1; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512dq < %s | FileCheck %s
2
3target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
4target triple = "x86_64-unknown-unknown"
5
6; Stack reload folding tests.
7;
8; By including a nop call with sideeffects we can force a partial register spill of the
9; relevant registers and check that the reload is correctly folded into the instruction.
10
11define <8 x double> @stack_fold_addpd_zmm(<8 x double> %a0, <8 x double> %a1) {
12  ;CHECK-LABEL: stack_fold_addpd_zmm
13  ;CHECK:       vaddpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
14  %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}"()
15  %2 = fadd <8 x double> %a0, %a1
16  ret <8 x double> %2
17}
18
19define <8 x double> @stack_fold_addpd_zmm_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
20  ;CHECK-LABEL: stack_fold_addpd_zmm_kz
21  ;CHECK:       vaddpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
22  %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}"()
23  %2 = fadd <8 x double> %a1, %a0
24  %3 = bitcast i8 %mask to <8 x i1>
25  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> zeroinitializer
26  ret <8 x double> %4
27}
28
29define <16 x float> @stack_fold_addps_zmm(<16 x float> %a0, <16 x float> %a1) {
30  ;CHECK-LABEL: stack_fold_addps_zmm
31  ;CHECK:       vaddps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
32  %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}"()
33  %2 = fadd <16 x float> %a0, %a1
34  ret <16 x float> %2
35}
36
37define <16 x float> @stack_fold_addps_zmm_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
38  ;CHECK-LABEL: stack_fold_addps_zmm_kz
39  ;CHECK:       vaddps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
40  %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}"()
41  %2 = fadd <16 x float> %a1, %a0
42  %3 = bitcast i16 %mask to <16 x i1>
43  %4 = select <16 x i1> %3, <16 x float> %2, <16 x float> zeroinitializer
44  ret <16 x float> %4
45}
46
47define double @stack_fold_addsd(double %a0, double %a1) {
48  ;CHECK-LABEL: stack_fold_addsd
49  ;CHECK:       vaddsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 8-byte Folded Reload
50  %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}"()
51  %2 = fadd double %a0, %a1
52  ret double %2
53}
54
55define <2 x double> @stack_fold_addsd_int(<2 x double> %a0, <2 x double> %a1) {
56  ;CHECK-LABEL: stack_fold_addsd_int
57  ;CHECK:       vaddsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
58  %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}"()
59  %2 = extractelement <2 x double> %a0, i32 0
60  %3 = extractelement <2 x double> %a1, i32 0
61  %4 = fadd double %2, %3
62  %5 = insertelement <2 x double> %a0, double %4, i32 0
63  ret <2 x double> %5
64}
65
66define float @stack_fold_addss(float %a0, float %a1) {
67  ;CHECK-LABEL: stack_fold_addss
68  ;CHECK:       vaddss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
69  %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}"()
70  %2 = fadd float %a0, %a1
71  ret float %2
72}
73
74define <4 x float> @stack_fold_addss_int(<4 x float> %a0, <4 x float> %a1) {
75  ;CHECK-LABEL: stack_fold_addss_int
76  ;CHECK:       vaddss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
77  %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}"()
78  %2 = extractelement <4 x float> %a0, i32 0
79  %3 = extractelement <4 x float> %a1, i32 0
80  %4 = fadd float %2, %3
81  %5 = insertelement <4 x float> %a0, float %4, i32 0
82  ret <4 x float> %5
83}
84
85define <8 x double> @stack_fold_andnpd_zmm(<8 x double> %a0, <8 x double> %a1) {
86  ;CHECK-LABEL: stack_fold_andnpd_zmm
87  ;CHECK:       vandnpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
88  %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}"()
89  %2 = bitcast <8 x double> %a0 to <8 x i64>
90  %3 = bitcast <8 x double> %a1 to <8 x i64>
91  %4 = xor <8 x i64> %2, <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>
92  %5 = and <8 x i64> %4, %3
93  %6 = bitcast <8 x i64> %5 to <8 x double>
94  ; fadd forces execution domain
95  %7 = fadd <8 x double> %6, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
96  ret <8 x double> %7
97}
98
99define <16 x float> @stack_fold_andnps_zmm(<16 x float> %a0, <16 x float> %a1) {
100  ;CHECK-LABEL: stack_fold_andnps_zmm
101  ;CHECK:       vandnps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
102  %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}"()
103  %2 = bitcast <16 x float> %a0 to <16 x i32>
104  %3 = bitcast <16 x float> %a1 to <16 x i32>
105  %4 = xor <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>
106  %5 = and <16 x i32> %4, %3
107  %6 = bitcast <16 x i32> %5 to <16 x float>
108  ; fadd forces execution domain
109  %7 = fadd <16 x float> %6, <float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0>
110  ret <16 x float> %7
111}
112
113define <8 x double> @stack_fold_andpd_zmm(<8 x double> %a0, <8 x double> %a1) {
