| /external/libhevc/common/arm/ |
| D | ihevc_quant_iquant_ssd_neon_intr.c | 203 q0 = vqmovn_s32(qtmp_0); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
204 q1 = vqmovn_s32(qtmp_1); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
205 q2 = vqmovn_s32(qtmp_2); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
206 q3 = vqmovn_s32(qtmp_3); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
261 iq0 = vqmovn_s32(iqtmp_0); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
262 iq1 = vqmovn_s32(iqtmp_1); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
263 iq2 = vqmovn_s32(iqtmp_2); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
264 iq3 = vqmovn_s32(iqtmp_3); in ihevc_quant_iquant_ssd_flat_scale_mat_neon()
480 q_00 = vcombine_s16(vqmovn_s32(q_tmp0), vqmovn_s32(q_tmp1)); in ihevc_q_iq_ssd_flat_scale_mat_var_rnd_fact_neon()
481 q_01 = vcombine_s16(vqmovn_s32(q_tmp2), vqmovn_s32(q_tmp3)); in ihevc_q_iq_ssd_flat_scale_mat_var_rnd_fact_neon()
[all …]
|
| /external/XNNPACK/src/q8-vadd/ |
| D | neon.c | 87 …const int16x8_t vacc0 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0_lo), vacc0_hi), vy_zero_point); in xnn_q8_vadd_ukernel__neon()
88 …const int16x8_t vacc1 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1_lo), vacc1_hi), vy_zero_point); in xnn_q8_vadd_ukernel__neon()
89 …const int16x8_t vacc2 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2_lo), vacc2_hi), vy_zero_point); in xnn_q8_vadd_ukernel__neon()
90 …const int16x8_t vacc3 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc3_lo), vacc3_hi), vy_zero_point); in xnn_q8_vadd_ukernel__neon()
140 …const int16x8_t vacc0 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0_lo), vqmovn_s32(vacc0_hi)), vy_ze… in xnn_q8_vadd_ukernel__neon()
141 …const int16x8_t vacc1 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc1_lo), vqmovn_s32(vacc1_hi)), vy_ze… in xnn_q8_vadd_ukernel__neon()
182 const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), vy_zero_point); in xnn_q8_vadd_ukernel__neon()
184 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), vy_zero_… in xnn_q8_vadd_ukernel__neon()
225 const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), vy_zero_point); in xnn_q8_vadd_ukernel__neon()
227 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), vy_zero_… in xnn_q8_vadd_ukernel__neon()
|
| /external/XNNPACK/src/requantization/ |
| D | q31-neon.c | 82 …const int16x8_t xy_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(x_scaled), y_scaled), vzero_poin… in xnn_requantize_q31__neon()
83 …const int16x8_t zw_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(z_scaled), w_scaled), vzero_poin… in xnn_requantize_q31__neon()
86 …const int16x8_t xy_packed = vqaddq_s16(vcombine_s16(vqmovn_s32(x_scaled), vqmovn_s32(y_scaled)), v… in xnn_requantize_q31__neon()
87 …const int16x8_t zw_packed = vqaddq_s16(vcombine_s16(vqmovn_s32(z_scaled), vqmovn_s32(w_scaled)), v… in xnn_requantize_q31__neon()
|
| D | gemmlowp-neon.c | 90 …const int16x8_t xy_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(x_scaled), y_scaled), vzero_poin… in xnn_requantize_gemmlowp__neon()
91 …const int16x8_t zw_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(z_scaled), w_scaled), vzero_poin… in xnn_requantize_gemmlowp__neon()
94 …const int16x8_t xy_packed = vqaddq_s16(vcombine_s16(vqmovn_s32(x_scaled), vqmovn_s32(y_scaled)), v… in xnn_requantize_gemmlowp__neon()
95 …const int16x8_t zw_packed = vqaddq_s16(vcombine_s16(vqmovn_s32(z_scaled), vqmovn_s32(w_scaled)), v… in xnn_requantize_gemmlowp__neon()
|
