/cmsis-nn-latest/Source/NNSupportFunctions/ |
D | arm_s8_to_s16_unordered_with_offset.c | 42 …8_to_s16_unordered_with_offset(const int8_t *src, int16_t *dst, int32_t block_size, int16_t offset) in arm_s8_to_s16_unordered_with_offset() argument 50 const int32_t offset_s16x2 = PKHBT(offset, offset, 16); in arm_s8_to_s16_unordered_with_offset() 68 *dst++ = (int16_t)*src++ + offset; in arm_s8_to_s16_unordered_with_offset()
|
D | arm_q7_to_q15_with_offset.c | 42 void arm_q7_to_q15_with_offset(const int8_t *src, int16_t *dst, int32_t block_size, int16_t offset) in arm_q7_to_q15_with_offset() argument 49 const int16x8_t source_offset = vdupq_n_s16(offset); in arm_q7_to_q15_with_offset() 76 const int32_t offset_q15x2 = PKHBT(offset, offset, 16); in arm_q7_to_q15_with_offset() 105 *dst++ = (int16_t)*src++ + offset; in arm_q7_to_q15_with_offset()
|
D | arm_nn_mat_mul_core_4x_s8.c | 48 const int32_t offset, in arm_nn_mat_mul_core_4x_s8() argument 67 const int8_t *ip_row_1 = row_base + offset; in arm_nn_mat_mul_core_4x_s8() 68 const int8_t *ip_row_2 = row_base + (2 * offset); in arm_nn_mat_mul_core_4x_s8() 69 const int8_t *ip_row_3 = row_base + (3 * offset); in arm_nn_mat_mul_core_4x_s8() 136 (void)offset; in arm_nn_mat_mul_core_4x_s8()
|
D | arm_nn_depthwise_conv_nt_t_s16.c | 64 for (int i_loop_cnt = 0, offset = 0; i_loop_cnt < loop_count; in arm_nn_depthwise_conv_nt_t_s16() local 65 num_ch_to_process -= 4, offset += 4, out += 4, i_loop_cnt++) in arm_nn_depthwise_conv_nt_t_s16() 67 const int8_t *rhs_0 = rhs + offset; in arm_nn_depthwise_conv_nt_t_s16() 68 const int16_t *lhs_0 = lhs + offset; in arm_nn_depthwise_conv_nt_t_s16() 69 const int16_t *lhs_1 = lhs + row_x_col * num_ch + offset; in arm_nn_depthwise_conv_nt_t_s16() 70 const int16_t *lhs_2 = lhs + (row_x_col * num_ch * 2) + offset; in arm_nn_depthwise_conv_nt_t_s16() 71 const int16_t *lhs_3 = lhs + (row_x_col * num_ch * 3) + offset; in arm_nn_depthwise_conv_nt_t_s16()
|
D | arm_nn_depthwise_conv_nt_t_padded_s8.c | 75 for (int i_loop_cnt = 0, offset = 0; i_loop_cnt < loop_count; in arm_nn_depthwise_conv_nt_t_padded_s8() local 76 num_ch_to_process -= 4, out += 4, offset += 4, i_loop_cnt++) in arm_nn_depthwise_conv_nt_t_padded_s8() 88 const int8_t *rhs_0 = rhs + offset; in arm_nn_depthwise_conv_nt_t_padded_s8() 89 const int8_t *lhs_0 = lhs + offset; in arm_nn_depthwise_conv_nt_t_padded_s8() 90 const int8_t *lhs_1 = lhs + row_x_col * CH_IN_BLOCK_MVE + offset; in arm_nn_depthwise_conv_nt_t_padded_s8() 91 const int8_t *lhs_2 = lhs + (row_x_col * CH_IN_BLOCK_MVE * 2) + offset; in arm_nn_depthwise_conv_nt_t_padded_s8() 92 const int8_t *lhs_3 = lhs + (row_x_col * CH_IN_BLOCK_MVE * 3) + offset; in arm_nn_depthwise_conv_nt_t_padded_s8()
|
