| /aosp_15_r20/external/XNNPACK/test/ |
| H A D | qs8-gemm-minmax-fp32.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qs8_gemm_minmax_fp32_ukernel_1x1c4__armsimd32
9 k-block: 4
10 - name: xnn_qs8_gemm_minmax_fp32_ukernel_1x2c4__armsimd32
12 k-block: 4
13 - name: xnn_qs8_gemm_minmax_fp32_ukernel_2x1c4__armsimd32
15 k-block: 4
16 - name: xnn_qs8_gemm_minmax_fp32_ukernel_2x2c4__armsimd32
18 k-block: 4
21 - name: xnn_qs8_gemm_minmax_fp32_ukernel_1x8c2__neon_mlal_dup
[all …]
|
| H A D | qc8-gemm-minmax-fp32.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qc8_gemm_minmax_fp32_ukernel_1x8__aarch32_neon_mlal_lane_cortex_a7
9 k-block: 8
11 - name: xnn_qc8_gemm_minmax_fp32_ukernel_1x8__aarch32_neon_mlal_lane_prfm_cortex_a7
13 k-block: 8
15 - name: xnn_qc8_gemm_minmax_fp32_ukernel_1x8__aarch32_neonv8_mlal_lane_cortex_a35
17 k-block: 8
19 - name: xnn_qc8_gemm_minmax_fp32_ukernel_1x8__aarch32_neonv8_mlal_lane_prfm_cortex_a35
21 k-block: 8
23 - name: xnn_qc8_gemm_minmax_fp32_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a7
[all …]
|
| H A D | qc8-igemm-minmax-fp32.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qc8_igemm_minmax_fp32_ukernel_1x8__aarch32_neon_mlal_lane_cortex_a7
9 k-block: 8
11 - name: xnn_qc8_igemm_minmax_fp32_ukernel_1x8__aarch32_neon_mlal_lane_prfm_cortex_a7
13 k-block: 8
15 - name: xnn_qc8_igemm_minmax_fp32_ukernel_1x8__aarch32_neonv8_mlal_lane_cortex_a35
17 k-block: 8
19 - name: xnn_qc8_igemm_minmax_fp32_ukernel_1x8__aarch32_neonv8_mlal_lane_prfm_cortex_a35
21 k-block: 8
23 - name: xnn_qc8_igemm_minmax_fp32_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a53
[all …]
|
| H A D | qs8-gemm-minmax-rndnu.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qs8_gemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_cortex_a7
9 k-block: 8
11 - name: xnn_qs8_gemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_prfm_cortex_a7
13 k-block: 8
15 - name: xnn_qs8_gemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a7
17 k-block: 8
19 - name: xnn_qs8_gemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a53
21 k-block: 8
23 - name: xnn_qs8_gemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld64
[all …]
|
| H A D | qs8-igemm-minmax-rndnu.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_cortex_a7
9 k-block: 8
11 - name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_prfm_cortex_a7
13 k-block: 8
15 - name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a53
17 k-block: 8
19 - name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld64
21 k-block: 8
23 - name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prfm_cortex_a53
[all …]
|
| H A D | qs8-igemm-minmax-fp32.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qs8_igemm_minmax_fp32_ukernel_1x1c4__armsimd32
9 k-block: 4
10 - name: xnn_qs8_igemm_minmax_fp32_ukernel_1x2c4__armsimd32
12 k-block: 4
13 - name: xnn_qs8_igemm_minmax_fp32_ukernel_2x1c4__armsimd32
15 k-block: 4
16 - name: xnn_qs8_igemm_minmax_fp32_ukernel_2x2c4__armsimd32
18 k-block: 4
21 - name: xnn_qs8_igemm_minmax_fp32_ukernel_1x8c2__neon_mlal_dup
[all …]
|
| H A D | f32-gemm-minmax.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_gemm_minmax_ukernel_4x4__aarch32_vfp_ld64
9 k-block: 2
11 - name: xnn_f32_gemm_minmax_ukernel_4x8__aarch32_neon_cortex_a7
13 k-block: 2
15 - name: xnn_f32_gemm_minmax_ukernel_4x8__aarch32_neon_cortex_a53
