| /cmsis-nn-latest/Source/NNSupportFunctions/ |
| D | arm_nn_vec_mat_mult_t_svdf_s8.c | 72 int32_t acc_1 = 0; in arm_nn_vec_mat_mult_t_svdf_s8() local
101 acc_1 = vmladavaq_s8(acc_1, ker_1, input); in arm_nn_vec_mat_mult_t_svdf_s8()
114 int32x4_t acc = {acc_0, acc_1, acc_2, 0}; in arm_nn_vec_mat_mult_t_svdf_s8()
171 int32_t acc_1 = 0; in arm_nn_vec_mat_mult_t_svdf_s8() local
199 acc_1 = SMLAD(ker_1, vec_1, acc_1); in arm_nn_vec_mat_mult_t_svdf_s8()
200 acc_1 = SMLAD(ker_0, vec_0, acc_1); in arm_nn_vec_mat_mult_t_svdf_s8()
214 acc_1 = SMLAD(ker_1, vec_1, acc_1); in arm_nn_vec_mat_mult_t_svdf_s8()
215 acc_1 = SMLAD(ker_0, vec_0, acc_1); in arm_nn_vec_mat_mult_t_svdf_s8()
229 acc_1 = SMLAD(ker_1, vec_1, acc_1); in arm_nn_vec_mat_mult_t_svdf_s8()
230 acc_1 = SMLAD(ker_0, vec_0, acc_1); in arm_nn_vec_mat_mult_t_svdf_s8()
[all …]
|
| D | arm_nn_vec_mat_mul_result_acc_s8_s16.c | 77 int32_t acc_1 = *effective_bias_ptr++; in arm_nn_vec_mat_mul_result_acc_s8_s16() local
94 acc_1 = vmladavaq_s8(acc_1, ker_1, input); in arm_nn_vec_mat_mul_result_acc_s8_s16()
109 int32x4_t acc = {acc_0, acc_1, acc_2, acc_3}; in arm_nn_vec_mat_mul_result_acc_s8_s16()
158 int32_t acc_1 = *effective_bias_ptr++; in arm_nn_vec_mat_mul_result_acc_s8_s16() local
185 acc_1 = SMLAD(ker_1, vec_1, acc_1); in arm_nn_vec_mat_mul_result_acc_s8_s16()
186 acc_1 = SMLAD(ker_0, vec_0, acc_1); in arm_nn_vec_mat_mul_result_acc_s8_s16()
195 acc_1 += lhs_temp * (*rhs_1); in arm_nn_vec_mat_mul_result_acc_s8_s16()
200 acc_1 = arm_nn_requantize(acc_1, dst_multiplier, dst_shift); in arm_nn_vec_mat_mul_result_acc_s8_s16()
209 acc_1 += *dst; in arm_nn_vec_mat_mul_result_acc_s8_s16()
210 acc_1 = MAX(acc_1, NN_Q15_MIN); in arm_nn_vec_mat_mul_result_acc_s8_s16()
[all …]
|
| D | arm_nn_vec_mat_mult_t_s8.c | 83 int32_t acc_1 = 0; in arm_nn_vec_mat_mult_t_s8() local
100 acc_1 = *bias++; in arm_nn_vec_mat_mult_t_s8()
119 acc_1 = vmladavaq_s8(acc_1, ker_1, input); in arm_nn_vec_mat_mult_t_s8()
135 int32x4_t acc = {acc_0, acc_1, acc_2, acc_3}; in arm_nn_vec_mat_mult_t_s8()
211 int32_t acc_1 = 0; in arm_nn_vec_mat_mult_t_s8()
215 acc_1 = *bias++; in arm_nn_vec_mat_mult_t_s8()
243 acc_1 = SMLAD(ker_1, vec_1, acc_1); in arm_nn_vec_mat_mult_t_s8()
244 acc_1 = SMLAD(ker_0, vec_0, acc_1); in arm_nn_vec_mat_mult_t_s8()
253 acc_1 += lhs_temp * (*rhs_1_ptr + rhs_offset); in arm_nn_vec_mat_mult_t_s8()
258 acc_1 = arm_nn_requantize(acc_1, dst_multiplier, dst_shift); in arm_nn_vec_mat_mult_t_s8()
[all …]
|
| D | arm_nn_vec_mat_mult_t_s16.c | 226 int32_t acc_1 = 0; in arm_nn_vec_mat_mult_t_s16() local
249 acc_1 = SMLAD(ker_0, vec_part_0, acc_1); in arm_nn_vec_mat_mult_t_s16()
250 acc_1 = SMLAD(ker_1, vec_part_1, acc_1); in arm_nn_vec_mat_mult_t_s16()
254 acc_64_1 += acc_1; in arm_nn_vec_mat_mult_t_s16()
|
| D | arm_nn_vec_mat_mul_result_acc_s16.c | 225 int32_t acc_1 = 0; in arm_nn_vec_mat_mul_result_acc_s16() local
249 acc_1 = SMLAD(ker_0, vec_part_0, acc_1); in arm_nn_vec_mat_mul_result_acc_s16()
250 acc_1 = SMLAD(ker_1, vec_part_1, acc_1); in arm_nn_vec_mat_mul_result_acc_s16()
254 acc_64_1 += acc_1; in arm_nn_vec_mat_mul_result_acc_s16()
|
| /cmsis-nn-latest/Source/ConvolutionFunctions/ |
| D | arm_depthwise_conv_s4.c | 95 int32_t acc_1 = 0; in depthwise_conv_s4_generic() local
131 acc_1 = bias[i_input_ch + 1]; in depthwise_conv_s4_generic()
151 acc_1 += (input[idx_0 + 1] + input_offset) * ker_val1; in depthwise_conv_s4_generic()
167 … acc_1 = arm_nn_requantize(acc_1, output_mult[i_input_ch + 1], output_shift[i_input_ch + 1]); in depthwise_conv_s4_generic()
168 acc_1 += output_offset; in depthwise_conv_s4_generic()
169 acc_1 = MAX(acc_1, output_activation_min); in depthwise_conv_s4_generic()
170 acc_1 = MIN(acc_1, output_activation_max); in depthwise_conv_s4_generic()
171 output[i_out++] = acc_1; in depthwise_conv_s4_generic()
279 int32_t acc_1 = 0; in depthwise_conv_s4_generic() local
315 acc_1 = bias[idx_out_ch + 1]; in depthwise_conv_s4_generic()
[all …]
|
| D | arm_nn_mat_mult_s8.c | 68 int32_t acc_1 = 0; in arm_nn_mat_mult_s8() local
99 acc_1 = vmladavaq_p_s16(acc_1, r0, c1, p); in arm_nn_mat_mult_s8()
104 int32x4_t res = {acc_0, acc_1, acc_2, acc_3}; in arm_nn_mat_mult_s8()
|
| D | arm_depthwise_conv_fast_s16.c | 359 int64_t acc_1 = sum_1; in arm_depthwise_conv_fast_s16() local
366 acc_1 += *bias++; in arm_depthwise_conv_fast_s16()
372 result = arm_nn_requantize_s64(acc_1, output_mult_1, *output_shift++); in arm_depthwise_conv_fast_s16()
|
