/external/libjpeg-turbo/simd/arm/ |
D | jcphuff-neon.c | 75 coefs1 = vshlq_s16(abs_coefs1, vdupq_n_s16(-Al)); in jsimd_encode_mcu_AC_first_prepare_neon() 76 coefs2 = vshlq_s16(abs_coefs2, vdupq_n_s16(-Al)); in jsimd_encode_mcu_AC_first_prepare_neon() 104 int16x8_t coefs2 = vdupq_n_s16(0); in jsimd_encode_mcu_AC_first_prepare_neon() 130 coefs1 = vshlq_s16(abs_coefs1, vdupq_n_s16(-Al)); in jsimd_encode_mcu_AC_first_prepare_neon() 131 coefs2 = vshlq_s16(abs_coefs2, vdupq_n_s16(-Al)); in jsimd_encode_mcu_AC_first_prepare_neon() 147 int16x8_t coefs = vdupq_n_s16(0); in jsimd_encode_mcu_AC_first_prepare_neon() 174 coefs = vshlq_s16(abs_coefs, vdupq_n_s16(-Al)); in jsimd_encode_mcu_AC_first_prepare_neon() 189 vst1q_s16(values_ptr, vdupq_n_s16(0)); in jsimd_encode_mcu_AC_first_prepare_neon() 190 vst1q_s16(diff_values_ptr, vdupq_n_s16(0)); in jsimd_encode_mcu_AC_first_prepare_neon() 207 uint8x8_t row0_eq0 = vmovn_u16(vceqq_s16(row0, vdupq_n_s16(0))); in jsimd_encode_mcu_AC_first_prepare_neon() [all …]
|
/external/libvpx/libvpx/vpx_dsp/arm/ |
D | quantize_neon.c | 41 const int16x8_t one = vdupq_n_s16(1); in vpx_quantize_b_neon() 42 const int16x8_t neg_one = vdupq_n_s16(-1); in vpx_quantize_b_neon() 99 const int16x8_t zbin = vdupq_n_s16(zbin_ptr[1]); in vpx_quantize_b_neon() 100 const int16x8_t round = vdupq_n_s16(round_ptr[1]); in vpx_quantize_b_neon() 101 const int16x8_t quant = vdupq_n_s16(quant_ptr[1]); in vpx_quantize_b_neon() 102 const int16x8_t quant_shift = vdupq_n_s16(quant_shift_ptr[1]); in vpx_quantize_b_neon() 103 const int16x8_t dequant = vdupq_n_s16(dequant_ptr[1]); in vpx_quantize_b_neon() 199 const int16x8_t one = vdupq_n_s16(1); in vpx_quantize_b_32x32_neon() 200 const int16x8_t neg_one = vdupq_n_s16(-1); in vpx_quantize_b_32x32_neon() 257 const int16x8_t zbin = vrshrq_n_s16(vdupq_n_s16(zbin_ptr[1]), 1); in vpx_quantize_b_32x32_neon() [all …]
|
D | highbd_idct4x4_add_neon.c | 39 const int16x8_t max = vdupq_n_s16((1 << bd) - 1); in vpx_highbd_idct4x4_1_add_neon() 45 const int16x8_t dc = vdupq_n_s16(a1); in vpx_highbd_idct4x4_1_add_neon() 53 const int16x8_t max = vdupq_n_s16((1 << bd) - 1); in vpx_highbd_idct4x4_16_add_neon()
|
D | highbd_idct32x32_add_neon.c | 69 const int16x8_t dc = vdupq_n_s16(a1); in vpx_highbd_idct32x32_1_add_neon() 73 const int16x8_t max = vdupq_n_s16((1 << bd) - 1); in vpx_highbd_idct32x32_1_add_neon()
|
D | idct4x4_1_add_neon.c | 39 const int16x8_t dc = vdupq_n_s16(a1); in vpx_idct4x4_1_add_neon()
|
/external/libhevc/encoder/arm/ |
