, Subtarget))
+      return Cvt;
+
   return SDValue();
 }
 
@@ -36992,9 +37025,41 @@
       return FILDChain;
     }
   }
+
+  // Attempt AVX512-only combines.
+  if (Subtarget.hasAVX512())
+    if (SDValue Cvt = combineIntToFP_AVX512(N, DAG, Subtarget))
+      return Cvt;
+
   return SDValue();
 }
 
+static SDValue combineFPToInt(SDNode *N, SelectionDAG &DAG,
+                              const X86Subtarget &Subtarget) {
+  SDValue Op0 = N->getOperand(0);
+  EVT VT = N->getValueType(0);
+  EVT InVT = Op0.getValueType();
+
+  // AVX512DQ 32-bit targets don't have float/double scalar i64 conversions,
+  // so perform it entirely as a vector.
+  if (!Subtarget.hasDQI() || VT != MVT::i64 ||
+      (InVT != MVT::f32 && InVT != MVT::f64) ||
+      DAG.getTargetLoweringInfo().isTypeLegal(VT))
+    return SDValue();
+
+  SDLoc DL(N);
+  MVT VecVT = Subtarget.hasVLX() ? MVT::v4i64 : MVT::v8i64;
+  MVT VecInVT =
+      MVT::getVectorVT(InVT.getSimpleVT(), VecVT.getVectorNumElements());
+
+  SDValue ZeroIdx = DAG.getIntPtrConstant(0, DL);
+  SDValue InVec =
+      DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, VecInVT,
+                  getZeroVector(VecInVT, Subtarget, DAG, DL), Op0, ZeroIdx);
+  SDValue CvtVec = DAG.getNode(N->getOpcode(), SDLoc(N), VecVT, InVec);
+  return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, CvtVec, ZeroIdx);
+}
+
 static SDValue combineSBB(SDNode *N, SelectionDAG &DAG) {
   if (SDValue Flags = combineCarryThroughADD(N->getOperand(2))) {
     MVT VT = N->getSimpleValueType(0);
@@ -37956,6 +38021,8 @@
   case ISD::MSTORE:         return combineMaskedStore(N, DAG, Subtarget);
   case ISD::SINT_TO_FP:     return combineSIntToFP(N, DAG, Subtarget);
   case ISD::UINT_TO_FP:     return combineUIntToFP(N, DAG, Subtarget);
+  case ISD::FP_TO_SINT:
+  case ISD::FP_TO_UINT:     return combineFPToInt(N, DAG, Subtarget);
   case ISD::FADD:
   case ISD::FSUB:           return combineFaddFsub(N, DAG, Subtarget);
   case ISD::FNEG:           return combineFneg(N, DAG, Subtarget);
Index: test/CodeGen/X86/avx512-regcall-NoMask.ll
===================================================================
--- test/CodeGen/X86/avx512-regcall-NoMask.ll
+++ test/CodeGen/X86/avx512-regcall-NoMask.ll
@@ -1149,30 +1149,28 @@
 define x86_regcallcc i32 @test_argRetMixTypes(double, float, i8 signext, i32, i64, i16 signext, i32*) #0 {
 ; X32-LABEL: test_argRetMixTypes:
 ; X32:       # %bb.0:
-; X32-NEXT:    pushl %ebp
-; X32-NEXT:    movl %esp, %ebp
-; X32-NEXT:    andl $-8, %esp
-; X32-NEXT:    subl $16, %esp
-; X32-NEXT:    vmovd %edx, %xmm2
-; X32-NEXT:    vpinsrd $1, %edi, %xmm2, %xmm2
-; X32-NEXT:    movl 8(%ebp), %edx
+; X32-NEXT:    pushl %ebx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
 ; X32-NEXT:    vcvtss2sd %xmm1, %xmm1, %xmm1
 ; X32-NEXT:    vaddsd %xmm0, %xmm1, %xmm0
-; X32-NEXT:    vcvtsi2sdl %eax, %xmm3, %xmm1
+; X32-NEXT:    vcvtsi2sdl %eax, %xmm2, %xmm1
 ; X32-NEXT:    vaddsd %xmm1, %xmm0, %xmm0
-; X32-NEXT:    vcvtsi2sdl %ecx, %xmm3, %xmm1
+; X32-NEXT:    vcvtsi2sdl %ecx, %xmm2, %xmm1
+; X32-NEXT:    vaddsd %xmm1, %xmm0, %xmm0
+; X32-NEXT:    vmovd %edi, %xmm1
+; X32-NEXT:    vmovd %edx, %xmm2
+; X32-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X32-NEXT:    vmovq {{.*#+}} xmm1 = xmm1[0],zero
+; X32-NEXT:    vmovdqa %xmm1, %xmm1
+; X32-NEXT:    vcvtqq2pd %ymm1, %ymm1
 ; X32-NEXT:    vaddsd %xmm1, %xmm0, %xmm0
-; X32-NEXT:    vmovq %xmm2, {{[0-9]+}}(%esp)
-; X32-NEXT:    fildll {{[0-9]+}}(%esp)
-; X32-NEXT:    fstpl (%esp)
-; X32-NEXT:    vaddsd (%esp), %xmm0, %xmm0
 ; X32-NEXT:    vcvtsi2sdl %esi, %xmm3, %xmm1
 ; X32-NEXT:    vaddsd %xmm1, %xmm0, %xmm0
-; X32-NEXT:    vcvtsi2sdl (%edx), %xmm3, %xmm1
+; X32-NEXT:    vcvtsi2sdl (%ebx), %xmm3, %xmm1
 ; X32-NEXT:    vaddsd %xmm1, %xmm0, %xmm0
 ; X32-NEXT:    vcvttsd2si %xmm0, %eax
-; X32-NEXT:    movl %ebp, %esp
-; X32-NEXT:    popl %ebp
+; X32-NEXT:    popl %ebx
+; X32-NEXT:    vzeroupper
 ; X32-NEXT:    retl
 ;
 ; WIN64-LABEL: test_argRetMixTypes:
Index: test/CodeGen/X86/scalar-fp-to-i64.ll
===================================================================
--- test/CodeGen/X86/scalar-fp-to-i64.ll
+++ test/CodeGen/X86/scalar-fp-to-i64.ll
@@ -35,55 +35,73 @@
 ; SSE2 (cvtts[ds]2si) and vanilla X87 (fnstcw+fist, 32-bit only).