17 k-block: 4
20 - name: xnn_f32_gemm_minmax_ukernel_4x8__aarch32_neon_cortex_a55
22 k-block: 4
25 - name: xnn_f32_gemm_minmax_ukernel_4x8__aarch32_neon_cortex_a75
[all …]
|
| H A D | f32-igemm-minmax.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch32_neon_cortex_a7
9 k-block: 2
11 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch32_neon_cortex_a53
13 k-block: 4
16 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch32_neon_cortex_a55
18 k-block: 4
21 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch32_neon_cortex_a75
23 k-block: 4
26 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch32_neon_ld64
[all …]
|
| H A D | qu8-gemm-minmax-fp32.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qu8_gemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_cortex_a55
9 k-block: 16
11 - name: xnn_qu8_gemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_ld128
13 k-block: 16
17 - name: xnn_qu8_gemm_minmax_fp32_ukernel_1x1c4__armsimd32
19 k-block: 4
20 - name: xnn_qu8_gemm_minmax_fp32_ukernel_1x2c4__armsimd32
22 k-block: 4
23 - name: xnn_qu8_gemm_minmax_fp32_ukernel_2x1c4__armsimd32
[all …]
|
| H A D | qu8-igemm-minmax-fp32.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qu8_igemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_cortex_a55
9 k-block: 16
11 - name: xnn_qu8_igemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_ld128
13 k-block: 16
17 - name: xnn_qu8_igemm_minmax_fp32_ukernel_1x1c4__armsimd32
19 k-block: 4
20 - name: xnn_qu8_igemm_minmax_fp32_ukernel_1x2c4__armsimd32
22 k-block: 4
23 - name: xnn_qu8_igemm_minmax_fp32_ukernel_2x1c4__armsimd32
[all …]
|
| H A D | f32-gemminc-minmax.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_cortex_a53
9 k-block: 8
12 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_cortex_a75
14 k-block: 8
17 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_ld64
19 k-block: 2
21 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_prfm_cortex_a75
23 k-block: 8
26 - name: xnn_f32_gemminc_minmax_ukernel_1x12__aarch64_neonfma_cortex_a53
[all …]
|
| H A D | f32-spmm-minmax.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_spmm_minmax_ukernel_4x1__neon
9 k-block: 1
10 - name: xnn_f32_spmm_minmax_ukernel_4x1__neon_pipelined
12 k-block: 1
13 - name: xnn_f32_spmm_minmax_ukernel_4x1__neon_x2
15 k-block: 2
16 - name: xnn_f32_spmm_minmax_ukernel_4x1__neonfma
18 k-block: 1
19 - name: xnn_f32_spmm_minmax_ukernel_4x1__neonfma_pipelined
[all …]
|
| H A D | qu8-gemm-minmax-rndnu.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qu8_gemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_cortex_a7
9 k-block: 8
11 - name: xnn_qu8_gemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_prfm_cortex_a7
13 k-block: 8
15 - name: xnn_qu8_gemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a7
17 k-block: 8
19 - name: xnn_qu8_gemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a53
21 k-block: 8
23 - name: xnn_qu8_gemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld64
[all …]
|
| H A D | qu8-igemm-minmax-rndnu.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_qu8_igemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_cortex_a7
9 k-block: 8
11 - name: xnn_qu8_igemm_minmax_rndnu_ukernel_1x8__aarch32_neon_mlal_lane_prfm_cortex_a7
13 k-block: 8
