1 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s
3 declare <16 x float> @llvm.x86.avx512.gather.dps.512 (<16 x float>, i8*, <16 x i32>, i16, i32)
4 declare void @llvm.x86.avx512.scatter.dps.512 (i8*, i16, <16 x i32>, <16 x float>, i32)
5 declare <8 x double> @llvm.x86.avx512.gather.dpd.512 (<8 x double>, i8*, <8 x i32>, i8, i32)
6 declare void @llvm.x86.avx512.scatter.dpd.512 (i8*, i8, <8 x i32>, <8 x double>, i32)
8 declare <8 x float> @llvm.x86.avx512.gather.qps.512 (<8 x float>, i8*, <8 x i64>, i8, i32)
9 declare void @llvm.x86.avx512.scatter.qps.512 (i8*, i8, <8 x i64>, <8 x float>, i32)
10 declare <8 x double> @llvm.x86.avx512.gather.qpd.512 (<8 x double>, i8*, <8 x i64>, i8, i32)
11 declare void @llvm.x86.avx512.scatter.qpd.512 (i8*, i8, <8 x i64>, <8 x double>, i32)
13 define void @gather_mask_dps(<16 x i32> %ind, <16 x float> %src, i16 %mask, i8* %base, i8* %stbuf) {
14 ; CHECK-LABEL: gather_mask_dps:
16 ; CHECK-NEXT: kmovw %edi, %k1
17 ; CHECK-NEXT: kmovw %k1, %k2
18 ; CHECK-NEXT: vgatherdps (%rsi,%zmm0,4), %zmm1 {%k2}
19 ; CHECK-NEXT: vpaddd {{.*}}(%rip), %zmm0, %zmm0
20 ; CHECK-NEXT: vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1}
22 %x = call <16 x float> @llvm.x86.avx512.gather.dps.512 (<16 x float> %src, i8* %base, <16 x i32>%ind, i16 %mask, i32 4)
23 %ind2 = add <16 x i32> %ind, <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
24 call void @llvm.x86.avx512.scatter.dps.512 (i8* %stbuf, i16 %mask, <16 x i32>%ind2, <16 x float> %x, i32 4)
28 define void @gather_mask_dpd(<8 x i32> %ind, <8 x double> %src, i8 %mask, i8* %base, i8* %stbuf) {
29 ; CHECK-LABEL: gather_mask_dpd:
31 ; CHECK-NEXT: kmovb %edi, %k1
32 ; CHECK-NEXT: kmovw %k1, %k2
33 ; CHECK-NEXT: vgatherdpd (%rsi,%ymm0,4), %zmm1 {%k2}
34 ; CHECK-NEXT: vpaddd {{.*}}(%rip), %ymm0, %ymm0
35 ; CHECK-NEXT: vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1}
37 %x = call <8 x double> @llvm.x86.avx512.gather.dpd.512 (<8 x double> %src, i8* %base, <8 x i32>%ind, i8 %mask, i32 4)
38 %ind2 = add <8 x i32> %ind, <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
39 call void @llvm.x86.avx512.scatter.dpd.512 (i8* %stbuf, i8 %mask, <8 x i32>%ind2, <8 x double> %x, i32 4)
43 define void @gather_mask_qps(<8 x i64> %ind, <8 x float> %src, i8 %mask, i8* %base, i8* %stbuf) {
44 ; CHECK-LABEL: gather_mask_qps:
46 ; CHECK-NEXT: kmovb %edi, %k1
47 ; CHECK-NEXT: kmovw %k1, %k2
48 ; CHECK-NEXT: vgatherqps (%rsi,%zmm0,4), %ymm1 {%k2}
49 ; CHECK-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
50 ; CHECK-NEXT: vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1}
52 %x = call <8 x float> @llvm.x86.avx512.gather.qps.512 (<8 x float> %src, i8* %base, <8 x i64>%ind, i8 %mask, i32 4)
53 %ind2 = add <8 x i64> %ind, <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>
54 call void @llvm.x86.avx512.scatter.qps.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind2, <8 x float> %x, i32 4)
58 define void @gather_mask_qpd(<8 x i64> %ind, <8 x double> %src, i8 %mask, i8* %base, i8* %stbuf) {
59 ; CHECK-LABEL: gather_mask_qpd:
61 ; CHECK-NEXT: kmovb %edi, %k1
62 ; CHECK-NEXT: kmovw %k1, %k2
63 ; CHECK-NEXT: vgatherqpd (%rsi,%zmm0,4), %zmm1 {%k2}
