| /external/XNNPACK/src/f32-argmaxpool/ |
| D | 9p8x-wasmsimd-c4.c | 75 v128_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4() local
78 const v128_t vm1 = wasm_f32x4_gt(vi1, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
79 vmax = wasm_v128_bitselect(vi1, vmax, vm1); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
82 const v128_t vm2 = wasm_f32x4_gt(vi2, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
83 vmax = wasm_v128_bitselect(vi2, vmax, vm2); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
86 const v128_t vm3 = wasm_f32x4_gt(vi3, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
87 vmax = wasm_v128_bitselect(vi3, vmax, vm3); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
90 const v128_t vm4 = wasm_f32x4_gt(vi4, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
91 vmax = wasm_v128_bitselect(vi4, vmax, vm4); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
94 const v128_t vm5 = wasm_f32x4_gt(vi5, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
[all …]
|
| D | 9p8x-neon-c4.c | 66 float32x4_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4() local
69 const uint32x4_t vm1 = vcgtq_f32(vi1, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
70 vmax = vbslq_f32(vm1, vi1, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
73 const uint32x4_t vm2 = vcgtq_f32(vi2, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
74 vmax = vbslq_f32(vm2, vi2, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
77 const uint32x4_t vm3 = vcgtq_f32(vi3, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
78 vmax = vbslq_f32(vm3, vi3, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
81 const uint32x4_t vm4 = vcgtq_f32(vi4, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
82 vmax = vbslq_f32(vm4, vi4, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
85 const uint32x4_t vm5 = vcgtq_f32(vi5, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
[all …]
|
| D | 9p8x-scalar-c1.c | 66 float vmax = vi0; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1() local
69 if (vi1 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
70 vmax = vi1; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
74 if (vi2 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
75 vmax = vi2; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
79 if (vi3 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
80 vmax = vi3; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
84 if (vi4 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
85 vmax = vi4; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
89 if (vi5 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
[all …]
|
| D | 9x-wasmsimd-c4.c | 95 v128_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4() local
98 const v128_t vm1 = wasm_f32x4_gt(vi1, vmax); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
99 vmax = wasm_v128_bitselect(vi1, vmax, vm1); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
102 const v128_t vm2 = wasm_f32x4_gt(vi2, vmax); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
103 vmax = wasm_v128_bitselect(vi2, vmax, vm2); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
106 const v128_t vm3 = wasm_f32x4_gt(vi3, vmax); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
107 vmax = wasm_v128_bitselect(vi3, vmax, vm3); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
110 const v128_t vm4 = wasm_f32x4_gt(vi4, vmax); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
111 vmax = wasm_v128_bitselect(vi4, vmax, vm4); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
114 const v128_t vm5 = wasm_f32x4_gt(vi5, vmax); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
[all …]
|
| D | 9p8x-sse2-c4.c | 75 __m128 vmax = vi0; in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4() local
78 const __m128i vm1 = _mm_castps_si128(_mm_cmpgt_ps(vi1, vmax)); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
79 vmax = _mm_max_ps(vi1, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
82 const __m128i vm2 = _mm_castps_si128(_mm_cmpgt_ps(vi2, vmax)); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
