[X86] Make a bunch of merge masked binops commutable for loading folding.

This primarily affects add/fadd/mul/fmul/and/or/xor/pmuludq/pmuldq/max/min/fmaxc/fminc/pmaddwd/pavg.

We already commuted the unmasked and zero masked versions.

I've added 512-bit stack folding tests for most of the instructions
affected. I've tested needing commuting and not commuting across
unmasked, merged masked, and zero masked. The 128/256 bit instructions
should behave similarly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@362746 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86InstrAVX512.td b/lib/Target/X86/X86InstrAVX512.td
index 8987b6a33bf8b42a1097a4e730062f57b32f95bc..b2b4ea29f11c363563b446bd8397206dfb8b50a7 100644 (file)
--- a/lib/Target/X86/X86InstrAVX512.td
+++ b/lib/Target/X86/X86InstrAVX512.td
@@ -277,10 +277,9 @@ multiclass AVX512_maskable<bits<8> O, Format F, X86VectorVTInfo _,
 multiclass AVX512_maskable_scalar<bits<8> O, Format F, X86VectorVTInfo _,
                            dag Outs, dag Ins, string OpcodeStr,
                            string AttSrcAsm, string IntelSrcAsm,
-                           dag RHS,
-                           bit IsCommutable = 0> :
+                           dag RHS> :
    AVX512_maskable<O, F, _, Outs, Ins, OpcodeStr, AttSrcAsm, IntelSrcAsm,
-                   RHS, IsCommutable, 0, IsCommutable, X86selects>;
+                   RHS, 0, 0, 0, X86selects>;
 
 // Similar to AVX512_maskable but in this case one of the source operands
 // ($src1) is already tied to $dst so we just use that for the preserved
@@ -4602,7 +4601,7 @@ multiclass avx512_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
                     (ins _.RC:$src1, _.RC:$src2), OpcodeStr,
                     "$src2, $src1", "$src1, $src2",
                     (_.VT (OpNode _.RC:$src1, _.RC:$src2)),
-                    IsCommutable>, AVX512BIBase, EVEX_4V,
+                    IsCommutable, IsCommutable>, AVX512BIBase, EVEX_4V,
                     Sched<[sched]>;
 
   defm rm : AVX512_maskable<opc, MRMSrcMem, _, (outs _.RC:$dst),
@@ -4838,7 +4837,7 @@ multiclass avx512_packs_rm<bits<8> opc, string OpcodeStr,
                             (_Dst.VT (OpNode
                                          (_Src.VT _Src.RC:$src1),
                                          (_Src.VT _Src.RC:$src2))),
-                            IsCommutable>,
+                            IsCommutable, IsCommutable>,
                             EVEX_CD8<_Src.EltSize, CD8VF>, EVEX_4V, Sched<[sched]>;
   defm rm : AVX512_maskable<opc, MRMSrcMem, _Dst, (outs _Dst.RC:$dst),
                         (ins _Src.RC:$src1, _Src.MemOp:$src2), OpcodeStr,
@@ -5531,13 +5530,13 @@ defm VMAXCSDZ : avx512_comutable_binop_s<0x5F, "vmaxsd", f64x_info, X86fmaxc,
 multiclass avx512_fp_packed<bits<8> opc, string OpcodeStr, SDPatternOperator OpNode,
                             X86VectorVTInfo _, X86FoldableSchedWrite sched,
                             bit IsCommutable,
-                            bit IsKZCommutable = IsCommutable> {
+                            bit IsKCommutable = IsCommutable> {
   let ExeDomain = _.ExeDomain, hasSideEffects = 0 in {
   defm rr: AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
                   (ins _.RC:$src1, _.RC:$src2), OpcodeStr##_.Suffix,
                   "$src2, $src1", "$src1, $src2",
-                  (_.VT (OpNode _.RC:$src1, _.RC:$src2)), IsCommutable, 0,
-                  IsKZCommutable>,
+                  (_.VT (OpNode _.RC:$src1, _.RC:$src2)), IsCommutable,
+                  IsKCommutable, IsKCommutable>,
                   EVEX_4V, Sched<[sched]>;
   let mayLoad = 1 in {
     defm rm: AVX512_maskable<opc, MRMSrcMem, _, (outs _.RC:$dst),

diff --git a/test/CodeGen/X86/stack-folding-fp-avx512.ll b/test/CodeGen/X86/stack-folding-fp-avx512.ll
index def0fe943efa77954f6fb0ad0c036162df89aebe..6e926e69849be0d02918891ca3b6d13c0c16cc4a 100644 (file)
--- a/test/CodeGen/X86/stack-folding-fp-avx512.ll
+++ b/test/CodeGen/X86/stack-folding-fp-avx512.ll
@@ -16,6 +16,28 @@ define <8 x double> @stack_fold_addpd_zmm(<8 x double> %a0, <8 x double> %a1) {
   ret <8 x double> %2
 }
 
+define <8 x double> @stack_fold_addpd_zmm_k(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) {
+  ;CHECK-LABEL: stack_fold_addpd_zmm_k:
+  ;CHECK:       vaddpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fadd <8 x double> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
+define <8 x double> @stack_fold_addpd_zmm_k_commuted(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) {
+  ;CHECK-LABEL: stack_fold_addpd_zmm_k_commuted:
+  ;CHECK:       vaddpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fadd <8 x double> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
 define <8 x double> @stack_fold_addpd_zmm_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
   ;CHECK-LABEL: stack_fold_addpd_zmm_kz
   ;CHECK:       vaddpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -34,6 +56,28 @@ define <16 x float> @stack_fold_addps_zmm(<16 x float> %a0, <16 x float> %a1) {
   ret <16 x float> %2
 }
 
+define <16 x float> @stack_fold_addps_zmm_k(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) {
+  ;CHECK-LABEL: stack_fold_addps_zmm_k:
+  ;CHECK:       vaddps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fadd <16 x float> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
+define <16 x float> @stack_fold_addps_zmm_k_commuted(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) {
+  ;CHECK-LABEL: stack_fold_addps_zmm_k_commuted:
+  ;CHECK:       vaddps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fadd <16 x float> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
 define <16 x float> @stack_fold_addps_zmm_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
   ;CHECK-LABEL: stack_fold_addps_zmm_kz
   ;CHECK:       vaddps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -311,6 +355,28 @@ define <8 x double> @stack_fold_maxpd_zmm_commutable(<8 x double> %a0, <8 x doub
   ret <8 x double> %2
 }
 
+define <8 x double> @stack_fold_maxpd_zmm_commutable_k(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_maxpd_zmm_commutable_k:
+  ;CHECK:       vmaxpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x double> @llvm.x86.avx512.max.pd.512(<8 x double> %a0, <8 x double> %a1, i32 4)
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
+define <8 x double> @stack_fold_maxpd_zmm_commutable_k_commuted(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_maxpd_zmm_commutable_k_commuted:
+  ;CHECK:       vmaxpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x double> @llvm.x86.avx512.max.pd.512(<8 x double> %a1, <8 x double> %a0, i32 4)
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
 define <8 x double> @stack_fold_maxpd_zmm_commutable_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) #1 {
   ;CHECK-LABEL: stack_fold_maxpd_zmm_commutable_kz
   ;CHECK:       vmaxpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -338,6 +404,28 @@ define <16 x float> @stack_fold_maxps_zmm_commutable(<16 x float> %a0, <16 x flo
   ret <16 x float> %2
 }
 
+define <16 x float> @stack_fold_maxps_zmm_commutable_k(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_maxps_zmm_commutable_k:
+  ;CHECK:       vmaxps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
+define <16 x float> @stack_fold_maxps_zmm_commutable_k_commuted(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_maxps_zmm_commutable_k_commuted:
+  ;CHECK:       vmaxps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a1, <16 x float> %a0, i32 4)
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
 define <16 x float> @stack_fold_maxps_zmm_commutable_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) #1 {
   ;CHECK-LABEL: stack_fold_maxps_zmm_commutable_kz
   ;CHECK:       vmaxps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -365,6 +453,28 @@ define <8 x double> @stack_fold_minpd_zmm_commutable(<8 x double> %a0, <8 x doub
   ret <8 x double> %2
 }
 
+define <8 x double> @stack_fold_minpd_zmm_commutable_k(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_minpd_zmm_commutable_k:
+  ;CHECK:       vminpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x double> @llvm.x86.avx512.min.pd.512(<8 x double> %a0, <8 x double> %a1, i32 4)
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
+define <8 x double> @stack_fold_minpd_zmm_commutable_k_commuted(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_minpd_zmm_commutable_k_commuted:
+  ;CHECK:       vminpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x double> @llvm.x86.avx512.min.pd.512(<8 x double> %a1, <8 x double> %a0, i32 4)
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
 define <8 x double> @stack_fold_minpd_zmm_commutable_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) #1 {
   ;CHECK-LABEL: stack_fold_minpd_zmm_commutable_kz
   ;CHECK:       vminpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -392,6 +502,28 @@ define <16 x float> @stack_fold_minps_zmm_commutable(<16 x float> %a0, <16 x flo
   ret <16 x float> %2
 }
 
+define <16 x float> @stack_fold_minps_zmm_commutable_k(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_minps_zmm_commutable_k:
+  ;CHECK:       vminps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
+define <16 x float> @stack_fold_minps_zmm_commutable_k_commuted(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) #1 {
+  ;CHECK-LABEL: stack_fold_minps_zmm_commutable_k_commuted:
+  ;CHECK:       vminps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a1, <16 x float> %a0, i32 4)
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
 define <16 x float> @stack_fold_minps_zmm_commutable_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) #1 {
   ;CHECK-LABEL: stack_fold_minps_zmm_commutable_kz
   ;CHECK:       vminps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -402,6 +534,86 @@ define <16 x float> @stack_fold_minps_zmm_commutable_kz(<16 x float> %a0, <16 x
   ret <16 x float> %4
 }
 
