| /cmsis-nn-latest/Source/ConcatenationFunctions/ |
| D | arm_concatenation_s8_z.c | 50 const uint16_t input_x, in arm_concatenation_s8_z() argument
58 const uint32_t input_copy_size = input_x * input_y * input_z; in arm_concatenation_s8_z()
59 const uint32_t output_stride = input_x * input_y * output_z; in arm_concatenation_s8_z()
61 output += offset_z * (input_x * input_y); in arm_concatenation_s8_z()
|
| D | arm_concatenation_s8_y.c | 50 const uint16_t input_x, in arm_concatenation_s8_y() argument
59 const uint32_t input_copy_size = input_x * input_y; in arm_concatenation_s8_y()
60 const uint32_t output_stride = input_x * output_y; in arm_concatenation_s8_y()
62 output += offset_y * input_x; in arm_concatenation_s8_y()
|
| D | arm_concatenation_s8_w.c | 50 const uint16_t input_x, in arm_concatenation_s8_w() argument
57 const uint32_t input_copy_size = input_x * input_y * input_z * input_w; in arm_concatenation_s8_w()
59 output += offset_w * (input_x * input_y * input_z); in arm_concatenation_s8_w()
|
| D | arm_concatenation_s8_x.c | 50 const uint16_t input_x, in arm_concatenation_s8_x() argument
67 arm_memcpy_s8(output, input, input_x); in arm_concatenation_s8_x()
68 input += input_x; in arm_concatenation_s8_x()
|
| /cmsis-nn-latest/Tests/UnitTest/TestCases/Common/ |
| D | conv2d_s4_weights_template.json | 17 input_x,
|
| D | dw_s4_weights_template.json | 17 input_x,
|
| /cmsis-nn-latest/Source/ConvolutionFunctions/ |
| D | arm_depthwise_conv_s4.c | 45 const int32_t input_x, in depthwise_conv_s4_generic() argument
106 … const int32_t end_min_x = (input_x - base_idx_x + dilation_x - 1) / dilation_x; in depthwise_conv_s4_generic()
112 ker_x_end = MIN(kernel_x, input_x - base_idx_x); in depthwise_conv_s4_generic()
138 int32_t idx_0 = (idx_y * input_x + idx_x) * input_ch + i_input_ch; in depthwise_conv_s4_generic()
198 … const int32_t end_min_x = (input_x - base_idx_x + dilation_x - 1) / dilation_x; in depthwise_conv_s4_generic()
204 ker_x_end = MIN(kernel_x, input_x - base_idx_x); in depthwise_conv_s4_generic()
229 … int32_t idx_0 = (idx_y * input_x + idx_x) * input_ch + i_input_ch; in depthwise_conv_s4_generic()
290 … const int32_t end_min_x = (input_x - base_idx_x + dilation_x - 1) / dilation_x; in depthwise_conv_s4_generic()
296 ker_x_end = MIN(kernel_x, input_x - base_idx_x); in depthwise_conv_s4_generic()
322 … int32_t idx_0 = (idx_y * input_x + idx_x) * input_ch + i_input_ch; in depthwise_conv_s4_generic()
[all …]
|
| D | arm_depthwise_conv_3x3_s8.c | 67 const int32_t input_x = input_dims->w; in arm_depthwise_conv_3x3_s8() local
117 …const int8_t *input_ptr = input + (in_h + ker_h_start) * (input_ch * input_x) + in_w * input_ch + … in arm_depthwise_conv_3x3_s8()
157 if ((input_x - in_w) >= 3) in arm_depthwise_conv_3x3_s8()
172 input_ptr += (input_ch * input_x); in arm_depthwise_conv_3x3_s8()
201 if ((input_x - in_w) >= 3) in arm_depthwise_conv_3x3_s8()
212 input_ptr += (input_ch * input_x); in arm_depthwise_conv_3x3_s8()
247 …const int8_t *input_ptr = input + (in_h + ker_h_start) * (input_ch * input_x) + in_w * input_ch + … in arm_depthwise_conv_3x3_s8()
259 if ((input_x - in_w) >= 3) in arm_depthwise_conv_3x3_s8()
264 input_ptr += (input_ch * input_x); in arm_depthwise_conv_3x3_s8()
|
| D | arm_depthwise_conv_s16.c | 44 const int32_t input_x, in depthwise_conv_s16_mult_4_s16() argument
92 … int32_t in_idx = (in_h + ker_h) * (input_ch * input_x) + in_w * input_ch + in_ch; in depthwise_conv_s16_mult_4_s16()
96 for (int32_t ker_w = ker_w_start; ker_w < MIN(kernel_x, input_x - in_w); in depthwise_conv_s16_mult_4_s16()
135 const uint16_t input_x, in depthwise_conv_s16_generic_s16() argument
184 … const int32_t end_min_x = (input_x - base_idx_x + dilation_x - 1) / dilation_x; in depthwise_conv_s16_generic_s16()
190 ker_x_end = MIN(kernel_x, input_x - base_idx_x); in depthwise_conv_s16_generic_s16()