15 - name: xnn_qu8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_cortex_a53
17 k-block: 8
19 - name: xnn_qu8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld64
21 k-block: 8
23 - name: xnn_qu8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prfm_cortex_a53
[all …]
|
| H A D | f32-igemm-relu.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_igemm_relu_ukernel_1x8__wasmsimd_loadsplat
8 k-block: 1
9 - name: xnn_f32_igemm_relu_ukernel_1x8__wasmsimd_splat
10 k-block: 4
11 - name: xnn_f32_igemm_relu_ukernel_1x8s4__wasmsimd
12 k-block: 4
13 - name: xnn_f32_igemm_relu_ukernel_3x8__wasmsimd_loadsplat
14 k-block: 1
15 - name: xnn_f32_igemm_relu_ukernel_3x8__wasmsimd_splat
[all …]
|
| H A D | f32-gemm-relu.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_gemm_relu_ukernel_1x8__wasmsimd_loadsplat
8 k-block: 1
9 - name: xnn_f32_gemm_relu_ukernel_1x8__wasmsimd_splat
10 k-block: 4
11 - name: xnn_f32_gemm_relu_ukernel_1x8s4__wasmsimd
12 k-block: 4
13 - name: xnn_f32_gemm_relu_ukernel_3x8__wasmsimd_loadsplat
14 k-block: 1
15 - name: xnn_f32_gemm_relu_ukernel_3x8__wasmsimd_splat
[all …]
|
| H A D | f32-gemm.yaml | 3 # This source code is licensed under the BSD-style license found in the
7 - name: xnn_f32_gemm_ukernel_4x4__aarch32_vfp_ld64
8 k-block: 2
11 - name: xnn_f32_gemm_ukernel_1x8__wasmsimd_loadsplat
12 k-block: 1
13 - name: xnn_f32_gemm_ukernel_1x8__wasmsimd_splat
14 k-block: 4
15 - name: xnn_f32_gemm_ukernel_1x8s4__wasmsimd
16 k-block: 4
17 - name: xnn_f32_gemm_ukernel_3x8__wasmsimd_loadsplat
[all …]
|
| /aosp_15_r20/external/boringssl/src/gen/bcm/ |
| H A D | aesv8-gcm-armv8-linux.S | 1 // This file is generated from a similarly-named Perl script in the BoringSSL
10 .arch armv8-a+crypto
18 stp x29, x30, [sp, #-128]!
32 ldr q31, [x19, #-16] // load round N-1 keys
37 … // special case vector load initial counter so we can start first AES block as quickly as poss…
38 sub x5, x5, #1 // byte_len - 1
44 fmov d2, x10 // CTR block 2
47 fmov d1, x10 // CTR block 1
49 aesmc v0.16b, v0.16b // AES block 0 - round 0
51 rev w9, w12 // CTR block 1
[all …]
|
| H A D | aesv8-gcm-armv8-win.S | 1 // This file is generated from a similarly-named Perl script in the BoringSSL
10 .arch armv8-a+crypto
20 stp x29, x30, [sp, #-128]!
34 ldr q31, [x19, #-16] // load round N-1 keys
39 … // special case vector load initial counter so we can start first AES block as quickly as poss…
40 sub x5, x5, #1 // byte_len - 1
46 fmov d2, x10 // CTR block 2
49 fmov d1, x10 // CTR block 1
51 aesmc v0.16b, v0.16b // AES block 0 - round 0
53 rev w9, w12 // CTR block 1
[all …]
|
| H A D | aesv8-gcm-armv8-apple.S | 1 // This file is generated from a similarly-named Perl script in the BoringSSL
18 stp x29, x30, [sp, #-128]!
32 ldr q31, [x19, #-16] // load round N-1 keys
37 … // special case vector load initial counter so we can start first AES block as quickly as poss…
38 sub x5, x5, #1 // byte_len - 1
44 fmov d2, x10 // CTR block 2
47 fmov d1, x10 // CTR block 1
49 aesmc v0.16b, v0.16b // AES block 0 - round 0
51 rev w9, w12 // CTR block 1
52 fmov d3, x10 // CTR block 3
[all …]
|
| /aosp_15_r20/external/cronet/third_party/boringssl/src/gen/bcm/ |
| H A D | aesv8-gcm-armv8-linux.S | 1 // This file is generated from a similarly-named Perl script in the BoringSSL
10 .arch armv8-a+crypto
18 stp x29, x30, [sp, #-128]!