114  ;CHECK-LABEL: stack_fold_andpd_zmm
115  ;CHECK:       vandpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
116  %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}"()
117  %2 = bitcast <8 x double> %a0 to <8 x i64>
118  %3 = bitcast <8 x double> %a1 to <8 x i64>
119  %4 = and <8 x i64> %2, %3
120  %5 = bitcast <8 x i64> %4 to <8 x double>
121  ; fadd forces execution domain
122  %6 = fadd <8 x double> %5, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
123  ret <8 x double> %6
124}
125
126define <16 x float> @stack_fold_andps_zmm(<16 x float> %a0, <16 x float> %a1) {
127  ;CHECK-LABEL: stack_fold_andps_zmm
128  ;CHECK:       vandps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
129  %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}"()
130  %2 = bitcast <16 x float> %a0 to <16 x i32>
131  %3 = bitcast <16 x float> %a1 to <16 x i32>
132  %4 = and <16 x i32> %2, %3
133  %5 = bitcast <16 x i32> %4 to <16 x float>
134  ; fadd forces execution domain
135  %6 = fadd <16 x float> %5, <float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0>
136  ret <16 x float> %6
137}
138
139define i8 @stack_fold_cmppd(<8 x double> %a0, <8 x double> %a1) {
140  ;CHECK-LABEL: stack_fold_cmppd
141  ;CHECK:       vcmpeqpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-9]}} {{.*#+}} 64-byte Folded Reload
142  %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}"()
143  %res = call i8 @llvm.x86.avx512.mask.cmp.pd.512(<8 x double> %a0, <8 x double> %a1, i32 0, i8 -1, i32 4)
144  ret i8 %res
145}
146declare i8 @llvm.x86.avx512.mask.cmp.pd.512(<8 x double> , <8 x double> , i32, i8, i32)
147
148define i16 @stack_fold_cmpps(<16 x float> %a0, <16 x float> %a1) {
149  ;CHECK-LABEL: stack_fold_cmpps
150  ;CHECK:       vcmpeqps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-9]*}} {{.*#+}} 64-byte Folded Reload
151  %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}"()
152  %res = call i16 @llvm.x86.avx512.mask.cmp.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0, i16 -1, i32 4)
153  ret i16 %res
154}
155declare i16 @llvm.x86.avx512.mask.cmp.ps.512(<16 x float> , <16 x float> , i32, i16, i32)
156
157define <2 x double> @stack_fold_divsd_int(<2 x double> %a0, <2 x double> %a1) {
158  ;CHECK-LABEL: stack_fold_divsd_int
159  ;CHECK:       vdivsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
160  %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}"()
161  %2 = extractelement <2 x double> %a0, i32 0
162  %3 = extractelement <2 x double> %a1, i32 0
163  %4 = fdiv double %2, %3
164  %5 = insertelement <2 x double> %a0, double %4, i32 0
165  ret <2 x double> %5
166}
167
168define float @stack_fold_divss(float %a0, float %a1) {
169  ;CHECK-LABEL: stack_fold_divss
170  ;CHECK:       vdivss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
171  %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}"()
172  %2 = fdiv float %a0, %a1
173  ret float %2
174}
175
176define <4 x float> @stack_fold_divss_int(<4 x float> %a0, <4 x float> %a1) {
177  ;CHECK-LABEL: stack_fold_divss_int
178  ;CHECK:       vdivss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
179  %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}"()
180  %2 = extractelement <4 x float> %a0, i32 0
181  %3 = extractelement <4 x float> %a1, i32 0
182  %4 = fdiv float %2, %3
183  %5 = insertelement <4 x float> %a0, float %4, i32 0
184  ret <4 x float> %5
185}
186
187define <4 x float> @stack_fold_insertps(<4 x float> %a0, <4 x float> %a1) {
188  ;CHECK-LABEL: stack_fold_insertps
189  ;CHECK:       vinsertps $17, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
190  ;CHECK-NEXT:                                                                              {{.*#+}} xmm0 = zero,mem[0],xmm0[2,3]
191  %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}"()
192  %2 = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 209)
193  ret <4 x float> %2
194}
195declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i8) nounwind readnone
196
197define <8 x double> @stack_fold_maxpd_zmm(<8 x double> %a0, <8 x double> %a1) #0 {
198  ;CHECK-LABEL: stack_fold_maxpd_zmm
199  ;CHECK:       vmaxpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
200  %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}"()
201  %2 = call <8 x double> @llvm.x86.avx512.mask.max.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> zeroinitializer, i8 -1, i32 4)
202  ret <8 x double> %2
203}
204declare <8 x double> @llvm.x86.avx512.mask.max.pd.512(<8 x double>, <8 x double>, <8 x double>, i8, i32) nounwind readnone
205
206define <8 x double> @stack_fold_maxpd_zmm_commutable(<8 x double> %a0, <8 x double> %a1) #1 {
207  ;CHECK-LABEL: stack_fold_maxpd_zmm_commutable
208  ;CHECK:       vmaxpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
209  %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}"()