D | ihevce_itrans_recon_neon.c | 66 a0 = vdupq_n_s16(dc_value); in ihevce_itrans_recon_dc_4x4_luma_neon() 95 a0 = vdupq_n_s16(dc_value); in ihevce_itrans_recon_dc_4x4_chroma_neon() 118 a0 = (dc_value >= 0) ? vqmovun_s16(vdupq_n_s16(dc_value)) in ihevce_itrans_recon_dc_8x8_luma_neon() 119 : vqmovun_s16(vabsq_s16(vdupq_n_s16(dc_value))); in ihevce_itrans_recon_dc_8x8_luma_neon() 146 a1 = (dc_value >= 0) ? vqmovun_s16(vdupq_n_s16(dc_value)) in ihevce_itrans_recon_dc_8x8_chroma_neon() 147 : vqmovun_s16(vabsq_s16(vdupq_n_s16(dc_value))); in ihevce_itrans_recon_dc_8x8_chroma_neon() 176 a0 = (dc_value >= 0) ? vqmovun_s16(vdupq_n_s16(dc_value)) in ihevce_itrans_recon_dc_16x16_luma_neon() 177 : vqmovun_s16(vabsq_s16(vdupq_n_s16(dc_value))); in ihevce_itrans_recon_dc_16x16_luma_neon() 201 a0 = (dc_value >= 0) ? vqmovun_s16(vdupq_n_s16(dc_value)) in ihevce_itrans_recon_dc_16x16_chroma_neon() 202 : vqmovun_s16(vabsq_s16(vdupq_n_s16(dc_value))); in ihevce_itrans_recon_dc_16x16_chroma_neon()
|
D | ihevce_scale_by_2_neon.c | 65 const int16x8_t prec = vdupq_n_s16(8192); in ihevce_horz_scale_neon_w16() 66 const int16x8_t inv_prec = vdupq_n_s16(64); in ihevce_horz_scale_neon_w16() 116 const int16x8_t prec = vdupq_n_s16(8192); in ihevce_vert_scale_neon_w16() 117 const int16x8_t inv_prec = vdupq_n_s16(64); in ihevce_vert_scale_neon_w16()
|
/external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
D | depthwiseconv_uint8.h | 49 vdupq_n_s16(filter_offset)); 63 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 92 const int16x8_t filter = vaddq_s16(filter_s16, vdupq_n_s16(filter_offset)); 113 input[i] = vaddq_s16(input[i], vdupq_n_s16(input_offset)); 139 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 160 const int16x8_t filter = vaddq_s16(filter_s16, vdupq_n_s16(filter_offset)); 174 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 232 filter[i] = vaddq_s16(filter_s16, vdupq_n_s16(filter_offset)); 327 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 399 input[i] = vaddq_s16(input[i], vdupq_n_s16(input_offset)); [all …]
|
/external/libaom/libaom/av1/encoder/arm/neon/ |
D | quantize_neon.c | 42 const int16x8_t v_zero = vdupq_n_s16(0); in av1_quantize_fp_neon() 43 const int16x8_t v_one = vdupq_n_s16(1); in av1_quantize_fp_neon() 44 int16x8_t v_eobmax_76543210 = vdupq_n_s16(-1); in av1_quantize_fp_neon() 139 const int16x8_t v_zero = vdupq_n_s16(0); in av1_quantize_lp_neon() 140 const int16x8_t v_one = vdupq_n_s16(1); in av1_quantize_lp_neon() 141 int16x8_t v_eobmax_76543210 = vdupq_n_s16(-1); in av1_quantize_lp_neon()
|
/external/libaom/libaom/av1/common/arm/ |
D | cfl_neon.c | 488 const int16x8_t alpha_sign = vdupq_n_s16(alpha_q3); in cfl_predict_lbd_neon() 489 const int16x8_t dc = vdupq_n_s16(*dst); in cfl_predict_lbd_neon() 522 return vreinterpretq_u16_s16(vmaxq_s16(vminq_s16(a, max), vdupq_n_s16(0))); in clampq_s16() 528 vmaxq_s16(vminq_s16(a.val[0], max), vdupq_n_s16(0))); in clamp2q_s16() 530 vmaxq_s16(vminq_s16(a.val[1], max), vdupq_n_s16(0))); in clamp2q_s16() 537 vmaxq_s16(vminq_s16(a.val[0], max), vdupq_n_s16(0))); in clamp4q_s16() 539 vmaxq_s16(vminq_s16(a.val[1], max), vdupq_n_s16(0))); in clamp4q_s16() 541 vmaxq_s16(vminq_s16(a.val[2], max), vdupq_n_s16(0))); in clamp4q_s16() 543 vmaxq_s16(vminq_s16(a.val[3], max), vdupq_n_s16(0))); in clamp4q_s16() 565 const int16x8_t alpha_sign = vdupq_n_s16(alpha_q3); in cfl_predict_hbd_neon() [all …]