D | arm_nn_depthwise_conv_nt_t_s8.c | 66 for (int i_loop_cnt = 0, offset = 0; i_loop_cnt < loop_count; in arm_nn_depthwise_conv_nt_t_s8() local 67 num_ch_to_process -= 4, offset += 4, out += 4, i_loop_cnt++) in arm_nn_depthwise_conv_nt_t_s8() 79 const int8_t *rhs_0 = rhs + offset; in arm_nn_depthwise_conv_nt_t_s8() 80 const int8_t *lhs_0 = lhs + offset; in arm_nn_depthwise_conv_nt_t_s8() 81 const int8_t *lhs_1 = lhs + row_x_col * CH_IN_BLOCK_MVE + offset; in arm_nn_depthwise_conv_nt_t_s8() 82 const int8_t *lhs_2 = lhs + (row_x_col * CH_IN_BLOCK_MVE * 2) + offset; in arm_nn_depthwise_conv_nt_t_s8() 83 const int8_t *lhs_3 = lhs + (row_x_col * CH_IN_BLOCK_MVE * 3) + offset; in arm_nn_depthwise_conv_nt_t_s8()
|
D | arm_nn_depthwise_conv_nt_t_s4.c | 70 for (int i_loop_cnt = 0, offset = 0; i_loop_cnt < loop_count; in arm_nn_depthwise_conv_nt_t_s4() local 71 num_ch_to_process -= 4, offset += 4, out += 4, i_loop_cnt++) in arm_nn_depthwise_conv_nt_t_s4() 83 const int8_t *rhs_0 = rhs + (offset >> 1); in arm_nn_depthwise_conv_nt_t_s4() 84 const int8_t *lhs_0 = lhs + offset; in arm_nn_depthwise_conv_nt_t_s4() 85 const int8_t *lhs_1 = lhs + row_x_col * S4_CH_IN_BLOCK_MVE + offset; in arm_nn_depthwise_conv_nt_t_s4() 86 const int8_t *lhs_2 = lhs + (row_x_col * S4_CH_IN_BLOCK_MVE * 2) + offset; in arm_nn_depthwise_conv_nt_t_s4() 87 const int8_t *lhs_3 = lhs + (row_x_col * S4_CH_IN_BLOCK_MVE * 3) + offset; in arm_nn_depthwise_conv_nt_t_s4()
|
/cmsis-nn-latest/Source/ConvolutionFunctions/ |
D | arm_nn_mat_mult_s8.c | 76 const int16x8_t offset = vdupq_x_n_s16(col_offset, p); in arm_nn_mat_mult_s8() local 81 c0 = vaddq_s16(c0, offset); in arm_nn_mat_mult_s8() 85 c1 = vaddq_s16(c1, offset); in arm_nn_mat_mult_s8() 89 c2 = vaddq_s16(c2, offset); in arm_nn_mat_mult_s8() 93 c3 = vaddq_s16(c3, offset); in arm_nn_mat_mult_s8() 136 const int16x8_t offset = vdupq_x_n_s16(col_offset, p); in arm_nn_mat_mult_s8() local 141 c0 = vaddq_s16(c0, offset); in arm_nn_mat_mult_s8()
|
D | arm_depthwise_conv_fast_s16.c | 160 …for (int i_loop_cnt = 0, offset = 0; i_loop_cnt < loop_count; num_ch_to_process -= 4, offset += 4,… in arm_depthwise_conv_fast_s16() local 162 const int8_t *row_0 = kernel + offset; in arm_depthwise_conv_fast_s16() 163 const int16_t *col_0 = lhs_buffer + (kernel_size * input_ch * i_buf) + offset; in arm_depthwise_conv_fast_s16() 185 in_requantize_0 += bias[offset]; in arm_depthwise_conv_fast_s16() 186 in_requantize_1 += bias[offset + 1]; in arm_depthwise_conv_fast_s16() 187 in_requantize_2 += bias[offset + 2]; in arm_depthwise_conv_fast_s16() 188 in_requantize_3 += bias[offset + 3]; in arm_depthwise_conv_fast_s16() 191 int32_t reduced_multiplier_0 = REDUCE_MULTIPLIER(output_mult[offset]); in arm_depthwise_conv_fast_s16() 192 int32_t reduced_multiplier_1 = REDUCE_MULTIPLIER(output_mult[offset + 1]); in arm_depthwise_conv_fast_s16() 193 int32_t reduced_multiplier_2 = REDUCE_MULTIPLIER(output_mult[offset + 2]); in arm_depthwise_conv_fast_s16() [all …]