210  %2 = call <8 x double> @llvm.x86.avx512.mask.max.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> zeroinitializer, i8 -1, i32 4)
211  ret <8 x double> %2
212}
213
214define <8 x double> @stack_fold_maxpd_zmm_commutable_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) #1 {
215  ;CHECK-LABEL: stack_fold_maxpd_zmm_commutable_kz
216  ;CHECK:       vmaxpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
217  %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}"()
218  %2 = call <8 x double> @llvm.x86.avx512.mask.max.pd.512(<8 x double> %a1, <8 x double> %a0, <8 x double> zeroinitializer, i8 %mask, i32 4)
219  ret <8 x double> %2
220}
221
222define <16 x float> @stack_fold_maxps_zmm(<16 x float> %a0, <16 x float> %a1) #0 {
223  ;CHECK-LABEL: stack_fold_maxps_zmm
224  ;CHECK:       vmaxps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
225  %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}"()
226  %2 = call <16 x float> @llvm.x86.avx512.mask.max.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> zeroinitializer, i16 -1, i32 4)
227  ret <16 x float> %2
228}
229declare <16 x float> @llvm.x86.avx512.mask.max.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32) nounwind readnone
230
231define <16 x float> @stack_fold_maxps_zmm_commutable(<16 x float> %a0, <16 x float> %a1) #1 {
232  ;CHECK-LABEL: stack_fold_maxps_zmm_commutable
233  ;CHECK:       vmaxps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
234  %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}"()
235  %2 = call <16 x float> @llvm.x86.avx512.mask.max.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> zeroinitializer, i16 -1, i32 4)
236  ret <16 x float> %2
237}
238
239define <16 x float> @stack_fold_maxps_zmm_commutable_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) #1 {
240  ;CHECK-LABEL: stack_fold_maxps_zmm_commutable_kz
241  ;CHECK:       vmaxps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
242  %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}"()
243  %2 = call <16 x float> @llvm.x86.avx512.mask.max.ps.512(<16 x float> %a1, <16 x float> %a0, <16 x float> zeroinitializer, i16 %mask, i32 4)
244  ret <16 x float> %2
245}
246
247define <8 x double> @stack_fold_minpd_zmm(<8 x double> %a0, <8 x double> %a1) #0 {
248  ;CHECK-LABEL: stack_fold_minpd_zmm
249  ;CHECK:       vminpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
250  %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}"()
251  %2 = call <8 x double> @llvm.x86.avx512.mask.min.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> zeroinitializer, i8 -1, i32 4)
252  ret <8 x double> %2
253}
254declare <8 x double> @llvm.x86.avx512.mask.min.pd.512(<8 x double>, <8 x double>, <8 x double>, i8, i32) nounwind readnone
255
256define <8 x double> @stack_fold_minpd_zmm_commutable(<8 x double> %a0, <8 x double> %a1) #1 {
257  ;CHECK-LABEL: stack_fold_minpd_zmm_commutable
258  ;CHECK:       vminpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
259  %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}"()
260  %2 = call <8 x double> @llvm.x86.avx512.mask.min.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> zeroinitializer, i8 -1, i32 4)
261  ret <8 x double> %2
262}
263
264define <8 x double> @stack_fold_minpd_zmm_commutable_kz(<8 x double> %a0, <8 x double> %a1, i8 %mask) #1 {
265  ;CHECK-LABEL: stack_fold_minpd_zmm_commutable_kz
266  ;CHECK:       vminpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
267  %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}"()
268  %2 = call <8 x double> @llvm.x86.avx512.mask.min.pd.512(<8 x double> %a1, <8 x double> %a0, <8 x double> zeroinitializer, i8 %mask, i32 4)
269  ret <8 x double> %2
270}
271
272define <16 x float> @stack_fold_minps_zmm(<16 x float> %a0, <16 x float> %a1) #0 {
273  ;CHECK-LABEL: stack_fold_minps_zmm
274  ;CHECK:       vminps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
275  %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}"()
276  %2 = call <16 x float> @llvm.x86.avx512.mask.min.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> zeroinitializer, i16 -1, i32 4)
277  ret <16 x float> %2
278}
279declare <16 x float> @llvm.x86.avx512.mask.min.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32) nounwind readnone
280
281define <16 x float> @stack_fold_minps_zmm_commutable(<16 x float> %a0, <16 x float> %a1) #1 {
282  ;CHECK-LABEL: stack_fold_minps_zmm_commutable
283  ;CHECK:       vminps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
284  %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}"()
285  %2 = call <16 x float> @llvm.x86.avx512.mask.min.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> zeroinitializer, i16 -1, i32 4)
286  ret <16 x float> %2
287}
288
289define <16 x float> @stack_fold_minps_zmm_commutable_kz(<16 x float> %a0, <16 x float> %a1, i16 %mask) #1 {
290  ;CHECK-LABEL: stack_fold_minps_zmm_commutable_kz