217 int32_t idx_0 = (idx_y * input_x + idx_x) * input_ch + i_input_ch; in depthwise_conv_s16_generic_s16()
234 input += (input_x * input_y * input_ch); in depthwise_conv_s16_generic_s16()
|
| D | arm_convolve_1_x_n_s4.c | 73 const uint16_t input_x = input_dims->w; in arm_convolve_1_x_n_s4() local
82 const int32_t total_pad = ((output_x - 1) * stride_x + kernel_x - input_x); in arm_convolve_1_x_n_s4()
164 const int32_t ker_end_idx = MIN(kernel_x, input_x - est_input_x_idx); in arm_convolve_1_x_n_s4()
177 input_data += (input_x * input_ch); in arm_convolve_1_x_n_s4()
|
| D | arm_depthwise_conv_s8.c | 48 const int32_t input_x, in depthwise_conv_s8_mult_4() argument
101 … int32_t in_idx = (in_h + ker_h) * (input_ch * input_x) + in_w * input_ch + in_ch; in depthwise_conv_s8_mult_4()
105 for (int32_t ker_w = ker_w_start; ker_w < MIN(kernel_x, input_x - in_w); in depthwise_conv_s8_mult_4()
156 const uint16_t input_x, in depthwise_conv_s8_generic() argument
210 … const int32_t end_min_x = (input_x - base_idx_x + dilation_x - 1) / dilation_x; in depthwise_conv_s8_generic()
216 ker_x_end = MIN(kernel_x, input_x - base_idx_x); in depthwise_conv_s8_generic()
243 int32_t idx_0 = (idx_y * input_x + idx_x) * input_ch + i_input_ch; in depthwise_conv_s8_generic()
262 input += (input_x * input_y * input_ch); in depthwise_conv_s8_generic()
|
| D | arm_convolve_s4.c | 71 const uint16_t input_x = input_dims->w; in arm_convolve_s4() local
120 if (k_y < 0 || k_y >= input_y || k_x < 0 || k_x >= input_x) in arm_convolve_s4()
126 … arm_memcpy_s8(im2col_buf, input_data + (k_y * input_x + k_x) * input_ch, input_ch); in arm_convolve_s4()
204 if (k_y < 0 || k_y >= input_y || k_x < 0 || k_x >= input_x) in arm_convolve_s4()
213 … input_data + (k_y * input_x + k_x) * input_ch, two_column_buf, input_ch, input_offset); in arm_convolve_s4()
325 input_data += (input_x * input_y * input_ch); in arm_convolve_s4()
|
| D | arm_convolve_s16.c | 71 const int32_t input_x = input_dims->w; in arm_convolve_s16() local
119 if (k_y < 0 || k_y >= input_y || k_x < 0 || k_x >= input_x) in arm_convolve_s16()
127 … (const int8_t *)(input_data + (k_y * input_x + k_x) * input_ch), in arm_convolve_s16()
282 input_data += (input_x * input_y * input_ch); in arm_convolve_s16()
|
| D | arm_convolve_1_x_n_s8.c | 72 const int32_t input_x = input_dims->w; in arm_convolve_1_x_n_s8() local
80 const int32_t total_pad = ((output_x - 1) * stride_x + kernel_x - input_x); in arm_convolve_1_x_n_s8()
210 input_data += (input_x * input_ch); in arm_convolve_1_x_n_s8()
|
| D | arm_depthwise_conv_fast_s16.c | 84 const int32_t input_x = input_dims->w; in arm_depthwise_conv_fast_s16() local
118 if (i_ker_y < 0 || i_ker_y >= input_y || i_ker_x < 0 || i_ker_x >= input_x) in arm_depthwise_conv_fast_s16()
125 … (int16_t *)(input + (i_ker_y * input_x + i_ker_x) * input_ch), in arm_depthwise_conv_fast_s16()
150 input += input_x * input_y * input_ch; in arm_depthwise_conv_fast_s16()
259 if (idx_x < 0 || idx_x >= input_x) in arm_depthwise_conv_fast_s16()
266 input + (idx_y * input_x + idx_x) * input_ch, in arm_depthwise_conv_fast_s16()
427 input += (input_x * input_y * input_ch); in arm_depthwise_conv_fast_s16()
|
| D | arm_convolve_s8.c | 70 const uint16_t input_x = input_dims->w; in arm_convolve_s8() local
144 if (k_y < 0 || k_y >= input_y || k_x < 0 || k_x >= input_x) in arm_convolve_s8()
151 … input_data + (k_y * input_x + k_x) * input_ch + i_group * kernel_ch, in arm_convolve_s8()
336 input_data += (input_x * input_y * input_ch); in arm_convolve_s8()
|
| D | arm_convolve_get_buffer_sizes_s4.c | 60 const int32_t input_x = input_dims->w; in arm_convolve_1_x_n_s4_get_buffer_size_mve() local
65 const int32_t total_pad = ((output_x - 1) * stride_x + kernel_x - input_x); in arm_convolve_1_x_n_s4_get_buffer_size_mve()
|
| D | arm_depthwise_conv_s8_opt.c | 78 const int32_t input_x = input_dims->w; in arm_depthwise_conv_s8_opt() local
