| /external/XNNPACK/src/f32-prelu/gen/ |
| D | avx512f-2x32.c | 82 const __m512 vi1 = _mm512_loadu_ps(i1); in xnn_f32_prelu_ukernel__avx512f_2x32() local
87 const __mmask16 vsign1 = _mm512_cmp_ps_mask(vi1, vzero, _CMP_LT_OQ); in xnn_f32_prelu_ukernel__avx512f_2x32()
88 const __m512 vacc1 = _mm512_mask_mul_ps(vi1, vsign1, vi1, vw); in xnn_f32_prelu_ukernel__avx512f_2x32()
105 const __m512 vi1 = _mm512_maskz_loadu_ps(vmask, i1); in xnn_f32_prelu_ukernel__avx512f_2x32() local
110 const __mmask16 vsign1 = _mm512_cmp_ps_mask(vi1, vzero, _CMP_LT_OQ); in xnn_f32_prelu_ukernel__avx512f_2x32()
111 const __m512 vacc1 = _mm512_mask_mul_ps(vi1, vsign1, vi1, vw); in xnn_f32_prelu_ukernel__avx512f_2x32()
|
| D | wasm-2x1.c | 49 float vi1 = *i1++; in xnn_f32_prelu_ukernel__wasm_2x1() local
53 float vacc1 = __builtin_wasm_max_f32(vi1, vzero); in xnn_f32_prelu_ukernel__wasm_2x1()
54 vi1 = __builtin_wasm_min_f32(vi1, vzero); in xnn_f32_prelu_ukernel__wasm_2x1()
57 vacc1 += vi1 * vw; in xnn_f32_prelu_ukernel__wasm_2x1()
|
| D | avx-2x16.c | 83 const __m256 vi1 = _mm256_loadu_ps(i1); in xnn_f32_prelu_ukernel__avx_2x16() local
87 const __m256 vprod1 = _mm256_mul_ps(vi1, vw); in xnn_f32_prelu_ukernel__avx_2x16()
90 const __m256 vacc1 = _mm256_blendv_ps(vi1, vprod1, vi1); in xnn_f32_prelu_ukernel__avx_2x16()
106 const __m256 vi1 = _mm256_maskload_ps(i1, vmask); in xnn_f32_prelu_ukernel__avx_2x16() local
110 const __m256 vprod1 = _mm256_mul_ps(vi1, vw); in xnn_f32_prelu_ukernel__avx_2x16()
113 __m256 vacc1 = _mm256_blendv_ps(vi1, vprod1, vi1); in xnn_f32_prelu_ukernel__avx_2x16()
|
| D | scalar-2x1.c | 48 const float vi1 = *i1++; in xnn_f32_prelu_ukernel__scalar_2x1() local
51 const float vacc1 = XNN_UNPREDICTABLE(vi1 < 0.0f) ? vi1 * vw : vi1; in xnn_f32_prelu_ukernel__scalar_2x1()
|
| D | wasm-2x4.c | 105 float vi1 = *i1++; in xnn_f32_prelu_ukernel__wasm_2x4() local
109 float vacc1 = __builtin_wasm_max_f32(vi1, vzero); in xnn_f32_prelu_ukernel__wasm_2x4()
110 vi1 = __builtin_wasm_min_f32(vi1, vzero); in xnn_f32_prelu_ukernel__wasm_2x4()
113 vacc1 += vi1 * vw; in xnn_f32_prelu_ukernel__wasm_2x4()
|
| /external/XNNPACK/src/f32-argmaxpool/ |
| D | 9p8x-scalar-c1.c | 57 const float vi1 = *i1++; 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()
139 const float vi1 = *i1++; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1() local
155 if (vi1 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
156 vmax = vi1; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
243 const float vi1 = *i1++; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1() local
259 if (vi1 > vmax) { in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
260 vmax = vi1; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
|
| D | 4x-wasmsimd-c4.c | 53 const v128_t vi1 = wasm_v128_load(i1); 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()
82 const v128_t vi1 = wasm_v128_load(i1); 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()
90 vmax = wasm_v128_bitselect(vi1, vmax, vm1); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
|
| D | 4x-neon-c4.c | 51 const float32x4_t vi1 = vld1q_f32(i1); i1 += 4; 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()
75 const float32x4_t vi1 = vld1q_f32(i1); 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()
83 vmax = vbslq_f32(vm1, vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
|
| D | 4x-sse2-c4.c | 52 const __m128 vi1 = _mm_loadu_ps(i1); 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()
81 const __m128 vi1 = _mm_loadu_ps(i1); 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()
89 vmax = _mm_max_ps(vi1, vmax); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
|
| /external/XNNPACK/src/f32-spmm/gen/ |
| D | 8x1-minmax-scalar-pipelined.c | 41 float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined() local
63 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined()
75 vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined()
123 float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined() local
137 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined()
145 vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined()
175 float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined() local
185 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined()
191 vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar_pipelined()
|
| D | 4x1-minmax-scalar-pipelined.c | 41 float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar_pipelined() local
55 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_4x1__scalar_pipelined()
63 vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar_pipelined()
95 float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar_pipelined() local
105 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_4x1__scalar_pipelined()
111 vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar_pipelined()
|
| D | 8x1-minmax-scalar.c | 53 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar() local