+define <8 x double> @stack_fold_mulpd_zmm(<8 x double> %a0, <8 x double> %a1) {
+  ;CHECK-LABEL: stack_fold_mulpd_zmm
+  ;CHECK:       vmulpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <8 x double> %a0, %a1
+  ret <8 x double> %2
+}
+
+define <8 x double> @stack_fold_mulpd_zmm_k(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) {
+  ;CHECK-LABEL: stack_fold_mulpd_zmm_k:
+  ;CHECK:       vmulpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <8 x double> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
+define <8 x double> @stack_fold_mulpd_zmm_k_commuted(<8 x double> %a0, <8 x double> %a1, i8 %mask, <8 x double>* %passthru) {
+  ;CHECK-LABEL: stack_fold_mulpd_zmm_k_commuted:
+  ;CHECK:       vmulpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <8 x double> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = load <8 x double>, <8 x double>* %passthru
+  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
+  ret <8 x double> %5
+}
+
+define <8 x double> @stack_fold_mulpd_zmm_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_mulpd_zmm_kz
+  ;CHECK:       vmulpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <8 x double> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> zeroinitializer
+  ret <8 x double> %4
+}
+
+define <16 x float> @stack_fold_mulps_zmm(<16 x float> %a0, <16 x float> %a1) {
+  ;CHECK-LABEL: stack_fold_mulps_zmm
+  ;CHECK:       vmulps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <16 x float> %a0, %a1
+  ret <16 x float> %2
+}
+
+define <16 x float> @stack_fold_mulps_zmm_k(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) {
+  ;CHECK-LABEL: stack_fold_mulps_zmm_k:
+  ;CHECK:       vmulps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <16 x float> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
+define <16 x float> @stack_fold_mulps_zmm_k_commuted(<16 x float> %a0, <16 x float> %a1, i16 %mask, <16 x float>* %passthru) {
+  ;CHECK-LABEL: stack_fold_mulps_zmm_k_commuted:
+  ;CHECK:       vmulps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <16 x float> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = load <16 x float>, <16 x float>* %passthru
+  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
+  ret <16 x float> %5
+}
+
+define <16 x float> @stack_fold_mulps_zmm_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_mulps_zmm_kz
+  ;CHECK:       vmulps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = fmul <16 x float> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x float> %2, <16 x float> zeroinitializer
+  ret <16 x float> %4
+}
+
 define double @stack_fold_mulsd(double %a0, double %a1) {
   ;CHECK-LABEL: stack_fold_mulsd
   ;CHECK:       vmulsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 8-byte Folded Reload

diff --git a/test/CodeGen/X86/stack-folding-int-avx512.ll b/test/CodeGen/X86/stack-folding-int-avx512.ll
index a899475c9d80fe7b2f67054f48541759d175493d..f8a2973f9aacc27604dba005334a40ac57ca5755 100644 (file)
--- a/test/CodeGen/X86/stack-folding-int-avx512.ll
+++ b/test/CodeGen/X86/stack-folding-int-avx512.ll
@@ -70,88 +70,122 @@ define <64 x i8> @stack_fold_pavgb(<64 x i8> %a0, <64 x i8> %a1) {
   ;CHECK-LABEL: stack_fold_pavgb
   ;CHECK:       vpavgb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <64 x i8> %a0 to <64 x i16>
-  %3 = zext <64 x i8> %a1 to <64 x i16>
-  %4 = add <64 x i16> %2, %3
-  %5 = add <64 x i16> %4, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %6 = lshr <64 x i16> %5, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %7 = trunc <64 x i16> %6 to <64 x i8>
-  ret <64 x i8> %7
+  %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a0, <64 x i8> %a1)
+  ret <64 x i8> %2
+}
+declare <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8>, <64 x i8>)
+
+define <64 x i8> @stack_fold_pavgb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pavgb_commuted:
+  ;CHECK:       vpavgb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a1, <64 x i8> %a0)
+  ret <64 x i8> %2
 }
 
-define <64 x i8> @stack_fold_pavgb_mask(<64 x i8>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+define <64 x i8> @stack_fold_pavgb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
   ;CHECK-LABEL: stack_fold_pavgb_mask
   ;CHECK:       vpavgb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = load <64 x i8>, <64 x i8>* %passthru
-  %3 = zext <64 x i8> %a0 to <64 x i16>
-  %4 = zext <64 x i8> %a1 to <64 x i16>
-  %5 = add <64 x i16> %3, %4
-  %6 = add <64 x i16> %5, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %7 = lshr <64 x i16> %6, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %8 = trunc <64 x i16> %7 to <64 x i8>
-  %9 = bitcast i64 %mask to <64 x i1>
-  %10 = select <64 x i1> %9, <64 x i8> %8, <64 x i8> %2
-  ret <64 x i8> %10
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_pavgb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pavgb_mask_commuted:
+  ;CHECK:       vpavgb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
 }
 
 define <64 x i8> @stack_fold_pavgb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
   ;CHECK-LABEL: stack_fold_pavgb_maskz
   ;CHECK:       vpavgb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <64 x i8> %a0 to <64 x i16>
-  %3 = zext <64 x i8> %a1 to <64 x i16>
-  %4 = add <64 x i16> %2, %3
-  %5 = add <64 x i16> %4, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %6 = lshr <64 x i16> %5, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %7 = trunc <64 x i16> %6 to <64 x i8>
-  %8 = bitcast i64 %mask to <64 x i1>
-  %9 = select <64 x i1> %8, <64 x i8> %7, <64 x i8> zeroinitializer
-  ret <64 x i8> %9
+  %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @stack_fold_pavgb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pavgb_maskz_commuted:
+  ;CHECK:       vpavgb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
 }
 
 define <32 x i16> @stack_fold_pavgw(<32 x i16> %a0, <32 x i16> %a1) {
   ;CHECK-LABEL: stack_fold_pavgw
   ;CHECK:       vpavgw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <32 x i16> %a0 to <32 x i32>
-  %3 = zext <32 x i16> %a1 to <32 x i32>
-  %4 = add <32 x i32> %2, %3
-  %5 = add <32 x i32> %4, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %6 = lshr <32 x i32> %5, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %7 = trunc <32 x i32> %6 to <32 x i16>
-  ret <32 x i16> %7
+  %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a0, <32 x i16> %a1)
+  ret <32 x i16> %2
+}
+declare <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16>, <32 x i16>)
+
+define <32 x i16> @stack_fold_pavgw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pavgw_commuted:
+  ;CHECK:       vpavgw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a1, <32 x i16> %a0)
+  ret <32 x i16> %2
 }
 
-define <32 x i16> @stack_fold_pavgw_mask(<32 x i16>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+define <32 x i16> @stack_fold_pavgw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
   ;CHECK-LABEL: stack_fold_pavgw_mask
   ;CHECK:       vpavgw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = load <32 x i16>, <32 x i16>* %passthru
-  %3 = zext <32 x i16> %a0 to <32 x i32>
-  %4 = zext <32 x i16> %a1 to <32 x i32>
-  %5 = add <32 x i32> %3, %4
-  %6 = add <32 x i32> %5, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %7 = lshr <32 x i32> %6, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %8 = trunc <32 x i32> %7 to <32 x i16>
-  %9 = bitcast i32 %mask to <32 x i1>
-  %10 = select <32 x i1> %9, <32 x i16> %8, <32 x i16> %2
-  ret <32 x i16> %10
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pavgw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pavgw_mask_commuted:
+  ;CHECK:       vpavgw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
 }
 
 define <32 x i16> @stack_fold_pavgw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
   ;CHECK-LABEL: stack_fold_pavgw_maskz
   ;CHECK:       vpavgw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <32 x i16> %a0 to <32 x i32>
-  %3 = zext <32 x i16> %a1 to <32 x i32>
-  %4 = add <32 x i32> %2, %3
-  %5 = add <32 x i32> %4, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %6 = lshr <32 x i32> %5, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %7 = trunc <32 x i32> %6 to <32 x i16>
-  %8 = bitcast i32 %mask to <32 x i1>
-  %9 = select <32 x i1> %8, <32 x i16> %7, <32 x i16> zeroinitializer
-  ret <32 x i16> %9
+  %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @stack_fold_pavgw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pavgw_maskz_commuted:
+  ;CHECK:       vpavgw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
 }
 
 define <4 x i32> @stack_fold_extracti32x4(<16 x i16> %a0, <16 x i32> %a1) {
@@ -415,6 +449,14 @@ define <64 x i8> @stack_fold_paddb(<64 x i8> %a0, <64 x i8> %a1) {
   ret <64 x i8> %2
 }
 
+define <64 x i8> @stack_fold_paddb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_paddb_commuted:
+  ;CHECK:       vpaddb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <64 x i8> %a1, %a0
+  ret <64 x i8> %2
+}
+
 define <64 x i8> @stack_fold_paddb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
   ;CHECK-LABEL: stack_fold_paddb_mask
   ;CHECK:       vpaddb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
@@ -427,6 +469,18 @@ define <64 x i8> @stack_fold_paddb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>*
   ret <64 x i8> %5
 }
 
+define <64 x i8> @stack_fold_paddb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddb_mask_commuted:
+  ;CHECK:       vpaddb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <64 x i8> %a1, %a0
+  %3 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
+}
+
 define <64 x i8> @stack_fold_paddb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
   ;CHECK-LABEL: stack_fold_paddb_maskz
   ;CHECK:       vpaddb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
@@ -437,6 +491,16 @@ define <64 x i8> @stack_fold_paddb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask
   ret <64 x i8> %4
 }
 
+define <64 x i8> @stack_fold_paddb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddb_maskz_commuted:
+  ;CHECK:       vpaddb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <64 x i8> %a1, %a0
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
+}
+
 define <16 x i32> @stack_fold_paddd(<16 x i32> %a0, <16 x i32> %a1) {
   ;CHECK-LABEL: stack_fold_paddd
   ;CHECK:       vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
@@ -445,6 +509,58 @@ define <16 x i32> @stack_fold_paddd(<16 x i32> %a0, <16 x i32> %a1) {
   ret <16 x i32> %2
 }
 
+define <16 x i32> @stack_fold_paddd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_paddd_commuted:
+  ;CHECK:       vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <16 x i32> %a1, %a0
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_paddd_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_paddd_mask
+  ;CHECK:       vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_paddd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_paddd_mask_commuted:
+  ;CHECK:       vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_paddd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_paddd_maskz
+  ;CHECK:       vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @stack_fold_paddd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_paddd_maskz_commuted:
+  ;CHECK:       vpaddd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
 define <8 x i64> @stack_fold_paddq(<8 x i64> %a0, <8 x i64> %a1) {
   ;CHECK-LABEL: stack_fold_paddq
   ;CHECK:       vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
@@ -453,514 +569,2030 @@ define <8 x i64> @stack_fold_paddq(<8 x i64> %a0, <8 x i64> %a1) {
   ret <8 x i64> %2
 }
 
-define <64 x i8> @stack_fold_paddsb(<64 x i8> %a0, <64 x i8> %a1) {
-  ;CHECK-LABEL: stack_fold_paddsb
-  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_paddq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_paddq_commuted:
+  ;CHECK:       vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
-  ret <64 x i8> %2
+  %2 = add <8 x i64> %a1, %a0
+  ret <8 x i64> %2
 }
 
-define <32 x i16> @stack_fold_paddsw(<32 x i16> %a0, <32 x i16> %a1) {
-  ;CHECK-LABEL: stack_fold_paddsw
-  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
-  ret <32 x i16> %2
+define <8 x i64> @stack_fold_paddq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_paddq_mask
+  ;CHECK:       vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
 }
 
-define <64 x i8> @stack_fold_paddusb(<64 x i8> %a0, <64 x i8> %a1) {
-  ;CHECK-LABEL: stack_fold_paddusb
-  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_paddq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_paddq_mask_commuted:
+  ;CHECK:       vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_paddq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_paddq_maskz
+  ;CHECK:       vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
-  ret <64 x i8> %2
+  %2 = add <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
 }
-declare <64 x i8> @llvm.x86.avx512.mask.paddus.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64) nounwind readnone
 
