Searched refs:sum_4 (Results 1 – 4 of 4) sorted by relevance
/cmsis-nn-3.7.0/Source/PoolingFunctions/ |
D | arm_avgpool_s8.c | 133 int32x4_t sum_4 = vdupq_n_s32(0); in arm_avgpool_s8() local 157 sum_4 = vaddq_s32(sum_4, temp_hi_hi); in arm_avgpool_s8() 185 p = vcmpgtq_n_s32(sum_4, 0); in arm_avgpool_s8() 186 sum_4 = vaddq_m_n_s32(sum_4, sum_4, half_count, p); in arm_avgpool_s8() 187 sum_4 = vsubq_m_n_s32(sum_4, sum_4, half_count, ~p); in arm_avgpool_s8() 194 sum_4[i] = sum_4[i] / count; in arm_avgpool_s8() 206 sum_4 = vmaxq_s32(sum_4, vdupq_n_s32(act_min)); in arm_avgpool_s8() 207 sum_4 = vminq_s32(sum_4, vdupq_n_s32(act_max)); in arm_avgpool_s8() 213 temp_hi = vmovntq_s32(temp_hi, sum_4); in arm_avgpool_s8()
|
/cmsis-nn-3.7.0/Source/ConvolutionFunctions/ |
D | arm_depthwise_conv_s4_opt.c | 326 int32_t sum_4 = 0; in arm_depthwise_conv_s4_opt() local 332 sum_4 = *bias++; in arm_depthwise_conv_s4_opt() 375 sum_4 = SMLAD(op_a, op_b, sum_4); in arm_depthwise_conv_s4_opt() 390 sum_4 += ker3 * col_pos[3]; in arm_depthwise_conv_s4_opt() 401 sum_4 += ker3 * col_pos[3 + input_ch]; in arm_depthwise_conv_s4_opt() 428 sum_4 += ker3 * col_pos[3]; in arm_depthwise_conv_s4_opt() 453 sum_4 = arm_nn_requantize(sum_4, *output_mult++, *output_shift++); in arm_depthwise_conv_s4_opt() 454 sum_4 += output_offset; in arm_depthwise_conv_s4_opt() 455 sum_4 = MAX(sum_4, output_activation_min); in arm_depthwise_conv_s4_opt() 456 sum_4 = MIN(sum_4, output_activation_max); in arm_depthwise_conv_s4_opt() [all …]
|
D | arm_depthwise_conv_s8_opt.c | 279 int32_t sum_4 = 0; in arm_depthwise_conv_s8_opt() local 285 sum_4 = *bias++; in arm_depthwise_conv_s8_opt() 328 sum_4 = SMLAD(op_a, op_b, sum_4); in arm_depthwise_conv_s8_opt() 341 sum_4 += row_pos[3] * col_pos[3]; in arm_depthwise_conv_s8_opt() 365 sum_4 = arm_nn_requantize(sum_4, *output_mult++, *output_shift++); in arm_depthwise_conv_s8_opt() 366 sum_4 += output_offset; in arm_depthwise_conv_s8_opt() 367 sum_4 = MAX(sum_4, output_activation_min); in arm_depthwise_conv_s8_opt() 368 sum_4 = MIN(sum_4, output_activation_max); in arm_depthwise_conv_s8_opt() 369 *output++ = (int8_t)sum_4; in arm_depthwise_conv_s8_opt()
|
D | arm_depthwise_conv_fast_s16.c | 290 int32_t sum_4 = 0; in arm_depthwise_conv_fast_s16() local 338 sum_4 = SMLAD(col_a, row_c, sum_4); in arm_depthwise_conv_fast_s16() 351 sum_4 += row_pos[3] * col_pos[3]; in arm_depthwise_conv_fast_s16() 362 int64_t acc_4 = sum_4; in arm_depthwise_conv_fast_s16()
|