64 ; CHECK-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
65 ; CHECK-NEXT: vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1}
67 %x = call <8 x double> @llvm.x86.avx512.gather.qpd.512 (<8 x double> %src, i8* %base, <8 x i64>%ind, i8 %mask, i32 4)
68 %ind2 = add <8 x i64> %ind, <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>
69 call void @llvm.x86.avx512.scatter.qpd.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind2, <8 x double> %x, i32 4)
73 ;; Integer Gather/Scatter
75 declare <16 x i32> @llvm.x86.avx512.gather.dpi.512 (<16 x i32>, i8*, <16 x i32>, i16, i32)
76 declare void @llvm.x86.avx512.scatter.dpi.512 (i8*, i16, <16 x i32>, <16 x i32>, i32)
77 declare <8 x i64> @llvm.x86.avx512.gather.dpq.512 (<8 x i64>, i8*, <8 x i32>, i8, i32)
78 declare void @llvm.x86.avx512.scatter.dpq.512 (i8*, i8, <8 x i32>, <8 x i64>, i32)
80 declare <8 x i32> @llvm.x86.avx512.gather.qpi.512 (<8 x i32>, i8*, <8 x i64>, i8, i32)
81 declare void @llvm.x86.avx512.scatter.qpi.512 (i8*, i8, <8 x i64>, <8 x i32>, i32)
82 declare <8 x i64> @llvm.x86.avx512.gather.qpq.512 (<8 x i64>, i8*, <8 x i64>, i8, i32)
83 declare void @llvm.x86.avx512.scatter.qpq.512 (i8*, i8, <8 x i64>, <8 x i64>, i32)
85 define void @gather_mask_dd(<16 x i32> %ind, <16 x i32> %src, i16 %mask, i8* %base, i8* %stbuf) {
86 ; CHECK-LABEL: gather_mask_dd:
88 ; CHECK-NEXT: kmovw %edi, %k1
89 ; CHECK-NEXT: kmovw %k1, %k2
90 ; CHECK-NEXT: vpgatherdd (%rsi,%zmm0,4), %zmm1 {%k2}
91 ; CHECK-NEXT: vpaddd {{.*}}(%rip), %zmm0, %zmm0
92 ; CHECK-NEXT: vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1}
94 %x = call <16 x i32> @llvm.x86.avx512.gather.dpi.512 (<16 x i32> %src, i8* %base, <16 x i32>%ind, i16 %mask, i32 4)
95 %ind2 = add <16 x i32> %ind, <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
96 call void @llvm.x86.avx512.scatter.dpi.512 (i8* %stbuf, i16 %mask, <16 x i32>%ind2, <16 x i32> %x, i32 4)
100 define void @gather_mask_qd(<8 x i64> %ind, <8 x i32> %src, i8 %mask, i8* %base, i8* %stbuf) {
101 ; CHECK-LABEL: gather_mask_qd:
103 ; CHECK-NEXT: kmovb %edi, %k1
104 ; CHECK-NEXT: kmovw %k1, %k2
105 ; CHECK-NEXT: vpgatherqd (%rsi,%zmm0,4), %ymm1 {%k2}
106 ; CHECK-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
107 ; CHECK-NEXT: vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1}
109 %x = call <8 x i32> @llvm.x86.avx512.gather.qpi.512 (<8 x i32> %src, i8* %base, <8 x i64>%ind, i8 %mask, i32 4)
110 %ind2 = add <8 x i64> %ind, <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>
111 call void @llvm.x86.avx512.scatter.qpi.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind2, <8 x i32> %x, i32 4)
115 define void @gather_mask_qq(<8 x i64> %ind, <8 x i64> %src, i8 %mask, i8* %base, i8* %stbuf) {
116 ; CHECK-LABEL: gather_mask_qq:
118 ; CHECK-NEXT: kmovb %edi, %k1
119 ; CHECK-NEXT: kmovw %k1, %k2
120 ; CHECK-NEXT: vpgatherqq (%rsi,%zmm0,4), %zmm1 {%k2}
121 ; CHECK-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
122 ; CHECK-NEXT: vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1}
124 %x = call <8 x i64> @llvm.x86.avx512.gather.qpq.512 (<8 x i64> %src, i8* %base, <8 x i64>%ind, i8 %mask, i32 4)
125 %ind2 = add <8 x i64> %ind, <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>