-define <32 x i16> @stack_fold_paddusw(<32 x i16> %a0, <32 x i16> %a1) {
-  ;CHECK-LABEL: stack_fold_paddusw
-  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_paddq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_paddq_maskz_commuted:
+  ;CHECK:       vpaddq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
-  ret <32 x i16> %2
+  %2 = add <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
 }
-declare <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
-define <32 x i16> @stack_fold_paddw(<32 x i16> %a0, <32 x i16> %a1) {
-  ;CHECK-LABEL: stack_fold_paddw
-  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <64 x i8> @stack_fold_paddsb(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_paddsb
+  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = add <32 x i16> %a0, %a1
-  ret <32 x i16> %2
+  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
+  ret <64 x i8> %2
 }
 
-define <64 x i8> @stack_fold_palignr(<64 x i8> %a0, <64 x i8> %a1) {
-  ;CHECK-LABEL: stack_fold_palignr
-  ;CHECK:       vpalignr $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
+define <64 x i8> @stack_fold_paddsb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_paddsb_commuted:
+  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   ret <64 x i8> %2
 }
 
-define <64 x i8> @stack_fold_palignr_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %passthru, i64 %mask) {
-  ;CHECK-LABEL: stack_fold_palignr_mask
-  ;CHECK:       vpalignr $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
+define <64 x i8> @stack_fold_paddsb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsb_mask
+  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
-  %4 = load <64 x i8>, <64 x i8>* %passthru
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }
 
-define <64 x i8> @stack_fold_palignr_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
-  ;CHECK-LABEL: stack_fold_palignr_maskz
-  ;CHECK:       vpalignr $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
+define <64 x i8> @stack_fold_paddsb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsb_mask_commuted:
+  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
-  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
-  ret <64 x i8> %4
-}
-
-define <16 x i32> @stack_fold_vpconflictd(<16 x i32> %a0) {
-  ;CHECK-LABEL: stack_fold_vpconflictd
-  ;CHECK:       vpconflictd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.conflict.d.512(<16 x i32> %a0)
-  ret <16 x i32> %2
-}
-declare <16 x i32> @llvm.x86.avx512.mask.conflict.d.512(<16 x i32>, <16 x i32>, i16) nounwind readonly
-
-define <8 x i64> @stack_fold_vpconflictq(<8 x i64> %a0) {
-  ;CHECK-LABEL: stack_fold_vpconflictq
-  ;CHECK:       vpconflictq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x i64> @llvm.x86.avx512.conflict.q.512(<8 x i64> %a0)
-  ret <8 x i64> %2
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
 }
-declare <8 x i64> @llvm.x86.avx512.mask.conflict.q.512(<8 x i64>, <8 x i64>, i8) nounwind readnone
 
-define i64 @stack_fold_pcmpeqb(<64 x i8> %a0, <64 x i8> %a1) {
-  ;CHECK-LABEL: stack_fold_pcmpeqb
-  ;CHECK:       vpcmpeqb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+define <64 x i8> @stack_fold_paddsb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsb_maskz
+  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = icmp eq <64 x i8> %a0, %a1
-  %3 = bitcast <64 x i1> %2 to i64
-  ret i64 %3
+  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
 }
 
-define i16 @stack_fold_pcmpeqd(<16 x i32> %a0, <16 x i32> %a1) {
-  ;CHECK-LABEL: stack_fold_pcmpeqd
-  ;CHECK:       vpcmpeqd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+define <64 x i8> @stack_fold_paddsb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsb_maskz_commuted:
+  ;CHECK:       vpaddsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = icmp eq <16 x i32> %a0, %a1
-  %3 = bitcast <16 x i1> %2 to i16
-  ret i16 %3
+  %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
 }
 
-define i8 @stack_fold_pcmpeqq(<8 x i64> %a0, <8 x i64> %a1) {
-  ;CHECK-LABEL: stack_fold_pcmpeqq
-  ;CHECK:       vpcmpeqq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_paddsw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_paddsw
+  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = icmp eq <8 x i64> %a0, %a1
-  %3 = bitcast <8 x i1> %2 to i8
-  ret i8 %3
+  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
+  ret <32 x i16> %2
 }
 
-define i32 @stack_fold_pcmpeqw(<32 x i16> %a0, <32 x i16> %a1) {
-  ;CHECK-LABEL: stack_fold_pcmpeqw
-  ;CHECK:       vpcmpeqw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_paddsw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_paddsw_commuted:
+  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = icmp eq <32 x i16> %a0, %a1
-  %3 = bitcast <32 x i1> %2 to i32
-  ret i32 %3
+  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_paddsw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsw_mask
+  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_paddsw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsw_mask_commuted:
+  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_paddsw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsw_maskz
+  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @stack_fold_paddsw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddsw_maskz_commuted:
+  ;CHECK:       vpaddsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <64 x i8> @stack_fold_paddusb(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_paddusb
+  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
+  ret <64 x i8> %2
+}
+
+define <64 x i8> @stack_fold_paddusb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_paddusb_commuted:
+  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
+  ret <64 x i8> %2
+}
+
+define <64 x i8> @stack_fold_paddusb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusb_mask
+  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_paddusb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusb_mask_commuted:
+  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <64 x i8>, <64 x i8>* %a2
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_paddusb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusb_maskz
+  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @stack_fold_paddusb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusb_maskz_commuted:
+  ;CHECK:       vpaddusb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
+}
+
+define <32 x i16> @stack_fold_paddusw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_paddusw
+  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_paddusw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_paddusw_commuted:
+  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_paddusw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusw_mask
+  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_paddusw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusw_mask_commuted:
+  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_paddusw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusw_maskz
+  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @stack_fold_paddusw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddusw_maskz_commuted:
+  ;CHECK:       vpaddusw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @stack_fold_paddw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_paddw
+  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <32 x i16> %a0, %a1
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_paddw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_paddw_commuted:
+  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <32 x i16> %a1, %a0
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_paddw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddw_mask
+  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <32 x i16> %a0, %a1
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_paddw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddw_mask_commuted:
+  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <32 x i16> %a1, %a0
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_paddw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddw_maskz
+  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <32 x i16> %a0, %a1
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @stack_fold_paddw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_paddw_maskz_commuted:
+  ;CHECK:       vpaddw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = add <32 x i16> %a1, %a0
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <64 x i8> @stack_fold_palignr(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_palignr
+  ;CHECK:       vpalignr $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
+  ret <64 x i8> %2
+}
+
+define <64 x i8> @stack_fold_palignr_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %passthru, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_palignr_mask
+  ;CHECK:       vpalignr $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = load <64 x i8>, <64 x i8>* %passthru
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_palignr_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_palignr_maskz
+  ;CHECK:       vpalignr $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
+}
+
+define <16 x i32> @stack_fold_pandd(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pandd
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <16 x i32> %a0, %a1
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pandd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pandd_commuted:
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <16 x i32> %a1, %a0
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pandd_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pandd_mask
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pandd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pandd_mask_commuted:
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pandd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pandd_maskz
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @stack_fold_pandd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pandd_maskz_commuted:
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <8 x i64> @stack_fold_pandq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pandq
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, %a1
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pandq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pandq_commuted:
+  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a1, %a0
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pandq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pandq_mask
+  ;CHECK:       vandpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pandq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pandq_mask_commuted:
+  ;CHECK:       vandpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pandq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pandq_maskz
+  ;CHECK:       vandpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @stack_fold_pandq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pandq_maskz_commuted:
+  ;CHECK:       vandpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <16 x i32> @stack_fold_vpconflictd(<16 x i32> %a0) {
+  ;CHECK-LABEL: stack_fold_vpconflictd
+  ;CHECK:       vpconflictd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.conflict.d.512(<16 x i32> %a0)
+  ret <16 x i32> %2
+}
+declare <16 x i32> @llvm.x86.avx512.mask.conflict.d.512(<16 x i32>, <16 x i32>, i16) nounwind readonly
+
+define <8 x i64> @stack_fold_vpconflictq(<8 x i64> %a0) {
+  ;CHECK-LABEL: stack_fold_vpconflictq
+  ;CHECK:       vpconflictq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x i64> @llvm.x86.avx512.conflict.q.512(<8 x i64> %a0)
+  ret <8 x i64> %2
+}
+declare <8 x i64> @llvm.x86.avx512.mask.conflict.q.512(<8 x i64>, <8 x i64>, i8) nounwind readnone
+
+define i64 @stack_fold_pcmpeqb(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pcmpeqb
+  ;CHECK:       vpcmpeqb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp eq <64 x i8> %a0, %a1
+  %3 = bitcast <64 x i1> %2 to i64
+  ret i64 %3
+}
+
+define i16 @stack_fold_pcmpeqd(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pcmpeqd
+  ;CHECK:       vpcmpeqd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp eq <16 x i32> %a0, %a1
+  %3 = bitcast <16 x i1> %2 to i16
+  ret i16 %3
+}
+
+define i8 @stack_fold_pcmpeqq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pcmpeqq
+  ;CHECK:       vpcmpeqq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp eq <8 x i64> %a0, %a1
+  %3 = bitcast <8 x i1> %2 to i8
+  ret i8 %3
+}
+
+define i32 @stack_fold_pcmpeqw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pcmpeqw
+  ;CHECK:       vpcmpeqw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp eq <32 x i16> %a0, %a1
+  %3 = bitcast <32 x i1> %2 to i32
+  ret i32 %3
 }
 