| D | precise-neon.c | 114 …const int16x8_t xy_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(x_scaled), y_scaled), vzero_poin… in xnn_requantize_precise__neon()
115 …const int16x8_t zw_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(z_scaled), w_scaled), vzero_poin… in xnn_requantize_precise__neon()
123 …const int16x8_t xy_packed = vqaddq_s16(vcombine_s16(vqmovn_s32(x_scaled), vqmovn_s32(y_scaled)), v… in xnn_requantize_precise__neon()
124 …const int16x8_t zw_packed = vqaddq_s16(vcombine_s16(vqmovn_s32(z_scaled), vqmovn_s32(w_scaled)), v… in xnn_requantize_precise__neon()
|
| D | fp32-neon.c | 73 …const int16x8_t xy_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(x_rounded), y_rounded), vzero_po… in xnn_requantize_fp32__neon()
74 …const int16x8_t zw_packed = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(z_rounded), w_rounded), vzero_po… in xnn_requantize_fp32__neon()
|
| /external/gemmlowp/internal/ |
| D | output_neon.h | 38 int16x4_t res_16 = vqmovn_s32(input.reg[0]);
58 vcombine_s16(vqmovn_s32(input.reg[0]), vqmovn_s32(input.reg[1]));
77 vcombine_s16(vqmovn_s32(input.reg[0]), vqmovn_s32(input.reg[1]));
79 vcombine_s16(vqmovn_s32(input.reg[2]), vqmovn_s32(input.reg[3]));
100 res_16[i] = vcombine_s16(vqmovn_s32(input.reg[2 * i]),
101 vqmovn_s32(input.reg[2 * i + 1]));
122 int16x4_t res_16 = vqmovn_s32(input.reg[0]);
142 vcombine_s16(vqmovn_s32(input.reg[0]), vqmovn_s32(input.reg[1]));
161 vcombine_s16(vqmovn_s32(input.reg[0]), vqmovn_s32(input.reg[1]));
163 vcombine_s16(vqmovn_s32(input.reg[2]), vqmovn_s32(input.reg[3]));
[all …]
|
| /external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
| D | depthwiseconv_uint8_transitional.h | 3414 vcombine_s16(vqmovn_s32(acc0), vqmovn_s32(acc1));
3416 vcombine_s16(vqmovn_s32(acc2), vqmovn_s32(acc3));
3490 vcombine_s16(vqmovn_s32(acc0), vqmovn_s32(acc1));
3492 vcombine_s16(vqmovn_s32(acc2), vqmovn_s32(acc3));
3558 vcombine_s16(vqmovn_s32(acc0), vqmovn_s32(acc1));
3560 vcombine_s16(vqmovn_s32(acc2), vqmovn_s32(acc3));
3613 vcombine_s16(vqmovn_s32(acc0), vqmovn_s32(acc1));
3615 vcombine_s16(vqmovn_s32(acc2), vqmovn_s32(acc3));
3713 vcombine_s16(vqmovn_s32(acc0), vqmovn_s32(acc1));
3715 vcombine_s16(vqmovn_s32(acc2), vqmovn_s32(acc3));
[all …]
|
| D | neon_tensor_utils.cc | 862 vcombine_s16(vqmovn_s32(temp_val.val[0]), vqmovn_s32(temp_val.val[1])); in NeonMatrixBatchVectorAccumulateImpl()
932 vcombine_s16(vqmovn_s32(temp_val.val[0]), vqmovn_s32(temp_val.val[1])); in NeonMatrixBatchVectorAccumulateImpl()
934 vcombine_s16(vqmovn_s32(temp_val.val[2]), vqmovn_s32(temp_val.val[3])); in NeonMatrixBatchVectorAccumulateImpl()
1451 vst1_s16(output + index, vqmovn_s32(val5_s32.val[0])); in NeonApplyLayerNorm()
1452 vst1_s16(output + index + 4, vqmovn_s32(val5_s32.val[1])); in NeonApplyLayerNorm()
1453 vst1_s16(output + index + 8, vqmovn_s32(val5_s32.val[2])); in NeonApplyLayerNorm()
1454 vst1_s16(output + index + 12, vqmovn_s32(val5_s32.val[3])); in NeonApplyLayerNorm()
1678 vst1_s16(output + index, vqmovn_s32(sum_0)); in NeonCwiseAdd()
1679 vst1_s16(output + index + 4, vqmovn_s32(sum_1)); in NeonCwiseAdd()
|
| D | legacy_optimized_ops.h | 973 const int16x4_t res16 = vqmovn_s32(reduced); in LegacyFullyConnectedAsGEMVWorkerImpl()
1334 const int16x4_t res16 = vqmovn_s32(reduced); in GEMVForLstmCell()
1626 const int16x4_t res16 = vqmovn_s32(reduced); in GEMVForLstmCellWithSymmetricRange()
1984 const int16x4_t res16 = vqmovn_s32(reduced); in LegacyInt8FullyConnectedAsGEMVWorkerImpl()
4150 vcombine_s16(vqmovn_s32(output_s32_0), vqmovn_s32(output_s32_1)); in Softmax()
4345 int16x8_t output_val_s16_0 = vcombine_s16(vqmovn_s32(output_val_s32_0), in Logistic()