83 vmax = _mm_max_ps(vi2, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
86 const __m128i vm3 = _mm_castps_si128(_mm_cmpgt_ps(vi3, vmax)); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
87 vmax = _mm_max_ps(vi3, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
90 const __m128i vm4 = _mm_castps_si128(_mm_cmpgt_ps(vi4, vmax)); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
91 vmax = _mm_max_ps(vi4, vmax); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
94 const __m128i vm5 = _mm_castps_si128(_mm_cmpgt_ps(vi5, vmax)); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
[all …]
|
| D | 9x-neon-c4.c | 85 float32x4_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_9x__neon_c4() local
88 const uint32x4_t vm1 = vcgtq_f32(vi1, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
89 vmax = vbslq_f32(vm1, vi1, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
92 const uint32x4_t vm2 = vcgtq_f32(vi2, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
93 vmax = vbslq_f32(vm2, vi2, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
96 const uint32x4_t vm3 = vcgtq_f32(vi3, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
97 vmax = vbslq_f32(vm3, vi3, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
100 const uint32x4_t vm4 = vcgtq_f32(vi4, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
101 vmax = vbslq_f32(vm4, vi4, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
104 const uint32x4_t vm5 = vcgtq_f32(vi5, vmax); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
[all …]
|
| D | 9x-sse2-c4.c | 94 __m128 vmax = vi0; in xnn_f32_argmaxpool_ukernel_9x__sse2_c4() local
97 const __m128i vm1 = _mm_castps_si128(_mm_cmpgt_ps(vi1, vmax)); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
98 vmax = _mm_max_ps(vi1, vmax); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
101 const __m128i vm2 = _mm_castps_si128(_mm_cmpgt_ps(vi2, vmax)); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
102 vmax = _mm_max_ps(vi2, vmax); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
105 const __m128i vm3 = _mm_castps_si128(_mm_cmpgt_ps(vi3, vmax)); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
106 vmax = _mm_max_ps(vi3, vmax); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
109 const __m128i vm4 = _mm_castps_si128(_mm_cmpgt_ps(vi4, vmax)); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
110 vmax = _mm_max_ps(vi4, vmax); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
113 const __m128i vm5 = _mm_castps_si128(_mm_cmpgt_ps(vi5, vmax)); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
[all …]
|
| D | 4x-wasmsimd-c4.c | 60 v128_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4() local
63 const v128_t vm1 = wasm_f32x4_gt(vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
64 vmax = wasm_v128_bitselect(vi1, vmax, vm1); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
67 const v128_t vm2 = wasm_f32x4_gt(vi2, vmax); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
68 vmax = wasm_v128_bitselect(vi2, vmax, vm2); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
71 const v128_t vm3 = wasm_f32x4_gt(vi3, vmax); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
72 vmax = wasm_v128_bitselect(vi3, vmax, vm3); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
75 wasm_v128_store(output, vmax); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
86 v128_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4() local
89 const v128_t vm1 = wasm_f32x4_gt(vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
[all …]
|
| D | 4x-sse2-c4.c | 59 __m128 vmax = vi0; in xnn_f32_argmaxpool_ukernel_4x__sse2_c4() local
62 const __m128i vm1 = _mm_castps_si128(_mm_cmpgt_ps(vi1, vmax)); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
63 vmax = _mm_max_ps(vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
66 const __m128i vm2 = _mm_castps_si128(_mm_cmpgt_ps(vi2, vmax)); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
67 vmax = _mm_max_ps(vi2, vmax); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
70 const __m128i vm3 = _mm_castps_si128(_mm_cmpgt_ps(vi3, vmax)); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