126 call void @llvm.x86.avx512.scatter.qpq.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind2, <8 x i64> %x, i32 4)
130 define void @gather_mask_dq(<8 x i32> %ind, <8 x i64> %src, i8 %mask, i8* %base, i8* %stbuf) {
131 ; CHECK-LABEL: gather_mask_dq:
133 ; CHECK-NEXT: kmovb %edi, %k1
134 ; CHECK-NEXT: kmovw %k1, %k2
135 ; CHECK-NEXT: vpgatherdq (%rsi,%ymm0,4), %zmm1 {%k2}
136 ; CHECK-NEXT: vpaddd {{.*}}(%rip), %ymm0, %ymm0
137 ; CHECK-NEXT: vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1}
139 %x = call <8 x i64> @llvm.x86.avx512.gather.dpq.512 (<8 x i64> %src, i8* %base, <8 x i32>%ind, i8 %mask, i32 4)
140 %ind2 = add <8 x i32> %ind, <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
141 call void @llvm.x86.avx512.scatter.dpq.512 (i8* %stbuf, i8 %mask, <8 x i32>%ind2, <8 x i64> %x, i32 4)
145 define void @gather_mask_dpd_execdomain(<8 x i32> %ind, <8 x double> %src, i8 %mask, i8* %base, <8 x double>* %stbuf) {
146 ; CHECK-LABEL: gather_mask_dpd_execdomain:
148 ; CHECK-NEXT: kmovb %edi, %k1
149 ; CHECK-NEXT: vgatherdpd (%rsi,%ymm0,4), %zmm1 {%k1}
150 ; CHECK-NEXT: vmovapd %zmm1, (%rdx)
152 %x = call <8 x double> @llvm.x86.avx512.gather.dpd.512 (<8 x double> %src, i8* %base, <8 x i32>%ind, i8 %mask, i32 4)
153 store <8 x double> %x, <8 x double>* %stbuf
157 define void @gather_mask_qpd_execdomain(<8 x i64> %ind, <8 x double> %src, i8 %mask, i8* %base, <8 x double>* %stbuf) {
158 ; CHECK-LABEL: gather_mask_qpd_execdomain:
160 ; CHECK-NEXT: kmovb %edi, %k1
161 ; CHECK-NEXT: vgatherqpd (%rsi,%zmm0,4), %zmm1 {%k1}
162 ; CHECK-NEXT: vmovapd %zmm1, (%rdx)
164 %x = call <8 x double> @llvm.x86.avx512.gather.qpd.512 (<8 x double> %src, i8* %base, <8 x i64>%ind, i8 %mask, i32 4)
165 store <8 x double> %x, <8 x double>* %stbuf
169 define <16 x float> @gather_mask_dps_execdomain(<16 x i32> %ind, <16 x float> %src, i16 %mask, i8* %base) {
170 ; CHECK-LABEL: gather_mask_dps_execdomain:
172 ; CHECK-NEXT: kmovw %edi, %k1
173 ; CHECK-NEXT: vgatherdps (%rsi,%zmm0,4), %zmm1 {%k1}
174 ; CHECK-NEXT: vmovaps %zmm1, %zmm0
176 %res = call <16 x float> @llvm.x86.avx512.gather.dps.512 (<16 x float> %src, i8* %base, <16 x i32>%ind, i16 %mask, i32 4)
177 ret <16 x float> %res;
180 define <8 x float> @gather_mask_qps_execdomain(<8 x i64> %ind, <8 x float> %src, i8 %mask, i8* %base) {
181 ; CHECK-LABEL: gather_mask_qps_execdomain:
183 ; CHECK-NEXT: kmovb %edi, %k1
184 ; CHECK-NEXT: vgatherqps (%rsi,%zmm0,4), %ymm1 {%k1}
185 ; CHECK-NEXT: vmovaps %zmm1, %zmm0
187 %res = call <8 x float> @llvm.x86.avx512.gather.qps.512 (<8 x float> %src, i8* %base, <8 x i64>%ind, i8 %mask, i32 4)
188 ret <8 x float> %res;
191 define void @scatter_mask_dpd_execdomain(<8 x i32> %ind, <8 x double>* %src, i8 %mask, i8* %base, i8* %stbuf) {
192 ; CHECK-LABEL: scatter_mask_dpd_execdomain:
194 ; CHECK-NEXT: kmovb %esi, %k1
195 ; CHECK-NEXT: vmovapd (%rdi), %zmm1
196 ; CHECK-NEXT: vscatterdpd %zmm1, (%rcx,%ymm0,4) {%k1}
198 %x = load <8 x double>, <8 x double>* %src, align 64
199 call void @llvm.x86.avx512.scatter.dpd.512 (i8* %stbuf, i8 %mask, <8 x i32>%ind, <8 x double> %x, i32 4)
203 define void @scatter_mask_qpd_execdomain(<8 x i64> %ind, <8 x double>* %src, i8 %mask, i8* %base, i8* %stbuf) {