|
D | jnt_convolve_neon.c | 104 const int16x8_t round_bits_vec = vdupq_n_s16(-round_bits); in compute_avg_8x1() 129 const int16x8_t zero = vdupq_n_s16(0); in compute_avg_4x4() 211 const int16x8_t zero = vdupq_n_s16(0); in compute_avg_8x4() 289 const int16x8_t round_bits_vec = vdupq_n_s16(-round_bits); in compute_avg_8x4() 437 const int16x8_t horiz_const = vdupq_n_s16((1 << (bd + FILTER_BITS - 2))); in dist_wtd_convolve_2d_horiz_neon() 438 const int16x8_t shift_round_0 = vdupq_n_s16(-(round_0)); in dist_wtd_convolve_2d_horiz_neon() 777 const int16x8_t dup_bits16x8 = vdupq_n_s16(bits); in av1_dist_wtd_convolve_2d_copy_neon() 1126 const int16x8_t round_offset128 = vdupq_n_s16(round_offset); in av1_dist_wtd_convolve_x_neon() 1128 const int16x8_t shift_round_0 = vdupq_n_s16(-conv_params->round_0 + 1); in av1_dist_wtd_convolve_x_neon() 1129 const int16x8_t horiz_const = vdupq_n_s16(bits); in av1_dist_wtd_convolve_x_neon() [all …]
|
/external/tensorflow/tensorflow/lite/kernels/internal/optimized/integer_ops/ |
D | add.h | 56 const int16x8_t input1_offset_dup = vdupq_n_s16(params.input1_offset); in AddElementwise() 57 const int16x8_t input2_offset_dup = vdupq_n_s16(params.input2_offset); in AddElementwise() 131 vdupq_n_s16(params.output_offset)); in AddElementwise() 133 vdupq_n_s16(params.output_offset)); in AddElementwise() 193 vaddq_s16(input1_val_s16, vdupq_n_s16(params.input1_offset)); in AddScalarBroadcast() 210 vaddq_s16(input2_val_s16, vdupq_n_s16(params.input2_offset)); in AddScalarBroadcast() 231 vdupq_n_s16(params.output_offset)); in AddScalarBroadcast()
|
D | mul.h | 47 const int16x8_t input1_offset_vector = vdupq_n_s16(params.input1_offset); in MulElementwise() 48 const int16x8_t input2_offset_vector = vdupq_n_s16(params.input2_offset); in MulElementwise() 49 const int16x8_t output_offset_vector = vdupq_n_s16(params.output_offset); in MulElementwise() 155 const auto input2_offset_vector = vdupq_n_s16(params.input2_offset); in MulSimpleBroadcast() 156 const auto output_offset_vector = vdupq_n_s16(params.output_offset); in MulSimpleBroadcast()
|
D | depthwise_conv.h | 72 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 121 input[i] = vaddq_s16(input[i], vdupq_n_s16(input_offset)); 147 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 181 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 328 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 397 input[i] = vaddq_s16(input[i], vdupq_n_s16(input_offset)); 422 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 501 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 559 const int16x8_t input = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); 651 input[i] = vaddq_s16(input_s16, vdupq_n_s16(input_offset)); [all …]