71 vmax = _mm_max_ps(vi3, vmax); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
74 _mm_storeu_ps(output, vmax); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
85 __m128 vmax = vi0; in xnn_f32_argmaxpool_ukernel_4x__sse2_c4() local
88 const __m128i vm1 = _mm_castps_si128(_mm_cmpgt_ps(vi1, vmax)); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
[all …]
|
| D | 4x-neon-c4.c | 55 float32x4_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_4x__neon_c4() local
58 const uint32x4_t vm1 = vcgtq_f32(vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
59 vmax = vbslq_f32(vm1, vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
62 const uint32x4_t vm2 = vcgtq_f32(vi2, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
63 vmax = vbslq_f32(vm2, vi2, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
66 const uint32x4_t vm3 = vcgtq_f32(vi3, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
67 vmax = vbslq_f32(vm3, vi3, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
70 vst1q_f32(output, vmax); output += 4; in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
79 float32x4_t vmax = vi0; in xnn_f32_argmaxpool_ukernel_4x__neon_c4() local
82 const uint32x4_t vm1 = vcgtq_f32(vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
[all …]
|
| D | 9x-scalar-c1.c | 84 float vmax = vi0; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1() local
87 if (vi1 > vmax) { in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
88 vmax = vi1; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
92 if (vi2 > vmax) { in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
93 vmax = vi2; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
97 if (vi3 > vmax) { in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
98 vmax = vi3; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
102 if (vi4 > vmax) { in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
103 vmax = vi4; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
107 if (vi5 > vmax) { in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
[all …]
|
| /external/XNNPACK/src/u8-rmax/ |
| D | sse2.c | 24 __m128i vmax = _mm_setzero_si128(); in xnn_u8_rmax_ukernel__sse2() local
28 vmax = _mm_max_epu8(vmax, vx); in xnn_u8_rmax_ukernel__sse2()
35 vmax = _mm_max_epu8(vmax, vx); in xnn_u8_rmax_ukernel__sse2()
37 vmax = _mm_max_epu8(vmax, _mm_unpackhi_epi64(vmax, vmax)); in xnn_u8_rmax_ukernel__sse2()
38 vmax = _mm_max_epu8(vmax, _mm_srli_epi64(vmax, 32)); in xnn_u8_rmax_ukernel__sse2()
39 vmax = _mm_max_epu8(vmax, _mm_srli_epi32(vmax, 16)); in xnn_u8_rmax_ukernel__sse2()
40 vmax = _mm_max_epu8(vmax, _mm_srli_epi16(vmax, 8)); in xnn_u8_rmax_ukernel__sse2()
41 *y = (uint8_t) _mm_cvtsi128_si32(vmax); in xnn_u8_rmax_ukernel__sse2()
43 uint8_t vmax = 0; in xnn_u8_rmax_ukernel__sse2() local
46 vmax = vx > vmax ? vx : vmax; in xnn_u8_rmax_ukernel__sse2()
[all …]
|
| D | neon.c | 24 uint8x16_t vmax = vmovq_n_u8(0); in xnn_u8_rmax_ukernel__neon() local
27 vmax = vmaxq_u8(vmax, vx); in xnn_u8_rmax_ukernel__neon()
34 vmax = vmaxq_u8(vmax, vx); in xnn_u8_rmax_ukernel__neon()
36 uint8x8_t vmax8 = vmax_u8(vget_low_u8(vmax), vget_high_u8(vmax)); in xnn_u8_rmax_ukernel__neon()
42 uint8x8_t vmax = vmov_n_u8(0); in xnn_u8_rmax_ukernel__neon() local
45 vmax = vmax_u8(vmax, vx); in xnn_u8_rmax_ukernel__neon()
47 vst1_lane_u8(y, vmax, 0); in xnn_u8_rmax_ukernel__neon()
|
| /external/llvm-project/llvm/test/MC/ARM/ |
| D | neont2-minmax-encoding.s | 5 vmax.s8 d1, d2, d3
6 vmax.s16 d4, d5, d6
7 vmax.s32 d7, d8, d9
8 vmax.u8 d10, d11, d12
9 vmax.u16 d13, d14, d15
10 vmax.u32 d16, d17, d18
11 vmax.f32 d19, d20, d21
13 vmax.s8 d2, d3
14 vmax.s16 d5, d6
15 vmax.s32 d8, d9
[all …]
|
| D | neon-minmax-encoding.s | 3 vmax.s8 d1, d2, d3
4 vmax.s16 d4, d5, d6
5 vmax.s32 d7, d8, d9
6 vmax.u8 d10, d11, d12
7 vmax.u16 d13, d14, d15
8 vmax.u32 d16, d17, d18
9 vmax.f32 d19, d20, d21
11 vmax.s8 d2, d3
12 vmax.s16 d5, d6
13 vmax.s32 d8, d9
[all …]
|
| /external/llvm/test/MC/ARM/ |
| D | neon-minmax-encoding.s | 3 vmax.s8 d1, d2, d3
4 vmax.s16 d4, d5, d6
5 vmax.s32 d7, d8, d9
6 vmax.u8 d10, d11, d12
7 vmax.u16 d13, d14, d15