|
D | arm_nn_depthwise_conv_s8_core.c | 69 const int32_t offset = idx * 4; in arm_nn_depthwise_conv_s8_core() local 70 const int8_t *row_0 = row + offset; in arm_nn_depthwise_conv_s8_core() 71 const int16_t *col_0 = col + offset; in arm_nn_depthwise_conv_s8_core() 72 const int16_t *col_1 = col + kernel_size * num_ch + offset; in arm_nn_depthwise_conv_s8_core()
|
D | arm_depthwise_conv_s8_opt.c | 169 for (int i_loop_cnt = 0, offset = i_ch * CH_IN_BLOCK_MVE; i_loop_cnt < loop_count; in arm_depthwise_conv_s8_opt() local 170 num_ch_to_process -= 4, offset += 4, i_loop_cnt++) in arm_depthwise_conv_s8_opt() 173 const int8_t *row_0 = kernel + offset; in arm_depthwise_conv_s8_opt() 177 out_0 = vldrwq_s32(&bias[offset]); in arm_depthwise_conv_s8_opt() 191 const int32x4_t mult = vldrwq_s32(&output_mult[offset]); in arm_depthwise_conv_s8_opt() 192 const int32x4_t shift = vldrwq_s32(&output_shift[offset]); in arm_depthwise_conv_s8_opt()
|
D | arm_depthwise_conv_s4_opt.c | 171 for (int i_loop_cnt = 0, offset = i_ch * S4_CH_IN_BLOCK_MVE; i_loop_cnt < loop_count; in arm_depthwise_conv_s4_opt() local 172 num_ch_to_process -= 4, offset += 4, i_loop_cnt++) in arm_depthwise_conv_s4_opt() 175 const int8_t *row_0 = kernel + (offset >> 1); in arm_depthwise_conv_s4_opt() 179 out_0 = vldrwq_s32(&bias[offset]); in arm_depthwise_conv_s4_opt() 233 const int32x4_t mult = vldrwq_s32(&output_mult[offset]); in arm_depthwise_conv_s4_opt() 234 const int32x4_t shift = vldrwq_s32(&output_shift[offset]); in arm_depthwise_conv_s4_opt()
|
/cmsis-nn-latest/Source/SoftmaxFunctions/ |
D | arm_softmax_s16.c | 77 const int16_t offset = saturated_symmetric_scaled_diff & 0x7f; in arm_softmax_s16() local 80 const int16_t delta = (slope * offset + 64) >> 7; in arm_softmax_s16() 99 const int16_t offset = symmetric_shifted_sum & 0x7f; in arm_softmax_s16() local 102 const int16_t delta = (slope * offset + 64) >> 7; in arm_softmax_s16()
|
/cmsis-nn-latest/Documentation/Doxygen/style_template/ |
D | navtree.js | 159 pos-docContent.offset().top, 164 scrollTop: pos + docContent.scrollTop() - docContent.offset().top 261 navtree.scrollTo('#selected',100,{offset:-windowHeight/2});
|
/cmsis-nn-latest/Include/ |
D | arm_nnsupportfunctions.h | 157 void arm_q7_to_q15_with_offset(const int8_t *src, int16_t *dst, int32_t block_size, int16_t offset); 182 …_to_s16_unordered_with_offset(const int8_t *src, int16_t *dst, int32_t block_size, int16_t offset); 408 const int32_t offset,
|