32 ldr q31, [x19, #-16] // load round N-1 keys
37 … // special case vector load initial counter so we can start first AES block as quickly as poss…
38 sub x5, x5, #1 // byte_len - 1
44 fmov d2, x10 // CTR block 2
47 fmov d1, x10 // CTR block 1
49 aesmc v0.16b, v0.16b // AES block 0 - round 0
51 rev w9, w12 // CTR block 1
[all …]
|
| H A D | aesv8-gcm-armv8-win.S | 1 // This file is generated from a similarly-named Perl script in the BoringSSL
10 .arch armv8-a+crypto
20 stp x29, x30, [sp, #-128]!
34 ldr q31, [x19, #-16] // load round N-1 keys
39 … // special case vector load initial counter so we can start first AES block as quickly as poss…
40 sub x5, x5, #1 // byte_len - 1
46 fmov d2, x10 // CTR block 2
49 fmov d1, x10 // CTR block 1
51 aesmc v0.16b, v0.16b // AES block 0 - round 0
53 rev w9, w12 // CTR block 1
[all …]
|
| H A D | aesv8-gcm-armv8-apple.S | 1 // This file is generated from a similarly-named Perl script in the BoringSSL
18 stp x29, x30, [sp, #-128]!
32 ldr q31, [x19, #-16] // load round N-1 keys
37 … // special case vector load initial counter so we can start first AES block as quickly as poss…
38 sub x5, x5, #1 // byte_len - 1
44 fmov d2, x10 // CTR block 2
47 fmov d1, x10 // CTR block 1
49 aesmc v0.16b, v0.16b // AES block 0 - round 0
51 rev w9, w12 // CTR block 1
52 fmov d3, x10 // CTR block 3
[all …]
|
| /aosp_15_r20/external/cronet/third_party/boringssl/src/crypto/fipsmodule/modes/asm/ |
| H A D | aesv8-gcm-armv8.pl | 19 # derived from https://github.com/ARM-software/AArch64cryptolib, original
23 # Approach - assume we don't want to reload constants, so reserve ~half of
34 # | CTR block 4k+8 | AES block 4k+4 | GHASH block 4k+0 |
37 # | CTR block 4k+9 | AES block 4k+5 | GHASH block 4k+1 |
40 # | CTR block 4k+10| AES block 4k+6 | GHASH block 4k+2 |
43 # | CTR block 4k+11| AES block 4k+7 | GHASH block 4k+3 |
50 # result for GHASH 4k+0
53 # EOR res_curr (4k+0), res_curr (4k+0), low_acc
55 # CTR block: Increment and byte reverse counter in scalar registers and transfer
65 # AES block:
[all …]
|
| /aosp_15_r20/external/boringssl/src/crypto/fipsmodule/modes/asm/ |
| H A D | aesv8-gcm-armv8.pl | 19 # derived from https://github.com/ARM-software/AArch64cryptolib, original
23 # Approach - assume we don't want to reload constants, so reserve ~half of
34 # | CTR block 4k+8 | AES block 4k+4 | GHASH block 4k+0 |
37 # | CTR block 4k+9 | AES block 4k+5 | GHASH block 4k+1 |
40 # | CTR block 4k+10| AES block 4k+6 | GHASH block 4k+2 |
43 # | CTR block 4k+11| AES block 4k+7 | GHASH block 4k+3 |
50 # result for GHASH 4k+0
53 # EOR res_curr (4k+0), res_curr (4k+0), low_acc
55 # CTR block: Increment and byte reverse counter in scalar registers and transfer
65 # AES block:
[all …]
|