8 vmax.u32 d16, d17, d18
9 vmax.f32 d19, d20, d21
11 vmax.s8 d2, d3
12 vmax.s16 d5, d6
13 vmax.s32 d8, d9
[all …]
|
| D | neont2-minmax-encoding.s | 5 vmax.s8 d1, d2, d3
6 vmax.s16 d4, d5, d6
7 vmax.s32 d7, d8, d9
8 vmax.u8 d10, d11, d12
9 vmax.u16 d13, d14, d15
10 vmax.u32 d16, d17, d18
11 vmax.f32 d19, d20, d21
13 vmax.s8 d2, d3
14 vmax.s16 d5, d6
15 vmax.s32 d8, d9
[all …]
|
| /external/capstone/suite/MC/ARM/ |
| D | neon-minmax-encoding.s.cs | 2 0x03,0x16,0x02,0xf2 = vmax.s8 d1, d2, d3
3 0x06,0x46,0x15,0xf2 = vmax.s16 d4, d5, d6
4 0x09,0x76,0x28,0xf2 = vmax.s32 d7, d8, d9
5 0x0c,0xa6,0x0b,0xf3 = vmax.u8 d10, d11, d12
6 0x0f,0xd6,0x1e,0xf3 = vmax.u16 d13, d14, d15
7 0xa2,0x06,0x61,0xf3 = vmax.u32 d16, d17, d18
8 0xa5,0x3f,0x44,0xf2 = vmax.f32 d19, d20, d21
9 0x03,0x26,0x02,0xf2 = vmax.s8 d2, d2, d3
10 0x06,0x56,0x15,0xf2 = vmax.s16 d5, d5, d6
11 0x09,0x86,0x28,0xf2 = vmax.s32 d8, d8, d9
[all …]
|
| D | neont2-minmax-encoding.s.cs | 2 0x02,0xef,0x03,0x16 = vmax.s8 d1, d2, d3
3 0x15,0xef,0x06,0x46 = vmax.s16 d4, d5, d6
4 0x28,0xef,0x09,0x76 = vmax.s32 d7, d8, d9
5 0x0b,0xff,0x0c,0xa6 = vmax.u8 d10, d11, d12
6 0x1e,0xff,0x0f,0xd6 = vmax.u16 d13, d14, d15
7 0x61,0xff,0xa2,0x06 = vmax.u32 d16, d17, d18
8 0x44,0xef,0xa5,0x3f = vmax.f32 d19, d20, d21
9 0x02,0xef,0x03,0x26 = vmax.s8 d2, d2, d3
10 0x15,0xef,0x06,0x56 = vmax.s16 d5, d5, d6
11 0x28,0xef,0x09,0x86 = vmax.s32 d8, d8, d9
[all …]
|
| /external/XNNPACK/src/f32-spmm/gen/ |
| D | 8x4-minmax-scalar.c | 32 const float vmax = params->scalar.max; in xnn_f32_spmm_minmax_ukernel_8x4__scalar() local
123 float vout0x0 = math_min_f32(vacc0x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
124 float vout1x0 = math_min_f32(vacc1x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
125 float vout2x0 = math_min_f32(vacc2x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
126 float vout3x0 = math_min_f32(vacc3x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
127 float vout4x0 = math_min_f32(vacc4x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
128 float vout5x0 = math_min_f32(vacc5x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
129 float vout6x0 = math_min_f32(vacc6x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
130 float vout7x0 = math_min_f32(vacc7x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
131 float vout0x1 = math_min_f32(vacc0x1, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
[all …]
|
| D | 8x2-minmax-scalar.c | 32 const float vmax = params->scalar.max; in xnn_f32_spmm_minmax_ukernel_8x2__scalar() local
89 float vout0x0 = math_min_f32(vacc0x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
90 float vout1x0 = math_min_f32(vacc1x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
91 float vout2x0 = math_min_f32(vacc2x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
92 float vout3x0 = math_min_f32(vacc3x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
93 float vout4x0 = math_min_f32(vacc4x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
94 float vout5x0 = math_min_f32(vacc5x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
95 float vout6x0 = math_min_f32(vacc6x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
96 float vout7x0 = math_min_f32(vacc7x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
97 float vout0x1 = math_min_f32(vacc0x1, vmax); in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
[all …]
|
| D | 32x1-minmax-wasmsimd-x86.c | 33 const v128_t vmax = wasm_v32x4_load_splat(¶ms->scalar.max); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86() local
73 v128_t vout0123 = wasm_v128_bitselect(vacc0123, vmax, wasm_f32x4_le(vacc0123, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
74 v128_t vout4567 = wasm_v128_bitselect(vacc4567, vmax, wasm_f32x4_le(vacc4567, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
75 v128_t vout89AB = wasm_v128_bitselect(vacc89AB, vmax, wasm_f32x4_le(vacc89AB, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
76 v128_t voutCDEF = wasm_v128_bitselect(vaccCDEF, vmax, wasm_f32x4_le(vaccCDEF, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