4346 vqmovn_s32(output_val_s32_1)); in Logistic()
4347 int16x8_t output_val_s16_1 = vcombine_s16(vqmovn_s32(output_val_s32_2), in Logistic()
4348 vqmovn_s32(output_val_s32_3)); in Logistic()
4547 int16x8_t output_val_s16_0 = vcombine_s16(vqmovn_s32(output_val_s32_0), in Tanh()
[all …]
|
| /external/XNNPACK/src/q8-igemm/ |
| D | 8x8-neon.c | 552 …const int16x8_t vacc0x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0x0123), vacc0x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
553 …const int16x8_t vacc1x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1x0123), vacc1x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
554 …const int16x8_t vacc2x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2x0123), vacc2x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
555 …const int16x8_t vacc3x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc3x0123), vacc3x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
556 …const int16x8_t vacc4x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc4x0123), vacc4x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
557 …const int16x8_t vacc5x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc5x0123), vacc5x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
558 …const int16x8_t vacc6x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc6x0123), vacc6x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
559 …const int16x8_t vacc7x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc7x0123), vacc7x4567), v… in xnn_q8_igemm_ukernel_8x8__neon()
567 vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0x0123), vqmovn_s32(vacc0x4567)), voutput_zero_point); in xnn_q8_igemm_ukernel_8x8__neon()
569 vqaddq_s16(vcombine_s16(vqmovn_s32(vacc1x0123), vqmovn_s32(vacc1x4567)), voutput_zero_point); in xnn_q8_igemm_ukernel_8x8__neon()
[all …]
|
| D | 4x8-neon.c | 352 …const int16x8_t vacc0x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0x0123), vacc0x4567), v… in xnn_q8_igemm_ukernel_4x8__neon()
353 …const int16x8_t vacc1x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1x0123), vacc1x4567), v… in xnn_q8_igemm_ukernel_4x8__neon()
354 …const int16x8_t vacc2x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2x0123), vacc2x4567), v… in xnn_q8_igemm_ukernel_4x8__neon()
355 …const int16x8_t vacc3x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc3x0123), vacc3x4567), v… in xnn_q8_igemm_ukernel_4x8__neon()
360 …const int16x8_t vacc0x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0x0123), vqmovn_s32(vacc0x… in xnn_q8_igemm_ukernel_4x8__neon()
361 …const int16x8_t vacc1x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc1x0123), vqmovn_s32(vacc1x… in xnn_q8_igemm_ukernel_4x8__neon()
362 …const int16x8_t vacc2x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc2x0123), vqmovn_s32(vacc2x… in xnn_q8_igemm_ukernel_4x8__neon()
363 …const int16x8_t vacc3x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc3x0123), vqmovn_s32(vacc3x… in xnn_q8_igemm_ukernel_4x8__neon()
|
| /external/XNNPACK/src/q8-gemm/ |
| D | 8x8-neon.c | 507 …const int16x8_t vacc0x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0x0123), vacc0x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
508 …const int16x8_t vacc1x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1x0123), vacc1x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
509 …const int16x8_t vacc2x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2x0123), vacc2x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