 define <16 x i32> @stack_fold_pcmpeqd_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
   ;CHECK-LABEL: stack_fold_pcmpeqd_mask
   ;CHECK:       vpcmpeqd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
-  %2 = load <16 x i32>, <16 x i32>* %a2
-  %3 = add <16 x i32> %a1, %2
+  ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
+  %2 = load <16 x i32>, <16 x i32>* %a2
+  %3 = add <16 x i32> %a1, %2
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = icmp eq <16 x i32> %3, %a0
+  %6 = and <16 x i1> %4, %5
+  %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
+  ret <16 x i32> %7
+}
+
+define <16 x i32> @stack_fold_pcmpeqd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
+  ;CHECK-LABEL: stack_fold_pcmpeqd_mask_commuted
+  ;CHECK:       vpcmpeqd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
+  %2 = load <16 x i32>, <16 x i32>* %a2
+  %3 = add <16 x i32> %a1, %2
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = icmp eq <16 x i32> %a0, %3
+  %6 = and <16 x i1> %4, %5
+  %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
+  ret <16 x i32> %7
+}
+
+define <16 x i32> @stack_fold_pcmpled_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
+  ;CHECK-LABEL: stack_fold_pcmpled_mask
+  ;CHECK:       vpcmpled {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
+  %2 = load <16 x i32>, <16 x i32>* %a2
+  %3 = add <16 x i32> %a1, %2
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = icmp sge <16 x i32> %a0, %3
+  %6 = and <16 x i1> %4, %5
+  %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
+  ret <16 x i32> %7
+}
+
+define i16 @stack_fold_pcmpleud(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pcmpleud
+  ;CHECK:       vpcmpleud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = load <16 x i32>, <16 x i32>* %a2
+  %3 = add <16 x i32> %a1, %2
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = icmp uge <16 x i32> %a0, %3
+  %6 = and <16 x i1> %5, %4
+  %7 = bitcast <16 x i1> %6 to i16
+  ret i16 %7
+}
+
+define <64 x i8> @stack_fold_permbvar(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_permbvar
+  ;CHECK:   vpermb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
+  ret <64 x i8> %2
+}
+declare <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8>, <64 x i8>) nounwind readonly
+
+define <64 x i8> @stack_fold_permbvar_mask(<64 x i8>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_permbvar_mask
+  ;CHECK:   vpermb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled above the asm block
+  %4 = load <64 x i8>, <64 x i8>* %passthru
+  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_permbvar_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_permbvar_maskz
+  ;CHECK:   vpermb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
+  %3 = bitcast i64 %mask to <64 x i1>
+  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
+  ret <64 x i8> %4
+}
+
+define <16 x i32> @stack_fold_permd(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_permd
+  ;CHECK:   vpermd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32> %a1, <16 x i32> %a0)
+  ; add forces execution domain
+  %3 = add <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  ret <16 x i32> %3
+}
+declare <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32>, <16 x i32>) nounwind readonly
+
+define <64 x i8> @stack_fold_vpermi2b(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermi2b
+  ;CHECK:       vpermi2b {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %x1, <64 x i8> %x0, <64 x i8> %x2)
+  ret <64 x i8> %2
+}
+
+define <16 x i32> @stack_fold_vpermi2d(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermi2d
+  ;CHECK:       vpermi2d {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
+  ret <16 x i32> %2
+}
+
+define <8 x i64> @stack_fold_vpermi2q(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermi2q
+  ;CHECK:       vpermi2q {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x2)
+  ret <8 x i64> %2
+}
+
+define <32 x i16> @stack_fold_vpermi2w(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermi2w
+  ;CHECK:       vpermi2w {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16> %x1, <32 x i16> %x0, <32 x i16> %x2)
+  ret <32 x i16> %2
+}
+
+define <8 x i64> @stack_fold_permq(<8 x i64> %a0) {
+  ;CHECK-LABEL: stack_fold_permq
+  ;CHECK:   vpermq $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
+  ; add forces execution domain
+  %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  ret <8 x i64> %3
+}
+
+define <8 x i64> @stack_fold_permq_mask(<8 x i64>* %passthru, <8 x i64> %a0, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_permq_mask
+  ;CHECK:   vpermq $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled above the asm block
+  %4 = load <8 x i64>, <8 x i64>* %passthru
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ; add forces execution domain
+  %6 = add <8 x i64> %5, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_permq_maskz(<8 x i64>* %passthru, <8 x i64> %a0, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_permq_maskz
+  ;CHECK:   vpermq $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @stack_fold_permqvar(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_permqvar
+  ;CHECK:   vpermq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %a1, <8 x i64> %a0)
+  ; add forces execution domain
+  %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  ret <8 x i64> %3
+}
+declare <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64>, <8 x i64>) nounwind readonly
+
+define <8 x i64> @stack_fold_permqvar_mask(<8 x i64>* %passthru, <8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_permqvar_mask
+  ;CHECK:   vpermq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %a1, <8 x i64> %a0)
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled above the asm block
+  %4 = load <8 x i64>, <8 x i64>* %passthru
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ; add forces execution domain
+  %6 = add <8 x i64> %5, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  ret <8 x i64> %6
+}
+
+define <64 x i8> @stack_fold_vpermt2b(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermt2b
+  ;CHECK:       vpermt2b {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2)
+  ret <64 x i8> %2
+}
+declare <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8>, <64 x i8>, <64 x i8>)
+
+define <16 x i32> @stack_fold_vpermt2d(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermt2d
+  ;CHECK:       vpermt2d {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
+  ret <16 x i32> %2
+}
+declare <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32>, <16 x i32>, <16 x i32>)
+
+define <8 x i64> @stack_fold_vpermt2q(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermt2q
+  ;CHECK:       vpermt2q {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2)
+  ret <8 x i64> %2
+}
+declare <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64>, <8 x i64>, <8 x i64>)
+
+define <32 x i16> @stack_fold_vpermt2w(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
+  ;CHECK-LABEL: stack_fold_vpermt2w
+  ;CHECK:       vpermt2w {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2)
+  ret <32 x i16> %2
+}
+declare <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16>, <32 x i16>, <32 x i16>)
+
+define <32 x i16> @stack_fold_permwvar(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_permwvar
+  ;CHECK:   vpermw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
+  ret <32 x i16> %2
+}
+declare <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16>, <32 x i16>) nounwind readonly
+
+define <32 x i16> @stack_fold_permwvar_mask(<32 x i16>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_permwvar_mask
+  ;CHECK:   vpermw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled above the asm block
+  %4 = load <32 x i16>, <32 x i16>* %passthru
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_permwvar_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_permwvar_maskz
+  ;CHECK:   vpermw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define i32 @stack_fold_pextrd(<4 x i32> %a0, <4 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pextrd
+  ;CHECK:       vpextrd $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
+  ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
+  ; add forces execution domain
+  %1 = add <4 x i32> %a0, %a1
+  %2 = extractelement <4 x i32> %1, i32 1
+  %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i32 %2
+}
+
+define i64 @stack_fold_pextrq(<2 x i64> %a0) {
+  ;CHECK-LABEL: stack_fold_pextrq
+  ;CHECK:       vpextrq $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 8-byte Folded Spill
+  ;CHECK:       movq    {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Reload
+  %1 = extractelement <2 x i64> %a0, i32 1
+  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i64 %1
+}
+
+define <16 x i8> @stack_fold_pinsrb(<16 x i8> %a0, i8 %a1) {
+  ;CHECK-LABEL: stack_fold_pinsrb
+  ;CHECK:       vpinsrb $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  %2 = insertelement <16 x i8> %a0, i8 %a1, i32 1
+  ret <16 x i8> %2
+}
+
+define <4 x i32> @stack_fold_pinsrd(<4 x i32> %a0, i32 %a1) {
+  ;CHECK-LABEL: stack_fold_pinsrd
+  ;CHECK:       vpinsrd $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  %2 = insertelement <4 x i32> %a0, i32 %a1, i32 1
+  ret <4 x i32> %2
+}
+
+define <2 x i64> @stack_fold_pinsrq(<2 x i64> %a0, i64 %a1) {
+  ;CHECK-LABEL: stack_fold_pinsrq
+  ;CHECK:       vpinsrq $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 8-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  %2 = insertelement <2 x i64> %a0, i64 %a1, i32 1
+  ret <2 x i64> %2
+}
+
+define <8 x i16> @stack_fold_pinsrw(<8 x i16> %a0, i16 %a1) {
+  ;CHECK-LABEL: stack_fold_pinsrw
+  ;CHECK:       vpinsrw $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  %2 = insertelement <8 x i16> %a0, i16 %a1, i32 1
+  ret <8 x i16> %2
+}
+
+define <16 x i32> @stack_fold_vplzcntd(<16 x i32> %a0) {
+  ;CHECK-LABEL: stack_fold_vplzcntd
+  ;CHECK:       vplzcntd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %a0, i1 false)
+  ret <16 x i32> %2
+}
+
+define <8 x i64> @stack_fold_vplzcntq(<8 x i64> %a0) {
+  ;CHECK-LABEL: stack_fold_vplzcntq
+  ;CHECK:       vplzcntq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <8 x i64> @llvm.ctlz.v8i64(<8 x i64> %a0, i1 false)
+  ret <8 x i64> %2
+}
+
+define <32 x i16> @stack_fold_pmaddubsw_zmm(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaddubsw_zmm
+  ;CHECK:       vpmaddubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
+  ret <32 x i16> %2
+}
+declare <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8>, <64 x i8>) nounwind readnone
+
+define <32 x i16> @stack_fold_pmaddubsw_zmm_mask(<32 x i16>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaddubsw_zmm_mask
+  ;CHECK:       vpmaddubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %passthru
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pmaddubsw_zmm_maskz(<64 x i8> %a0, <64 x i8> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaddubsw_zmm_maskz
+  ;CHECK:       vpmaddubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
+
+define <16 x i32> @stack_fold_pmaddwd_zmm(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaddwd_zmm
+  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
+  ret <16 x i32> %2
+}
+declare <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16>, <32 x i16>) nounwind readnone
+
+define <16 x i32> @stack_fold_pmaddwd_zmm_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_commuted:
+  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a1, <32 x i16> %a0)
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pmaddwd_zmm_mask(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_mask
+  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %passthru
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pmaddwd_zmm_mask_commuted(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_mask_commuted:
+  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %passthru
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pmaddwd_zmm_maskz(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_maskz
+  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @stack_fold_pmaddwd_zmm_maskz_commuted(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_maskz_commuted:
+  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a1, <32 x i16> %a0)
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <64 x i8> @stack_fold_pmaxsb(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsb
+  ;CHECK:       vpmaxsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  ret <64 x i8> %3
+}
+
+define <64 x i8> @stack_fold_pmaxsb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsb_commuted:
+  ;CHECK:       vpmaxsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  ret <64 x i8> %3
+}
+
+define <64 x i8> @stack_fold_pmaxsb_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsb_mask:
+  ;CHECK:       vpmaxsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
+}
+
+define <64 x i8> @stack_fold_pmaxsb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsb_mask_commuted:
+  ;CHECK:       vpmaxsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
+}
+
+define <64 x i8> @stack_fold_pmaxsb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsb_maskz:
+  ;CHECK:       vpmaxsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_pmaxsb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsb_maskz_commuted:
+  ;CHECK:       vpmaxsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
+}
+
+define <16 x i32> @stack_fold_pmaxsd(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsd
+  ;CHECK:       vpmaxsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  ret <16 x i32> %3
+}
+
+define <16 x i32> @stack_fold_pmaxsd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsd_commuted:
+  ;CHECK:       vpmaxsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  ret <16 x i32> %3
+}
+
+define <16 x i32> @stack_fold_pmaxsd_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsd_mask:
+  ;CHECK:       vpmaxsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  %4 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
+}
+
+define <16 x i32> @stack_fold_pmaxsd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsd_mask_commuted:
+  ;CHECK:       vpmaxsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  %4 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
+}
+
+define <16 x i32> @stack_fold_pmaxsd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsd_maskz:
+  ;CHECK:       vpmaxsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pmaxsd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsd_maskz_commuted:
+  ;CHECK:       vpmaxsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
+}
+
+define <8 x i64> @stack_fold_pmaxsq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsq
+  ;CHECK:       vpmaxsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  ret <8 x i64> %3
+}
+
+define <8 x i64> @stack_fold_pmaxsq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsq_commuted:
+  ;CHECK:       vpmaxsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  ret <8 x i64> %3
+}
+
+define <8 x i64> @stack_fold_pmaxsq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsq_mask:
+  ;CHECK:       vpmaxsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_pmaxsq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsq_mask_commuted:
+  ;CHECK:       vpmaxsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_pmaxsq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsq_maskz:
+  ;CHECK:       vpmaxsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pmaxsq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsq_maskz_commuted:
+  ;CHECK:       vpmaxsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
+}
+
+define <32 x i16> @stack_fold_pmaxsw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsw
+  ;CHECK:       vpmaxsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  ret <32 x i16> %3
+}
+
+define <32 x i16> @stack_fold_pmaxsw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxsw_commuted:
+  ;CHECK:       vpmaxsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  ret <32 x i16> %3
+}
+
+define <32 x i16> @stack_fold_pmaxsw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsw_mask:
+  ;CHECK:       vpmaxsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
+}
+
+define <32 x i16> @stack_fold_pmaxsw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxsw_mask_commuted:
+  ;CHECK:       vpmaxsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
+}
+
+define <32 x i16> @stack_fold_pmaxsw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsw_maskz:
+  ;CHECK:       vpmaxsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pmaxsw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxsw_maskz_commuted:
+  ;CHECK:       vpmaxsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp sgt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
+}
+
+define <64 x i8> @stack_fold_pmaxub(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxub
+  ;CHECK:       vpmaxub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  ret <64 x i8> %3
+}
+
+define <64 x i8> @stack_fold_pmaxub_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxub_commuted:
+  ;CHECK:       vpmaxub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  ret <64 x i8> %3
+}
+
+define <64 x i8> @stack_fold_pmaxub_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxub_mask:
+  ;CHECK:       vpmaxub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
+}
+
+define <64 x i8> @stack_fold_pmaxub_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxub_mask_commuted:
+  ;CHECK:       vpmaxub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
+}
+
+define <64 x i8> @stack_fold_pmaxub_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxub_maskz:
+  ;CHECK:       vpmaxub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_pmaxub_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxub_maskz_commuted:
+  ;CHECK:       vpmaxub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
+}
+
+define <16 x i32> @stack_fold_pmaxud(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxud
+  ;CHECK:       vpmaxud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  ret <16 x i32> %3
+}
+
+define <16 x i32> @stack_fold_pmaxud_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxud_commuted:
+  ;CHECK:       vpmaxud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  ret <16 x i32> %3
+}
+
+define <16 x i32> @stack_fold_pmaxud_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxud_mask:
+  ;CHECK:       vpmaxud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
-  %5 = icmp eq <16 x i32> %3, %a0
-  %6 = and <16 x i1> %4, %5
-  %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
-  ret <16 x i32> %7
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
 }
 