204 ; CHECK-LABEL: scatter_mask_qpd_execdomain:
206 ; CHECK-NEXT: kmovb %esi, %k1
207 ; CHECK-NEXT: vmovapd (%rdi), %zmm1
208 ; CHECK-NEXT: vscatterqpd %zmm1, (%rcx,%zmm0,4) {%k1}
210 %x = load <8 x double>, <8 x double>* %src, align 64
211 call void @llvm.x86.avx512.scatter.qpd.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind, <8 x double> %x, i32 4)
215 define void @scatter_mask_dps_execdomain(<16 x i32> %ind, <16 x float>* %src, i16 %mask, i8* %base, i8* %stbuf) {
216 ; CHECK-LABEL: scatter_mask_dps_execdomain:
218 ; CHECK-NEXT: kmovw %esi, %k1
219 ; CHECK-NEXT: vmovaps (%rdi), %zmm1
220 ; CHECK-NEXT: vscatterdps %zmm1, (%rcx,%zmm0,4) {%k1}
222 %x = load <16 x float>, <16 x float>* %src, align 64
223 call void @llvm.x86.avx512.scatter.dps.512 (i8* %stbuf, i16 %mask, <16 x i32>%ind, <16 x float> %x, i32 4)
227 define void @scatter_mask_qps_execdomain(<8 x i64> %ind, <8 x float>* %src, i8 %mask, i8* %base, i8* %stbuf) {
228 ; CHECK-LABEL: scatter_mask_qps_execdomain:
230 ; CHECK-NEXT: kmovb %esi, %k1
231 ; CHECK-NEXT: vmovaps (%rdi), %ymm1
232 ; CHECK-NEXT: vscatterqps %ymm1, (%rcx,%zmm0,4) {%k1}
234 %x = load <8 x float>, <8 x float>* %src, align 32
235 call void @llvm.x86.avx512.scatter.qps.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind, <8 x float> %x, i32 4)
239 define void @gather_qps(<8 x i64> %ind, <8 x float> %src, i8* %base, i8* %stbuf) {
240 ; CHECK-LABEL: gather_qps:
242 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
243 ; CHECK-NEXT: kxnorw %k2, %k2, %k2
244 ; CHECK-NEXT: vgatherqps (%rdi,%zmm0,4), %ymm1 {%k2}
245 ; CHECK-NEXT: vpaddq {{.*}}(%rip), %zmm0, %zmm0
246 ; CHECK-NEXT: vscatterqps %ymm1, (%rsi,%zmm0,4) {%k1}
248 %x = call <8 x float> @llvm.x86.avx512.gather.qps.512 (<8 x float> %src, i8* %base, <8 x i64>%ind, i8 -1, i32 4)
249 %ind2 = add <8 x i64> %ind, <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>
250 call void @llvm.x86.avx512.scatter.qps.512 (i8* %stbuf, i8 -1, <8 x i64>%ind2, <8 x float> %x, i32 4)
254 declare void @llvm.x86.avx512.gatherpf.qps.512(i8, <8 x i64>, i8* , i32, i32);
255 declare void @llvm.x86.avx512.scatterpf.qps.512(i8, <8 x i64>, i8* , i32, i32);
256 define void @prefetch(<8 x i64> %ind, i8* %base) {
257 ; CHECK-LABEL: prefetch:
259 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
260 ; CHECK-NEXT: vgatherpf0qps (%rdi,%zmm0,4) {%k1}
261 ; CHECK-NEXT: vgatherpf1qps (%rdi,%zmm0,4) {%k1}
262 ; CHECK-NEXT: vscatterpf0qps (%rdi,%zmm0,2) {%k1}
263 ; CHECK-NEXT: vscatterpf1qps (%rdi,%zmm0,2) {%k1}
265 call void @llvm.x86.avx512.gatherpf.qps.512(i8 -1, <8 x i64> %ind, i8* %base, i32 4, i32 0)
266 call void @llvm.x86.avx512.gatherpf.qps.512(i8 -1, <8 x i64> %ind, i8* %base, i32 4, i32 1)
267 call void @llvm.x86.avx512.scatterpf.qps.512(i8 -1, <8 x i64> %ind, i8* %base, i32 2, i32 0)
268 call void @llvm.x86.avx512.scatterpf.qps.512(i8 -1, <8 x i64> %ind, i8* %base, i32 2, i32 1)
273 declare <2 x double> @llvm.x86.avx512.gather3div2.df(<2 x double>, i8*, <2 x i64>, i8, i32)
275 define <2 x double>@test_int_x86_avx512_gather3div2_df(<2 x double> %x0, i8* %x1, <2 x i64> %x2, i8 %x3) {
276 ; CHECK-LABEL: test_int_x86_avx512_gather3div2_df:
278 ; CHECK-NEXT: kmovb %esi, %k1
279 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
280 ; CHECK-NEXT: vgatherqpd (%rdi,%xmm1,4), %xmm2 {%k1}
281 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
282 ; CHECK-NEXT: vgatherqpd (%rdi,%xmm1,0), %xmm0 {%k1}
283 ; CHECK-NEXT: vaddpd %xmm0, %xmm2, %xmm0
285 %res = call <2 x double> @llvm.x86.avx512.gather3div2.df(<2 x double> %x0, i8* %x1, <2 x i64> %x2, i8 %x3, i32 4)
286 %res1 = call <2 x double> @llvm.x86.avx512.gather3div2.df(<2 x double> %x0, i8* %x1, <2 x i64> %x2, i8 -1, i32 0)
287 %res2 = fadd <2 x double> %res, %res1
288 ret <2 x double> %res2
291 declare <4 x i32> @llvm.x86.avx512.gather3div2.di(<2 x i64>, i8*, <2 x i64>, i8, i32)
293 define <4 x i32>@test_int_x86_avx512_gather3div2_di(<2 x i64> %x0, i8* %x1, <2 x i64> %x2, i8 %x3) {
294 ; CHECK-LABEL: test_int_x86_avx512_gather3div2_di:
296 ; CHECK-NEXT: kmovb %esi, %k1
297 ; CHECK-NEXT: vpgatherqq (%rdi,%xmm1,8), %xmm0 {%k1}
298 ; CHECK-NEXT: vpaddd %xmm0, %xmm0, %xmm0
300 %res = call <4 x i32> @llvm.x86.avx512.gather3div2.di(<2 x i64> %x0, i8* %x1, <2 x i64> %x2, i8 %x3, i32 8)
301 %res1 = call <4 x i32> @llvm.x86.avx512.gather3div2.di(<2 x i64> %x0, i8* %x1, <2 x i64> %x2, i8 %x3, i32 8)
302 %res2 = add <4 x i32> %res, %res1
306 declare <4 x double> @llvm.x86.avx512.gather3div4.df(<4 x double>, i8*, <4 x i64>, i8, i32)
308 define <4 x double>@test_int_x86_avx512_gather3div4_df(<4 x double> %x0, i8* %x1, <4 x i64> %x2, i8 %x3) {
309 ; CHECK-LABEL: test_int_x86_avx512_gather3div4_df:
311 ; CHECK-NEXT: kmovb %esi, %k1
312 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
313 ; CHECK-NEXT: vgatherqpd (%rdi,%ymm1,4), %ymm2 {%k1}
314 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
315 ; CHECK-NEXT: vgatherqpd (%rdi,%ymm1,0), %ymm0 {%k1}
316 ; CHECK-NEXT: vaddpd %ymm0, %ymm2, %ymm0
318 %res = call <4 x double> @llvm.x86.avx512.gather3div4.df(<4 x double> %x0, i8* %x1, <4 x i64> %x2, i8 %x3, i32 4)
319 %res1 = call <4 x double> @llvm.x86.avx512.gather3div4.df(<4 x double> %x0, i8* %x1, <4 x i64> %x2, i8 -1, i32 0)
320 %res2 = fadd <4 x double> %res, %res1
321 ret <4 x double> %res2
324 declare <8 x i32> @llvm.x86.avx512.gather3div4.di(<4 x i64>, i8*, <4 x i64>, i8, i32)
326 define <8 x i32>@test_int_x86_avx512_gather3div4_di(<4 x i64> %x0, i8* %x1, <4 x i64> %x2, i8 %x3) {
327 ; CHECK-LABEL: test_int_x86_avx512_gather3div4_di:
329 ; CHECK-NEXT: kmovb %esi, %k1
330 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
331 ; CHECK-NEXT: vpgatherqq (%rdi,%ymm1,8), %ymm2 {%k1}
332 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
333 ; CHECK-NEXT: vpgatherqq (%rdi,%ymm1,8), %ymm0 {%k1}
334 ; CHECK-NEXT: vpaddd %ymm0, %ymm2, %ymm0
336 %res = call <8 x i32> @llvm.x86.avx512.gather3div4.di(<4 x i64> %x0, i8* %x1, <4 x i64> %x2, i8 %x3, i32 8)
337 %res1 = call <8 x i32> @llvm.x86.avx512.gather3div4.di(<4 x i64> %x0, i8* %x1, <4 x i64> %x2, i8 -1, i32 8)
338 %res2 = add <8 x i32> %res, %res1
342 declare <4 x float> @llvm.x86.avx512.gather3div4.sf(<4 x float>, i8*, <2 x i64>, i8, i32)
344 define <4 x float>@test_int_x86_avx512_gather3div4_sf(<4 x float> %x0, i8* %x1, <2 x i64> %x2, i8 %x3) {