291  ;CHECK:       vminps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 64-byte Folded Reload
292  %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}"()
293  %2 = call <16 x float> @llvm.x86.avx512.mask.min.ps.512(<16 x float> %a1, <16 x float> %a0, <16 x float> zeroinitializer, i16 %mask, i32 4)
294  ret <16 x float> %2
295}
296
297define double @stack_fold_mulsd(double %a0, double %a1) {
298  ;CHECK-LABEL: stack_fold_mulsd
299  ;CHECK:       vmulsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 8-byte Folded Reload
300  %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}"()
301  %2 = fmul double %a0, %a1
302  ret double %2
303}
304
305define <2 x double> @stack_fold_mulsd_int(<2 x double> %a0, <2 x double> %a1) {
306  ;CHECK-LABEL: stack_fold_mulsd_int
307  ;CHECK:       vmulsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
308  %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}"()
309  %2 = extractelement <2 x double> %a0, i32 0
310  %3 = extractelement <2 x double> %a1, i32 0
311  %4 = fmul double %2, %3
312  %5 = insertelement <2 x double> %a0, double %4, i32 0
313  ret <2 x double> %5
314}
315
316define float @stack_fold_mulss(float %a0, float %a1) {
317  ;CHECK-LABEL: stack_fold_mulss
318  ;CHECK:       vmulss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
319  %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}"()
320  %2 = fmul float %a0, %a1
321  ret float %2
322}
323
324define <4 x float> @stack_fold_mulss_int(<4 x float> %a0, <4 x float> %a1) {
325  ;CHECK-LABEL: stack_fold_mulss_int
326  ;CHECK:       vmulss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
327  %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}"()
328  %2 = extractelement <4 x float> %a0, i32 0
329  %3 = extractelement <4 x float> %a1, i32 0
330  %4 = fmul float %2, %3
331  %5 = insertelement <4 x float> %a0, float %4, i32 0
332  ret <4 x float> %5
333}
334
335define <8 x double> @stack_fold_orpd_zmm(<8 x double> %a0, <8 x double> %a1) #0 {
336  ;CHECK-LABEL: stack_fold_orpd_zmm
337  ;CHECK:       vorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
338  %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}"()
339  %2 = bitcast <8 x double> %a0 to <8 x i64>
340  %3 = bitcast <8 x double> %a1 to <8 x i64>
341  %4 = or <8 x i64> %2, %3
342  %5 = bitcast <8 x i64> %4 to <8 x double>
343  ; fadd forces execution domain
344  %6 = fadd <8 x double> %5, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
345  ret <8 x double> %6
346}
347
348define <16 x float> @stack_fold_orps_zmm(<16 x float> %a0, <16 x float> %a1) #0 {
349  ;CHECK-LABEL: stack_fold_orps_zmm
350  ;CHECK:       vorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
351  %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}"()
352  %2 = bitcast <16 x float> %a0 to <16 x i32>
353  %3 = bitcast <16 x float> %a1 to <16 x i32>
354  %4 = or <16 x i32> %2, %3
355  %5 = bitcast <16 x i32> %4 to <16 x float>
356  ; fadd forces execution domain
357  %6 = fadd <16 x float> %5, <float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0>
358  ret <16 x float> %6
359}
360
361define <8 x double> @stack_fold_subpd_zmm(<8 x double> %a0, <8 x double> %a1) {
362  ;CHECK-LABEL: stack_fold_subpd_zmm
363  ;CHECK:       vsubpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
364  %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}"()
365  %2 = fsub <8 x double> %a0, %a1
366  ret <8 x double> %2
367}
368
369define <16 x float> @stack_fold_subps_zmm(<16 x float> %a0, <16 x float> %a1) {
370  ;CHECK-LABEL: stack_fold_subps_zmm
371  ;CHECK:       vsubps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
372  %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}"()
373  %2 = fsub <16 x float> %a0, %a1
374  ret <16 x float> %2
375}
376
377define double @stack_fold_subsd(double %a0, double %a1) {
378  ;CHECK-LABEL: stack_fold_subsd
379  ;CHECK:       vsubsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 8-byte Folded Reload
380  %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}"()
381  %2 = fsub double %a0, %a1
382  ret double %2
383}
384
385define <2 x double> @stack_fold_subsd_int(<2 x double> %a0, <2 x double> %a1) {
386  ;CHECK-LABEL: stack_fold_subsd_int
387  ;CHECK:       vsubsd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
388  %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}"()
389  %2 = extractelement <2 x double> %a0, i32 0
390  %3 = extractelement <2 x double> %a1, i32 0
391  %4 = fsub double %2, %3
392  %5 = insertelement <2 x double> %a0, double %4, i32 0
393  ret <2 x double> %5
394}
395
396define float @stack_fold_subss(float %a0, float %a1) {
397  ;CHECK-LABEL: stack_fold_subss
398  ;CHECK:       vsubss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 4-byte Folded Reload
399  %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}"()
400  %2 = fsub float %a0, %a1
401  ret float %2
402}
403
404define <4 x float> @stack_fold_subss_int(<4 x float> %a0, <4 x float> %a1) {