 
 define i64 @f_to_u64(float %a) nounwind {
-; AVX512_32_WIN-LABEL: f_to_u64:
-; AVX512_32_WIN:       # %bb.0:
-; AVX512_32_WIN-NEXT:    pushl %ebp
-; AVX512_32_WIN-NEXT:    movl %esp, %ebp
-; AVX512_32_WIN-NEXT:    andl $-8, %esp
-; AVX512_32_WIN-NEXT:    subl $16, %esp
-; AVX512_32_WIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX512_32_WIN-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; AVX512_32_WIN-NEXT:    vcmpltss %xmm1, %xmm0, %k1
-; AVX512_32_WIN-NEXT:    vsubss %xmm1, %xmm0, %xmm2
-; AVX512_32_WIN-NEXT:    vmovss %xmm0, %xmm0, %xmm2 {%k1}
-; AVX512_32_WIN-NEXT:    vmovss %xmm2, {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    flds {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    fisttpll (%esp)
-; AVX512_32_WIN-NEXT:    xorl %edx, %edx
-; AVX512_32_WIN-NEXT:    vucomiss %xmm0, %xmm1
-; AVX512_32_WIN-NEXT:    setbe %dl
-; AVX512_32_WIN-NEXT:    shll $31, %edx
-; AVX512_32_WIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
-; AVX512_32_WIN-NEXT:    movl (%esp), %eax
-; AVX512_32_WIN-NEXT:    movl %ebp, %esp
-; AVX512_32_WIN-NEXT:    popl %ebp
-; AVX512_32_WIN-NEXT:    retl
-;
-; AVX512_32_LIN-LABEL: f_to_u64:
-; AVX512_32_LIN:       # %bb.0:
-; AVX512_32_LIN-NEXT:    subl $20, %esp
-; AVX512_32_LIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX512_32_LIN-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; AVX512_32_LIN-NEXT:    vcmpltss %xmm1, %xmm0, %k1
-; AVX512_32_LIN-NEXT:    vsubss %xmm1, %xmm0, %xmm2
-; AVX512_32_LIN-NEXT:    vmovss %xmm0, %xmm0, %xmm2 {%k1}
-; AVX512_32_LIN-NEXT:    vmovss %xmm2, {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    flds {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    fisttpll (%esp)
-; AVX512_32_LIN-NEXT:    xorl %edx, %edx
-; AVX512_32_LIN-NEXT:    vucomiss %xmm0, %xmm1
-; AVX512_32_LIN-NEXT:    setbe %dl
-; AVX512_32_LIN-NEXT:    shll $31, %edx
-; AVX512_32_LIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
-; AVX512_32_LIN-NEXT:    movl (%esp), %eax
-; AVX512_32_LIN-NEXT:    addl $20, %esp
-; AVX512_32_LIN-NEXT:    retl
+; AVX512DQ_32_WIN-LABEL: f_to_u64:
+; AVX512DQ_32_WIN:       # %bb.0:
+; AVX512DQ_32_WIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512DQ_32_WIN-NEXT:    vcvttps2uqq %ymm0, %zmm0
+; AVX512DQ_32_WIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_WIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_WIN-NEXT:    vzeroupper
+; AVX512DQ_32_WIN-NEXT:    retl
+;
+; AVX512DQ_32_LIN-LABEL: f_to_u64:
+; AVX512DQ_32_LIN:       # %bb.0:
+; AVX512DQ_32_LIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512DQ_32_LIN-NEXT:    vcvttps2uqq %ymm0, %zmm0
+; AVX512DQ_32_LIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_LIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_LIN-NEXT:    vzeroupper
+; AVX512DQ_32_LIN-NEXT:    retl
 ;
 ; AVX512_64-LABEL: f_to_u64:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvttss2usi %xmm0, %rax
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32_WIN-LABEL: f_to_u64:
+; AVX512F_32_WIN:       # %bb.0:
+; AVX512F_32_WIN-NEXT:    pushl %ebp
+; AVX512F_32_WIN-NEXT:    movl %esp, %ebp
+; AVX512F_32_WIN-NEXT:    andl $-8, %esp
+; AVX512F_32_WIN-NEXT:    subl $16, %esp
+; AVX512F_32_WIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512F_32_WIN-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX512F_32_WIN-NEXT:    vcmpltss %xmm1, %xmm0, %k1
+; AVX512F_32_WIN-NEXT:    vsubss %xmm1, %xmm0, %xmm2
+; AVX512F_32_WIN-NEXT:    vmovss %xmm0, %xmm0, %xmm2 {%k1}
+; AVX512F_32_WIN-NEXT:    vmovss %xmm2, {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    flds {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_WIN-NEXT:    xorl %edx, %edx
+; AVX512F_32_WIN-NEXT:    vucomiss %xmm0, %xmm1
+; AVX512F_32_WIN-NEXT:    setbe %dl
+; AVX512F_32_WIN-NEXT:    shll $31, %edx
+; AVX512F_32_WIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_WIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_WIN-NEXT:    movl %ebp, %esp
+; AVX512F_32_WIN-NEXT:    popl %ebp
+; AVX512F_32_WIN-NEXT:    retl
+;
+; AVX512F_32_LIN-LABEL: f_to_u64:
+; AVX512F_32_LIN:       # %bb.0:
+; AVX512F_32_LIN-NEXT:    subl $20, %esp
+; AVX512F_32_LIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512F_32_LIN-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX512F_32_LIN-NEXT:    vcmpltss %xmm1, %xmm0, %k1
