| /external/XNNPACK/src/xnnpack/ |
| D | math.h | 48 inline static size_t round_down_po2(size_t n, size_t q) { in round_down_po2() function
55 return round_down_po2(n + q - 1, q); in round_up_po2()
|
| /external/XNNPACK/src/ |
| D | packing.c | 29 const size_t skc = round_down_po2(kc, skr); in xnn_pack_f32_gemm_goi_w()
45 …k[(nr_block_start + nr_block_offset) * kc + round_down_po2(kr_block_start, skr) + ((kr_block_start… in xnn_pack_f32_gemm_goi_w()
83 const size_t skc = round_down_po2(kc, skr); in xnn_pack_f16_gemm_goi_w()
99 …k[(nr_block_start + nr_block_offset) * kc + round_down_po2(kr_block_start, skr) + ((kr_block_start… in xnn_pack_f16_gemm_goi_w()
299 const size_t skc = round_down_po2(kc, skr); in xnn_pack_f32_gemm_io_w()
314 …k[(round_down_po2(kr_block_start, skr) + ((kr_block_start + nr_block_offset * kr) & sr_mask) + kr_… in xnn_pack_f32_gemm_io_w()
346 const size_t skc = round_down_po2(kc, skr); in xnn_pack_f16_gemm_io_w()
361 …k[(round_down_po2(kr_block_start, skr) + ((kr_block_start + nr_block_offset * kr) & sr_mask) + kr_… in xnn_pack_f16_gemm_io_w()
443 const size_t skc = round_down_po2(kc, skr); in xnn_pack_f32_conv_goki_w()
460 …k[((nr_block_start + nr_block_offset) * ks + ki) * kc + round_down_po2(kr_block_start, skr) + ((kr… in xnn_pack_f32_conv_goki_w()
[all …]
|
| /external/XNNPACK/src/operators/ |
| D | convolution-nchw.c | 241 for (size_t oc = 0; oc < round_down_po2(group_output_channels, 4); oc += 4) { in xnn_create_convolution2d_nchw_f32()
253 …for (size_t oc = round_down_po2(group_output_channels, 4); oc < round_down_po2(group_output_channe… in xnn_create_convolution2d_nchw_f32()
262 for (size_t oc = round_down_po2(group_output_channels, 2); oc < group_output_channels; oc++) { in xnn_create_convolution2d_nchw_f32()
326 …for (size_t ocb = 0; ocb < round_down_po2(group_output_channels, output_channels_block_size); ocb … in xnn_create_convolution2d_nchw_f32()
363 …for (size_t oc = round_down_po2(group_output_channels, output_channels_block_size); oc < group_out… in xnn_create_convolution2d_nchw_f32()
|
| /external/XNNPACK/src/f32-dwconv2d-chw/gen/ |
| D | 3x3s2p1-minmax-neonfma-2x4-acc2.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_2x4_acc2()
44 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_2x4_acc2()
|
| D | 3x3s2p1-minmax-neon-2x4.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neon_2x4()
44 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neon_2x4()
|
| D | 3x3s2p1-minmax-neonfma-2x4.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_2x4()
44 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_2x4()
|
| D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-2x4.c | 56 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_2x4()
57 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_2x4()
|
| D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-2x4-acc2.c | 56 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_2x4_acc2()
57 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_2x4_acc2()
|
| D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-2x4-acc2.c | 56 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_2x4_acc2()
57 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_2x4_acc2()
|
| D | 3x3s2p1-minmax-wasmsimd-arm-splat-2x4-acc2.c | 46 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_2x4_acc2()
47 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_2x4_acc2()
|
| D | 3x3s2p1-minmax-wasmsimd-arm-splat-2x4.c | 46 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_2x4()
47 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_2x4()
|
| D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-2x4.c | 56 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_2x4()
57 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_2x4()
|
| D | 3x3s2p1-minmax-scalar-2x1.c | 46 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__scalar_2x1()
|
| D | 3x3s2p1-minmax-scalar-2x1-acc2.c | 46 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__scalar_2x1_acc2()
|
| D | 3x3s2p1-minmax-sse-2x4-acc2.c | 50 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_2x4_acc2()
51 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_2x4_acc2()
|
| D | 3x3s2p1-minmax-sse-2x4.c | 50 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_2x4()
51 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_2x4()
|
| D | 3x3s2p1-minmax-wasmsimd-x86-splat-2x4.c | 46 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_2x4()
47 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_2x4()
|
| D | 3x3s2p1-minmax-wasmsimd-x86-splat-2x4-acc2.c | 46 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_2x4_acc2()
47 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_2x4_acc2()
|
| D | 3x3s2p1-minmax-neonfma-3x4.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_3x4()
44 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_3x4()
|
| D | 3x3s2p1-minmax-neon-3x4.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neon_3x4()
44 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neon_3x4()
|
| D | 3x3s2p1-minmax-neonfma-1x4.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_1x4()
|
| D | 3x3s2p1-minmax-neonfma-1x4-acc2.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_1x4_acc2()
|
| D | 3x3s2p1-minmax-neonfma-1x4-acc4.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neonfma_1x4_acc4()
|
| D | 3x3s2p1-minmax-neon-1x4-acc2.c | 43 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__neon_1x4_acc2()
|
| /external/XNNPACK/src/f32-dwconv2d-chw/ |
| D | 3x3s2p1-sse.c.in | 48 …const size_t input_decrement = round_down_po2(input_width, 4 /* SIMD output width */ * 2 /* subsam…
50 …const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ +…
|