|
/external/gemmlowp/fixedpoint/ |
D | fixedpoint_neon.h | 74 return veorq_s16(a, vdupq_n_s16(-1)); 114 return vshlq_s16(a, vdupq_n_s16(offset)); 134 return vshlq_s16(a, vdupq_n_s16(-offset)); 176 return MaskIfEqual(a, vdupq_n_s16(0)); 289 const int16x8_t shift_vec = vdupq_n_s16(-exponent); 346 return vdupq_n_s16(x);
|
/external/libgav1/libgav1/src/dsp/arm/ |
D | intrapred_cfl_neon.cc | 45 const int16x8_t val_v = vdupq_n_s16(static_cast<int16_t>(val)); in BlockSubtract() 283 const int16x8_t dc = vdupq_n_s16(dst[0]); in CflIntraPredictor4xN_NEON() 302 const int16x8_t dc = vdupq_n_s16(dst[0]); in CflIntraPredictor8xN_NEON() 317 const int16x8_t dc = vdupq_n_s16(dst[0]); in CflIntraPredictor16xN_NEON() 335 const int16x8_t dc = vdupq_n_s16(dst[0]); in CflIntraPredictor32xN_NEON() 618 const int16x8_t averages = vdupq_n_s16(static_cast<int16_t>(average_sum)); in CflSubsampler444_8xH_NEON() 714 const int16x8_t averages = vdupq_n_s16(static_cast<int16_t>(average_sum)); in CflSubsampler444_WxH_NEON() 717 int16x8_t ssamples_ext = vdupq_n_s16(0); in CflSubsampler444_WxH_NEON() 933 const int16x8_t averages = vdupq_n_s16(static_cast<int16_t>(average_sum)); in CflSubsampler420Impl_8xH_NEON() 1039 const int16x8_t averages = vdupq_n_s16(static_cast<int16_t>(average_sum)); in CflSubsampler420Impl_WxH_NEON() [all …]
|
D | motion_vector_search_neon.cc | 57 const int16x8_t projection_mv_clamp = vdupq_n_s16(kProjectionMvClamp); in ProjectionClip() 102 const int16x8_t kRoundDownMask = vdupq_n_s16(1); in LowPrecision() 110 const int16x8_t kRoundDownMask = vdupq_n_s16(7); in ForceInteger() 113 const int16x8_t mv1 = vaddq_s16(mv0, vdupq_n_s16(3)); in ForceInteger()
|
D | film_grain_neon.cc | 656 remainder = vandq_s16(remainder, vdupq_n_s16(3)); in GetScalingFactors() 700 const int16x8_t floor = vdupq_n_s16(min_value); in BlendNoiseWithImageLuma_NEON() 701 const int16x8_t ceiling = vdupq_n_s16(max_luma); in BlendNoiseWithImageLuma_NEON() 705 const int16x8_t scaling_shift_vect16 = vdupq_n_s16(-scaling_shift); in BlendNoiseWithImageLuma_NEON() 770 const int16x8_t floor = vdupq_n_s16(min_value); in BlendChromaPlaneWithCfl_NEON() 771 const int16x8_t ceiling = vdupq_n_s16(max_chroma); in BlendChromaPlaneWithCfl_NEON() 777 const int16x8_t scaling_shift_vect16 = vdupq_n_s16(-scaling_shift); in BlendChromaPlaneWithCfl_NEON() 905 const int16x8_t floor = vdupq_n_s16(min_value); in BlendChromaPlane8bpp_NEON() 906 const int16x8_t ceiling = vdupq_n_s16(max_chroma); in BlendChromaPlane8bpp_NEON() 910 const int16x8_t scaling_shift_vect = vdupq_n_s16(-scaling_shift); in BlendChromaPlane8bpp_NEON() [all …]