+; AVX512F_32_LIN-NEXT:    vsubss %xmm1, %xmm0, %xmm2
+; AVX512F_32_LIN-NEXT:    vmovss %xmm0, %xmm0, %xmm2 {%k1}
+; AVX512F_32_LIN-NEXT:    vmovss %xmm2, {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    flds {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_LIN-NEXT:    xorl %edx, %edx
+; AVX512F_32_LIN-NEXT:    vucomiss %xmm0, %xmm1
+; AVX512F_32_LIN-NEXT:    setbe %dl
+; AVX512F_32_LIN-NEXT:    shll $31, %edx
+; AVX512F_32_LIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_LIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_LIN-NEXT:    addl $20, %esp
+; AVX512F_32_LIN-NEXT:    retl
+;
 ; SSE3_32_WIN-LABEL: f_to_u64:
 ; SSE3_32_WIN:       # %bb.0:
 ; SSE3_32_WIN-NEXT:    pushl %ebp
@@ -300,39 +318,57 @@
 }
 
 define i64 @f_to_s64(float %a) nounwind {
-; AVX512_32_WIN-LABEL: f_to_s64:
-; AVX512_32_WIN:       # %bb.0:
-; AVX512_32_WIN-NEXT:    pushl %ebp
-; AVX512_32_WIN-NEXT:    movl %esp, %ebp
-; AVX512_32_WIN-NEXT:    andl $-8, %esp
-; AVX512_32_WIN-NEXT:    subl $16, %esp
-; AVX512_32_WIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX512_32_WIN-NEXT:    vmovss %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    flds {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    fisttpll (%esp)
-; AVX512_32_WIN-NEXT:    movl (%esp), %eax
-; AVX512_32_WIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512_32_WIN-NEXT:    movl %ebp, %esp
-; AVX512_32_WIN-NEXT:    popl %ebp
-; AVX512_32_WIN-NEXT:    retl
-;
-; AVX512_32_LIN-LABEL: f_to_s64:
-; AVX512_32_LIN:       # %bb.0:
-; AVX512_32_LIN-NEXT:    subl $20, %esp
-; AVX512_32_LIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX512_32_LIN-NEXT:    vmovss %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    flds {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    fisttpll (%esp)
-; AVX512_32_LIN-NEXT:    movl (%esp), %eax
-; AVX512_32_LIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512_32_LIN-NEXT:    addl $20, %esp
-; AVX512_32_LIN-NEXT:    retl
+; AVX512DQ_32_WIN-LABEL: f_to_s64:
+; AVX512DQ_32_WIN:       # %bb.0:
+; AVX512DQ_32_WIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512DQ_32_WIN-NEXT:    vcvttps2qq %ymm0, %zmm0
+; AVX512DQ_32_WIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_WIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_WIN-NEXT:    vzeroupper
+; AVX512DQ_32_WIN-NEXT:    retl
+;
+; AVX512DQ_32_LIN-LABEL: f_to_s64:
+; AVX512DQ_32_LIN:       # %bb.0:
+; AVX512DQ_32_LIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512DQ_32_LIN-NEXT:    vcvttps2qq %ymm0, %zmm0
+; AVX512DQ_32_LIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_LIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_LIN-NEXT:    vzeroupper
+; AVX512DQ_32_LIN-NEXT:    retl
 ;
 ; AVX512_64-LABEL: f_to_s64:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvttss2si %xmm0, %rax
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32_WIN-LABEL: f_to_s64:
+; AVX512F_32_WIN:       # %bb.0:
+; AVX512F_32_WIN-NEXT:    pushl %ebp
+; AVX512F_32_WIN-NEXT:    movl %esp, %ebp
+; AVX512F_32_WIN-NEXT:    andl $-8, %esp
+; AVX512F_32_WIN-NEXT:    subl $16, %esp
+; AVX512F_32_WIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512F_32_WIN-NEXT:    vmovss %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    flds {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_WIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_WIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_WIN-NEXT:    movl %ebp, %esp
+; AVX512F_32_WIN-NEXT:    popl %ebp
+; AVX512F_32_WIN-NEXT:    retl
+;
+; AVX512F_32_LIN-LABEL: f_to_s64:
+; AVX512F_32_LIN:       # %bb.0:
+; AVX512F_32_LIN-NEXT:    subl $20, %esp
+; AVX512F_32_LIN-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512F_32_LIN-NEXT:    vmovss %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    flds {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_LIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_LIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_LIN-NEXT:    addl $20, %esp
+; AVX512F_32_LIN-NEXT:    retl
+;
 ; SSE3_32_WIN-LABEL: f_to_s64:
 ; SSE3_32_WIN:       # %bb.0:
 ; SSE3_32_WIN-NEXT:    pushl %ebp
@@ -451,55 +487,73 @@
 }
 
