/external/XNNPACK/test/ |
D | dwconv-spchw-microkernel-tester.h | 154 inline DWConvSpCHWMicrokernelTester& output_tuple_stride(uint32_t output_tuple_stride) { in output_tuple_stride() argument 155 assert(output_tuple_stride != 0); in output_tuple_stride() 156 this->output_tuple_stride_ = output_tuple_stride; in output_tuple_stride() 160 inline uint32_t output_tuple_stride() const { in output_tuple_stride() function 225 ASSERT_EQ(0, output_tuple_stride() % output_tuple_size()); 235 (output_width() - 1) / output_tuple_size() * output_tuple_stride() + output_tuple_size()); 288 input_tuple_stride() * sizeof(float), output_tuple_stride() * sizeof(float), 297 …output[y * output_width_stride() + x / output_tuple_size() * output_tuple_stride() + x % output_tu…
|
D | f32-dwconv-spchw.cc | 148 TEST(F32_DWCONV_SPCHW_3X3P1__SSE, output_tuple_stride) { in TEST() argument 161 .output_tuple_stride(5 * 4) in TEST() 326 TEST(F32_DWCONV_SPCHW_3X3S2P1__SSE, output_tuple_stride) { in TEST() argument 340 .output_tuple_stride(5 * 4) in TEST() 499 TEST(F32_DWCONV_SPCHW_3X3P1__NEONFMA, output_tuple_stride) { in TEST() argument 512 .output_tuple_stride(5 * 4) in TEST() 678 TEST(F32_DWCONV_SPCHW_3X3S2P1__NEONFMA, output_tuple_stride) { in TEST() argument 692 .output_tuple_stride(5 * 4) in TEST() 867 TEST(F32_DWCONV_SPCHW_5X5P2__NEONFMA, output_tuple_stride) { in TEST() argument 880 .output_tuple_stride(5 * 4) in TEST() [all …]
|
/external/XNNPACK/src/f32-dwconv-spchw/ |
D | 3x3s2p1-scalar.c | 19 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__scalar() argument 27 const size_t output_width_increment = output_width_stride - (n/2) * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__scalar() 76 *output0 = voutput; output0 = (float *) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__scalar()
|
D | 3x3p1-scalar.c | 19 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_3x3p1__scalar() argument 27 const size_t output_width_increment = output_width_stride - (n - 1) * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3p1__scalar() 79 *output0 = voutput; output0 = (float *) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__scalar()
|
D | 3x3p1-neonfma.c | 21 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma() argument 33 const size_t output_width_increment = 3 * output_width_stride - (n - 1) / 4 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma() 35 …nst size_t output_width_increment_single = output_width_stride - (n - 1) / 4 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma() 147 vst1q_f32(output0, vo0); output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma() 148 vst1q_f32(output1, vo1); output1 = (float*) ((uintptr_t) output1 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma() 149 vst1q_f32(output2, vo2); output2 = (float*) ((uintptr_t) output2 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma() 313 vst1q_f32(output0, vo); output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__neonfma()
|
D | 5x5p2-scalar.c | 19 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar() argument 30 const size_t output_width_increment_single = output_width_stride - (n - 1) * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar() 139 *output0 = voutput; output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar() 168 *output0 = voutput; output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__scalar()
|
D | 3x3s2p1-neonfma.c | 21 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__neonfma() argument 34 const size_t output_width_increment = output_width_stride - n / 8 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__neonfma() 100 vst1q_f32(output, vo); output = (float*) ((uintptr_t) output + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__neonfma()
|
D | 5x5s2p2-scalar.c | 19 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar() argument 30 const size_t output_width_increment = output_width_stride - (n - 1) / 2 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar() 128 *output0 = voutput; output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__scalar()
|
D | 3x3p1-sse.c | 21 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_3x3p1__sse() argument 33 const size_t output_width_increment = output_width_stride - (n - 1) / 4 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3p1__sse() 137 output = (float*) ((uintptr_t) output + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3p1__sse()
|
D | 3x3s2p1-sse.c | 21 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__sse() argument 34 const size_t output_width_increment = output_width_stride - n / 8 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__sse() 113 output = (float*) ((uintptr_t) output + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__sse()
|
D | 5x5s2p2-neonfma.c | 21 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__neonfma() argument 34 …nst size_t output_width_increment_single = output_width_stride - (n + 1) / 8 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__neonfma() 189 output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__neonfma() 353 output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5s2p2__neonfma()
|
D | 5x5p2-neonfma.c | 21 size_t output_tuple_stride, in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() argument 33 …nst size_t output_width_increment_single = output_width_stride - (n - 1) / 4 * output_tuple_stride; in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 246 vst1q_f32(output0, vo0); output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 247 vst1q_f32(output1, vo1); output1 = (float*) ((uintptr_t) output1 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 248 vst1q_f32(output2, vo2); output2 = (float*) ((uintptr_t) output2 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 435 vst1q_f32(output0, vo0); output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 436 vst1q_f32(output1, vo1); output1 = (float*) ((uintptr_t) output1 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 437 vst1q_f32(output2, vo2); output2 = (float*) ((uintptr_t) output2 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 795 vst1q_f32(output0, vo0); output0 = (float*) ((uintptr_t) output0 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() 796 vst1q_f32(output1, vo1); output1 = (float*) ((uintptr_t) output1 + output_tuple_stride); in xnn_f32_dwconv_spchw_ukernel_5x5p2__neonfma() [all …]
|
/external/XNNPACK/src/xnnpack/ |
D | dwconv.h | 155 size_t output_tuple_stride, \
|
D | params.h | 672 size_t output_tuple_stride, 684 size_t output_tuple_stride,
|
D | compute.h | 275 size_t output_tuple_stride; member
|
/external/XNNPACK/src/ |
D | convolution-nchw.c | 671 ….output_tuple_stride = convolution_op->ukernel.dwconv2d.output_width_tile << log2_output_element_s… in setup_convolution2d_nchw()
|
D | operator-run.c | 267 context->output_tuple_stride, in xnn_compute_dwconv2d_spchw()
|