345 ; CHECK-LABEL: test_int_x86_avx512_gather3div4_sf:
347 ; CHECK-NEXT: kmovb %esi, %k1
348 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
349 ; CHECK-NEXT: vgatherqps (%rdi,%xmm1,4), %xmm2 {%k1}
350 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
351 ; CHECK-NEXT: vgatherqps (%rdi,%xmm1,0), %xmm0 {%k1}
352 ; CHECK-NEXT: vaddps %xmm0, %xmm2, %xmm0
354 %res = call <4 x float> @llvm.x86.avx512.gather3div4.sf(<4 x float> %x0, i8* %x1, <2 x i64> %x2, i8 %x3, i32 4)
355 %res1 = call <4 x float> @llvm.x86.avx512.gather3div4.sf(<4 x float> %x0, i8* %x1, <2 x i64> %x2, i8 -1, i32 0)
356 %res2 = fadd <4 x float> %res, %res1
357 ret <4 x float> %res2
360 declare <4 x i32> @llvm.x86.avx512.gather3div4.si(<4 x i32>, i8*, <2 x i64>, i8, i32)
362 define <4 x i32>@test_int_x86_avx512_gather3div4_si(<4 x i32> %x0, i8* %x1, <2 x i64> %x2, i8 %x3) {
363 ; CHECK-LABEL: test_int_x86_avx512_gather3div4_si:
365 ; CHECK-NEXT: kmovb %esi, %k1
366 ; CHECK-NEXT: kxnorw %k2, %k2, %k2
367 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
368 ; CHECK-NEXT: vpgatherqd (%rdi,%xmm1,4), %xmm2 {%k2}
369 ; CHECK-NEXT: vpgatherqd (%rdi,%xmm1,4), %xmm0 {%k1}
370 ; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0
372 %res = call <4 x i32> @llvm.x86.avx512.gather3div4.si(<4 x i32> %x0, i8* %x1, <2 x i64> %x2, i8 -1, i32 4)
373 %res1 = call <4 x i32> @llvm.x86.avx512.gather3div4.si(<4 x i32> %x0, i8* %x1, <2 x i64> %x2, i8 %x3, i32 4)
374 %res2 = add <4 x i32> %res, %res1
378 declare <4 x float> @llvm.x86.avx512.gather3div8.sf(<4 x float>, i8*, <4 x i64>, i8, i32)
380 define <4 x float>@test_int_x86_avx512_gather3div8_sf(<4 x float> %x0, i8* %x1, <4 x i64> %x2, i8 %x3) {
381 ; CHECK-LABEL: test_int_x86_avx512_gather3div8_sf:
383 ; CHECK-NEXT: kmovb %esi, %k1
384 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
385 ; CHECK-NEXT: vgatherqps (%rdi,%ymm1,4), %xmm2 {%k1}
386 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
387 ; CHECK-NEXT: vgatherqps (%rdi,%ymm1,0), %xmm0 {%k1}
388 ; CHECK-NEXT: vaddps %xmm0, %xmm2, %xmm0
390 %res = call <4 x float> @llvm.x86.avx512.gather3div8.sf(<4 x float> %x0, i8* %x1, <4 x i64> %x2, i8 %x3, i32 4)
391 %res1 = call <4 x float> @llvm.x86.avx512.gather3div8.sf(<4 x float> %x0, i8* %x1, <4 x i64> %x2, i8 -1, i32 0)
392 %res2 = fadd <4 x float> %res, %res1
393 ret <4 x float> %res2
396 declare <4 x i32> @llvm.x86.avx512.gather3div8.si(<4 x i32>, i8*, <4 x i64>, i8, i32)
398 define <4 x i32>@test_int_x86_avx512_gather3div8_si(<4 x i32> %x0, i8* %x1, <4 x i64> %x2, i8 %x3) {
399 ; CHECK-LABEL: test_int_x86_avx512_gather3div8_si:
401 ; CHECK-NEXT: kmovb %esi, %k1
402 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
403 ; CHECK-NEXT: kmovw %k1, %k2
404 ; CHECK-NEXT: vpgatherqd (%rdi,%ymm1,4), %xmm2 {%k2}
405 ; CHECK-NEXT: vpgatherqd (%rdi,%ymm1,2), %xmm0 {%k1}
406 ; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0
408 %res = call <4 x i32> @llvm.x86.avx512.gather3div8.si(<4 x i32> %x0, i8* %x1, <4 x i64> %x2, i8 %x3, i32 4)
409 %res1 = call <4 x i32> @llvm.x86.avx512.gather3div8.si(<4 x i32> %x0, i8* %x1, <4 x i64> %x2, i8 %x3, i32 2)
410 %res2 = add <4 x i32> %res, %res1
414 declare <2 x double> @llvm.x86.avx512.gather3siv2.df(<2 x double>, i8*, <4 x i32>, i8, i32)