-define <16 x i32> @stack_fold_pcmpeqd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
-  ;CHECK-LABEL: stack_fold_pcmpeqd_mask_commuted
-  ;CHECK:       vpcmpeqd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
-  %2 = load <16 x i32>, <16 x i32>* %a2
-  %3 = add <16 x i32> %a1, %2
+define <16 x i32> @stack_fold_pmaxud_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxud_mask_commuted:
+  ;CHECK:       vpmaxud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
-  %5 = icmp eq <16 x i32> %a0, %3
-  %6 = and <16 x i1> %4, %5
-  %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
-  ret <16 x i32> %7
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
 }
 
-define <16 x i32> @stack_fold_pcmpled_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
-  ;CHECK-LABEL: stack_fold_pcmpled_mask
-  ;CHECK:       vpcmpled {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
-  %2 = load <16 x i32>, <16 x i32>* %a2
-  %3 = add <16 x i32> %a1, %2
+define <16 x i32> @stack_fold_pmaxud_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxud_maskz:
+  ;CHECK:       vpmaxud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
-  %5 = icmp sge <16 x i32> %a0, %3
-  %6 = and <16 x i1> %4, %5
-  %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
-  ret <16 x i32> %7
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
 }
 
-define i16 @stack_fold_pcmpleud(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
-  ;CHECK-LABEL: stack_fold_pcmpleud
-  ;CHECK:       vpcmpleud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = load <16 x i32>, <16 x i32>* %a2
-  %3 = add <16 x i32> %a1, %2
+define <16 x i32> @stack_fold_pmaxud_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxud_maskz_commuted:
+  ;CHECK:       vpmaxud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
-  %5 = icmp uge <16 x i32> %a0, %3
-  %6 = and <16 x i1> %5, %4
-  %7 = bitcast <16 x i1> %6 to i16
-  ret i16 %7
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
 }
 
-define <64 x i8> @stack_fold_permbvar(<64 x i8> %a0, <64 x i8> %a1) {
-  ;CHECK-LABEL: stack_fold_permbvar
-  ;CHECK:   vpermb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_pmaxuq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxuq
+  ;CHECK:       vpmaxuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
-  ret <64 x i8> %2
+  %2 = icmp ugt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  ret <8 x i64> %3
 }
-declare <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8>, <64 x i8>) nounwind readonly
 
-define <64 x i8> @stack_fold_permbvar_mask(<64 x i8>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
-  ;CHECK-LABEL: stack_fold_permbvar_mask
-  ;CHECK:   vpermb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_pmaxuq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxuq_commuted:
+  ;CHECK:       vpmaxuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
-  %3 = bitcast i64 %mask to <64 x i1>
-  ; load needed to keep the operation from being scheduled above the asm block
-  %4 = load <64 x i8>, <64 x i8>* %passthru
-  %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
+  %2 = icmp ugt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  ret <8 x i64> %3
+}
+
+define <8 x i64> @stack_fold_pmaxuq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxuq_mask:
+  ;CHECK:       vpmaxuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_pmaxuq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxuq_mask_commuted:
+  ;CHECK:       vpmaxuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_pmaxuq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxuq_maskz:
+  ;CHECK:       vpmaxuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pmaxuq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxuq_maskz_commuted:
+  ;CHECK:       vpmaxuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
+}
+
+define <32 x i16> @stack_fold_pmaxuw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxuw
+  ;CHECK:       vpmaxuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  ret <32 x i16> %3
+}
+
+define <32 x i16> @stack_fold_pmaxuw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmaxuw_commuted:
+  ;CHECK:       vpmaxuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  ret <32 x i16> %3
+}
+
+define <32 x i16> @stack_fold_pmaxuw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxuw_mask:
+  ;CHECK:       vpmaxuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
+}
+
+define <32 x i16> @stack_fold_pmaxuw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pmaxuw_mask_commuted:
+  ;CHECK:       vpmaxuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
+}
+
+define <32 x i16> @stack_fold_pmaxuw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxuw_maskz:
+  ;CHECK:       vpmaxuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pmaxuw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmaxuw_maskz_commuted:
+  ;CHECK:       vpmaxuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ugt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
+}
+
+define <64 x i8> @stack_fold_pminsb(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsb
+  ;CHECK:       vpminsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  ret <64 x i8> %3
+}
+
+define <64 x i8> @stack_fold_pminsb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsb_commuted:
+  ;CHECK:       vpminsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  ret <64 x i8> %3
+}
+
+define <64 x i8> @stack_fold_pminsb_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsb_mask:
+  ;CHECK:       vpminsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
+}
+
+define <64 x i8> @stack_fold_pminsb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsb_mask_commuted:
+  ;CHECK:       vpminsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
+}
+
+define <64 x i8> @stack_fold_pminsb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsb_maskz:
+  ;CHECK:       vpminsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
+}
+
+define <64 x i8> @stack_fold_pminsb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsb_maskz_commuted:
+  ;CHECK:       vpminsb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }
 
-define <64 x i8> @stack_fold_permbvar_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
-  ;CHECK-LABEL: stack_fold_permbvar_maskz
-  ;CHECK:   vpermb {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+define <16 x i32> @stack_fold_pminsd(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsd
+  ;CHECK:       vpminsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  ret <16 x i32> %3
+}
+
+define <16 x i32> @stack_fold_pminsd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsd_commuted:
+  ;CHECK:       vpminsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  ret <16 x i32> %3
+}
+
+define <16 x i32> @stack_fold_pminsd_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsd_mask:
+  ;CHECK:       vpminsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  %4 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
+}
+
+define <16 x i32> @stack_fold_pminsd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsd_mask_commuted:
+  ;CHECK:       vpminsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  %4 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
+}
+
+define <16 x i32> @stack_fold_pminsd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsd_maskz:
+  ;CHECK:       vpminsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pminsd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsd_maskz_commuted:
+  ;CHECK:       vpminsd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
+}
+
+define <8 x i64> @stack_fold_pminsq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsq
+  ;CHECK:       vpminsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  ret <8 x i64> %3
+}
+
+define <8 x i64> @stack_fold_pminsq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsq_commuted:
+  ;CHECK:       vpminsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
-  %3 = bitcast i64 %mask to <64 x i1>
-  %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
-  ret <64 x i8> %4
+  %2 = icmp slt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  ret <8 x i64> %3
 }
 
-define <16 x i32> @stack_fold_permd(<16 x i32> %a0, <16 x i32> %a1) {
-  ;CHECK-LABEL: stack_fold_permd
-  ;CHECK:   vpermd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_pminsq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsq_mask:
+  ;CHECK:       vpminsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32> %a1, <16 x i32> %a0)
-  ; add forces execution domain
-  %3 = add <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  ret <16 x i32> %3
+  %2 = icmp slt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
 }
-declare <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32>, <16 x i32>) nounwind readonly
 
-define <64 x i8> @stack_fold_vpermi2b(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermi2b
-  ;CHECK:       vpermi2b {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %x1, <64 x i8> %x0, <64 x i8> %x2)
-  ret <64 x i8> %2
+define <8 x i64> @stack_fold_pminsq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsq_mask_commuted:
+  ;CHECK:       vpminsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
 }
 
-define <16 x i32> @stack_fold_vpermi2d(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermi2d
-  ;CHECK:       vpermi2d {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
-  ret <16 x i32> %2
+define <8 x i64> @stack_fold_pminsq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsq_maskz:
+  ;CHECK:       vpminsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
 }
 
-define <8 x i64> @stack_fold_vpermi2q(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermi2q
-  ;CHECK:       vpermi2q {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x2)
-  ret <8 x i64> %2
+define <8 x i64> @stack_fold_pminsq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsq_maskz_commuted:
+  ;CHECK:       vpminsq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
 }
 
-define <32 x i16> @stack_fold_vpermi2w(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermi2w
-  ;CHECK:       vpermi2w {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16> %x1, <32 x i16> %x0, <32 x i16> %x2)
-  ret <32 x i16> %2
+define <32 x i16> @stack_fold_pminsw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsw
+  ;CHECK:       vpminsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  ret <32 x i16> %3
 }
 