405  ;CHECK-LABEL: stack_fold_subss_int
406  ;CHECK:       vsubss {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
407  %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}"()
408  %2 = extractelement <4 x float> %a0, i32 0
409  %3 = extractelement <4 x float> %a1, i32 0
410  %4 = fsub float %2, %3
411  %5 = insertelement <4 x float> %a0, float %4, i32 0
412  ret <4 x float> %5
413}
414
415define <8 x double> @stack_fold_xorpd_zmm(<8 x double> %a0, <8 x double> %a1) #0 {
416  ;CHECK-LABEL: stack_fold_xorpd_zmm
417  ;CHECK:       vxorpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
418  %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}"()
419  %2 = bitcast <8 x double> %a0 to <8 x i64>
420  %3 = bitcast <8 x double> %a1 to <8 x i64>
421  %4 = xor <8 x i64> %2, %3
422  %5 = bitcast <8 x i64> %4 to <8 x double>
423  ; fadd forces execution domain
424  %6 = fadd <8 x double> %5, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
425  ret <8 x double> %6
426}
427
428define <16 x float> @stack_fold_xorps_zmm(<16 x float> %a0, <16 x float> %a1) #0 {
429  ;CHECK-LABEL: stack_fold_xorps_zmm
430  ;CHECK:       vxorps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
431  %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}"()
432  %2 = bitcast <16 x float> %a0 to <16 x i32>
433  %3 = bitcast <16 x float> %a1 to <16 x i32>
434  %4 = xor <16 x i32> %2, %3
435  %5 = bitcast <16 x i32> %4 to <16 x float>
436  ; fadd forces execution domain
437  %6 = fadd <16 x float> %5, <float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0, float 0x0>
438  ret <16 x float> %6
439}
440
441define i32 @stack_fold_extractps(<4 x float> %a0) {
442  ;CHECK-LABEL: stack_fold_extractps
443  ;CHECK:       vextractps $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
444  ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
445  %1 = extractelement <4 x float> %a0, i32 1
446  %2 = bitcast float %1 to i32
447  %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}"()
448  ret i32 %2
449}
450
451define <4 x float> @stack_fold_extracti32x4(<16 x float> %a0, <16 x float> %a1) {
452  ;CHECK-LABEL: stack_fold_extracti32x4
453  ;CHECK:       vextractf32x4 $3, {{%zmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 16-byte Folded Spill
454  %1 = shufflevector <16 x float> %a0, <16 x float> %a1, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
455  %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}"()
456  ret <4 x float> %1
457}
458
459define <2 x double> @stack_fold_extractf64x2(<8 x double> %a0, <8 x double> %a1) {
460  ;CHECK-LABEL: stack_fold_extractf64x2
461  ;CHECK:       vextractf64x2 $3, {{%zmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 16-byte Folded Spill
462  %1 = shufflevector <8 x double> %a0, <8 x double> %a1, <2 x i32> <i32 6, i32 7>
463  %2 = tail call <2 x double> 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}"()
464  ret <2 x double> %1
465}
466
467define <8 x float> @stack_fold_extracti32x8(<16 x float> %a0, <16 x float> %a1) {
468  ;CHECK-LABEL: stack_fold_extracti32x8
469  ;CHECK:       vextractf32x8 $1, {{%zmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 32-byte Folded Spill
470  %1 = shufflevector <16 x float> %a0, <16 x float> %a1, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
471  %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}"()
472  ret <8 x float> %1
473}
474
475define <4 x double> @stack_fold_extractf64x4(<8 x double> %a0, <8 x double> %a1) {
476  ;CHECK-LABEL: stack_fold_extractf64x4
477  ;CHECK:       vextractf64x4 $1, {{%zmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 32-byte Folded Spill
478  %1 = shufflevector <8 x double> %a0, <8 x double> %a1, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
479  %2 = tail call <2 x double> 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}"()
480  ret <4 x double> %1
481}
482
483define <16 x float> @stack_fold_insertf32x8(<8 x float> %a0, <8 x float> %a1) {
484  ;CHECK-LABEL: stack_fold_insertf32x8
485  ;CHECK:       vinsertf32x8 $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
486  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
487  %2 = shufflevector <8 x float> %a0, <8 x float> %a1, <16 x i32> <i32 0, 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>
488  ret <16 x float> %2
489}
490
491define <8 x double> @stack_fold_insertf64x4(<4 x double> %a0, <4 x double> %a1) {
492  ;CHECK-LABEL: stack_fold_insertf64x4
493  ;CHECK:       vinsertf64x4 $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
494  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
495  %2 = shufflevector <4 x double> %a0, <4 x double> %a1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
496  ret <8 x double> %2
497}
498
499define <8 x double> @stack_fold_insertf64x4_mask(<8 x double> %passthru, <4 x double> %a0, <4 x double> %a1, i8 %mask) {
500  ;CHECK-LABEL: stack_fold_insertf64x4_mask
