Commit | Line | Data |
---|---|---|
d6aab7a1 XG |
1 | # Check 64bit AVX512{BF16,VL} instructions |
2 | ||
3 | .allow_index_reg | |
4 | .text | |
5 | _start: | |
6 | vcvtne2ps2bf16 %ymm28, %ymm29, %ymm30 #AVX512{BF16,VL} | |
7 | vcvtne2ps2bf16 %xmm28, %xmm29, %xmm30 #AVX512{BF16,VL} | |
8 | vcvtne2ps2bf16 0x10000000(%rbp, %r14, 8), %ymm29, %ymm30{%k7} #AVX512{BF16,VL} MASK_ENABLING | |
9 | vcvtne2ps2bf16 (%r9){1to8}, %ymm29, %ymm30 #AVX512{BF16,VL} BROADCAST_EN | |
10 | vcvtne2ps2bf16 4064(%rcx), %ymm29, %ymm30 #AVX512{BF16,VL} Disp8 | |
11 | vcvtne2ps2bf16 -4096(%rdx){1to8}, %ymm29, %ymm30{%k7}{z} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
12 | vcvtne2ps2bf16 0x10000000(%rbp, %r14, 8), %xmm29, %xmm30{%k7} #AVX512{BF16,VL} MASK_ENABLING | |
13 | vcvtne2ps2bf16 (%r9){1to4}, %xmm29, %xmm30 #AVX512{BF16,VL} BROADCAST_EN | |
14 | vcvtne2ps2bf16 2032(%rcx), %xmm29, %xmm30 #AVX512{BF16,VL} Disp8 | |
15 | vcvtne2ps2bf16 -2048(%rdx){1to4}, %xmm29, %xmm28{%k7}{z} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
16 | vcvtneps2bf16 %xmm29, %xmm30 #AVX512{BF16,VL} | |
17 | vcvtneps2bf16 %ymm29, %xmm30 #AVX512{BF16,VL} | |
18 | vcvtneps2bf16x 0x10000000(%rbp, %r14, 8), %xmm30{%k7} #AVX512{BF16,VL} MASK_ENABLING | |
19 | vcvtneps2bf16 (%r9){1to4}, %xmm21 #AVX512{BF16,VL} BROADCAST_EN | |
20 | vcvtneps2bf16x 2032(%rcx), %xmm30 #AVX512{BF16,VL} Disp8 | |
21 | vcvtneps2bf16 -2048(%rdx){1to4}, %xmm29{%k7}{z} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
22 | vcvtneps2bf16 (%r9){1to8}, %xmm22 #AVX512{BF16,VL} BROADCAST_EN | |
23 | vcvtneps2bf16y 4064(%rcx), %xmm23 #AVX512{BF16,VL} Disp8 | |
24 | vcvtneps2bf16 -4096(%rdx){1to8}, %xmm27{%k7}{z} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
25 | vdpbf16ps %ymm28, %ymm29, %ymm30 #AVX512{BF16,VL} | |
26 | vdpbf16ps %xmm28, %xmm29, %xmm30 #AVX512{BF16,VL} | |
27 | vdpbf16ps 0x10000000(%rbp, %r14, 8), %ymm29, %ymm30{%k7} #AVX512{BF16,VL} MASK_ENABLING | |
28 | vdpbf16ps (%r9){1to8}, %ymm29, %ymm30 #AVX512{BF16,VL} BROADCAST_EN | |
29 | vdpbf16ps 4064(%rcx), %ymm29, %ymm30 #AVX512{BF16,VL} Disp8 | |
30 | vdpbf16ps -4096(%rdx){1to8}, %ymm29, %ymm30{%k7}{z} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
31 | vdpbf16ps 0x10000000(%rbp, %r14, 8), %xmm29, %xmm30{%k7} #AVX512{BF16,VL} MASK_ENABLING | |
32 | vdpbf16ps (%r9){1to4}, %xmm29, %xmm30 #AVX512{BF16,VL} BROADCAST_EN | |
33 | vdpbf16ps 2032(%rcx), %xmm29, %xmm30 #AVX512{BF16,VL} Disp8 | |
34 | vdpbf16ps -2048(%rdx){1to4}, %xmm29, %xmm30{%k7}{z} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
35 | ||
36 | .intel_syntax noprefix | |