63 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
122 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar() local
132 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
190 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar() local
196 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
227 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar() local
233 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
271 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x1__scalar() local
275 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x1__scalar()
[all …]
|
| D | 8x2-minmax-scalar.c | 61 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x2__scalar() local
72 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
80 vacc1x1 += vi1 * vw1; in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
164 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x2__scalar() local
174 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
236 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x2__scalar() local
243 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
247 vacc1x1 += vi1 * vw1; in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
291 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x2__scalar() local
297 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x2__scalar()
[all …]
|
| D | 8x4-minmax-scalar.c | 77 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x4__scalar() local
90 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
98 vacc1x1 += vi1 * vw1; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
106 vacc1x2 += vi1 * vw2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
114 vacc1x3 += vi1 * vw3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
248 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x4__scalar() local
258 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
328 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_8x4__scalar() local
337 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
341 vacc1x1 += vi1 * vw1; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
[all …]
|
| D | 4x1-minmax-scalar.c | 49 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar() local
55 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_4x1__scalar()
90 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar() local
96 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_4x1__scalar()
136 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar() local
140 vacc1x0 += vi1 * vw0; in xnn_f32_spmm_minmax_ukernel_4x1__scalar()
161 const float vi1 = input[1]; in xnn_f32_spmm_minmax_ukernel_4x1__scalar() local
165 vacc1 += vi1 * vw; in xnn_f32_spmm_minmax_ukernel_4x1__scalar()
|
| /external/XNNPACK/src/f32-maxpool/ |
| D | 9p8x-minmax-wasmsimd-x86-c4.c | 80 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4() local
97 const v128_t vmax01 = wasm_v128_bitselect(vi1, vi0, wasm_f32x4_lt(vi0, vi1)); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4()
119 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4() local
136 const v128_t vmax01 = wasm_v128_bitselect(vi1, vi0, wasm_f32x4_lt(vi0, vi1)); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4()
207 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4() local
223 const v128_t vmax01 = wasm_v128_bitselect(vi1, vi0, wasm_f32x4_lt(vi0, vi1)); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4()
244 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4() local
253 const v128_t vmax01 = wasm_v128_bitselect(vi1, vi0, wasm_f32x4_lt(vi0, vi1)); in xnn_f32_maxpool_minmax_ukernel_9p8x__wasmsimd_x86_c4()
|
| D | 9p8x-minmax-sse-c4.c | 80 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4() local
97 const __m128 vmax018 = _mm_max_ps(_mm_max_ps(vi0, vi1), vi8); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4()
113 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4() local
130 const __m128 vmax018 = _mm_max_ps(_mm_max_ps(vi0, vi1), vi8); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4()
196 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4() local
212 const __m128 vmax01 = _mm_max_ps(_mm_max_ps(vi0, vi1), vo); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4()
227 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4() local
236 const __m128 vmax01 = _mm_max_ps(_mm_max_ps(vi0, vi1), vo); in xnn_f32_maxpool_minmax_ukernel_9p8x__sse_c4()
|
| /external/XNNPACK/src/f32-gavgpool/ |
| D | 7p7x-minmax-scalar-c1.c | 38 const float vi1 = *i1++; in xnn_f32_gavgpool_minmax_ukernel_7p7x__scalar_c1() local
45 const float vsum01 = vi0 + vi1; in xnn_f32_gavgpool_minmax_ukernel_7p7x__scalar_c1()
70 const float vi1 = *i1++; in xnn_f32_gavgpool_minmax_ukernel_7p7x__scalar_c1() local
78 const float vsum01 = vi0 + vi1; in xnn_f32_gavgpool_minmax_ukernel_7p7x__scalar_c1()
124 const float vi1 = *i1++; in xnn_f32_gavgpool_minmax_ukernel_7p7x__scalar_c1() local
132 const float vsum01 = vi0 + vi1; in xnn_f32_gavgpool_minmax_ukernel_7p7x__scalar_c1()
