/external/libaom/libaom/av1/common/arm/ |
D | av1_inv_txfm_neon.c | 293 x[0] = vqaddq_s16(s0, s4); in iadst8_neon() 294 x[1] = vqaddq_s16(s1, s5); in iadst8_neon() 295 x[2] = vqaddq_s16(s2, s6); in iadst8_neon() 296 x[3] = vqaddq_s16(s3, s7); in iadst8_neon() 311 x[0] = vqaddq_s16(s0, s2); in iadst8_neon() 312 x[1] = vqaddq_s16(s1, s3); in iadst8_neon() 315 x[4] = vqaddq_s16(s4, s6); in iadst8_neon() 316 x[5] = vqaddq_s16(s5, s7); in iadst8_neon() 405 step2[4] = vqaddq_s16(step1[4], step1[5]); in idct8_neon() 408 step2[7] = vqaddq_s16(step1[7], step1[6]); in idct8_neon() [all …]
|
/external/libvpx/libvpx/vp8/common/arm/neon/ |
D | idct_blk_neon.c | 140 q10 = vqaddq_s16(q2, q3); in idct_dequant_full_2x_neon() 146 q4 = vqaddq_s16(q4, q8); in idct_dequant_full_2x_neon() 147 q5 = vqaddq_s16(q5, q9); in idct_dequant_full_2x_neon() 150 q3 = vqaddq_s16(q7, q4); in idct_dequant_full_2x_neon() 152 q4 = vqaddq_s16(q10, q3); in idct_dequant_full_2x_neon() 153 q5 = vqaddq_s16(q11, q2); in idct_dequant_full_2x_neon() 170 q2 = vqaddq_s16(q2tmp2.val[0], q2tmp3.val[0]); in idct_dequant_full_2x_neon() 176 q10 = vqaddq_s16(q2tmp2.val[1], q10); in idct_dequant_full_2x_neon() 177 q11 = vqaddq_s16(q2tmp3.val[1], q11); in idct_dequant_full_2x_neon() 180 q9 = vqaddq_s16(q9, q10); in idct_dequant_full_2x_neon() [all …]
|
D | sixtappredict_neon.c | 149 d0 = vqaddq_s16(vreinterpretq_s16_u16(c2), vreinterpretq_s16_u16(c0)); in yonly4x4() 150 d1 = vqaddq_s16(vreinterpretq_s16_u16(c3), vreinterpretq_s16_u16(c1)); in yonly4x4() 264 vqaddq_s16(vreinterpretq_s16_u16(d0123), vreinterpretq_s16_u16(d0123_a)); in vp8_sixtap_predict4x4_neon() 266 vqaddq_s16(vreinterpretq_s16_u16(d4567), vreinterpretq_s16_u16(d4567_a)); in vp8_sixtap_predict4x4_neon() 343 vqaddq_s16(vreinterpretq_s16_u16(d0123), vreinterpretq_s16_u16(d0123_a)); in vp8_sixtap_predict4x4_neon() 345 vqaddq_s16(vreinterpretq_s16_u16(d4567), vreinterpretq_s16_u16(d4567_a)); in vp8_sixtap_predict4x4_neon() 346 e89 = vqaddq_s16(vreinterpretq_s16_u16(d89), vreinterpretq_s16_u16(d89_a)); in vp8_sixtap_predict4x4_neon() 381 d0 = vqaddq_s16(vreinterpretq_s16_u16(c2), vreinterpretq_s16_u16(c0)); in vp8_sixtap_predict4x4_neon() 382 d1 = vqaddq_s16(vreinterpretq_s16_u16(c3), vreinterpretq_s16_u16(c1)); in vp8_sixtap_predict4x4_neon() 479 q7s16 = vqaddq_s16(q7s16, q3s16); in vp8_sixtap_predict8x4_neon() [all …]
|
D | shortidct4x4llm_neon.c | 51 q4s16 = vqaddq_s16(q4s16, q2s16); in vp8_short_idct4x4llm_neon() 80 q4s16 = vqaddq_s16(q4s16, q2s16); in vp8_short_idct4x4llm_neon()
|
/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/webp/src/dsp/ |
D | upsampling_neon.c | 150 const int16x8_t R1 = vqaddq_s16(Y1, R_Rounder); \ 151 const int16x8_t G2 = vqaddq_s16(Y1, G_Rounder); \ 152 const int16x8_t B1 = vqaddq_s16(Y1, B_Rounder); \ 153 const int16x8_t R2 = vqaddq_s16(R0, R1); \ 154 const int16x8_t G3 = vqaddq_s16(G0, G1); \ 155 const int16x8_t B2 = vqaddq_s16(B0, B1); \ 157 const int16x8_t B3 = vqaddq_s16(B2, U0); \
|
/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/libaom/libaom/av1/encoder/arm/neon/ |
D | quantize_neon.c | 58 const int16x8_t v_tmp = vqaddq_s16(v_abs, v_round); in av1_quantize_fp_neon() 84 const int16x8_t v_tmp = vqaddq_s16(v_abs, v_round); in av1_quantize_fp_neon() 156 const int16x8_t v_tmp = vqaddq_s16(v_abs, v_round); in av1_quantize_lp_neon() 181 const int16x8_t v_tmp = vqaddq_s16(v_abs, v_round); in av1_quantize_lp_neon()
|