 define i64 @d_to_u64(double %a) nounwind {
-; AVX512_32_WIN-LABEL: d_to_u64:
-; AVX512_32_WIN:       # %bb.0:
-; AVX512_32_WIN-NEXT:    pushl %ebp
-; AVX512_32_WIN-NEXT:    movl %esp, %ebp
-; AVX512_32_WIN-NEXT:    andl $-8, %esp
-; AVX512_32_WIN-NEXT:    subl $16, %esp
-; AVX512_32_WIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX512_32_WIN-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
-; AVX512_32_WIN-NEXT:    vcmpltsd %xmm1, %xmm0, %k1
-; AVX512_32_WIN-NEXT:    vsubsd %xmm1, %xmm0, %xmm2
-; AVX512_32_WIN-NEXT:    vmovsd %xmm0, %xmm0, %xmm2 {%k1}
-; AVX512_32_WIN-NEXT:    vmovsd %xmm2, {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    fldl {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    fisttpll (%esp)
-; AVX512_32_WIN-NEXT:    xorl %edx, %edx
-; AVX512_32_WIN-NEXT:    vucomisd %xmm0, %xmm1
-; AVX512_32_WIN-NEXT:    setbe %dl
-; AVX512_32_WIN-NEXT:    shll $31, %edx
-; AVX512_32_WIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
-; AVX512_32_WIN-NEXT:    movl (%esp), %eax
-; AVX512_32_WIN-NEXT:    movl %ebp, %esp
-; AVX512_32_WIN-NEXT:    popl %ebp
-; AVX512_32_WIN-NEXT:    retl
-;
-; AVX512_32_LIN-LABEL: d_to_u64:
-; AVX512_32_LIN:       # %bb.0:
-; AVX512_32_LIN-NEXT:    subl $20, %esp
-; AVX512_32_LIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX512_32_LIN-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
-; AVX512_32_LIN-NEXT:    vcmpltsd %xmm1, %xmm0, %k1
-; AVX512_32_LIN-NEXT:    vsubsd %xmm1, %xmm0, %xmm2
-; AVX512_32_LIN-NEXT:    vmovsd %xmm0, %xmm0, %xmm2 {%k1}
-; AVX512_32_LIN-NEXT:    vmovsd %xmm2, {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    fldl {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    fisttpll (%esp)
-; AVX512_32_LIN-NEXT:    xorl %edx, %edx
-; AVX512_32_LIN-NEXT:    vucomisd %xmm0, %xmm1
-; AVX512_32_LIN-NEXT:    setbe %dl
-; AVX512_32_LIN-NEXT:    shll $31, %edx
-; AVX512_32_LIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
-; AVX512_32_LIN-NEXT:    movl (%esp), %eax
-; AVX512_32_LIN-NEXT:    addl $20, %esp
-; AVX512_32_LIN-NEXT:    retl
+; AVX512DQ_32_WIN-LABEL: d_to_u64:
+; AVX512DQ_32_WIN:       # %bb.0:
+; AVX512DQ_32_WIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32_WIN-NEXT:    vcvttpd2uqq %zmm0, %zmm0
+; AVX512DQ_32_WIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_WIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_WIN-NEXT:    vzeroupper
+; AVX512DQ_32_WIN-NEXT:    retl
+;
+; AVX512DQ_32_LIN-LABEL: d_to_u64:
+; AVX512DQ_32_LIN:       # %bb.0:
+; AVX512DQ_32_LIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32_LIN-NEXT:    vcvttpd2uqq %zmm0, %zmm0
+; AVX512DQ_32_LIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_LIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_LIN-NEXT:    vzeroupper
+; AVX512DQ_32_LIN-NEXT:    retl
 ;
 ; AVX512_64-LABEL: d_to_u64:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvttsd2usi %xmm0, %rax
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32_WIN-LABEL: d_to_u64:
+; AVX512F_32_WIN:       # %bb.0:
+; AVX512F_32_WIN-NEXT:    pushl %ebp
+; AVX512F_32_WIN-NEXT:    movl %esp, %ebp
+; AVX512F_32_WIN-NEXT:    andl $-8, %esp
+; AVX512F_32_WIN-NEXT:    subl $16, %esp
+; AVX512F_32_WIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512F_32_WIN-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX512F_32_WIN-NEXT:    vcmpltsd %xmm1, %xmm0, %k1
+; AVX512F_32_WIN-NEXT:    vsubsd %xmm1, %xmm0, %xmm2
+; AVX512F_32_WIN-NEXT:    vmovsd %xmm0, %xmm0, %xmm2 {%k1}
+; AVX512F_32_WIN-NEXT:    vmovsd %xmm2, {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    fldl {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_WIN-NEXT:    xorl %edx, %edx
+; AVX512F_32_WIN-NEXT:    vucomisd %xmm0, %xmm1
+; AVX512F_32_WIN-NEXT:    setbe %dl
+; AVX512F_32_WIN-NEXT:    shll $31, %edx
+; AVX512F_32_WIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_WIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_WIN-NEXT:    movl %ebp, %esp
+; AVX512F_32_WIN-NEXT:    popl %ebp
+; AVX512F_32_WIN-NEXT:    retl
+;
+; AVX512F_32_LIN-LABEL: d_to_u64:
+; AVX512F_32_LIN:       # %bb.0:
+; AVX512F_32_LIN-NEXT:    subl $20, %esp
+; AVX512F_32_LIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512F_32_LIN-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX512F_32_LIN-NEXT:    vcmpltsd %xmm1, %xmm0, %k1