37 | vcvtne2ps2bf16 ymm30, ymm29, ymm28 #AVX512{BF16,VL} | |
38 | vcvtne2ps2bf16 xmm30, xmm29, xmm28 #AVX512{BF16,VL} | |
39 | vcvtne2ps2bf16 ymm30{k7}, ymm29, YMMWORD PTR [rbp+r14*8+0x10000000] #AVX512{BF16,VL} MASK_ENABLING | |
40 | vcvtne2ps2bf16 ymm30, ymm29, DWORD PTR [r9]{1to8} #AVX512{BF16,VL} BROADCAST_EN | |
41 | vcvtne2ps2bf16 ymm30, ymm29, YMMWORD PTR [rcx+4064] #AVX512{BF16,VL} Disp8 | |
42 | vcvtne2ps2bf16 ymm30{k7}{z}, ymm29, DWORD PTR [rdx-4096]{1to8} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
43 | vcvtne2ps2bf16 xmm30{k7}, xmm29, XMMWORD PTR [rbp+r14*8+0x10000000] #AVX512{BF16,VL} MASK_ENABLING | |
44 | vcvtne2ps2bf16 xmm30, xmm29, DWORD PTR [r9]{1to4} #AVX512{BF16,VL} BROADCAST_EN | |
45 | vcvtne2ps2bf16 xmm30, xmm29, XMMWORD PTR [rcx+2032] #AVX512{BF16,VL} Disp8 | |
46 | vcvtne2ps2bf16 xmm30{k7}{z}, xmm29, DWORD PTR [rdx-2048]{1to4} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
47 | vcvtneps2bf16 xmm30, xmm29 #AVX512{BF16,VL} | |
48 | vcvtneps2bf16 xmm30, ymm29 #AVX512{BF16,VL} | |
49 | vcvtneps2bf16 xmm30{k7}, XMMWORD PTR [rbp+r14*8+0x10000000] #AVX512{BF16,VL} MASK_ENABLING | |
50 | vcvtneps2bf16 xmm25, DWORD PTR [r9]{1to4} #AVX512{BF16,VL} BROADCAST_EN | |
51 | vcvtneps2bf16 xmm30, XMMWORD PTR [rcx+2032] #AVX512{BF16,VL} Disp8 | |
52 | vcvtneps2bf16 xmm30{k7}{z}, DWORD PTR [rdx-2048]{1to4} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
53 | vcvtneps2bf16 xmm24, DWORD PTR [r9]{1to8} #AVX512{BF16,VL} BROADCAST_EN | |
54 | vcvtneps2bf16 xmm30, YMMWORD PTR [rcx+4064] #AVX512{BF16,VL} Disp8 | |
55 | vcvtneps2bf16 xmm30{k7}{z}, DWORD PTR [rdx-4096]{1to8} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
56 | vdpbf16ps ymm30, ymm29, ymm28 #AVX512{BF16,VL} | |
57 | vdpbf16ps xmm30, xmm29, xmm28 #AVX512{BF16,VL} | |
58 | vdpbf16ps ymm30{k7}, ymm29, YMMWORD PTR [rbp+r14*8+0x10000000] #AVX512{BF16,VL} MASK_ENABLING | |
59 | vdpbf16ps ymm30, ymm29, DWORD PTR [r9]{1to8} #AVX512{BF16,VL} BROADCAST_EN | |
60 | vdpbf16ps ymm30, ymm29, YMMWORD PTR [rcx+4064] #AVX512{BF16,VL} Disp8 | |
61 | vdpbf16ps ymm30{k7}{z}, ymm29, DWORD PTR [rdx-4096]{1to8} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL | |
62 | vdpbf16ps xmm30{k7}, xmm29, XMMWORD PTR [rbp+r14*8+0x10000000] #AVX512{BF16,VL} MASK_ENABLING | |
63 | vdpbf16ps xmm30, xmm29, DWORD PTR [r9]{1to4} #AVX512{BF16,VL} BROADCAST_EN | |
64 | vdpbf16ps xmm30, xmm29, XMMWORD PTR [rcx+2032] #AVX512{BF16,VL} Disp8 | |
65 | vdpbf16ps xmm30{k7}{z}, xmm29, DWORD PTR [rdx-2048]{1to4} #AVX512{BF16,VL} Disp8 BROADCAST_EN MASK_ENABLING ZEROCTL |