501  ;CHECK:       vinsertf64x4 $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {{.*#+}} 32-byte Folded Reload
502  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
503  %2 = shufflevector <4 x double> %a0, <4 x double> %a1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
504  %3 = bitcast i8 %mask to <8 x i1>
505  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> %passthru
506  ret <8 x double> %4
507}
508
509define <8 x double> @stack_fold_insertf64x4_maskz(<4 x double> %a0, <4 x double> %a1, i8 %mask) {
510  ;CHECK-LABEL: stack_fold_insertf64x4_maskz
511  ;CHECK:       vinsertf64x4 $1, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[1-7]}}} {z} {{.*#+}} 32-byte Folded Reload
512  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
513  %2 = shufflevector <4 x double> %a0, <4 x double> %a1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
514  %3 = bitcast i8 %mask to <8 x i1>
515  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> zeroinitializer
516  ret <8 x double> %4
517}
518
519define <16 x float> @stack_fold_vpermt2ps(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2) {
520  ;CHECK-LABEL: stack_fold_vpermt2ps
521  ;CHECK:       vpermt2ps {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
522  %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}"()
523  %res = call <16 x float> @llvm.x86.avx512.mask.vpermi2var.ps.512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 -1)
524  ret <16 x float> %res
525}
526declare <16 x float> @llvm.x86.avx512.mask.vpermi2var.ps.512(<16 x float>, <16 x i32>, <16 x float>, i16)
527
528define <16 x float> @stack_fold_vpermi2ps(<16 x i32> %x0, <16 x float> %x1, <16 x float> %x2) {
529  ;CHECK-LABEL: stack_fold_vpermi2ps
530  ;CHECK:       vpermi2ps {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
531  %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}"()
532  %res = call <16 x float> @llvm.x86.avx512.mask.vpermt2var.ps.512(<16 x i32> %x0, <16 x float> %x1, <16 x float> %x2, i16 -1)
533  ret <16 x float> %res
534}
535declare <16 x float> @llvm.x86.avx512.mask.vpermt2var.ps.512(<16 x i32>, <16 x float>, <16 x float>, i16)
536
537define <16 x float> @stack_fold_vpermi2ps_mask(<16 x float> %x0, <16 x i32>* %x1, <16 x float> %x2, i16 %mask) {
538  ;CHECK-LABEL: stack_fold_vpermi2ps_mask
539  ;CHECK:       vpermi2ps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
540  %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}"()
541  %x1b = load <16 x i32>, <16 x i32>* %x1
542  %res = call <16 x float> @llvm.x86.avx512.mask.vpermi2var.ps.512(<16 x float> %x0, <16 x i32> %x1b, <16 x float> %x2, i16 %mask)
543  ret <16 x float> %res
544}
545
546define <16 x float> @stack_fold_vpermt2ps_mask(<16 x i32>* %x0, <16 x float> %x1, <16 x float> %x2, i16 %mask) {
547  ;CHECK-LABEL: stack_fold_vpermt2ps_mask
548  ;CHECK:       vpermt2ps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
549  %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}"()
550  %x0b = load <16 x i32>, <16 x i32>* %x0
551  %res = call <16 x float> @llvm.x86.avx512.mask.vpermt2var.ps.512(<16 x i32> %x0b, <16 x float> %x1, <16 x float> %x2, i16 %mask)
552  ret <16 x float> %res
553}
554
555define <16 x float> @stack_fold_vpermt2ps_maskz(<16 x i32>* %x0, <16 x float> %x1, <16 x float> %x2, i16 %mask) {
556  ;CHECK-LABEL: stack_fold_vpermt2ps_maskz
557  ;CHECK:       vpermt2ps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
558  %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}"()
559  %x0b = load <16 x i32>, <16 x i32>* %x0
560  %res = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> %x0b, <16 x float> %x1, <16 x float> %x2, i16 %mask)
561  ret <16 x float> %res
562}
563declare <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32>, <16 x float>, <16 x float>, i16)
564
565define <8 x double> @stack_fold_vpermt2pd(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2) {
566  ;CHECK-LABEL: stack_fold_vpermt2pd
567  ;CHECK:       vpermt2pd {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
568  %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}"()
569  %res = call <8 x double> @llvm.x86.avx512.mask.vpermi2var.pd.512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 -1)
570  ret <8 x double> %res
571}
572declare <8 x double> @llvm.x86.avx512.mask.vpermi2var.pd.512(<8 x double>, <8 x i64>, <8 x double>, i8)
573
574define <8 x double> @stack_fold_vpermi2pd(<8 x i64> %x0, <8 x double> %x1, <8 x double> %x2) {
575  ;CHECK-LABEL: stack_fold_vpermi2pd
576  ;CHECK:       vpermi2pd {{-?[0-9]*}}(%rsp), %zmm1, %zmm0 # 64-byte Folded Reload
577  %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}"()
578  %res = call <8 x double> @llvm.x86.avx512.mask.vpermt2var.pd.512(<8 x i64> %x0, <8 x double> %x1, <8 x double> %x2, i8 -1)
579  ret <8 x double> %res
580}
581declare <8 x double> @llvm.x86.avx512.mask.vpermt2var.pd.512(<8 x i64>, <8 x double>, <8 x double>, i8)
