| /dports/misc/ncnn/ncnn-20211208/src/layer/riscv/ |
| H A D | convolutiondepthwise_5x5_packn.h | 196 r0 += packn; in convdw5x5s1_packn_rvv()
197 r1 += packn; in convdw5x5s1_packn_rvv()
198 r2 += packn; in convdw5x5s1_packn_rvv()
199 r3 += packn; in convdw5x5s1_packn_rvv()
200 r4 += packn; in convdw5x5s1_packn_rvv()
204 r0 += 4 * packn + w * packn; in convdw5x5s1_packn_rvv()
205 r1 += 4 * packn + w * packn; in convdw5x5s1_packn_rvv()
206 r2 += 4 * packn + w * packn; in convdw5x5s1_packn_rvv()
207 r3 += 4 * packn + w * packn; in convdw5x5s1_packn_rvv()
208 r4 += 4 * packn + w * packn; in convdw5x5s1_packn_rvv()
[all …]
|
| H A D | convolutiondepthwise_5x5_packn_fp16s.h | 196 r0 += packn; in convdw5x5s1_packn_fp16sa_rvv()
197 r1 += packn; in convdw5x5s1_packn_fp16sa_rvv()
198 r2 += packn; in convdw5x5s1_packn_fp16sa_rvv()
199 r3 += packn; in convdw5x5s1_packn_fp16sa_rvv()
200 r4 += packn; in convdw5x5s1_packn_fp16sa_rvv()
204 r0 += 4 * packn + w * packn; in convdw5x5s1_packn_fp16sa_rvv()
205 r1 += 4 * packn + w * packn; in convdw5x5s1_packn_fp16sa_rvv()
206 r2 += 4 * packn + w * packn; in convdw5x5s1_packn_fp16sa_rvv()
207 r3 += 4 * packn + w * packn; in convdw5x5s1_packn_fp16sa_rvv()
208 r4 += 4 * packn + w * packn; in convdw5x5s1_packn_fp16sa_rvv()
[all …]
|
| H A D | convolution_sgemm_packn.h | 37 tmp.create(2 * maxk, inch, size / 2 + size % 2, 4u * packn, packn, opt.workspace_allocator); in im2col_sgemm_packn_rvv()
39 tmp.create(maxk, inch, size, 4u * packn, packn, opt.workspace_allocator); in im2col_sgemm_packn_rvv()
245 kptr0 += packn; in im2col_sgemm_packn_rvv()
257 outptr0 += packn * 8; in im2col_sgemm_packn_rvv()
291 kptr0 += packn; in im2col_sgemm_packn_rvv()
299 outptr0 += packn * 4; in im2col_sgemm_packn_rvv()
325 kptr0 += packn; in im2col_sgemm_packn_rvv()
331 outptr0 += packn * 2; in im2col_sgemm_packn_rvv()
353 kptr0 += packn; in im2col_sgemm_packn_rvv()
358 outptr0 += packn; in im2col_sgemm_packn_rvv()
[all …]
|
| H A D | convolutiondepthwise_3x3_packn.h | 191 r0 += packn; in convdw3x3s1_packn_rvv()
192 r1 += packn; in convdw3x3s1_packn_rvv()
193 r2 += packn; in convdw3x3s1_packn_rvv()
194 r3 += packn; in convdw3x3s1_packn_rvv()
197 r0 += 2 * packn + w * packn; in convdw3x3s1_packn_rvv()
198 r1 += 2 * packn + w * packn; in convdw3x3s1_packn_rvv()
199 r2 += 2 * packn + w * packn; in convdw3x3s1_packn_rvv()
200 r3 += 2 * packn + w * packn; in convdw3x3s1_packn_rvv()
290 r0 += packn; in convdw3x3s1_packn_rvv()
291 r1 += packn; in convdw3x3s1_packn_rvv()
[all …]
|
| H A D | convolution_sgemm_packn_fp16s.h | 17 const int packn = csrr_vlenb() / 2; in im2col_sgemm_packn_fp16sa_rvv() local
37 tmp.create(2 * maxk, inch, size / 2 + size % 2, 2u * packn, packn, opt.workspace_allocator); in im2col_sgemm_packn_fp16sa_rvv()
39 tmp.create(maxk, inch, size, 2u * packn, packn, opt.workspace_allocator); in im2col_sgemm_packn_fp16sa_rvv()
277 kptr0 += packn; in im2col_sgemm_packn_fp16sa_rvv()
289 outptr0 += packn * 8; in im2col_sgemm_packn_fp16sa_rvv()
323 kptr0 += packn; in im2col_sgemm_packn_fp16sa_rvv()
331 outptr0 += packn * 4; in im2col_sgemm_packn_fp16sa_rvv()
357 kptr0 += packn; in im2col_sgemm_packn_fp16sa_rvv()
385 kptr0 += packn; in im2col_sgemm_packn_fp16sa_rvv()
