| /cmsis-nn-3.6.0-3.5.0/Source/ConvolutionFunctions/ |
| D | arm_convolve_s8.c | 89 const int32_t rhs_cols = kernel_x * kernel_y * input_ch; in arm_convolve_s8() local
105 int8_t *im2col_buf = (int8_t *)buffer_a + rhs_cols * 2; in arm_convolve_s8()
151 rhs_cols, in arm_convolve_s8()
156 rhs_cols); in arm_convolve_s8()
166 im2col_buf - rhs_cols, im2col_buf_start_s16, rhs_cols, (int16_t)input_offset); in arm_convolve_s8()
168 …arm_q7_to_q15_with_offset(im2col_buf - rhs_cols, im2col_buf_start_s16, rhs_cols, (int16_t)input_of… in arm_convolve_s8()
170 im2col_buf_start_s16 += rhs_cols; in arm_convolve_s8()
182 rhs_cols, in arm_convolve_s8()
188 im2col_buf = (int8_t *)buffer_a + rhs_cols * 2; in arm_convolve_s8()
212 rhs_cols, in arm_convolve_s8()
[all …]
|
| D | arm_convolve_1x1_s8.c | 72 const int32_t rhs_cols = input_dims->c; in arm_convolve_1x1_s8() local
74 const int32_t input_inc = input_dims->w * conv_params->stride.h * rhs_cols; in arm_convolve_1x1_s8()
82 input_data = input_data_ref + (i_batch * rhs_cols * input_dims->w * input_dims->h); in arm_convolve_1x1_s8()
94 rhs_cols, in arm_convolve_1x1_s8()
99 rhs_cols * stride_w); in arm_convolve_1x1_s8()
|
| D | arm_convolve_1x1_s8_fast.c | 74 const int32_t rhs_cols = input_dims->c; in arm_convolve_1x1_s8_fast() local
84 rhs_cols, in arm_convolve_1x1_s8_fast()
89 rhs_cols); in arm_convolve_1x1_s8_fast()
|
| D | arm_convolve_fast_s16.c | 84 const int32_t rhs_cols = input_ch * kernel_y * kernel_x; in arm_convolve_fast_s16() local
128 if (two_column_buf == buffer_a + 2 * rhs_cols) in arm_convolve_fast_s16()
137 rhs_cols, in arm_convolve_fast_s16()
162 int32_t col_count = rhs_cols >> 2; in arm_convolve_fast_s16()
179 col_count = rhs_cols & 0x3; in arm_convolve_fast_s16()
218 (void)rhs_cols; in arm_convolve_fast_s16()
|
| D | arm_convolve_1_x_n_s8.c | 103 const int32_t rhs_cols = kernel_x * input_dims->c; in arm_convolve_1_x_n_s8() local
147 rhs_cols, in arm_convolve_1_x_n_s8()
|
| /cmsis-nn-3.6.0-3.5.0/Source/NNSupportFunctions/ |
| D | arm_nn_vec_mat_mul_result_acc_s8.c | 52 const int32_t rhs_cols, in arm_nn_vec_mat_mul_result_acc_s8() argument
59 const int8_t *lhs = lhs_in + i_batch * rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
73 const int8_t *rhs_1 = rhs + rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
74 const int8_t *rhs_2 = rhs + 2 * rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
75 const int8_t *rhs_3 = rhs + 3 * rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
77 int32_t col_cnt = rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
104 rhs += 4 * rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
129 int32_t col_cnt = rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
143 rhs += rhs_cols; in arm_nn_vec_mat_mul_result_acc_s8()
164 const int32_t col_loop_cnt = rhs_cols / 4; in arm_nn_vec_mat_mul_result_acc_s8()
[all …]
|
| D | arm_nn_vec_mat_mult_t_s16.c | 56 const int32_t rhs_cols, in arm_nn_vec_mat_mult_t_s16() argument
63 int32_t rhs_cols_fast = rhs_cols; in arm_nn_vec_mat_mult_t_s16()
65 if (rhs_cols > MAX_COL_COUNT) in arm_nn_vec_mat_mult_t_s16()