121 if (i_ker_y < 0 || i_ker_y >= input_y || i_ker_x < 0 || i_ker_x >= input_x) in arm_depthwise_conv_s8_opt()
128 input_slice + (i_ker_y * input_x + i_ker_x) * input_ch, in arm_depthwise_conv_s8_opt()
247 if (idx_x < 0 || idx_x >= input_x) in arm_depthwise_conv_s8_opt()
253 … arm_q7_to_q15_with_offset((int8_t *)input + (idx_y * input_x + idx_x) * input_ch, in arm_depthwise_conv_s8_opt()
|
| D | arm_convolve_get_buffer_sizes_s8.c | 60 const int32_t input_x = input_dims->w; in arm_convolve_1_x_n_s8_get_buffer_size_mve() local
65 const int32_t total_pad = ((output_x - 1) * stride_x + kernel_x - input_x); in arm_convolve_1_x_n_s8_get_buffer_size_mve()
|
| D | arm_depthwise_conv_s4_opt.c | 79 const int32_t input_x = input_dims->w; in arm_depthwise_conv_s4_opt() local
122 if (i_ker_y < 0 || i_ker_y >= input_y || i_ker_x < 0 || i_ker_x >= input_x) in arm_depthwise_conv_s4_opt()
129 input_slice + (i_ker_y * input_x + i_ker_x) * input_ch, in arm_depthwise_conv_s4_opt()
289 if (idx_x < 0 || idx_x >= input_x) in arm_depthwise_conv_s4_opt()
295 … arm_q7_to_q15_with_offset((int8_t *)input + (idx_y * input_x + idx_x) * input_ch, in arm_depthwise_conv_s4_opt()
|
| /cmsis-nn-latest/Source/PoolingFunctions/ |
| D | arm_avgpool_s8.c | 84 const int32_t input_x = input_dims->w; in arm_avgpool_s8() local
96 const int32_t batch_input = input_x * input_y * ch_src; in arm_avgpool_s8()
115 const int32_t k_x_end = MIN(i_x * stride_x - pad_x + kernel_x, input_x); in arm_avgpool_s8()
142 const int8_t *src_inner = src_base + (ch_src * (k_x + k_y * input_x)); in arm_avgpool_s8()
245 const int32_t input_x = input_dims->w; in arm_avgpool_s8() local
272 const int32_t batch_size = input_x * input_y * ch_src; in arm_avgpool_s8()
289 const int32_t kernel_x_end = MIN(kernel_x, input_x - idx_x); in arm_avgpool_s8()
297 … const int8_t *start = src + ch_src * (k_x + idx_x + (k_y + idx_y) * input_x); in arm_avgpool_s8()
336 const int32_t batch_input = input_x * input_y * ch_src; in arm_avgpool_s8()
353 if (k_y >= 0 && k_x >= 0 && k_y < input_y && k_x < input_x) in arm_avgpool_s8()
[all …]
|
| D | arm_avgpool_s16.c | 82 const int32_t input_x = input_dims->w; in arm_avgpool_s16() local
94 const int32_t batch_input = input_x * input_y * ch_src; in arm_avgpool_s16()
117 const int32_t k_x_end = MIN(i_x * stride_x - pad_x + kernel_x, input_x); in arm_avgpool_s16()
139 const int16_t *src_inner = src_base + (ch_src * (k_x + k_y * input_x)); in arm_avgpool_s16()
225 const int32_t kernel_x_end = MIN(kernel_x, input_x - idx_x); in arm_avgpool_s16()
233 … const int16_t *start = src + ch_src * (k_x + idx_x + (k_y + idx_y) * input_x); in arm_avgpool_s16()
286 const int32_t kernel_x_end = MIN(kernel_x, input_x - base_idx_x); in arm_avgpool_s16()
297 … sum += src[i_ch_in + ch_src * (k_x + base_idx_x + (k_y + base_idx_y) * input_x)]; in arm_avgpool_s16()
|
| D | arm_max_pool_s16.c | 155 const int32_t input_x = input_dims->w; in arm_max_pool_s16() local
167 const int32_t batch_size = input_x * input_y * channel_in; in arm_max_pool_s16()
190 const int32_t kernel_x_end = MIN(kernel_x, input_x - base_idx_x); in arm_max_pool_s16()
198 … const int16_t *start = src + channel_in * (k_x + base_idx_x + (k_y + base_idx_y) * input_x); in arm_max_pool_s16()
|
| D | arm_max_pool_s8.c | 172 const int32_t input_x = input_dims->w; in arm_max_pool_s8() local
184 const int32_t batch_size = input_x * input_y * channel_in; in arm_max_pool_s8()
206 const int32_t kernel_x_end = MIN(kernel_x, input_x - base_idx_x); in arm_max_pool_s8()
214 … const int8_t *start = src + channel_in * (k_x + base_idx_x + (k_y + base_idx_y) * input_x); in arm_max_pool_s8()
|
| /cmsis-nn-latest/Include/ |
| D | arm_nnfunctions.h | 2238 const uint16_t input_x,
2285 const uint16_t input_x,
2332 const uint16_t input_x,
2375 const uint16_t input_x,
|