390 outptr0 += packn; in im2col_sgemm_packn_fp16sa_rvv()
[all …]
|
| H A D | convolutiondepthwise_3x3_packn_fp16s.h | 191 r0 += packn; in convdw3x3s1_packn_fp16sa_rvv()
192 r1 += packn; in convdw3x3s1_packn_fp16sa_rvv()
193 r2 += packn; in convdw3x3s1_packn_fp16sa_rvv()
194 r3 += packn; in convdw3x3s1_packn_fp16sa_rvv()
197 r0 += 2 * packn + w * packn; in convdw3x3s1_packn_fp16sa_rvv()
198 r1 += 2 * packn + w * packn; in convdw3x3s1_packn_fp16sa_rvv()
199 r2 += 2 * packn + w * packn; in convdw3x3s1_packn_fp16sa_rvv()
200 r3 += 2 * packn + w * packn; in convdw3x3s1_packn_fp16sa_rvv()
290 r0 += packn; in convdw3x3s1_packn_fp16sa_rvv()
291 r1 += packn; in convdw3x3s1_packn_fp16sa_rvv()
[all …]
|
| H A D | crop_riscv.cpp | 122 int out_elempack = _outw % packn == 0 ? packn : 1; in forward()
131 if (_woffset % packn == 0 && out_elempack == packn) in forward()
148 int out_elempack = _outh % packn == 0 ? packn : 1; in forward()
157 if (_hoffset % packn == 0 && out_elempack == packn) in forward()
174 int out_elempack = _outc % packn == 0 ? packn : 1; in forward()
183 if (_coffset % packn == 0 && out_elempack == packn) in forward()
216 int out_elempack = _outc % packn == 0 ? packn : 1; in forward()
225 if (_coffset % packn == 0 && out_elempack == packn) in forward()
312 int out_elempack = _outw % packn == 0 ? packn : 1; in forward()
338 int out_elempack = _outh % packn == 0 ? packn : 1; in forward()
[all …]
|
| H A D | interp_bicubic_packn_fp16s.h | 24 Mat rowsbuf0(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn_fp16s()
25 Mat rowsbuf1(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn_fp16s()
26 Mat rowsbuf2(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn_fp16s()
27 Mat rowsbuf3(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn_fp16s()
233 Dp += packn; in resize_bicubic_image_packn_fp16s()
234 rows0p += packn; in resize_bicubic_image_packn_fp16s()
253 Mat rowsbuf0(w, (size_t)packn * 2u, packn); in resize_bicubic_image_packn_fp16sa()
254 Mat rowsbuf1(w, (size_t)packn * 2u, packn); in resize_bicubic_image_packn_fp16sa()
255 Mat rowsbuf2(w, (size_t)packn * 2u, packn); in resize_bicubic_image_packn_fp16sa()
256 Mat rowsbuf3(w, (size_t)packn * 2u, packn); in resize_bicubic_image_packn_fp16sa()
[all …]
|
| H A D | convolution_3x3_packn.h | 17 const int packn = csrr_vlenb() / 4; in conv3x3s1_winograd64_transform_kernel_packn_rvv() local
72 …kernel_tm_packn.create(inch / packn, 64, outch / packn, (size_t)4u * packn * packn, packn * packn); in conv3x3s1_winograd64_transform_kernel_packn_rvv()
74 for (int q = 0; q + (packn - 1) < outch; q += packn) in conv3x3s1_winograd64_transform_kernel_packn_rvv()
82 for (int p = 0; p + (packn - 1) < inch; p += packn) in conv3x3s1_winograd64_transform_kernel_packn_rvv()
100 const int packn = csrr_vlenb() / 4; in conv3x3s1_winograd64_packn_rvv() local
172 float tmp[8][8][packn]; in conv3x3s1_winograd64_packn_rvv()
434 tmpptr += packn; in conv3x3s1_winograd64_packn_rvv()
635 float tmp[6][8][packn]; in conv3x3s1_winograd64_packn_rvv()
800 …kernel_tm_packn.create(inch / packn, 36, outch / packn, (size_t)4u * packn * packn, packn * packn); in conv3x3s1_winograd42_transform_kernel_packn_rvv()