582
583define <8 x double> @stack_fold_permpd(<8 x double> %a0) {
584  ;CHECK-LABEL: stack_fold_permpd
585  ;CHECK:   vpermpd $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
586  %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}"()
587  %2 = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
588  ; fadd forces execution domain
589  %3 = fadd <8 x double> %2, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
590  ret <8 x double> %3
591}
592
593define <8 x double> @stack_fold_permpd_mask(<8 x double>* %passthru, <8 x double> %a0, i8 %mask) {
594  ;CHECK-LABEL: stack_fold_permpd_mask
595  ;CHECK:   vpermpd $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
596  %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}"()
597  %2 = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
598  %3 = bitcast i8 %mask to <8 x i1>
599  ; load needed to keep the operation from being scheduled above the asm block
600  %4 = load <8 x double>, <8 x double>* %passthru
601  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
602  ; fadd forces execution domain
603  %6 = fadd <8 x double> %5, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
604  ret <8 x double> %6
605}
606
607define <8 x double> @stack_fold_permpd_maskz(<8 x double> %a0, i8 %mask) {
608  ;CHECK-LABEL: stack_fold_permpd_maskz
609  ;CHECK:   vpermpd $235, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
610  %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}"()
611  %2 = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
612  %3 = bitcast i8 %mask to <8 x i1>
613  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> zeroinitializer
614  ret <8 x double> %4
615}
616
617define <8 x double> @stack_fold_permpdvar(<8 x i64> %a0, <8 x double> %a1) {
618  ;CHECK-LABEL: stack_fold_permpdvar
619  ;CHECK:   vpermpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
620  %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}"()
621  %2 = call <8 x double> @llvm.x86.avx512.mask.permvar.df.512(<8 x double> %a1, <8 x i64> %a0, <8 x double> undef, i8 -1)
622  ; fadd forces execution domain
623  %3 = fadd <8 x double> %2, <double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0, double 0x0>
624  ret <8 x double> %3
625}
626declare <8 x double> @llvm.x86.avx512.mask.permvar.df.512(<8 x double>, <8 x i64>, <8 x double>, i8) nounwind readonly
627
628define <16 x float> @stack_fold_permps(<16 x i32> %a0, <16 x float> %a1) {
629  ;CHECK-LABEL: stack_fold_permps
630  ;CHECK:       vpermps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
631  %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}"()
632  %2 = call <16 x float> @llvm.x86.avx512.mask.permvar.sf.512(<16 x float> %a1, <16 x i32> %a0, <16 x float> undef, i16 -1)
633  ret <16 x float> %2
634}
635declare <16 x float> @llvm.x86.avx512.mask.permvar.sf.512(<16 x float>, <16 x i32>, <16 x float>, i16) nounwind readonly
636
637define <8 x double> @stack_fold_permilpd_zmm(<8 x double> %a0) {
638  ;CHECK-LABEL: stack_fold_permilpd_zmm
639  ;CHECK:   vpermilpd $85, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
640  %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}"()
641  %2 = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
642  ret <8 x double> %2
643}
644
645define <8 x double> @stack_fold_permilpd_zmm_mask(<8 x double>* %passthru, <8 x double> %a0, i8 %mask) {
646  ;CHECK-LABEL: stack_fold_permilpd_zmm_mask
647  ;CHECK:   vpermilpd $85, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
648  %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}"()
649  %2 = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
650  %3 = bitcast i8 %mask to <8 x i1>
651  ; load needed to keep the operation from being scheduled above the asm block
652  %4 = load <8 x double>, <8 x double>* %passthru
653  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
654  ret <8 x double> %5
655}
656
657define <8 x double> @stack_fold_permilpd_zmm_maskz(<8 x double> %a0, i8 %mask) {
658  ;CHECK-LABEL: stack_fold_permilpd_zmm_maskz
659  ;CHECK:   vpermilpd $85, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
660  %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}"()
661  %2 = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
662  %3 = bitcast i8 %mask to <8 x i1>
663  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> zeroinitializer
664  ret <8 x double> %4
665}
666
667define <8 x double> @stack_fold_permilpdvar_zmm(<8 x double> %a0, <8 x i64> %a1) {
668  ;CHECK-LABEL: stack_fold_permilpdvar_zmm
669  ;CHECK:       vpermilpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
670  %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}"()
671  %2 = call <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double> %a0, <8 x i64> %a1, <8 x double> undef, i8 -1)
672  ret <8 x double> %2
673}
674declare <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double>, <8 x i64>, <8 x double>, i8) nounwind readnone