|
| D | 7p7x-minmax-wasm-c1.c | 38 const float vi1 = *i1++; in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasm_c1() local
45 const float vsum01 = vi0 + vi1; in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasm_c1()
70 const float vi1 = *i1++; in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasm_c1() local
78 const float vsum01 = vi0 + vi1; in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasm_c1()
124 const float vi1 = *i1++; in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasm_c1() local
132 const float vsum01 = vi0 + vi1; in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasm_c1()
|
| D | 7p7x-minmax-neon-c4.c | 40 const float32x4_t vi1 = vld1q_f32(i1); i1 += 4; in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4() local
47 const float32x4_t vsum01 = vaddq_f32(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4()
71 const float32x4_t vi1 = vld1q_f32(i1); i1 += 4; in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4() local
79 const float32x4_t vsum01 = vaddq_f32(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4()
125 const float32x4_t vi1 = vld1q_f32(i1); i1 += 4; in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4() local
133 const float32x4_t vsum01 = vaddq_f32(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4()
153 const float32x4_t vi1 = vld1q_f32(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4() local
161 const float32x4_t vsum01 = vaddq_f32(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__neon_c4()
|
| D | 7p7x-minmax-wasmsimd-arm-c4.c | 41 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4() local
54 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4()
80 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4() local
94 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4()
141 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4() local
156 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4()
177 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4() local
185 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_arm_c4()
|
| D | 7p7x-minmax-sse-c4.c | 41 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4() local
54 const __m128 vsum01 = _mm_add_ps(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4()
79 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4() local
93 const __m128 vsum01 = _mm_add_ps(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4()
140 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4() local
155 const __m128 vsum01 = _mm_add_ps(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4()
176 const __m128 vi1 = _mm_loadu_ps(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4() local
184 const __m128 vsum01 = _mm_add_ps(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__sse_c4()
|
| D | 7p7x-minmax-wasmsimd-x86-c4.c | 41 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4() local
54 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4()
80 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4() local
94 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4()
141 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4() local
156 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4()
177 const v128_t vi1 = wasm_v128_load(i1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4() local
185 const v128_t vsum01 = wasm_f32x4_add(vi0, vi1); in xnn_f32_gavgpool_minmax_ukernel_7p7x__wasmsimd_x86_c4()
|
| /external/XNNPACK/src/f32-gavgpool-cw/ |
| D | neon-x4.c | 43 const float32x4_t vi1 = vld1q_f32(i1); i1 += 4; in xnn_f32_gavgpool_cw_ukernel__neon_x4() local
48 vsum1 = vaddq_f32(vsum1, vi1); in xnn_f32_gavgpool_cw_ukernel__neon_x4()
56 float32x4_t vi1 = vld1q_f32(i1); i1 = (const float*) ((uintptr_t) i1 + n); in xnn_f32_gavgpool_cw_ukernel__neon_x4() local
61 vi1 = vreinterpretq_f32_u32(vandq_u32(vmask, vreinterpretq_u32_f32(vi1))); in xnn_f32_gavgpool_cw_ukernel__neon_x4()
66 vsum1 = vaddq_f32(vsum1, vi1); in xnn_f32_gavgpool_cw_ukernel__neon_x4()
|
| /external/XNNPACK/src/f16-gavgpool/ |
| D | 7p7x-minmax-neonfp16arith-c8.c | 41 const float16x8_t vi1 = vld1q_f16(i1); i1 += 8; in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8() local
48 const float16x8_t vsum01 = vaddq_f16(vi0, vi1); in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8()
72 const float16x8_t vi1 = vld1q_f16(i1); i1 += 8; in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8() local
80 const float16x8_t vsum01 = vaddq_f16(vi0, vi1); in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8()
126 const float16x8_t vi1 = vld1q_f16(i1); i1 += 8; in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8() local
134 const float16x8_t vsum01 = vaddq_f16(vi0, vi1); in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8()
154 const float16x8_t vi1 = vld1q_f16(i1); in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8() local
162 const float16x8_t vsum01 = vaddq_f16(vi0, vi1); in xnn_f16_gavgpool_minmax_ukernel_7p7x__neonfp16arith_c8()
|