802 for (int q = 0; q + (packn - 1) < outch; q += packn) in conv3x3s1_winograd42_transform_kernel_packn_rvv()
[all …]
|
| H A D | convolution_sgemm_packnto1.h | 38 tmp.create(maxk, inch, size, 4u * packn, packn, opt.workspace_allocator); in im2col_sgemm_packnto1_rvv()
196 int p = pp * packn; in im2col_sgemm_packnto1_rvv()
358 const float* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_rvv()
450 const float* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_rvv()
510 const float* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_rvv()
554 const float* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_rvv()
600 kernel_tm.create(packn * packn * maxk, inch / packn, outch / packn + outch % packn); in convolution_im2col_sgemm_transform_kernel_packnto1_rvv()
603 for (; q + (packn - 1) < outch; q += packn) in convolution_im2col_sgemm_transform_kernel_packnto1_rvv()
607 for (int p = 0; p + (packn - 1) < inch; p += packn) in convolution_im2col_sgemm_transform_kernel_packnto1_rvv()
629 float* g00 = kernel_tm.channel(q / packn + q % packn); in convolution_im2col_sgemm_transform_kernel_packnto1_rvv()
[all …]
|
| H A D | interp_bicubic_packn.h | 17 const int packn = csrr_vlenb() / 4; in resize_bicubic_image_packn() local
24 Mat rowsbuf0(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn()
25 Mat rowsbuf1(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn()
26 Mat rowsbuf2(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn()
27 Mat rowsbuf3(w, (size_t)packn * 4u, packn); in resize_bicubic_image_packn()
57 int sx = xofs[dx] * packn; in resize_bicubic_image_packn()
233 Dp += packn; in resize_bicubic_image_packn()
234 rows0p += packn; in resize_bicubic_image_packn()
235 rows1p += packn; in resize_bicubic_image_packn()
236 rows2p += packn; in resize_bicubic_image_packn()
[all …]
|
| H A D | convolution_3x3_packn_fp16s.h | 17 const int packn = csrr_vlenb() / 2; in conv3x3s1_winograd64_transform_kernel_packn_fp16sa_rvv() local
72 …kernel_tm_packn.create(inch / packn, 64, outch / packn, (size_t)2u * packn * packn, packn * packn); in conv3x3s1_winograd64_transform_kernel_packn_fp16sa_rvv()
74 for (int q = 0; q + (packn - 1) < outch; q += packn) in conv3x3s1_winograd64_transform_kernel_packn_fp16sa_rvv()
82 for (int p = 0; p + (packn - 1) < inch; p += packn) in conv3x3s1_winograd64_transform_kernel_packn_fp16sa_rvv()
100 const int packn = csrr_vlenb() / 2; in conv3x3s1_winograd64_packn_fp16sa_rvv() local
172 __fp16 tmp[8][8][packn]; in conv3x3s1_winograd64_packn_fp16sa_rvv()
434 tmpptr += packn; in conv3x3s1_winograd64_packn_fp16sa_rvv()
492 k0 += packn; in conv3x3s1_winograd64_packn_fp16sa_rvv()
800 …kernel_tm_packn.create(inch / packn, 36, outch / packn, (size_t)2u * packn * packn, packn * packn); in conv3x3s1_winograd42_transform_kernel_packn_fp16sa_rvv()
802 for (int q = 0; q + (packn - 1) < outch; q += packn) in conv3x3s1_winograd42_transform_kernel_packn_fp16sa_rvv()
[all …]
|
| H A D | convolution_sgemm_packnto1_fp16s.h | 38 tmp.create(maxk, inch, size, 2u * packn, packn, opt.workspace_allocator); in im2col_sgemm_packnto1_fp16sa_rvv()
196 int p = pp * packn; in im2col_sgemm_packnto1_fp16sa_rvv()