80 const int8_t *rhs_ptr_1 = rhs + rhs_cols; in arm_nn_vec_mat_mult_t_s16()
81 const int8_t *rhs_ptr_2 = rhs + rhs_cols * 2; in arm_nn_vec_mat_mult_t_s16()
82 const int8_t *rhs_ptr_3 = rhs + rhs_cols * 3; in arm_nn_vec_mat_mult_t_s16()
119 if (rhs_cols > MAX_COL_COUNT) in arm_nn_vec_mat_mult_t_s16()
121 for (int i_rhs_cols = MAX_COL_COUNT; i_rhs_cols < rhs_cols; i_rhs_cols++) in arm_nn_vec_mat_mult_t_s16()
164 rhs += 4 * rhs_cols; in arm_nn_vec_mat_mult_t_s16()
199 if (rhs_cols > MAX_COL_COUNT) in arm_nn_vec_mat_mult_t_s16()
[all …]
|
| D | arm_nn_vec_mat_mult_t_s8.c | 66 const int32_t rhs_cols, in arm_nn_vec_mat_mult_t_s8() argument
82 const int32_t col_loop_cnt = (rhs_cols + 15) / 16; in arm_nn_vec_mat_mult_t_s8()
86 const int8_t *rhs_1 = rhs + rhs_cols; in arm_nn_vec_mat_mult_t_s8()
87 const int8_t *rhs_2 = rhs + 2 * rhs_cols; in arm_nn_vec_mat_mult_t_s8()
99 uint32_t col_cnt = (uint32_t)rhs_cols; in arm_nn_vec_mat_mult_t_s8()
125 rhs += 3 * rhs_cols; in arm_nn_vec_mat_mult_t_s8()
152 const int32_t col_loop_cnt = (rhs_cols + 15) / 16; in arm_nn_vec_mat_mult_t_s8()
156 uint32_t col_cnt = (uint32_t)rhs_cols; in arm_nn_vec_mat_mult_t_s8()
171 rhs += rhs_cols; in arm_nn_vec_mat_mult_t_s8()
204 const int32_t col_loop_cnt = rhs_cols / 4; in arm_nn_vec_mat_mult_t_s8()
[all …]
|
| D | arm_nn_vec_mat_mult_t_svdf_s8.c | 56 const int32_t rhs_cols, in arm_nn_vec_mat_mult_t_svdf_s8() argument
61 if (rhs_cols < 0 || (NN_Q31_MAX - rhs_cols) < 16 || dst_offset < 0) in arm_nn_vec_mat_mult_t_svdf_s8()
75 const int32_t col_loop_cnt = (rhs_cols + 15) / 16; in arm_nn_vec_mat_mult_t_svdf_s8()
79 const int8_t *rhs_1 = rhs + rhs_cols; in arm_nn_vec_mat_mult_t_svdf_s8()
80 const int8_t *rhs_2 = rhs + 2 * rhs_cols; in arm_nn_vec_mat_mult_t_svdf_s8()
86 uint32_t col_cnt = (uint32_t)rhs_cols; in arm_nn_vec_mat_mult_t_svdf_s8()
112 rhs += 3 * rhs_cols; in arm_nn_vec_mat_mult_t_svdf_s8()
131 const int32_t col_loop_cnt = (rhs_cols + 15) / 16; in arm_nn_vec_mat_mult_t_svdf_s8()
135 uint32_t col_cnt = (uint32_t)rhs_cols; in arm_nn_vec_mat_mult_t_svdf_s8()
150 rhs += rhs_cols; in arm_nn_vec_mat_mult_t_svdf_s8()
[all …]
|
| D | arm_nn_mat_mult_nt_t_s8.c | 56 const int32_t rhs_cols, in arm_nn_mat_mult_nt_t_s8() argument
79 const int8_t *col_base = rhs + i * rhs_cols; in arm_nn_mat_mult_nt_t_s8()
83 for (int j = 0; j < rhs_cols; j++) in arm_nn_mat_mult_nt_t_s8()
126 : [cnt] "r"(rhs_cols) in arm_nn_mat_mult_nt_t_s8()
160 const int8_t *col_base = rhs + i * rhs_cols; in arm_nn_mat_mult_nt_t_s8()
164 for (int j = 0; j < rhs_cols; j++) in arm_nn_mat_mult_nt_t_s8()
184 : [cnt] "r"(rhs_cols) in arm_nn_mat_mult_nt_t_s8()
223 const int32_t rhs_off0 = rhs_cols - 4; in arm_nn_mat_mult_nt_t_s8()
234 for (int32_t x = 0; x < rhs_cols; ++x) in arm_nn_mat_mult_nt_t_s8()
237 lhs_offset_contribution1 += rhs[x + rhs_cols]; in arm_nn_mat_mult_nt_t_s8()
[all …]
|
| /cmsis-nn-3.6.0-3.5.0/Include/ |
| D | arm_nnsupportfunctions.h | 366 const int32_t rhs_cols,
403 const int32_t rhs_cols,
432 const int32_t rhs_cols,
464 const int32_t rhs_cols,
1294 const int32_t rhs_cols,
|