+; AVX512F_32_LIN-NEXT:    vsubsd %xmm1, %xmm0, %xmm2
+; AVX512F_32_LIN-NEXT:    vmovsd %xmm0, %xmm0, %xmm2 {%k1}
+; AVX512F_32_LIN-NEXT:    vmovsd %xmm2, {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    fldl {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_LIN-NEXT:    xorl %edx, %edx
+; AVX512F_32_LIN-NEXT:    vucomisd %xmm0, %xmm1
+; AVX512F_32_LIN-NEXT:    setbe %dl
+; AVX512F_32_LIN-NEXT:    shll $31, %edx
+; AVX512F_32_LIN-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_LIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_LIN-NEXT:    addl $20, %esp
+; AVX512F_32_LIN-NEXT:    retl
+;
 ; SSE3_32_WIN-LABEL: d_to_u64:
 ; SSE3_32_WIN:       # %bb.0:
 ; SSE3_32_WIN-NEXT:    pushl %ebp
@@ -716,39 +770,57 @@
 }
 
 define i64 @d_to_s64(double %a) nounwind {
-; AVX512_32_WIN-LABEL: d_to_s64:
-; AVX512_32_WIN:       # %bb.0:
-; AVX512_32_WIN-NEXT:    pushl %ebp
-; AVX512_32_WIN-NEXT:    movl %esp, %ebp
-; AVX512_32_WIN-NEXT:    andl $-8, %esp
-; AVX512_32_WIN-NEXT:    subl $16, %esp
-; AVX512_32_WIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX512_32_WIN-NEXT:    vmovsd %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    fldl {{[0-9]+}}(%esp)
-; AVX512_32_WIN-NEXT:    fisttpll (%esp)
-; AVX512_32_WIN-NEXT:    movl (%esp), %eax
-; AVX512_32_WIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512_32_WIN-NEXT:    movl %ebp, %esp
-; AVX512_32_WIN-NEXT:    popl %ebp
-; AVX512_32_WIN-NEXT:    retl
-;
-; AVX512_32_LIN-LABEL: d_to_s64:
-; AVX512_32_LIN:       # %bb.0:
-; AVX512_32_LIN-NEXT:    subl $20, %esp
-; AVX512_32_LIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX512_32_LIN-NEXT:    vmovsd %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    fldl {{[0-9]+}}(%esp)
-; AVX512_32_LIN-NEXT:    fisttpll (%esp)
-; AVX512_32_LIN-NEXT:    movl (%esp), %eax
-; AVX512_32_LIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512_32_LIN-NEXT:    addl $20, %esp
-; AVX512_32_LIN-NEXT:    retl
+; AVX512DQ_32_WIN-LABEL: d_to_s64:
+; AVX512DQ_32_WIN:       # %bb.0:
+; AVX512DQ_32_WIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32_WIN-NEXT:    vcvttpd2qq %zmm0, %zmm0
+; AVX512DQ_32_WIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_WIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_WIN-NEXT:    vzeroupper
+; AVX512DQ_32_WIN-NEXT:    retl
+;
+; AVX512DQ_32_LIN-LABEL: d_to_s64:
+; AVX512DQ_32_LIN:       # %bb.0:
+; AVX512DQ_32_LIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32_LIN-NEXT:    vcvttpd2qq %zmm0, %zmm0
+; AVX512DQ_32_LIN-NEXT:    vmovd %xmm0, %eax
+; AVX512DQ_32_LIN-NEXT:    vpextrd $1, %xmm0, %edx
+; AVX512DQ_32_LIN-NEXT:    vzeroupper
+; AVX512DQ_32_LIN-NEXT:    retl
 ;
 ; AVX512_64-LABEL: d_to_s64:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvttsd2si %xmm0, %rax
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32_WIN-LABEL: d_to_s64:
+; AVX512F_32_WIN:       # %bb.0:
+; AVX512F_32_WIN-NEXT:    pushl %ebp
+; AVX512F_32_WIN-NEXT:    movl %esp, %ebp
+; AVX512F_32_WIN-NEXT:    andl $-8, %esp
+; AVX512F_32_WIN-NEXT:    subl $16, %esp
+; AVX512F_32_WIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512F_32_WIN-NEXT:    vmovsd %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    fldl {{[0-9]+}}(%esp)
+; AVX512F_32_WIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_WIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_WIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_WIN-NEXT:    movl %ebp, %esp
+; AVX512F_32_WIN-NEXT:    popl %ebp
+; AVX512F_32_WIN-NEXT:    retl
+;
+; AVX512F_32_LIN-LABEL: d_to_s64:
+; AVX512F_32_LIN:       # %bb.0:
+; AVX512F_32_LIN-NEXT:    subl $20, %esp
+; AVX512F_32_LIN-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512F_32_LIN-NEXT:    vmovsd %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    fldl {{[0-9]+}}(%esp)
+; AVX512F_32_LIN-NEXT:    fisttpll (%esp)
+; AVX512F_32_LIN-NEXT:    movl (%esp), %eax
+; AVX512F_32_LIN-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F_32_LIN-NEXT:    addl $20, %esp
+; AVX512F_32_LIN-NEXT:    retl
+;
 ; SSE3_32_WIN-LABEL: d_to_s64:
 ; SSE3_32_WIN:       # %bb.0:
 ; SSE3_32_WIN-NEXT:    pushl %ebp
Index: test/CodeGen/X86/scalar-int-to-fp.ll