358 const __fp16* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_fp16sa_rvv()
450 const __fp16* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_fp16sa_rvv()
510 const __fp16* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_fp16sa_rvv()
554 const __fp16* kptr0 = kernel.channel(p / packn + p % packn); in im2col_sgemm_packnto1_fp16sa_rvv()
600 kernel_tm.create(packn * packn * maxk, inch / packn, outch / packn + outch % packn, (size_t)2u); in convolution_im2col_sgemm_transform_kernel_packnto1_fp16sa_rvv()
603 for (; q + (packn - 1) < outch; q += packn) in convolution_im2col_sgemm_transform_kernel_packnto1_fp16sa_rvv()
607 for (int p = 0; p + (packn - 1) < inch; p += packn) in convolution_im2col_sgemm_transform_kernel_packnto1_fp16sa_rvv()
629 __fp16* g00 = kernel_tm.channel(q / packn + q % packn); in convolution_im2col_sgemm_transform_kernel_packnto1_fp16sa_rvv()
[all …]
|
| H A D | padding_packn.h | 18 …const int packn = csrr_vlenb() / sizeof(T); …
19 …const word_type vl = vsetvl_e##SEW##m##LMUL(packn); …
30 …outptr += packn; …
39 …outptr += packn; …
45 …ptr += packn; …
46 …outptr += packn; …
51 …outptr += packn; …
60 …outptr += packn; …
172 …v##VT##m##LMUL##_t _p = vle##SEW##_v_##TSEW##m##LMUL(ptr0 - packn * 2 - x * packn, vl); …
196 …v##VT##m##LMUL##_t _p = vle##SEW##_v_##TSEW##m##LMUL(ptr - packn * 2 - x * packn, vl); …
[all …]
|
| H A D | padding_riscv.cpp | 116 int out_elempack = outw % packn == 0 ? packn : 1; in forward()
137 int out_elempack = outh % packn == 0 ? packn : 1; in forward()
140 if (top % packn == 0 && out_elempack == packn && type == 0) in forward()
159 int out_elempack = outc % packn == 0 ? packn : 1; in forward()
255 out_elempack = top_blob_unpacked.c % packn == 0 ? packn : 1; in forward()
286 int out_elempack = outw % packn == 0 ? packn : 1; in forward_bf16s_fp16s()
327 int out_elempack = outh % packn == 0 ? packn : 1; in forward_bf16s_fp16s()
369 int out_elempack = outc % packn == 0 ? packn : 1; in forward_bf16s_fp16s()
536 int out_elempack = outw % packn == 0 ? packn : 1; in forward_int8()
557 int out_elempack = outh % packn == 0 ? packn : 1; in forward_int8()
[all …]
|
| H A D | interp_bilinear_packn_fp16s.h | 17 const int packn = csrr_vlenb() / 2; in resize_bilinear_image_packn_fp16s() local
24 Mat rowsbuf0(w, (size_t)packn * 4u, packn); in resize_bilinear_image_packn_fp16s()
25 Mat rowsbuf1(w, (size_t)packn * 4u, packn); in resize_bilinear_image_packn_fp16s()
113 Dp += packn; in resize_bilinear_image_packn_fp16s()
114 rows0p += packn; in resize_bilinear_image_packn_fp16s()
115 rows1p += packn; in resize_bilinear_image_packn_fp16s()
131 Mat rowsbuf0(w, (size_t)packn * 2u, packn); in resize_bilinear_image_packn_fp16sa()
132 Mat rowsbuf1(w, (size_t)packn * 2u, packn); in resize_bilinear_image_packn_fp16sa()
220 Dp += packn; in resize_bilinear_image_packn_fp16sa()
221 rows0p += packn; in resize_bilinear_image_packn_fp16sa()
[all …]
|
| H A D | convolution1d_riscv.cpp | 52 const int packn = csrr_vlenb() / 4; in create_pipeline() local
62 elempack = num_input % packn == 0 ? packn : 1; in create_pipeline()
63 out_elempack = num_output % packn == 0 ? packn : 1; in create_pipeline()
146 out_elempack = num_output % packn == 0 ? packn : 1; in forward()