510 …const int16x8_t vacc3x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc3x0123), vacc3x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
511 …const int16x8_t vacc4x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc4x0123), vacc4x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
512 …const int16x8_t vacc5x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc5x0123), vacc5x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
513 …const int16x8_t vacc6x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc6x0123), vacc6x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
514 …const int16x8_t vacc7x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc7x0123), vacc7x4567), v… in xnn_q8_gemm_ukernel_8x8__neon()
522 vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0x0123), vqmovn_s32(vacc0x4567)), voutput_zero_point); in xnn_q8_gemm_ukernel_8x8__neon()
524 vqaddq_s16(vcombine_s16(vqmovn_s32(vacc1x0123), vqmovn_s32(vacc1x4567)), voutput_zero_point); in xnn_q8_gemm_ukernel_8x8__neon()
[all …]
|
| D | 4x8-neon.c | 315 …const int16x8_t vacc0x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0x0123), vacc0x4567), v… in xnn_q8_gemm_ukernel_4x8__neon()
316 …const int16x8_t vacc1x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1x0123), vacc1x4567), v… in xnn_q8_gemm_ukernel_4x8__neon()
317 …const int16x8_t vacc2x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2x0123), vacc2x4567), v… in xnn_q8_gemm_ukernel_4x8__neon()
318 …const int16x8_t vacc3x01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc3x0123), vacc3x4567), v… in xnn_q8_gemm_ukernel_4x8__neon()
323 …const int16x8_t vacc0x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0x0123), vqmovn_s32(vacc0x… in xnn_q8_gemm_ukernel_4x8__neon()
324 …const int16x8_t vacc1x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc1x0123), vqmovn_s32(vacc1x… in xnn_q8_gemm_ukernel_4x8__neon()
325 …const int16x8_t vacc2x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc2x0123), vqmovn_s32(vacc2x… in xnn_q8_gemm_ukernel_4x8__neon()
326 …const int16x8_t vacc3x01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc3x0123), vqmovn_s32(vacc3x… in xnn_q8_gemm_ukernel_4x8__neon()
|
| /external/XNNPACK/src/q8-avgpool/ |
| D | up9-neon.c | 138 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_avgpool_ukernel_up9__neon()
143 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_avgpool_ukernel_up9__neon()
211 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_avgpool_ukernel_up9__neon()
216 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_avgpool_ukernel_up9__neon()
|
| D | mp9p8q-neon.c | 218 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_avgpool_ukernel_mp9p8q__neon()
223 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_avgpool_ukernel_mp9p8q__neon()
292 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_avgpool_ukernel_mp9p8q__neon()
297 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_avgpool_ukernel_mp9p8q__neon()
|
| /external/XNNPACK/src/q8-gavgpool/ |
| D | up7-neon.c | 120 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_gavgpool_ukernel_up7__neon()
125 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_gavgpool_ukernel_up7__neon()
190 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_gavgpool_ukernel_up7__neon()
195 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_gavgpool_ukernel_up7__neon()
|
| D | mp7p7q-neon.c | 196 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_gavgpool_ukernel_mp7p7q__neon()
201 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_gavgpool_ukernel_mp7p7q__neon()
268 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_gavgpool_ukernel_mp7p7q__neon()