|
D | intrapred_directional_neon.cc | 444 vaddq_s16(offset_y, vdupq_n_s16(kPositiveIndexOffset)); in DirectionalZone2FromLeftCol_WxH() 448 const int16x8_t shift_masked = vandq_s16(shift_upsampled, vdupq_n_s16(0x3f)); in DirectionalZone2FromLeftCol_WxH() 568 left_y = vaddq_s16(left_y, vdupq_n_s16(-ystep_remainder)); in DirectionalZone2_4xH() 685 const int16x8_t increment_left8 = vdupq_n_s16(ystep_remainder8); in DirectionalZone2_8() 691 left_y = vaddq_s16(left_y, vdupq_n_s16(-ystep_remainder)); in DirectionalZone2_8() 1018 const int16x8_t max_base_x = vdupq_n_s16(max_base_index); in DirectionalZone1_WxH() 1024 const int16x8_t block_step = vdupq_n_s16(base_step8); in DirectionalZone1_WxH() 1043 int16x8_t base_x = vaddq_s16(vdupq_n_s16(top_base_x), index_offset); in DirectionalZone1_WxH() 1080 const int16x8_t max_base_x = vdupq_n_s16(max_base_index); in DirectionalZone1_Large() 1086 const int16x8_t block_step = vdupq_n_s16(base_step8); in DirectionalZone1_Large() [all …]
|
D | motion_field_projection_neon.cc | 69 const int16x8_t projection_mv_clamp = vdupq_n_s16(kProjectionMvClamp); in MvProjectionClip() 180 const int16x8_t d_sign = vdupq_n_s16(dst_sign); in MotionFieldProjectionKernel_NEON() 231 const int16x8_t position = vaddq_s16(pos, vdupq_n_s16(x8)); in MotionFieldProjectionKernel_NEON() 289 const int16x8_t position = vaddq_s16(pos, vdupq_n_s16(x8)); in MotionFieldProjectionKernel_NEON()
|
/external/libvpx/libvpx/vp9/encoder/arm/neon/ |
D | vp9_quantize_neon.c | 54 const int16x8_t v_zero = vdupq_n_s16(0); in vp9_quantize_fp_neon() 55 const int16x8_t v_one = vdupq_n_s16(1); in vp9_quantize_fp_neon() 56 int16x8_t v_eobmax_76543210 = vdupq_n_s16(-1); in vp9_quantize_fp_neon() 146 const int16x8_t one = vdupq_n_s16(1); in vp9_quantize_fp_32x32_neon() 147 const int16x8_t neg_one = vdupq_n_s16(-1); in vp9_quantize_fp_32x32_neon()
|
/external/webp/src/dsp/ |
D | yuv_neon.c | 107 DST_s16 = vaddq_s16(tmp3, vdupq_n_s16(CST)); \ 186 const int16x8_t zero = vdupq_n_s16(0); in SharpYUVUpdateY_NEON() 187 const int16x8_t max = vdupq_n_s16(MAX_Y); in SharpYUVUpdateY_NEON() 233 const int16x8_t max = vdupq_n_s16(MAX_Y); in SharpYUVFilterRow_NEON() 234 const int16x8_t zero = vdupq_n_s16(0); in SharpYUVFilterRow_NEON()
|
D | upsampling_neon.c | 211 const int16x8_t R_Rounder = vdupq_n_s16(-14234); \ 212 const int16x8_t G_Rounder = vdupq_n_s16(8708); \ 213 const int16x8_t B_Rounder = vdupq_n_s16(-17685); \
|
D | quant.h | 32 const int16x8_t tst_ones = vdupq_n_s16(-1); in IsFlat()
|
/external/libvpx/libvpx/vp8/common/arm/neon/ |
D | idct_blk_neon.c | 27 q2Add.val[0] = vdupq_n_s16((int16_t)a0); in idct_dequant_0_2x_neon() 28 q2Add.val[1] = vdupq_n_s16((int16_t)a1); in idct_dequant_0_2x_neon() 71 int16x8_t qEmpty = vdupq_n_s16(0); in idct_dequant_full_2x_neon()
|