-define <8 x i64> @stack_fold_permq(<8 x i64> %a0) {
-  ;CHECK-LABEL: stack_fold_permq
-  ;CHECK:   vpermq $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
-  ; add forces execution domain
-  %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
-  ret <8 x i64> %3
+define <32 x i16> @stack_fold_pminsw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pminsw_commuted:
+  ;CHECK:       vpminsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  ret <32 x i16> %3
 }
 
-define <8 x i64> @stack_fold_permq_mask(<8 x i64>* %passthru, <8 x i64> %a0, i8 %mask) {
-  ;CHECK-LABEL: stack_fold_permq_mask
-  ;CHECK:   vpermq $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
-  %3 = bitcast i8 %mask to <8 x i1>
-  ; load needed to keep the operation from being scheduled above the asm block
-  %4 = load <8 x i64>, <8 x i64>* %passthru
-  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
-  ; add forces execution domain
-  %6 = add <8 x i64> %5, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
-  ret <8 x i64> %6
+define <32 x i16> @stack_fold_pminsw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsw_mask:
+  ;CHECK:       vpminsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
 }
 
-define <8 x i64> @stack_fold_permq_maskz(<8 x i64>* %passthru, <8 x i64> %a0, i8 %mask) {
-  ;CHECK-LABEL: stack_fold_permq_maskz
-  ;CHECK:   vpermq $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
-  %3 = bitcast i8 %mask to <8 x i1>
-  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
-  ret <8 x i64> %4
+define <32 x i16> @stack_fold_pminsw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminsw_mask_commuted:
+  ;CHECK:       vpminsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp slt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
 }
 
-define <8 x i64> @stack_fold_permqvar(<8 x i64> %a0, <8 x i64> %a1) {
-  ;CHECK-LABEL: stack_fold_permqvar
-  ;CHECK:   vpermq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminsw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsw_maskz:
+  ;CHECK:       vpminsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %a1, <8 x i64> %a0)
-  ; add forces execution domain
-  %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
-  ret <8 x i64> %3
+  %2 = icmp slt <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
 }
-declare <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64>, <8 x i64>) nounwind readonly
 
-define <8 x i64> @stack_fold_permqvar_mask(<8 x i64>* %passthru, <8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
-  ;CHECK-LABEL: stack_fold_permqvar_mask
-  ;CHECK:   vpermq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminsw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pminsw_maskz_commuted:
+  ;CHECK:       vpminsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %a1, <8 x i64> %a0)
-  %3 = bitcast i8 %mask to <8 x i1>
-  ; load needed to keep the operation from being scheduled above the asm block
-  %4 = load <8 x i64>, <8 x i64>* %passthru
-  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
-  ; add forces execution domain
-  %6 = add <8 x i64> %5, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
-  ret <8 x i64> %6
+  %2 = icmp slt <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
 }
 
-define <64 x i8> @stack_fold_vpermt2b(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermt2b
-  ;CHECK:       vpermt2b {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2)
-  ret <64 x i8> %2
+define <64 x i8> @stack_fold_pminub(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pminub
+  ;CHECK:       vpminub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  ret <64 x i8> %3
 }
-declare <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8>, <64 x i8>, <64 x i8>)
 
-define <16 x i32> @stack_fold_vpermt2d(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermt2d
-  ;CHECK:       vpermt2d {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
-  ret <16 x i32> %2
+define <64 x i8> @stack_fold_pminub_commuted(<64 x i8> %a0, <64 x i8> %a1) {
+  ;CHECK-LABEL: stack_fold_pminub_commuted:
+  ;CHECK:       vpminub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  ret <64 x i8> %3
 }
-declare <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32>, <16 x i32>, <16 x i32>)
 
-define <8 x i64> @stack_fold_vpermt2q(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermt2q
-  ;CHECK:       vpermt2q {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2)
-  ret <8 x i64> %2
+define <64 x i8> @stack_fold_pminub_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminub_mask:
+  ;CHECK:       vpminub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
 }
-declare <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64>, <8 x i64>, <8 x i64>)
 
-define <32 x i16> @stack_fold_vpermt2w(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
-  ;CHECK-LABEL: stack_fold_vpermt2w
-  ;CHECK:       vpermt2w {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2)
-  ret <32 x i16> %2
+define <64 x i8> @stack_fold_pminub_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminub_mask_commuted:
+  ;CHECK:       vpminub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <64 x i8>, <64 x i8>* %passthru
+  %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
+  ret <64 x i8> %6
 }
-declare <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16>, <32 x i16>, <32 x i16>)
 
-define <32 x i16> @stack_fold_permwvar(<32 x i16> %a0, <32 x i16> %a1) {
-  ;CHECK-LABEL: stack_fold_permwvar
-  ;CHECK:   vpermw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <64 x i8> @stack_fold_pminub_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pminub_maskz:
+  ;CHECK:       vpminub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
-  ret <32 x i16> %2
+  %2 = icmp ult <64 x i8> %a0, %a1
+  %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
 }
-declare <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16>, <32 x i16>) nounwind readonly
 
-define <32 x i16> @stack_fold_permwvar_mask(<32 x i16>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
-  ;CHECK-LABEL: stack_fold_permwvar_mask
-  ;CHECK:   vpermw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+define <64 x i8> @stack_fold_pminub_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+  ;CHECK-LABEL: stack_fold_pminub_maskz_commuted:
+  ;CHECK:       vpminub {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
-  %3 = bitcast i32 %mask to <32 x i1>
-  ; load needed to keep the operation from being scheduled above the asm block
-  %4 = load <32 x i16>, <32 x i16>* %passthru
-  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
-  ret <32 x i16> %5
+  %2 = icmp ult <64 x i8> %a1, %a0
+  %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
+  %4 = bitcast i64 %mask to <64 x i1>
+  %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
+  ret <64 x i8> %5
 }
 
-define <32 x i16> @stack_fold_permwvar_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
-  ;CHECK-LABEL: stack_fold_permwvar_maskz
-  ;CHECK:   vpermw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+define <16 x i32> @stack_fold_pminud(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pminud
+  ;CHECK:       vpminud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
-  %3 = bitcast i32 %mask to <32 x i1>
-  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
-  ret <32 x i16> %4
+  %2 = icmp ult <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  ret <16 x i32> %3
 }
 
-define i32 @stack_fold_pextrd(<4 x i32> %a0, <4 x i32> %a1) {
-  ;CHECK-LABEL: stack_fold_pextrd
-  ;CHECK:       vpextrd $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
-  ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
-  ; add forces execution domain
-  %1 = add <4 x i32> %a0, %a1
-  %2 = extractelement <4 x i32> %1, i32 1
-  %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  ret i32 %2
+define <16 x i32> @stack_fold_pminud_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pminud_commuted:
+  ;CHECK:       vpminud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  ret <16 x i32> %3
 }
 
-define i64 @stack_fold_pextrq(<2 x i64> %a0) {
-  ;CHECK-LABEL: stack_fold_pextrq
-  ;CHECK:       vpextrq $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 8-byte Folded Spill
-  ;CHECK:       movq    {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Reload
-  %1 = extractelement <2 x i64> %a0, i32 1
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  ret i64 %1
+define <16 x i32> @stack_fold_pminud_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminud_mask:
+  ;CHECK:       vpminud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  %4 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
 }
 
-define <16 x i8> @stack_fold_pinsrb(<16 x i8> %a0, i8 %a1) {
-  ;CHECK-LABEL: stack_fold_pinsrb
-  ;CHECK:       vpinsrb $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  %2 = insertelement <16 x i8> %a0, i8 %a1, i32 1
-  ret <16 x i8> %2
+define <16 x i32> @stack_fold_pminud_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminud_mask_commuted:
+  ;CHECK:       vpminud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  %4 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <16 x i32>, <16 x i32>* %passthru
+  %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
+  ret <16 x i32> %6
 }
 
-define <4 x i32> @stack_fold_pinsrd(<4 x i32> %a0, i32 %a1) {
-  ;CHECK-LABEL: stack_fold_pinsrd
-  ;CHECK:       vpinsrd $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  %2 = insertelement <4 x i32> %a0, i32 %a1, i32 1
-  ret <4 x i32> %2
+define <16 x i32> @stack_fold_pminud_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pminud_maskz:
+  ;CHECK:       vpminud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <16 x i32> %a0, %a1
+  %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
 }
 
-define <2 x i64> @stack_fold_pinsrq(<2 x i64> %a0, i64 %a1) {
-  ;CHECK-LABEL: stack_fold_pinsrq
-  ;CHECK:       vpinsrq $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 8-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  %2 = insertelement <2 x i64> %a0, i64 %a1, i32 1
-  ret <2 x i64> %2
+define <16 x i32> @stack_fold_pminud_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pminud_maskz_commuted:
+  ;CHECK:       vpminud {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <16 x i32> %a1, %a0
+  %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
+  %4 = bitcast i16 %mask to <16 x i1>
+  %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
+  ret <16 x i32> %5
 }
 
-define <8 x i16> @stack_fold_pinsrw(<8 x i16> %a0, i16 %a1) {
-  ;CHECK-LABEL: stack_fold_pinsrw
-  ;CHECK:       vpinsrw $1, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  %2 = insertelement <8 x i16> %a0, i16 %a1, i32 1
-  ret <8 x i16> %2
+define <8 x i64> @stack_fold_pminuq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pminuq
+  ;CHECK:       vpminuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  ret <8 x i64> %3
 }
 
-define <16 x i32> @stack_fold_vplzcntd(<16 x i32> %a0) {
-  ;CHECK-LABEL: stack_fold_vplzcntd
-  ;CHECK:       vplzcntd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %a0, i1 false)
-  ret <16 x i32> %2
+define <8 x i64> @stack_fold_pminuq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pminuq_commuted:
+  ;CHECK:       vpminuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  ret <8 x i64> %3
 }
 
-define <8 x i64> @stack_fold_vplzcntq(<8 x i64> %a0) {
-  ;CHECK-LABEL: stack_fold_vplzcntq
-  ;CHECK:       vplzcntq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x i64> @llvm.ctlz.v8i64(<8 x i64> %a0, i1 false)
-  ret <8 x i64> %2
+define <8 x i64> @stack_fold_pminuq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminuq_mask:
+  ;CHECK:       vpminuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_pminuq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminuq_mask_commuted:
+  ;CHECK:       vpminuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <8 x i64>, <8 x i64>* %passthru
+  %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
+  ret <8 x i64> %6
+}
+
+define <8 x i64> @stack_fold_pminuq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pminuq_maskz:
+  ;CHECK:       vpminuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <8 x i64> %a0, %a1
+  %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
 }
 
-define <32 x i16> @stack_fold_pmaddubsw_zmm(<64 x i8> %a0, <64 x i8> %a1) {
-  ;CHECK-LABEL: stack_fold_pmaddubsw_zmm
-  ;CHECK:       vpmaddubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <8 x i64> @stack_fold_pminuq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pminuq_maskz_commuted:
+  ;CHECK:       vpminuq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
-  ret <32 x i16> %2
+  %2 = icmp ult <8 x i64> %a1, %a0
+  %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
+  %4 = bitcast i8 %mask to <8 x i1>
+  %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
+  ret <8 x i64> %5
 }
-declare <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8>, <64 x i8>) nounwind readnone
 