675
676define <8 x double> @stack_fold_permilpdvar_zmm_mask(<8 x double>* %passthru, <8 x double> %a0, <8 x i64> %a1, i8 %mask) {
677  ;CHECK-LABEL: stack_fold_permilpdvar_zmm_mask
678  ;CHECK:       vpermilpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
679  %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}"()
680  %2 = call <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double> %a0, <8 x i64> %a1, <8 x double> undef, i8 -1)
681  %3 = bitcast i8 %mask to <8 x i1>
682  ; load needed to keep the operation from being scheduled above the asm block
683  %4 = load <8 x double>, <8 x double>* %passthru
684  %5 = select <8 x i1> %3, <8 x double> %2, <8 x double> %4
685  ret <8 x double> %5
686}
687
688define <8 x double> @stack_fold_permilpdvar_zmm_maskz(<8 x double> %a0, <8 x i64> %a1, i8 %mask) {
689  ;CHECK-LABEL: stack_fold_permilpdvar_zmm_maskz
690  ;CHECK:       vpermilpd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
691  %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}"()
692  %2 = call <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double> %a0, <8 x i64> %a1, <8 x double> undef, i8 -1)
693  %3 = bitcast i8 %mask to <8 x i1>
694  %4 = select <8 x i1> %3, <8 x double> %2, <8 x double> zeroinitializer
695  ret <8 x double> %4
696}
697
698define <16 x float> @stack_fold_permilps_zmm(<16 x float> %a0) {
699  ;CHECK-LABEL: stack_fold_permilps_zmm
700  ;CHECK:   vpermilps $27, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
701  %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}"()
702  %2 = shufflevector <16 x float> %a0, <16 x float> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
703  ret <16 x float> %2
704}
705
706define <16 x float> @stack_fold_permilps_zmm_mask(<16 x float>* %passthru, <16 x float> %a0, i16 %mask) {
707  ;CHECK-LABEL: stack_fold_permilps_zmm_mask
708  ;CHECK:   vpermilps $27, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
709  %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}"()
710  %2 = shufflevector <16 x float> %a0, <16 x float> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
711  %3 = bitcast i16 %mask to <16 x i1>
712  ; load needed to keep the operation from being scheduled above the asm block
713  %4 = load <16 x float>, <16 x float>* %passthru
714  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
715  ret <16 x float> %5
716}
717
718define <16 x float> @stack_fold_permilps_zmm_maskz(<16 x float> %a0, i16 %mask) {
719  ;CHECK-LABEL: stack_fold_permilps_zmm_maskz
720  ;CHECK:   vpermilps $27, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
721  %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}"()
722  %2 = shufflevector <16 x float> %a0, <16 x float> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
723  %3 = bitcast i16 %mask to <16 x i1>
724  %4 = select <16 x i1> %3, <16 x float> %2, <16 x float> zeroinitializer
725  ret <16 x float> %4
726}
727
728define <16 x float> @stack_fold_permilpsvar_zmm(<16 x float> %a0, <16 x i32> %a1) {
729  ;CHECK-LABEL: stack_fold_permilpsvar_zmm
730  ;CHECK:       vpermilps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
731  %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}"()
732  %2 = call <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float> %a0, <16 x i32> %a1, <16 x float> undef, i16 -1)
733  ret <16 x float> %2
734}
735declare <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float>, <16 x i32>, <16 x float>, i16) nounwind readnone
736
737define <16 x float> @stack_fold_permilpsvar_zmm_mask(<16 x float>* %passthru, <16 x float> %a0, <16 x i32> %a1, i16 %mask) {
738  ;CHECK-LABEL: stack_fold_permilpsvar_zmm_mask
739  ;CHECK:       vpermilps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
740  %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}"()
741  %2 = call <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float> %a0, <16 x i32> %a1, <16 x float> undef, i16 -1)
742  %3 = bitcast i16 %mask to <16 x i1>
743  ; load needed to keep the operation from being scheduled above the asm block
744  %4 = load <16 x float>, <16 x float>* %passthru
745  %5 = select <16 x i1> %3, <16 x float> %2, <16 x float> %4
746  ret <16 x float> %5
747}
748
749define <16 x float> @stack_fold_permilpsvar_zmm_maskz(<16 x float> %a0, <16 x i32> %a1, i16 %mask) {
750  ;CHECK-LABEL: stack_fold_permilpsvar_zmm_maskz
751  ;CHECK:       vpermilps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
752  %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}"()
753  %2 = call <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float> %a0, <16 x i32> %a1, <16 x float> undef, i16 -1)
754  %3 = bitcast i16 %mask to <16 x i1>
755  %4 = select <16 x i1> %3, <16 x float> %2, <16 x float> zeroinitializer
756  ret <16 x float> %4
757}
758
759attributes #0 = { "unsafe-fp-math"="false" }
760attributes #1 = { "unsafe-fp-math"="true" }
761