===================================================================
--- test/CodeGen/X86/scalar-int-to-fp.ll
+++ test/CodeGen/X86/scalar-int-to-fp.ll
@@ -298,32 +298,43 @@
 }
 
 define float @u64_to_f(i64 %a) nounwind {
-; AVX512_32-LABEL: u64_to_f:
-; AVX512_32:       # %bb.0:
-; AVX512_32-NEXT:    pushl %ebp
-; AVX512_32-NEXT:    movl %esp, %ebp
-; AVX512_32-NEXT:    andl $-8, %esp
-; AVX512_32-NEXT:    subl $16, %esp
-; AVX512_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX512_32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    xorl %eax, %eax
-; AVX512_32-NEXT:    cmpl $0, 12(%ebp)
-; AVX512_32-NEXT:    setns %al
-; AVX512_32-NEXT:    fildll {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
-; AVX512_32-NEXT:    fstps {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX512_32-NEXT:    vmovss %xmm0, (%esp)
-; AVX512_32-NEXT:    flds (%esp)
-; AVX512_32-NEXT:    movl %ebp, %esp
-; AVX512_32-NEXT:    popl %ebp
-; AVX512_32-NEXT:    retl
+; AVX512DQ_32-LABEL: u64_to_f:
+; AVX512DQ_32:       # %bb.0:
+; AVX512DQ_32-NEXT:    pushl %eax
+; AVX512DQ_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32-NEXT:    vcvtuqq2ps %zmm0, %ymm0
+; AVX512DQ_32-NEXT:    vmovss %xmm0, (%esp)
+; AVX512DQ_32-NEXT:    flds (%esp)
+; AVX512DQ_32-NEXT:    popl %eax
+; AVX512DQ_32-NEXT:    vzeroupper
+; AVX512DQ_32-NEXT:    retl
 ;
 ; AVX512_64-LABEL: u64_to_f:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvtusi2ssq %rdi, %xmm0, %xmm0
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32-LABEL: u64_to_f:
+; AVX512F_32:       # %bb.0:
+; AVX512F_32-NEXT:    pushl %ebp
+; AVX512F_32-NEXT:    movl %esp, %ebp
+; AVX512F_32-NEXT:    andl $-8, %esp
+; AVX512F_32-NEXT:    subl $16, %esp
+; AVX512F_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512F_32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    xorl %eax, %eax
+; AVX512F_32-NEXT:    cmpl $0, 12(%ebp)
+; AVX512F_32-NEXT:    setns %al
+; AVX512F_32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX512F_32-NEXT:    fstps {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; AVX512F_32-NEXT:    vmovss %xmm0, (%esp)
+; AVX512F_32-NEXT:    flds (%esp)
+; AVX512F_32-NEXT:    movl %ebp, %esp
+; AVX512F_32-NEXT:    popl %ebp
+; AVX512F_32-NEXT:    retl
+;
 ; SSE2_32-LABEL: u64_to_f:
 ; SSE2_32:       # %bb.0:
 ; SSE2_32-NEXT:    pushl %ebp
@@ -386,20 +397,31 @@
 }
 
 define float @s64_to_f(i64 %a) nounwind {
-; AVX512_32-LABEL: s64_to_f:
-; AVX512_32:       # %bb.0:
-; AVX512_32-NEXT:    pushl %eax
-; AVX512_32-NEXT:    fildll {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    fstps (%esp)
-; AVX512_32-NEXT:    flds (%esp)
-; AVX512_32-NEXT:    popl %eax
-; AVX512_32-NEXT:    retl
+; AVX512DQ_32-LABEL: s64_to_f:
+; AVX512DQ_32:       # %bb.0:
+; AVX512DQ_32-NEXT:    pushl %eax
+; AVX512DQ_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32-NEXT:    vcvtqq2ps %zmm0, %ymm0
+; AVX512DQ_32-NEXT:    vmovss %xmm0, (%esp)
+; AVX512DQ_32-NEXT:    flds (%esp)
+; AVX512DQ_32-NEXT:    popl %eax
+; AVX512DQ_32-NEXT:    vzeroupper
+; AVX512DQ_32-NEXT:    retl
 ;
 ; AVX512_64-LABEL: s64_to_f:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvtsi2ssq %rdi, %xmm0, %xmm0
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32-LABEL: s64_to_f:
+; AVX512F_32:       # %bb.0:
+; AVX512F_32-NEXT:    pushl %eax
+; AVX512F_32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    fstps (%esp)
+; AVX512F_32-NEXT:    flds (%esp)
+; AVX512F_32-NEXT:    popl %eax
+; AVX512F_32-NEXT:    retl
+;
 ; SSE2_32-LABEL: s64_to_f:
 ; SSE2_32:       # %bb.0:
 ; SSE2_32-NEXT:    pushl %eax
@@ -423,25 +445,24 @@
 }
 
 define float @s64_to_f_2(i64 %a) nounwind {
-; AVX512_32-LABEL: s64_to_f_2:
-; AVX512_32:       # %bb.0:
-; AVX512_32-NEXT:    pushl %ebp
-; AVX512_32-NEXT:    movl %esp, %ebp
-; AVX512_32-NEXT:    andl $-8, %esp
-; AVX512_32-NEXT:    subl $16, %esp
-; AVX512_32-NEXT:    movl 8(%ebp), %eax
-; AVX512_32-NEXT:    movl 12(%ebp), %ecx
-; AVX512_32-NEXT:    addl $5, %eax
-; AVX512_32-NEXT:    adcl $0, %ecx
-; AVX512_32-NEXT:    vmovd %eax, %xmm0
-; AVX512_32-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
-; AVX512_32-NEXT:    vmovq %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    fildll {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    fstps {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    flds {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    movl %ebp, %esp