159 if (elempack == packn && out_elempack == packn) in forward()
353 elempack = num_input % packn == 0 ? packn : 1; in create_pipeline_fp16s()
354 out_elempack = num_output % packn == 0 ? packn : 1; in create_pipeline_fp16s()
415 int out_elempack = (opt.use_packing_layout && num_output % packn == 0) ? packn : 1; in forward_fp16s()
425 if (elempack == packn && out_elempack == packn) in forward_fp16s()
636 int out_elempack = (opt.use_packing_layout && num_output % packn == 0) ? packn : 1; in forward_fp16sa()
[all …]
|
| H A D | innerproduct_riscv.cpp | 219 if (elempack == packn) in forward()
256 int p = pp * packn; in forward()
338 out_elempack = num_output % packn == 0 ? packn : 1; in create_pipeline_fp16s()
389 if (elempack == packn && num_output_elempack == packn) in forward_fp16s()
548 int out_elempack = opt.use_packing_layout && num_output % packn == 0 ? packn : 1; in forward_fp16s()
555 if (out_elempack == packn) in forward_fp16s()
581 kptr += packn; in forward_fp16s()
651 if (elempack == packn && num_output_elempack == packn) in forward_fp16sa()
810 int out_elempack = opt.use_packing_layout && num_output % packn == 0 ? packn : 1; in forward_fp16sa()
817 if (out_elempack == packn) in forward_fp16sa()
[all …]
|
| H A D | deconvolution_packn_fp16s.h | 17 const int packn = csrr_vlenb() / 2; in deconvolution_packn_fp16s_rvv() local
18 const word_type vl = vsetvl_e16m1(packn); in deconvolution_packn_fp16s_rvv()
83 for (int l = 0; l < packn; l++) in deconvolution_packn_fp16s_rvv()
86 … vfloat16m1_t _w0 = vle16_v_f16m1(kptr + k * packn * packn + packn * l, vl); in deconvolution_packn_fp16s_rvv()
92 kptr += maxk * packn * packn; in deconvolution_packn_fp16s_rvv()
100 outptr += outw * packn; in deconvolution_packn_fp16s_rvv()
107 const int packn = csrr_vlenb() / 2; in deconvolution_packn_fp16sa_rvv() local
108 const word_type vl = vsetvl_e16m1(packn); in deconvolution_packn_fp16sa_rvv()
176 … vfloat16m1_t _w0 = vle16_v_f16m1(kptr + k * packn * packn + packn * l, vl); in deconvolution_packn_fp16sa_rvv()
182 kptr += maxk * packn * packn; in deconvolution_packn_fp16sa_rvv()
[all …]
|
| H A D | interp_bilinear_packn.h | 17 const int packn = csrr_vlenb() / 4; in resize_bilinear_image_packn() local
18 const word_type vl = vsetvl_e32m1(packn); in resize_bilinear_image_packn()
24 Mat rowsbuf0(w, (size_t)packn * 4u, packn); in resize_bilinear_image_packn()
25 Mat rowsbuf1(w, (size_t)packn * 4u, packn); in resize_bilinear_image_packn()
52 int sx = xofs[dx] * packn; in resize_bilinear_image_packn()
59 vse32_v_f32m1(rows1p + dx * packn, _rows1, vl); in resize_bilinear_image_packn()
76 int sx = xofs[dx] * packn; in resize_bilinear_image_packn()
87 vse32_v_f32m1(rows0p + dx * packn, _rows0, vl); in resize_bilinear_image_packn()
113 Dp += packn; in resize_bilinear_image_packn()
114 rows0p += packn; in resize_bilinear_image_packn()
[all …]
|
| H A D | convolution_7x7_pack1ton.h | 17 const int packn = csrr_vlenb() / 4; in conv7x7s2_pack1ton_rvv() local
18 const word_type vl = vsetvl_e32m1(packn); in conv7x7s2_pack1ton_rvv()
79 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
146 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