/external/libvpx/libvpx/vpx_dsp/arm/ |
D | highbd_idct_neon.h | 29 const int16x8_t b = vqaddq_s16(res, a); in highbd_idct4x4_1_add_kernel1() 123 d0_s16 = vqaddq_s16(a[0], vreinterpretq_s16_u16(d0)); in highbd_add8x8() 124 d1_s16 = vqaddq_s16(a[1], vreinterpretq_s16_u16(d1)); in highbd_add8x8() 125 d2_s16 = vqaddq_s16(a[2], vreinterpretq_s16_u16(d2)); in highbd_add8x8() 126 d3_s16 = vqaddq_s16(a[3], vreinterpretq_s16_u16(d3)); in highbd_add8x8() 127 d4_s16 = vqaddq_s16(a[4], vreinterpretq_s16_u16(d4)); in highbd_add8x8() 128 d5_s16 = vqaddq_s16(a[5], vreinterpretq_s16_u16(d5)); in highbd_add8x8() 129 d6_s16 = vqaddq_s16(a[6], vreinterpretq_s16_u16(d6)); in highbd_add8x8() 130 d7_s16 = vqaddq_s16(a[7], vreinterpretq_s16_u16(d7)); in highbd_add8x8()
|
D | quantize_neon.c | 67 const int16x8_t rounded = vqaddq_s16(coeff_abs, round); in vpx_quantize_b_neon() 117 const int16x8_t rounded = vqaddq_s16(coeff_abs, round); in vpx_quantize_b_neon() 227 const int16x8_t rounded = vqaddq_s16(coeff_abs, round); in vpx_quantize_b_32x32_neon() 275 const int16x8_t rounded = vqaddq_s16(coeff_abs, round); in vpx_quantize_b_32x32_neon()
|
D | idct_neon.h | 47 return vqaddq_s16(a, b); in final_add() 672 out[0] = vqaddq_s16(step2[0], step2[15]); in idct16x16_add_stage7() 673 out[1] = vqaddq_s16(step2[1], step2[14]); in idct16x16_add_stage7() 674 out[2] = vqaddq_s16(step2[2], step2[13]); in idct16x16_add_stage7() 675 out[3] = vqaddq_s16(step2[3], step2[12]); in idct16x16_add_stage7() 676 out[4] = vqaddq_s16(step2[4], step2[11]); in idct16x16_add_stage7() 677 out[5] = vqaddq_s16(step2[5], step2[10]); in idct16x16_add_stage7() 678 out[6] = vqaddq_s16(step2[6], step2[9]); in idct16x16_add_stage7() 679 out[7] = vqaddq_s16(step2[7], step2[8]); in idct16x16_add_stage7() 802 const int16x8_t res0 = vqaddq_s16(a, vreinterpretq_s16_u16(s)); in highbd_idct16x16_add8x1()
|
/external/libvpx/libvpx/vp9/encoder/arm/neon/ |
D | vp9_quantize_neon.c | 75 const int16x8_t v_tmp = vqaddq_s16(v_abs, v_round); in vp9_quantize_fp_neon() 100 const int16x8_t v_tmp = vqaddq_s16(v_abs, v_round); in vp9_quantize_fp_neon() 165 int16x8_t qcoeff = vqaddq_s16(coeff_abs, round); in vp9_quantize_fp_32x32_neon() 222 int16x8_t qcoeff = vqaddq_s16(coeff_abs, round); in vp9_quantize_fp_32x32_neon()
|
/external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
D | depthwiseconv_uint8_transitional.h | 3417 acc_s16_0_1 = vqaddq_s16(acc_s16_0_1, output_offset_vec); 3418 acc_s16_2_3 = vqaddq_s16(acc_s16_2_3, output_offset_vec); 3493 acc_s16_0_1 = vqaddq_s16(acc_s16_0_1, output_offset_vec); 3494 acc_s16_2_3 = vqaddq_s16(acc_s16_2_3, output_offset_vec); 3561 acc_s16_0_1 = vqaddq_s16(acc_s16_0_1, output_offset_vec); 3562 acc_s16_2_3 = vqaddq_s16(acc_s16_2_3, output_offset_vec); 3616 acc_s16_0_1 = vqaddq_s16(acc_s16_0_1, output_offset_vec); 3617 acc_s16_2_3 = vqaddq_s16(acc_s16_2_3, output_offset_vec); 3716 acc_s16_0_1 = vqaddq_s16(acc_s16_0_1, output_offset_vec); 3717 acc_s16_2_3 = vqaddq_s16(acc_s16_2_3, output_offset_vec); [all …]
|
/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()
|
/external/gemmlowp/fixedpoint/ |
D | fixedpoint_neon.h | 291 const int16x8_t fixed_up_x = vqaddq_s16(x, fixup); 307 const int16x8_t fixed_up_x = vqaddq_s16(x, fixup); 334 const int16x8_t fixed_up_x = vqaddq_s16(x, fixup); 352 return vqaddq_s16(a, b);
|
/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/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()
|