| /external/XNNPACK/src/f32-dwconv-spchw/ |
| D | 5x5p2-scalar.c | 81 float vi0x2 = *i0; i0 = (const float*) ((uintptr_t) i0 + input_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar() local
108 …const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2 + vw4 * vi0x3 + vw5 * vi0x4; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
110 vi0x1 = vi0x2; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
111 vi0x2 = vi0x3; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
142 const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2 + vw4 * vi0x3; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
144 vi0x1 = vi0x2; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
145 vi0x2 = vi0x3; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
173 const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
|
| D | 5x5s2p2-scalar.c | 81 float vi0x2 = *i0; i0 = (const float*) ((uintptr_t) i0 + input_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar() local
102 …const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2 + vw4 * vi0x3 + vw5 * vi0x4; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar()
103 vi0x0 = vi0x2; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar()
105 vi0x2 = vi0x4; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar()
137 const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2 + vw4 * vi0x3; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar()
151 const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar()
|
| D | 3x3s2p1-scalar.c | 60 const float vi0x2 = *i0; i0 = (const float*) ((uintptr_t) i0 + input_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__scalar() local
64 const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__scalar()
65 vi0x0 = vi0x2; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__scalar()
|
| D | 3x3p1-scalar.c | 60 const float vi0x2 = *i0; i0 = (const float*) ((uintptr_t) i0 + input_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__scalar() local
64 const float vrow0_accum = vw1 * vi0x0 + vw2 * vi0x1 + vw3 * vi0x2; in xnn_f32_dwconv_spchw_ukernel_3x3p1__scalar()
66 vi0x1 = vi0x2; in xnn_f32_dwconv_spchw_ukernel_3x3p1__scalar()
|
| /external/XNNPACK/src/f32-conv-hwc/ |
| D | 3x3s2p1c3x8-neonfma-2x2.c | 144 const float32x4_t vi0x2 = vld1q_f32(i0); i0 += 4; in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2() local
155 vo0x1c0123 = vfmaq_laneq_f32(vo0x1c0123, vk00c1x0123, vi0x2, 0); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
157 vo0x1c4567 = vfmaq_laneq_f32(vo0x1c4567, vk00c1x4567, vi0x2, 0); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
194 vo0x1c0123 = vfmaq_laneq_f32(vo0x1c0123, vk00c2x0123, vi0x2, 1); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
196 vo0x1c4567 = vfmaq_laneq_f32(vo0x1c4567, vk00c2x4567, vi0x2, 1); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
233 vo0x1c0123 = vfmaq_laneq_f32(vo0x1c0123, vk01c0x0123, vi0x2, 2); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
235 vo0x1c4567 = vfmaq_laneq_f32(vo0x1c4567, vk01c0x4567, vi0x2, 2); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
272 vo0x1c0123 = vfmaq_laneq_f32(vo0x1c0123, vk01c1x0123, vi0x2, 3); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
274 vo0x1c4567 = vfmaq_laneq_f32(vo0x1c4567, vk01c1x4567, vi0x2, 3); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
391 vo0x0c0123 = vfmaq_laneq_f32(vo0x0c0123, vk02c1x0123, vi0x2, 0); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x8__neonfma_2x2()
[all …]
|
| D | 3x3s2p1c3x4-neonfma-2x2.c | 121 const float32x4_t vi0x2 = vld1q_f32(i0); i0 += 4; in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2() local
129 vo0x1 = vfmaq_laneq_f32(vo0x1, vk00c1, vi0x2, 0); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
150 vo0x1 = vfmaq_laneq_f32(vo0x1, vk00c2, vi0x2, 1); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
171 vo0x1 = vfmaq_laneq_f32(vo0x1, vk01c0, vi0x2, 2); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
192 vo0x1 = vfmaq_laneq_f32(vo0x1, vk01c1, vi0x2, 3); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
260 vo0x0 = vfmaq_laneq_f32(vo0x0, vk02c1, vi0x2, 0); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
281 vo0x0 = vfmaq_laneq_f32(vo0x0, vk02c2, vi0x2, 1); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
398 float32x4_t vi0x2 = vmovq_n_f32(0.0f); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2() local
405 vi0x2 = vld1q_f32(i0 + 4); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
414 vo0x1 = vfmaq_laneq_f32(vo0x1, vk00c1, vi0x2, 0); in xnn_f32_conv_hwc_ukernel_3x3s2p1c3x4__neonfma_2x2()
[all …]
|
| /external/XNNPACK/src/f32-prelu/gen/ |
| D | scalar-2x4.c | 57 const float vi0x2 = i0[2]; in xnn_f32_prelu_ukernel__scalar_2x4() local
68 float vacc0x2 = signbit(vi0x2) ? vi0x2 * vw2 : vi0x2; in xnn_f32_prelu_ukernel__scalar_2x4()
|
| D | wasm-2x4.c | 57 const float vi0x2 = i0[2]; in xnn_f32_prelu_ukernel__wasm_2x4() local
68 float vacc0x2 = signbit(vi0x2) ? vi0x2 * vw2 : vi0x2; in xnn_f32_prelu_ukernel__wasm_2x4()
|
| /external/XNNPACK/src/f32-conv-hwc2spchw/ |
| D | 3x3s2p1c3x4-neonfma-2x2.c | 140 const float32x4_t vi0x2 = vld1q_f32(i0); i0 += 4; in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2() local
148 vo0x1 = vfmaq_laneq_f32(vo0x1, vk00c1, vi0x2, 0); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
169 vo0x1 = vfmaq_laneq_f32(vo0x1, vk00c2, vi0x2, 1); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
190 vo0x1 = vfmaq_laneq_f32(vo0x1, vk01c0, vi0x2, 2); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
211 vo0x1 = vfmaq_laneq_f32(vo0x1, vk01c1, vi0x2, 3); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
279 vo0x0 = vfmaq_laneq_f32(vo0x0, vk02c1, vi0x2, 0); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
300 vo0x0 = vfmaq_laneq_f32(vo0x0, vk02c2, vi0x2, 1); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
394 float32x4_t vi0x2 = vmovq_n_f32(0.0f); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2() local
401 vi0x2 = vld1q_f32(i0 + 4); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
410 vo0x1 = vfmaq_laneq_f32(vo0x1, vk00c1, vi0x2, 0); in xnn_f32_conv_hwc2spchw_ukernel_3x3s2p1c3x4__neonfma_2x2()
[all …]
|