213 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
280 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
347 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
414 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
574 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
613 kptr += packn * 7; in conv7x7s2_pack1ton_rvv()
[all …]
|
| H A D | convolution_7x7_pack1ton_fp16s.h | 17 const int packn = csrr_vlenb() / 2; in conv7x7s2_pack1ton_fp16sa_rvv() local
18 const word_type vl = vsetvl_e16m1(packn); in conv7x7s2_pack1ton_fp16sa_rvv()
79 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
146 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
213 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
280 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
347 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
414 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
574 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
613 kptr += packn * 7; in conv7x7s2_pack1ton_fp16sa_rvv()
[all …]
|
| H A D | deconvolution_riscv.cpp | 93 elempack = num_input % packn == 0 ? packn : 1; in create_pipeline()
94 out_elempack = num_output % packn == 0 ? packn : 1; in create_pipeline()
133 if (elempack == packn && out_elempack == packn) in create_pipeline()
199 out_elempack = num_output % packn == 0 ? packn : 1; in forward()
220 if (elempack == packn && out_elempack == packn) in forward()
335 elempack = num_input % packn == 0 ? packn : 1; in create_pipeline_fp16s()
336 out_elempack = num_output % packn == 0 ? packn : 1; in create_pipeline_fp16s()
390 if (elempack == packn && out_elempack == packn) in create_pipeline_fp16s()
434 int out_elempack = (opt.use_packing_layout && num_output % packn == 0) ? packn : 1; in forward_fp16s()
450 if (elempack == packn && out_elempack == packn) in forward_fp16s()
[all …]
|
| H A D | convolution_sgemm.h | 35 if (size >= packn) in im2col_sgemm_rvv()
36 tmp.create(packn * maxk, inch, size / packn + size % packn, 4u, 1, opt.workspace_allocator); in im2col_sgemm_rvv()
45 int i = ii * packn; in im2col_sgemm_rvv()
67 float* tmpptr = tmp.channel(i / packn + i % packn); in im2col_sgemm_rvv()
127 for (; i + (packn - 1) < size; i += packn) in im2col_sgemm_rvv()
167 outptr0 += packn; in im2col_sgemm_rvv()
178 const float* tmpptr = tmp.channel(i / packn + i % packn); in im2col_sgemm_rvv()
242 for (; i + (packn - 1) < size; i += packn) in im2col_sgemm_rvv()
277 const float* tmpptr = tmp.channel(i / packn + i % packn); in im2col_sgemm_rvv()
319 for (; i + (packn - 1) < size; i += packn) in im2col_sgemm_rvv()
[all …]
|
| H A D | convolution_3x3_pack1ton.h | 17 const int packn = csrr_vlenb() / 4; in conv3x3s1_pack1ton_rvv() local
18 const word_type vl = vsetvl_e32m1(packn); in conv3x3s1_pack1ton_rvv()
157 outptr0 += packn * 8; in conv3x3s1_pack1ton_rvv()
214 outptr0 += packn * 4; in conv3x3s1_pack1ton_rvv()
249 outptr0 += packn * 2; in conv3x3s1_pack1ton_rvv()
273 outptr0 += packn; in conv3x3s1_pack1ton_rvv()
285 k0 += 9 * packn; in conv3x3s1_pack1ton_rvv()
292 const int packn = csrr_vlenb() / 4; in conv3x3s2_pack1ton_rvv() local
435 outptr0 += packn * 8; in conv3x3s2_pack1ton_rvv()
551 outptr0 += packn; in conv3x3s2_pack1ton_rvv()
[all …]
|