273 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_gavgpool_ukernel_mp7p7q__neon()
|
| /external/tensorflow/tensorflow/lite/kernels/internal/optimized/integer_ops/ |
| D | mul.h | 76 const auto p1_narrowed = vqmovn_s32(p1); in MulElementwise()
77 const auto p2_narrowed = vqmovn_s32(p2); in MulElementwise()
146 const auto p1_narrowed = vqmovn_s32(p1); in MulSimpleBroadcast()
147 const auto p2_narrowed = vqmovn_s32(p2); in MulSimpleBroadcast()
|
| D | softmax.h | 203 vcombine_s16(vqmovn_s32(output_s32_0), vqmovn_s32(output_s32_1)); in Softmax()
|
| /external/XNNPACK/src/q8-dwconv/ |
| D | up8x9-neon.c | 201 …const int16x8_t vacc0 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0_lo), vacc0_hi), voutput_zero_p… in xnn_q8_dwconv_ukernel_up8x9__neon()
202 …const int16x8_t vacc1 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1_lo), vacc1_hi), voutput_zero_p… in xnn_q8_dwconv_ukernel_up8x9__neon()
203 …const int16x8_t vacc2 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2_lo), vacc2_hi), voutput_zero_p… in xnn_q8_dwconv_ukernel_up8x9__neon()
359 …const int16x8_t vacc0 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0_lo), vacc0_hi), voutput_zero_p… in xnn_q8_dwconv_ukernel_up8x9__neon()
360 …const int16x8_t vacc1 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1_lo), vacc1_hi), voutput_zero_p… in xnn_q8_dwconv_ukernel_up8x9__neon()
361 …const int16x8_t vacc2 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc2_lo), vacc2_hi), voutput_zero_p… in xnn_q8_dwconv_ukernel_up8x9__neon()
499 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_dwconv_ukernel_up8x9__neon()
501 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_dwconv_ukernel_up8x9__neon()
590 …const int16x8_t vacc = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc_lo), vacc_hi), voutput_zero_poin… in xnn_q8_dwconv_ukernel_up8x9__neon()
592 …const int16x8_t vacc = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc_lo), vqmovn_s32(vacc_hi)), voutput_… in xnn_q8_dwconv_ukernel_up8x9__neon()
|
| /external/libaom/libaom/av1/common/arm/ |
| D | jnt_convolve_neon.c | 48 tmp0 = vqmovn_s32(dst0); in compute_avg_4x1()
95 tmp0 = vqmovn_s32(dst0); in compute_avg_8x1()
96 tmp2 = vqmovn_s32(dst2); in compute_avg_8x1()
159 tmp0 = vqmovn_s32(dst0); in compute_avg_4x4()
160 tmp1 = vqmovn_s32(dst1); in compute_avg_4x4()
161 tmp2 = vqmovn_s32(dst2); in compute_avg_4x4()
162 tmp3 = vqmovn_s32(dst3); in compute_avg_4x4()
263 tmp0 = vqmovn_s32(dst0); in compute_avg_8x4()
264 tmp1 = vqmovn_s32(dst1); in compute_avg_8x4()
265 tmp2 = vqmovn_s32(dst2); in compute_avg_8x4()
[all …]
|
| /external/tensorflow/tensorflow/lite/kernels/ |
| D | cpu_backend_gemm_custom_gemv.h | 234 const int16x4_t res16 = vqmovn_s32(src); in ClampAndStore()
250 const int16x4_t res16 = vqmovn_s32(src); in ClampAndStore()
266 int16x4_t res16 = vqmovn_s32(src); in ClampAndStore()
|
| /external/libaom/libaom/aom_dsp/simd/ |
| D | v64_intrinsics_arm.h | 316 return vreinterpret_s64_s16(vqmovn_s32(vpaddlq_s16(t))); in v64_madd_us8()
446 return vreinterpret_s64_s16(vqmovn_s32( in v64_pack_s32_s16()
|
| /external/libgav1/libgav1/src/dsp/arm/ |
| D | inverse_transform_neon.cc | 1829 vcombine_s16(vqmovn_s32(shift_lo), vqmovn_s32(shift_hi))); in Identity4_NEON()
1863 vst1_lane_s16(&dst[0], vqmovn_s32(dst_0), 0); in Identity4DcOnly()
2049 vst1_lane_s16(&dst[0], vqmovn_s32(dst_0), 0); in Identity8DcOnly()
2070 vcombine_s16(vqmovn_s32(shift_lo), vqmovn_s32(shift_hi))); in Identity16Row_NEON()
2096 vst1_lane_s16(&dst[0], vqmovn_s32(dst_0), 0); in Identity16DcOnly()
|