-define <32 x i16> @stack_fold_pmaddubsw_zmm_mask(<32 x i16>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i32 %mask) {
-  ;CHECK-LABEL: stack_fold_pmaddubsw_zmm_mask
-  ;CHECK:       vpmaddubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminuw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pminuw
+  ;CHECK:       vpminuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
-  %3 = bitcast i32 %mask to <32 x i1>
-  ; load needed to keep the operation from being scheduled about the asm block
-  %4 = load <32 x i16>, <32 x i16>* %passthru
-  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
-  ret <32 x i16> %5
+  %2 = icmp ult <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  ret <32 x i16> %3
 }
 
-define <32 x i16> @stack_fold_pmaddubsw_zmm_maskz(<64 x i8> %a0, <64 x i8> %a1, i32 %mask) {
-  ;CHECK-LABEL: stack_fold_pmaddubsw_zmm_maskz
-  ;CHECK:       vpmaddubsw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminuw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pminuw_commuted:
+  ;CHECK:       vpminuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
-  %3 = bitcast i32 %mask to <32 x i1>
-  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
-  ret <32 x i16> %4
+  %2 = icmp ult <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  ret <32 x i16> %3
 }
 
-define <16 x i32> @stack_fold_pmaddwd_zmm(<32 x i16> %a0, <32 x i16> %a1) {
-  ;CHECK-LABEL: stack_fold_pmaddwd_zmm
-  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminuw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminuw_mask:
+  ;CHECK:       vpminuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
-  ret <16 x i32> %2
+  %2 = icmp ult <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
 }
-declare <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16>, <32 x i16>) nounwind readnone
 
-define <16 x i32> @stack_fold_pmaddwd_zmm_mask(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
-  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_mask
-  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminuw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
+  ;CHECK-LABEL: stack_fold_pminuw_mask_commuted:
+  ;CHECK:       vpminuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
-  %3 = bitcast i16 %mask to <16 x i1>
+  %2 = icmp ult <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
-  %4 = load <16 x i32>, <16 x i32>* %passthru
-  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
-  ret <16 x i32> %5
+  %5 = load <32 x i16>, <32 x i16>* %passthru
+  %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
+  ret <32 x i16> %6
 }
 
-define <16 x i32> @stack_fold_pmaddwd_zmm_maskz(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
-  ;CHECK-LABEL: stack_fold_pmaddwd_zmm_maskz
-  ;CHECK:       vpmaddwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+define <32 x i16> @stack_fold_pminuw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pminuw_maskz:
+  ;CHECK:       vpminuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
-  %3 = bitcast i16 %mask to <16 x i1>
-  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
-  ret <16 x i32> %4
+  %2 = icmp ult <32 x i16> %a0, %a1
+  %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pminuw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pminuw_maskz_commuted:
+  ;CHECK:       vpminuw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = icmp ult <32 x i16> %a1, %a0
+  %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
+  %4 = bitcast i32 %mask to <32 x i1>
+  %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
+  ret <32 x i16> %5
 }
 
 define <16 x i8> @stack_fold_vpmovdb(<16 x i32> %a0) {
@@ -1139,111 +2771,450 @@ define <16 x i8> @stack_fold_vpmovusdb(<16 x i32> %a0) {
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i8> %1
 }
-declare <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32>, <16 x i8>, i16)
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32>, <16 x i8>, i16)
+
+define <16 x i16> @stack_fold_vpmovusdw(<16 x i32> %a0) {
+  ;CHECK-LABEL: stack_fold_vpmovusdw
+  ;CHECK:       vpmovusdw %zmm0, {{-?[0-9]*}}(%rsp) # 32-byte Folded Spill
+  %1 = call <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32> %a0, <16 x i16> undef, i16 -1)
+  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  ret <16 x i16> %1
+}
+declare <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32>, <16 x i16>, i16)
+
+define <8 x i32> @stack_fold_vpmovusqd(<8 x i64> %a0) {
+  ;CHECK-LABEL: stack_fold_vpmovusqd
+  ;CHECK:       vpmovusqd %zmm0, {{-?[0-9]*}}(%rsp) # 32-byte Folded Spill
+  %1 = call <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64> %a0, <8 x i32> undef, i8 -1)
+  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  ret <8 x i32> %1
+}
+declare <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64>, <8 x i32>, i8)
+
+define <8 x i16> @stack_fold_vpmovusqw(<8 x i64> %a0) {
+  ;CHECK-LABEL: stack_fold_vpmovusqw
+  ;CHECK:       vpmovusqw %zmm0, {{-?[0-9]*}}(%rsp) # 16-byte Folded Spill
+  %1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64> %a0, <8 x i16> undef, i8 -1)
+  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  ret <8 x i16> %1
+}
+declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64>, <8 x i16>, i8)
+
+define <32 x i8> @stack_fold_vpmovuswb(<32 x i16> %a0) {
+  ;CHECK-LABEL: stack_fold_vpmovuswb
+  ;CHECK:       vpmovuswb %zmm0, {{-?[0-9]*}}(%rsp) # 32-byte Folded Spill
+  %1 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %a0, <32 x i8> undef, i32 -1)
+  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  ret <32 x i8> %1
+}
+declare <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16>, <32 x i8>, i32)
+
+define <16 x i32> @stack_fold_pmovzxbd_zmm(<16 x i8> %a0) {
+  ;CHECK-LABEL: stack_fold_pmovzxbd_zmm
+  ;CHECK:       vpmovzxbd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <16 x i8> %a0 to <16 x i32>
+  ret <16 x i32> %2
+}
+
+define <8 x i64> @stack_fold_pmovzxbq_zmm(<16 x i8> %a0) {
+  ;CHECK-LABEL: stack_fold_pmovzxbq_zmm
+  ;CHECK:       vpmovzxbq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %3 = zext <8 x i8> %2 to <8 x i64>
+  ret <8 x i64> %3
+}
+
+define <32 x i16> @stack_fold_pmovzxbw_zmm(<32 x i8> %a0) {
+  ;CHECK-LABEL: stack_fold_pmovzxbw_zmm
+  ;CHECK:       vpmovzxbw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <32 x i8> %a0 to <32 x i16>
+  ret <32 x i16> %2
+}
+
+define <8 x i64> @stack_fold_pmovzxdq_zmm(<8 x i32> %a0) {
+  ;CHECK-LABEL: stack_fold_pmovzxdq_zmm
+  ;CHECK:       vpmovzxdq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <8 x i32> %a0 to <8 x i64>
+  ret <8 x i64> %2
+}
+
+define <16 x i32> @stack_fold_pmovzxwd_zmm(<16 x i16> %a0) {
+  ;CHECK-LABEL: stack_fold_pmovzxwd_zmm
+  ;CHECK:       vpmovzxwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <16 x i16> %a0 to <16 x i32>
+  ret <16 x i32> %2
+}
+
+define <8 x i64> @stack_fold_pmovzxwq_zmm(<8 x i16> %a0) {
+  ;CHECK-LABEL: stack_fold_pmovzxwq_zmm
+  ;CHECK:       vpmovzxwq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <8 x i16> %a0 to <8 x i64>
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pmovzxwq_mask_zmm(<8 x i64> %passthru, <8 x i16> %a0, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmovzxwq_mask_zmm
+  ;CHECK:       vpmovzxwq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 16-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <8 x i16> %a0 to <8 x i64>
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %passthru
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @stack_fold_pmovzxwq_maskz_zmm(<8 x i16> %a0, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmovzxwq_maskz_zmm
+  ;CHECK:       vpmovzxwq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 16-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = zext <8 x i16> %a0 to <8 x i64>
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <16 x i32> @stack_fold_pmulld(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pmulld
+  ;CHECK:       vpmulld {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <16 x i32> %a0, %a1
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pmulld_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pmulld_commuted:
+  ;CHECK:       vpmulld {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <16 x i32> %a1, %a0
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pmulld_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmulld_mask
+  ;CHECK:       vpmulld {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pmulld_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmulld_mask_commuted:
+  ;CHECK:       vpmulld {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pmulld_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmulld_maskz
+  ;CHECK:       vpmulld {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @stack_fold_pmulld_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pmulld_maskz_commuted:
+  ;CHECK:       vpmulld {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <8 x i64> @stack_fold_pmullq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmullq
+  ;CHECK:       vpmullq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <8 x i64> %a0, %a1
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pmullq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmullq_commuted:
+  ;CHECK:       vpmullq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <8 x i64> %a1, %a0
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pmullq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullq_mask
+  ;CHECK:       vpmullq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pmullq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullq_mask_commuted:
+  ;CHECK:       vpmullq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pmullq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullq_maskz
+  ;CHECK:       vpmullq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @stack_fold_pmullq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullq_maskz_commuted:
+  ;CHECK:       vpmullq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <32 x i16> @stack_fold_pmullw(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmullw
+  ;CHECK:       vpmullw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <32 x i16> %a0, %a1
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_pmullw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
+  ;CHECK-LABEL: stack_fold_pmullw_commuted:
+  ;CHECK:       vpmullw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <32 x i16> %a1, %a0
+  ret <32 x i16> %2
+}
+
+define <32 x i16> @stack_fold_pmullw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullw_mask
+  ;CHECK:       vpmullw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <32 x i16> %a0, %a1
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pmullw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullw_mask_commuted:
+  ;CHECK:       vpmullw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <32 x i16> %a1, %a0
+  %3 = bitcast i32 %mask to <32 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <32 x i16>, <32 x i16>* %a2
+  %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
+  ret <32 x i16> %5
+}
+
+define <32 x i16> @stack_fold_pmullw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullw_maskz
+  ;CHECK:       vpmullw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <32 x i16> %a0, %a1
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
+}
 
-define <16 x i16> @stack_fold_vpmovusdw(<16 x i32> %a0) {
-  ;CHECK-LABEL: stack_fold_vpmovusdw
-  ;CHECK:       vpmovusdw %zmm0, {{-?[0-9]*}}(%rsp) # 32-byte Folded Spill
-  %1 = call <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32> %a0, <16 x i16> undef, i16 -1)
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ret <16 x i16> %1
+define <32 x i16> @stack_fold_pmullw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+  ;CHECK-LABEL: stack_fold_pmullw_maskz_commuted:
+  ;CHECK:       vpmullw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = mul <32 x i16> %a1, %a0
+  %3 = bitcast i32 %mask to <32 x i1>
+  %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
+  ret <32 x i16> %4
 }
-declare <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32>, <16 x i16>, i16)
 