416 define <2 x double>@test_int_x86_avx512_gather3siv2_df(<2 x double> %x0, i8* %x1, <4 x i32> %x2, i8 %x3) {
417 ; CHECK-LABEL: test_int_x86_avx512_gather3siv2_df:
419 ; CHECK-NEXT: kmovb %esi, %k1
420 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
421 ; CHECK-NEXT: vgatherdpd (%rdi,%xmm1,4), %xmm2 {%k1}
422 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
423 ; CHECK-NEXT: vgatherdpd (%rdi,%xmm1,0), %xmm0 {%k1}
424 ; CHECK-NEXT: vaddpd %xmm0, %xmm2, %xmm0
426 %res = call <2 x double> @llvm.x86.avx512.gather3siv2.df(<2 x double> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 4)
427 %res1 = call <2 x double> @llvm.x86.avx512.gather3siv2.df(<2 x double> %x0, i8* %x1, <4 x i32> %x2, i8 -1, i32 0)
428 %res2 = fadd <2 x double> %res, %res1
429 ret <2 x double> %res2
432 declare <4 x i32> @llvm.x86.avx512.gather3siv2.di(<2 x i64>, i8*, <4 x i32>, i8, i32)
434 define <4 x i32>@test_int_x86_avx512_gather3siv2_di(<2 x i64> %x0, i8* %x1, <4 x i32> %x2, i8 %x3) {
435 ; CHECK-LABEL: test_int_x86_avx512_gather3siv2_di:
437 ; CHECK-NEXT: kmovb %esi, %k1
438 ; CHECK-NEXT: vpgatherdq (%rdi,%xmm1,8), %xmm0 {%k1}
439 ; CHECK-NEXT: vpaddd %xmm0, %xmm0, %xmm0
441 %res = call <4 x i32> @llvm.x86.avx512.gather3siv2.di(<2 x i64> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 8)
442 %res1 = call <4 x i32> @llvm.x86.avx512.gather3siv2.di(<2 x i64> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 8)
443 %res2 = add <4 x i32> %res, %res1
447 declare <4 x double> @llvm.x86.avx512.gather3siv4.df(<4 x double>, i8*, <4 x i32>, i8, i32)
449 define <4 x double>@test_int_x86_avx512_gather3siv4_df(<4 x double> %x0, i8* %x1, <4 x i32> %x2, i8 %x3) {
450 ; CHECK-LABEL: test_int_x86_avx512_gather3siv4_df:
452 ; CHECK-NEXT: kmovb %esi, %k1
453 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
454 ; CHECK-NEXT: vgatherdpd (%rdi,%xmm1,4), %ymm2 {%k1}
455 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
456 ; CHECK-NEXT: vgatherdpd (%rdi,%xmm1,0), %ymm0 {%k1}
457 ; CHECK-NEXT: vaddpd %ymm0, %ymm2, %ymm0
459 %res = call <4 x double> @llvm.x86.avx512.gather3siv4.df(<4 x double> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 4)
460 %res1 = call <4 x double> @llvm.x86.avx512.gather3siv4.df(<4 x double> %x0, i8* %x1, <4 x i32> %x2, i8 -1, i32 0)
461 %res2 = fadd <4 x double> %res, %res1
462 ret <4 x double> %res2
465 declare <8 x i32> @llvm.x86.avx512.gather3siv4.di(<4 x i64>, i8*, <4 x i32>, i8, i32)
467 define <8 x i32>@test_int_x86_avx512_gather3siv4_di(<4 x i64> %x0, i8* %x1, <4 x i32> %x2, i8 %x3) {
468 ; CHECK-LABEL: test_int_x86_avx512_gather3siv4_di:
470 ; CHECK-NEXT: kmovb %esi, %k1
471 ; CHECK-NEXT: vpgatherdq (%rdi,%xmm1,8), %ymm0 {%k1}
472 ; CHECK-NEXT: vpaddd %ymm0, %ymm0, %ymm0
474 %res = call <8 x i32> @llvm.x86.avx512.gather3siv4.di(<4 x i64> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 8)
475 %res1 = call <8 x i32> @llvm.x86.avx512.gather3siv4.di(<4 x i64> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 8)
476 %res2 = add <8 x i32> %res, %res1
480 declare <4 x float> @llvm.x86.avx512.gather3siv4.sf(<4 x float>, i8*, <4 x i32>, i8, i32)