-; AVX512_32-NEXT:    popl %ebp
-; AVX512_32-NEXT:    retl
+; AVX512DQ_32-LABEL: s64_to_f_2:
+; AVX512DQ_32:       # %bb.0:
+; AVX512DQ_32-NEXT:    pushl %eax
+; AVX512DQ_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512DQ_32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512DQ_32-NEXT:    addl $5, %eax
+; AVX512DQ_32-NEXT:    adcl $0, %ecx
+; AVX512DQ_32-NEXT:    vmovd %ecx, %xmm0
+; AVX512DQ_32-NEXT:    vmovd %eax, %xmm1
+; AVX512DQ_32-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX512DQ_32-NEXT:    vmovq {{.*#+}} xmm0 = xmm0[0],zero
+; AVX512DQ_32-NEXT:    vmovdqa %xmm0, %xmm0
+; AVX512DQ_32-NEXT:    vcvtqq2ps %zmm0, %ymm0
+; AVX512DQ_32-NEXT:    vmovss %xmm0, (%esp)
+; AVX512DQ_32-NEXT:    flds (%esp)
+; AVX512DQ_32-NEXT:    popl %eax
+; AVX512DQ_32-NEXT:    vzeroupper
+; AVX512DQ_32-NEXT:    retl
 ;
 ; AVX512_64-LABEL: s64_to_f_2:
 ; AVX512_64:       # %bb.0:
@@ -449,6 +470,26 @@
 ; AVX512_64-NEXT:    vcvtsi2ssq %rdi, %xmm0, %xmm0
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32-LABEL: s64_to_f_2:
+; AVX512F_32:       # %bb.0:
+; AVX512F_32-NEXT:    pushl %ebp
+; AVX512F_32-NEXT:    movl %esp, %ebp
+; AVX512F_32-NEXT:    andl $-8, %esp
+; AVX512F_32-NEXT:    subl $16, %esp
+; AVX512F_32-NEXT:    movl 8(%ebp), %eax
+; AVX512F_32-NEXT:    movl 12(%ebp), %ecx
+; AVX512F_32-NEXT:    addl $5, %eax
+; AVX512F_32-NEXT:    adcl $0, %ecx
+; AVX512F_32-NEXT:    vmovd %eax, %xmm0
+; AVX512F_32-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
+; AVX512F_32-NEXT:    vmovq %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    fstps {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    flds {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    movl %ebp, %esp
+; AVX512F_32-NEXT:    popl %ebp
+; AVX512F_32-NEXT:    retl
+;
 ; SSE2_32-LABEL: s64_to_f_2:
 ; SSE2_32:       # %bb.0:
 ; SSE2_32-NEXT:    pushl %ebp
@@ -498,27 +539,42 @@
 }
 
 define double @u64_to_d(i64 %a) nounwind {
-; AVX512_32-LABEL: u64_to_d:
-; AVX512_32:       # %bb.0:
-; AVX512_32-NEXT:    pushl %ebp
-; AVX512_32-NEXT:    movl %esp, %ebp
-; AVX512_32-NEXT:    andl $-8, %esp
-; AVX512_32-NEXT:    subl $8, %esp
-; AVX512_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX512_32-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; AVX512_32-NEXT:    vsubpd {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX512_32-NEXT:    vhaddpd %xmm0, %xmm0, %xmm0
-; AVX512_32-NEXT:    vmovlpd %xmm0, (%esp)
-; AVX512_32-NEXT:    fldl (%esp)
-; AVX512_32-NEXT:    movl %ebp, %esp
-; AVX512_32-NEXT:    popl %ebp
-; AVX512_32-NEXT:    retl
+; AVX512DQ_32-LABEL: u64_to_d:
+; AVX512DQ_32:       # %bb.0:
+; AVX512DQ_32-NEXT:    pushl %ebp
+; AVX512DQ_32-NEXT:    movl %esp, %ebp
+; AVX512DQ_32-NEXT:    andl $-8, %esp
+; AVX512DQ_32-NEXT:    subl $8, %esp
+; AVX512DQ_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32-NEXT:    vcvtuqq2pd %zmm0, %zmm0
+; AVX512DQ_32-NEXT:    vmovlps %xmm0, (%esp)
+; AVX512DQ_32-NEXT:    fldl (%esp)
+; AVX512DQ_32-NEXT:    movl %ebp, %esp
+; AVX512DQ_32-NEXT:    popl %ebp
+; AVX512DQ_32-NEXT:    vzeroupper
+; AVX512DQ_32-NEXT:    retl
 ;
 ; AVX512_64-LABEL: u64_to_d:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvtusi2sdq %rdi, %xmm0, %xmm0
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32-LABEL: u64_to_d:
+; AVX512F_32:       # %bb.0:
+; AVX512F_32-NEXT:    pushl %ebp
+; AVX512F_32-NEXT:    movl %esp, %ebp
+; AVX512F_32-NEXT:    andl $-8, %esp
+; AVX512F_32-NEXT:    subl $8, %esp
+; AVX512F_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512F_32-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; AVX512F_32-NEXT:    vsubpd {{\.LCPI.*}}, %xmm0, %xmm0
+; AVX512F_32-NEXT:    vhaddpd %xmm0, %xmm0, %xmm0
+; AVX512F_32-NEXT:    vmovlpd %xmm0, (%esp)
+; AVX512F_32-NEXT:    fldl (%esp)
+; AVX512F_32-NEXT:    movl %ebp, %esp
+; AVX512F_32-NEXT:    popl %ebp
+; AVX512F_32-NEXT:    retl
+;
 ; SSE2_32-LABEL: u64_to_d:
 ; SSE2_32:       # %bb.0:
 ; SSE2_32-NEXT:    pushl %ebp
@@ -570,24 +626,39 @@
 }
 
 define double @s64_to_d(i64 %a) nounwind {
-; AVX512_32-LABEL: s64_to_d:
-; AVX512_32:       # %bb.0:
-; AVX512_32-NEXT:    pushl %ebp
-; AVX512_32-NEXT:    movl %esp, %ebp
-; AVX512_32-NEXT:    andl $-8, %esp
-; AVX512_32-NEXT:    subl $8, %esp
-; AVX512_32-NEXT:    fildll 8(%ebp)
-; AVX512_32-NEXT:    fstpl (%esp)
-; AVX512_32-NEXT:    fldl (%esp)
-; AVX512_32-NEXT:    movl %ebp, %esp
-; AVX512_32-NEXT:    popl %ebp
-; AVX512_32-NEXT:    retl
+; AVX512DQ_32-LABEL: s64_to_d:
+; AVX512DQ_32:       # %bb.0:
+; AVX512DQ_32-NEXT:    pushl %ebp
+; AVX512DQ_32-NEXT:    movl %esp, %ebp
+; AVX512DQ_32-NEXT:    andl $-8, %esp
+; AVX512DQ_32-NEXT:    subl $8, %esp
+; AVX512DQ_32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512DQ_32-NEXT:    vcvtqq2pd %zmm0, %zmm0
+; AVX512DQ_32-NEXT:    vmovlps %xmm0, (%esp)
+; AVX512DQ_32-NEXT:    fldl (%esp)
+; AVX512DQ_32-NEXT:    movl %ebp, %esp
+; AVX512DQ_32-NEXT:    popl %ebp
+; AVX512DQ_32-NEXT:    vzeroupper
+; AVX512DQ_32-NEXT:    retl
 ;
 ; AVX512_64-LABEL: s64_to_d:
 ; AVX512_64:       # %bb.0:
 ; AVX512_64-NEXT:    vcvtsi2sdq %rdi, %xmm0, %xmm0
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32-LABEL: s64_to_d:
+; AVX512F_32:       # %bb.0:
+; AVX512F_32-NEXT:    pushl %ebp
+; AVX512F_32-NEXT:    movl %esp, %ebp
+; AVX512F_32-NEXT:    andl $-8, %esp
+; AVX512F_32-NEXT:    subl $8, %esp
+; AVX512F_32-NEXT:    fildll 8(%ebp)
+; AVX512F_32-NEXT:    fstpl (%esp)
+; AVX512F_32-NEXT:    fldl (%esp)
+; AVX512F_32-NEXT:    movl %ebp, %esp
+; AVX512F_32-NEXT:    popl %ebp
+; AVX512F_32-NEXT:    retl
+;
 ; SSE2_32-LABEL: s64_to_d:
 ; SSE2_32:       # %bb.0:
 ; SSE2_32-NEXT:    pushl %ebp
@@ -615,25 +686,28 @@
 }
 
 define double @s64_to_d_2(i64 %a) nounwind {
-; AVX512_32-LABEL: s64_to_d_2:
-; AVX512_32:       # %bb.0:
-; AVX512_32-NEXT:    pushl %ebp
-; AVX512_32-NEXT:    movl %esp, %ebp
-; AVX512_32-NEXT:    andl $-8, %esp
-; AVX512_32-NEXT:    subl $16, %esp
-; AVX512_32-NEXT:    movl 8(%ebp), %eax
-; AVX512_32-NEXT:    movl 12(%ebp), %ecx
-; AVX512_32-NEXT:    addl $5, %eax
-; AVX512_32-NEXT:    adcl $0, %ecx
-; AVX512_32-NEXT:    vmovd %eax, %xmm0
-; AVX512_32-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
-; AVX512_32-NEXT:    vmovq %xmm0, {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    fildll {{[0-9]+}}(%esp)
-; AVX512_32-NEXT:    fstpl (%esp)
-; AVX512_32-NEXT:    fldl (%esp)
-; AVX512_32-NEXT:    movl %ebp, %esp
-; AVX512_32-NEXT:    popl %ebp
-; AVX512_32-NEXT:    retl
+; AVX512DQ_32-LABEL: s64_to_d_2:
+; AVX512DQ_32:       # %bb.0:
+; AVX512DQ_32-NEXT:    pushl %ebp
+; AVX512DQ_32-NEXT:    movl %esp, %ebp
+; AVX512DQ_32-NEXT:    andl $-8, %esp
+; AVX512DQ_32-NEXT:    subl $8, %esp
+; AVX512DQ_32-NEXT:    movl 8(%ebp), %eax
+; AVX512DQ_32-NEXT:    movl 12(%ebp), %ecx
+; AVX512DQ_32-NEXT:    addl $5, %eax
+; AVX512DQ_32-NEXT:    adcl $0, %ecx
+; AVX512DQ_32-NEXT:    vmovd %ecx, %xmm0
+; AVX512DQ_32-NEXT:    vmovd %eax, %xmm1
+; AVX512DQ_32-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX512DQ_32-NEXT:    vmovq {{.*#+}} xmm0 = xmm0[0],zero
+; AVX512DQ_32-NEXT:    vmovdqa %xmm0, %xmm0
+; AVX512DQ_32-NEXT:    vcvtqq2pd %zmm0, %zmm0
+; AVX512DQ_32-NEXT:    vmovlps %xmm0, (%esp)
+; AVX512DQ_32-NEXT:    fldl (%esp)
+; AVX512DQ_32-NEXT:    movl %ebp, %esp
+; AVX512DQ_32-NEXT:    popl %ebp
+; AVX512DQ_32-NEXT:    vzeroupper
+; AVX512DQ_32-NEXT:    retl
 ;
 ; AVX512_64-LABEL: s64_to_d_2:
 ; AVX512_64:       # %bb.0:
@@ -641,6 +715,26 @@
 ; AVX512_64-NEXT:    vcvtsi2sdq %rdi, %xmm0, %xmm0
 ; AVX512_64-NEXT:    retq
 ;
+; AVX512F_32-LABEL: s64_to_d_2:
+; AVX512F_32:       # %bb.0:
+; AVX512F_32-NEXT:    pushl %ebp
+; AVX512F_32-NEXT:    movl %esp, %ebp
+; AVX512F_32-NEXT:    andl $-8, %esp
+; AVX512F_32-NEXT:    subl $16, %esp
+; AVX512F_32-NEXT:    movl 8(%ebp), %eax
+; AVX512F_32-NEXT:    movl 12(%ebp), %ecx
+; AVX512F_32-NEXT:    addl $5, %eax
+; AVX512F_32-NEXT:    adcl $0, %ecx
+; AVX512F_32-NEXT:    vmovd %eax, %xmm0
+; AVX512F_32-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
+; AVX512F_32-NEXT:    vmovq %xmm0, {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX512F_32-NEXT:    fstpl (%esp)
+; AVX512F_32-NEXT:    fldl (%esp)
+; AVX512F_32-NEXT:    movl %ebp, %esp
+; AVX512F_32-NEXT:    popl %ebp
+; AVX512F_32-NEXT:    retl
+;
 ; SSE2_32-LABEL: s64_to_d_2:
 ; SSE2_32:       # %bb.0:
 ; SSE2_32-NEXT:    pushl %ebp