-define <8 x i32> @stack_fold_vpmovusqd(<8 x i64> %a0) {
-  ;CHECK-LABEL: stack_fold_vpmovusqd
-  ;CHECK:       vpmovusqd %zmm0, {{-?[0-9]*}}(%rsp) # 32-byte Folded Spill
-  %1 = call <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64> %a0, <8 x i32> undef, i8 -1)
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ret <8 x i32> %1
+define <8 x i64> @stack_fold_pmuldq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmuldq
+  ;CHECK:       vpmuldq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %6 = mul <8 x i64> %3, %5
+  ret <8 x i64> %6
 }
-declare <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64>, <8 x i32>, i8)
 
-define <8 x i16> @stack_fold_vpmovusqw(<8 x i64> %a0) {
-  ;CHECK-LABEL: stack_fold_vpmovusqw
-  ;CHECK:       vpmovusqw %zmm0, {{-?[0-9]*}}(%rsp) # 16-byte Folded Spill
-  %1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64> %a0, <8 x i16> undef, i8 -1)
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ret <8 x i16> %1
+define <8 x i64> @stack_fold_pmuldq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmuldq_commuted:
+  ;CHECK:       vpmuldq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %6 = mul <8 x i64> %5, %3
+  ret <8 x i64> %6
 }
-declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64>, <8 x i16>, i8)
 
-define <32 x i8> @stack_fold_vpmovuswb(<32 x i16> %a0) {
-  ;CHECK-LABEL: stack_fold_vpmovuswb
-  ;CHECK:       vpmovuswb %zmm0, {{-?[0-9]*}}(%rsp) # 32-byte Folded Spill
-  %1 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %a0, <32 x i8> undef, i32 -1)
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  ret <32 x i8> %1
+define <8 x i64> @stack_fold_pmuldq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuldq_mask
+  ;CHECK:       vpmuldq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %6 = mul <8 x i64> %3, %5
+  %7 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %8 = load <8 x i64>, <8 x i64>* %a2
+  %9 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> %8
+  ret <8 x i64> %9
 }
-declare <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16>, <32 x i8>, i32)
 
-define <16 x i32> @stack_fold_pmovzxbd_zmm(<16 x i8> %a0) {
-  ;CHECK-LABEL: stack_fold_pmovzxbd_zmm
-  ;CHECK:       vpmovzxbd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <16 x i8> %a0 to <16 x i32>
-  ret <16 x i32> %2
+define <8 x i64> @stack_fold_pmuldq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuldq_mask_commuted:
+  ;CHECK:       vpmuldq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %6 = mul <8 x i64> %5, %3
+  %7 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %8 = load <8 x i64>, <8 x i64>* %a2
+  %9 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> %8
+  ret <8 x i64> %9
 }
 
-define <8 x i64> @stack_fold_pmovzxbq_zmm(<16 x i8> %a0) {
-  ;CHECK-LABEL: stack_fold_pmovzxbq_zmm
-  ;CHECK:       vpmovzxbq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  %3 = zext <8 x i8> %2 to <8 x i64>
-  ret <8 x i64> %3
+define <8 x i64> @stack_fold_pmuldq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuldq_maskz
+  ;CHECK:       vpmuldq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %6 = mul <8 x i64> %3, %5
+  %7 = bitcast i8 %mask to <8 x i1>
+  %8 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> zeroinitializer
+  ret <8 x i64> %8
+}
+
+define <8 x i64> @stack_fold_pmuldq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuldq_maskz_commuted:
+  ;CHECK:       vpmuldq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %6 = mul <8 x i64> %5, %3
+  %7 = bitcast i8 %mask to <8 x i1>
+  %8 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> zeroinitializer
+  ret <8 x i64> %8
 }
 
-define <32 x i16> @stack_fold_pmovzxbw_zmm(<32 x i8> %a0) {
-  ;CHECK-LABEL: stack_fold_pmovzxbw_zmm
-  ;CHECK:       vpmovzxbw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <32 x i8> %a0 to <32 x i16>
-  ret <32 x i16> %2
+
+
+
+define <8 x i64> @stack_fold_pmuludq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmuludq
+  ;CHECK:       vpmuludq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %4 = mul <8 x i64> %2, %3
+  ret <8 x i64> %4
 }
 
-define <8 x i64> @stack_fold_pmovzxdq_zmm(<8 x i32> %a0) {
-  ;CHECK-LABEL: stack_fold_pmovzxdq_zmm
-  ;CHECK:       vpmovzxdq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <8 x i32> %a0 to <8 x i64>
-  ret <8 x i64> %2
+define <8 x i64> @stack_fold_pmuludq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pmuludq_commuted:
+  ;CHECK:       vpmuludq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %4 = mul <8 x i64> %3, %2
+  ret <8 x i64> %4
 }
 
-define <16 x i32> @stack_fold_pmovzxwd_zmm(<16 x i16> %a0) {
-  ;CHECK-LABEL: stack_fold_pmovzxwd_zmm
-  ;CHECK:       vpmovzxwd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <16 x i16> %a0 to <16 x i32>
-  ret <16 x i32> %2
+define <8 x i64> @stack_fold_pmuludq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuludq_mask
+  ;CHECK:       vpmuludq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %4 = mul <8 x i64> %2, %3
+  %5 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %6 = load <8 x i64>, <8 x i64>* %a2
+  %7 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> %6
+  ret <8 x i64> %7
 }
 
-define <8 x i64> @stack_fold_pmovzxwq_zmm(<8 x i16> %a0) {
-  ;CHECK-LABEL: stack_fold_pmovzxwq_zmm
-  ;CHECK:       vpmovzxwq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <8 x i16> %a0 to <8 x i64>
-  ret <8 x i64> %2
+define <8 x i64> @stack_fold_pmuludq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuludq_mask_commuted:
+  ;CHECK:       vpmuludq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %4 = mul <8 x i64> %3, %2
+  %5 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %6 = load <8 x i64>, <8 x i64>* %a2
+  %7 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> %6
+  ret <8 x i64> %7
 }
 
-define <8 x i64> @stack_fold_pmovzxwq_mask_zmm(<8 x i64> %passthru, <8 x i16> %a0, i8 %mask) {
-  ;CHECK-LABEL: stack_fold_pmovzxwq_mask_zmm
-  ;CHECK:       vpmovzxwq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 16-byte Folded Reload
+define <8 x i64> @stack_fold_pmuludq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuludq_maskz
+  ;CHECK:       vpmuludq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <8 x i16> %a0 to <8 x i64>
-  %3 = bitcast i8 %mask to <8 x i1>
-  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %passthru
-  ret <8 x i64> %4
+  %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %4 = mul <8 x i64> %2, %3
+  %5 = bitcast i8 %mask to <8 x i1>
+  %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
+  ret <8 x i64> %6
 }
 
-define <8 x i64> @stack_fold_pmovzxwq_maskz_zmm(<8 x i16> %a0, i8 %mask) {
-  ;CHECK-LABEL: stack_fold_pmovzxwq_maskz_zmm
-  ;CHECK:       vpmovzxwq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 16-byte Folded Reload
-  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = zext <8 x i16> %a0 to <8 x i64>
-  %3 = bitcast i8 %mask to <8 x i1>
-  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
-  ret <8 x i64> %4
+define <8 x i64> @stack_fold_pmuludq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pmuludq_maskz_commuted:
+  ;CHECK:       vpmuludq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+  %4 = mul <8 x i64> %3, %2
+  %5 = bitcast i8 %mask to <8 x i1>
+  %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
+  ret <8 x i64> %6
 }
 
 define <16 x i32> @stack_fold_vpopcntd(<16 x i32> %a0) {
@@ -1264,6 +3235,126 @@ define <8 x i64> @stack_fold_vpopcntq(<8 x i64> %a0) {
 }
 declare <8 x i64> @llvm.ctpop.v8i64(<8 x i64>) nounwind readnone
 
+define <16 x i32> @stack_fold_pord(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pord
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <16 x i32> %a0, %a1
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pord_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pord_commuted:
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <16 x i32> %a1, %a0
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pord_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pord_mask
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pord_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pord_mask_commuted:
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pord_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pord_maskz
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @stack_fold_pord_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pord_maskz_commuted:
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <8 x i64> @stack_fold_porq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_porq
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <8 x i64> %a0, %a1
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_porq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_porq_commuted:
+  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <8 x i64> %a1, %a0
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_porq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_porq_mask
+  ;CHECK:       vorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_porq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_porq_mask_commuted:
+  ;CHECK:       vorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_porq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_porq_maskz
+  ;CHECK:       vorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @stack_fold_porq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_porq_maskz_commuted:
+  ;CHECK:       vorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = or <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
 define <8 x i64> @stack_fold_psadbw(<64 x i8> %a0, <64 x i8> %a1) {
   ;CHECK-LABEL: stack_fold_psadbw
   ;CHECK:       vpsadbw {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
@@ -1875,6 +3966,126 @@ define <64 x i8> @stack_fold_punpckhbw_maskz_zmm(<64 x i8> %a0, <64 x i8> %a1, i
   ret <64 x i8> %4
 }
 
+define <16 x i32> @stack_fold_pxord(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pxord
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <16 x i32> %a0, %a1
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pxord_commuted(<16 x i32> %a0, <16 x i32> %a1) {
+  ;CHECK-LABEL: stack_fold_pxord_commuted:
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <16 x i32> %a1, %a0
+  ret <16 x i32> %2
+}
+
+define <16 x i32> @stack_fold_pxord_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pxord_mask
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pxord_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pxord_mask_commuted:
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <16 x i32>, <16 x i32>* %a2
+  %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
+  ret <16 x i32> %5
+}
+
+define <16 x i32> @stack_fold_pxord_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pxord_maskz
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <16 x i32> %a0, %a1
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @stack_fold_pxord_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+  ;CHECK-LABEL: stack_fold_pxord_maskz_commuted:
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <16 x i32> %a1, %a0
+  %3 = bitcast i16 %mask to <16 x i1>
+  %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
+  ret <16 x i32> %4
+}
+
+define <8 x i64> @stack_fold_pxorq(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pxorq
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <8 x i64> %a0, %a1
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pxorq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
+  ;CHECK-LABEL: stack_fold_pxorq_commuted:
+  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <8 x i64> %a1, %a0
+  ret <8 x i64> %2
+}
+
+define <8 x i64> @stack_fold_pxorq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pxorq_mask
+  ;CHECK:       vxorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pxorq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pxorq_mask_commuted:
+  ;CHECK:       vxorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  ; load needed to keep the operation from being scheduled about the asm block
+  %4 = load <8 x i64>, <8 x i64>* %a2
+  %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
+  ret <8 x i64> %5
+}
+
+define <8 x i64> @stack_fold_pxorq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pxorq_maskz
+  ;CHECK:       vxorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <8 x i64> %a0, %a1
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @stack_fold_pxorq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+  ;CHECK-LABEL: stack_fold_pxorq_maskz_commuted:
+  ;CHECK:       vxorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %2 = xor <8 x i64> %a1, %a0
+  %3 = bitcast i8 %mask to <8 x i1>
+  %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
+  ret <8 x i64> %4
+}
+
 declare <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8>, <64 x i8>)
 declare <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16>, <32 x i16>)
 declare <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8>, <64 x i8>)