482 define <4 x float>@test_int_x86_avx512_gather3siv4_sf(<4 x float> %x0, i8* %x1, <4 x i32> %x2, i8 %x3) {
483 ; CHECK-LABEL: test_int_x86_avx512_gather3siv4_sf:
485 ; CHECK-NEXT: kmovb %esi, %k1
486 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
487 ; CHECK-NEXT: vgatherdps (%rdi,%xmm1,4), %xmm2 {%k1}
488 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
489 ; CHECK-NEXT: vgatherdps (%rdi,%xmm1,0), %xmm0 {%k1}
490 ; CHECK-NEXT: vaddps %xmm0, %xmm2, %xmm0
492 %res = call <4 x float> @llvm.x86.avx512.gather3siv4.sf(<4 x float> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 4)
493 %res1 = call <4 x float> @llvm.x86.avx512.gather3siv4.sf(<4 x float> %x0, i8* %x1, <4 x i32> %x2, i8 -1, i32 0)
494 %res2 = fadd <4 x float> %res, %res1
495 ret <4 x float> %res2
498 declare <4 x i32> @llvm.x86.avx512.gather3siv4.si(<4 x i32>, i8*, <4 x i32>, i8, i32)
500 define <4 x i32>@test_int_x86_avx512_gather3siv4_si(<4 x i32> %x0, i8* %x1, <4 x i32> %x2, i8 %x3) {
501 ; CHECK-LABEL: test_int_x86_avx512_gather3siv4_si:
503 ; CHECK-NEXT: kmovb %esi, %k1
504 ; CHECK-NEXT: kxnorw %k2, %k2, %k2
505 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
506 ; CHECK-NEXT: vpgatherdd (%rdi,%xmm1,4), %xmm2 {%k2}
507 ; CHECK-NEXT: vpgatherdd (%rdi,%xmm1,0), %xmm0 {%k1}
508 ; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0
510 %res = call <4 x i32> @llvm.x86.avx512.gather3siv4.si(<4 x i32> %x0, i8* %x1, <4 x i32> %x2, i8 -1, i32 4)
511 %res1 = call <4 x i32> @llvm.x86.avx512.gather3siv4.si(<4 x i32> %x0, i8* %x1, <4 x i32> %x2, i8 %x3, i32 0)
512 %res2 = add <4 x i32> %res, %res1
516 declare <8 x float> @llvm.x86.avx512.gather3siv8.sf(<8 x float>, i8*, <8 x i32>, i8, i32)
518 define <8 x float>@test_int_x86_avx512_gather3siv8_sf(<8 x float> %x0, i8* %x1, <8 x i32> %x2, i8 %x3) {
519 ; CHECK-LABEL: test_int_x86_avx512_gather3siv8_sf:
521 ; CHECK-NEXT: kmovb %esi, %k1
522 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
523 ; CHECK-NEXT: vgatherdps (%rdi,%ymm1,4), %ymm2 {%k1}
524 ; CHECK-NEXT: kxnorw %k1, %k1, %k1
525 ; CHECK-NEXT: vgatherdps (%rdi,%ymm1,0), %ymm0 {%k1}
526 ; CHECK-NEXT: vaddps %ymm0, %ymm2, %ymm0
528 %res = call <8 x float> @llvm.x86.avx512.gather3siv8.sf(<8 x float> %x0, i8* %x1, <8 x i32> %x2, i8 %x3, i32 4)
529 %res1 = call <8 x float> @llvm.x86.avx512.gather3siv8.sf(<8 x float> %x0, i8* %x1, <8 x i32> %x2, i8 -1, i32 0)
530 %res2 = fadd <8 x float> %res, %res1
531 ret <8 x float> %res2
534 declare <8 x i32> @llvm.x86.avx512.gather3siv8.si(<8 x i32>, i8*, <8 x i32>, i8, i32)
536 define <8 x i32>@test_int_x86_avx512_gather3siv8_si(<8 x i32> %x0, i8* %x1, <8 x i32> %x2, i8 %x3) {
537 ; CHECK-LABEL: test_int_x86_avx512_gather3siv8_si:
539 ; CHECK-NEXT: kmovb %esi, %k1
540 ; CHECK-NEXT: vmovaps %zmm0, %zmm2
541 ; CHECK-NEXT: kmovw %k1, %k2
542 ; CHECK-NEXT: vpgatherdd (%rdi,%ymm1,4), %ymm2 {%k2}
543 ; CHECK-NEXT: vpgatherdd (%rdi,%ymm1,0), %ymm0 {%k1}
544 ; CHECK-NEXT: vpaddd %ymm0, %ymm2, %ymm0
546 %res = call <8 x i32> @llvm.x86.avx512.gather3siv8.si(<8 x i32> %x0, i8* %x1, <8 x i32> %x2, i8 %x3, i32 4)
547 %res1 = call <8 x i32> @llvm.x86.avx512.gather3siv8.si(<8 x i32> %x0, i8* %x1, <8 x i32> %x2, i8 %x3, i32 0)
548 %res2 = add <8 x i32> %res, %res1