77 v128_t voutGHIJ = wasm_v128_bitselect(vaccGHIJ, vmax, wasm_f32x4_le(vaccGHIJ, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
78 v128_t voutKLMN = wasm_v128_bitselect(vaccKLMN, vmax, wasm_f32x4_le(vaccKLMN, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
79 v128_t voutOPQR = wasm_v128_bitselect(vaccOPQR, vmax, wasm_f32x4_le(vaccOPQR, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
80 v128_t voutSTUV = wasm_v128_bitselect(vaccSTUV, vmax, wasm_f32x4_le(vaccSTUV, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
131 v128_t vout0123 = wasm_v128_bitselect(vacc0123, vmax, wasm_f32x4_le(vacc0123, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86()
[all …]
|
| D | 32x4-minmax-neonfma.c | 33 const float32x4_t vmax = vld1q_dup_f32(¶ms->scalar.max); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma() local
124 float32x4_t vout0123n0 = vminq_f32(vacc0123n0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
125 float32x4_t vout4567n0 = vminq_f32(vacc4567n0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
126 float32x4_t vout89ABn0 = vminq_f32(vacc89ABn0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
127 float32x4_t voutCDEFn0 = vminq_f32(vaccCDEFn0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
128 float32x4_t voutGHIJn0 = vminq_f32(vaccGHIJn0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
129 float32x4_t voutKLMNn0 = vminq_f32(vaccKLMNn0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
130 float32x4_t voutOPQRn0 = vminq_f32(vaccOPQRn0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
131 float32x4_t voutSTUVn0 = vminq_f32(vaccSTUVn0, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
132 float32x4_t vout0123n1 = vminq_f32(vacc0123n1, vmax); in xnn_f32_spmm_minmax_ukernel_32x4__neonfma()
[all …]
|
| D | 32x1-minmax-wasmsimd-x86-pipelined.c | 33 const v128_t vmax = wasm_v32x4_load_splat(¶ms->scalar.max); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined() local
87 v128_t vout0123 = wasm_v128_bitselect(vacc0123, vmax, wasm_f32x4_le(vacc0123, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
88 v128_t vout4567 = wasm_v128_bitselect(vacc4567, vmax, wasm_f32x4_le(vacc4567, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
89 v128_t vout89AB = wasm_v128_bitselect(vacc89AB, vmax, wasm_f32x4_le(vacc89AB, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
90 v128_t voutCDEF = wasm_v128_bitselect(vaccCDEF, vmax, wasm_f32x4_le(vaccCDEF, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
91 v128_t voutGHIJ = wasm_v128_bitselect(vaccGHIJ, vmax, wasm_f32x4_le(vaccGHIJ, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
92 v128_t voutKLMN = wasm_v128_bitselect(vaccKLMN, vmax, wasm_f32x4_le(vaccKLMN, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
93 v128_t voutOPQR = wasm_v128_bitselect(vaccOPQR, vmax, wasm_f32x4_le(vaccOPQR, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
94 v128_t voutSTUV = wasm_v128_bitselect(vaccSTUV, vmax, wasm_f32x4_le(vaccSTUV, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
145 v128_t vout0123 = wasm_v128_bitselect(vacc0123, vmax, wasm_f32x4_le(vacc0123, vmax)); in xnn_f32_spmm_minmax_ukernel_32x1__wasmsimd_x86_pipelined()
[all …]
|
| D | 8x1-minmax-scalar.c | 32 const float vmax = params->scalar.max; in xnn_f32_spmm_minmax_ukernel_8x1__scalar() local
72 float vout0x0 = math_min_f32(vacc0x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
73 float vout1x0 = math_min_f32(vacc1x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
74 float vout2x0 = math_min_f32(vacc2x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
75 float vout3x0 = math_min_f32(vacc3x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
76 float vout4x0 = math_min_f32(vacc4x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
77 float vout5x0 = math_min_f32(vacc5x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
78 float vout6x0 = math_min_f32(vacc6x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
79 float vout7x0 = math_min_f32(vacc7x0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
141 float vout0 = math_min_f32(vacc0, vmax); in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
[all …]
|