diff options
Diffstat (limited to 'third_party/aom/aom_dsp/aom_dsp_rtcd_defs.pl')
| -rwxr-xr-x | third_party/aom/aom_dsp/aom_dsp_rtcd_defs.pl | 149 |
1 files changed, 75 insertions, 74 deletions
diff --git a/third_party/aom/aom_dsp/aom_dsp_rtcd_defs.pl b/third_party/aom/aom_dsp/aom_dsp_rtcd_defs.pl index 8047cbc09..0c0356870 100755 --- a/third_party/aom/aom_dsp/aom_dsp_rtcd_defs.pl +++ b/third_party/aom/aom_dsp/aom_dsp_rtcd_defs.pl @@ -40,12 +40,26 @@ foreach $w (@block_widths) { push @block_sizes, [$w, $h] if ($w <= 2*$h && $h <= 2*$w) ; } } +if (aom_config("CONFIG_EXT_PARTITION_TYPES")) { + push @block_sizes, [4, 16]; + push @block_sizes, [16, 4]; + push @block_sizes, [8, 32]; + push @block_sizes, [32, 8]; +} @tx_dims = (2, 4, 8, 16, 32); if (aom_config("CONFIG_TX64X64") eq "yes") { push @tx_dims, '64'; } +@tx_sizes = (); +foreach $w (@tx_dims) { + push @tx_sizes, [$w, $w]; + foreach $h (@tx_dims) { + push @tx_sizes, [$w, $h] if ($w >=4 && $h >=4 && ($w == 2*$h || $h == 2*$w)); + } +} + @pred_names = qw/dc dc_top dc_left dc_128 v h d207e d63e d45e d117 d135 d153/; if (aom_config("CONFIG_ALT_INTRA") eq "yes") { push @pred_names, qw/paeth smooth/; @@ -60,9 +74,8 @@ if (aom_config("CONFIG_ALT_INTRA") eq "yes") { # Intra prediction # -foreach $dim (@tx_dims) { - $w = ${dim}; - $h = ${dim}; +foreach (@tx_sizes) { + ($w, $h) = @$_; foreach $pred_name (@pred_names) { add_proto "void", "aom_${pred_name}_predictor_${w}x${h}", "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"; @@ -142,20 +155,27 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { # # Sub Pixel Filters # -add_proto qw/void aom_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_scaled_2d/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_scaled_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_scaled_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_scaled_avg_2d/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_scaled_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; -add_proto qw/void aom_scaled_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_scaled_2d/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_scaled_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_scaled_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_scaled_avg_2d/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_scaled_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_scaled_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"; + +add_proto qw/void aom_convolve8_horiz_scale/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int subpel_x, int x_step_q4, const int16_t *filter_y, int subpel_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_vert_scale/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int subpel_x, int x_step_q4, const int16_t *filter_y, int subpel_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_avg_horiz_scale/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int subpel_x, int x_step_q4, const int16_t *filter_y, int subpel_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_avg_vert_scale/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int subpel_x, int x_step_q4, const int16_t *filter_y, int subpel_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_scale/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int subpel_x, int x_step_q4, const int16_t *filter_y, int subpel_y, int y_step_q4, int w, int h"; +add_proto qw/void aom_convolve8_avg_scale/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int subpel_x, int x_step_q4, const int16_t *filter_y, int subpel_y, int y_step_q4, int w, int h"; specialize qw/aom_convolve_copy sse2 /; specialize qw/aom_convolve_avg sse2 /; @@ -334,24 +354,15 @@ if ((aom_config("CONFIG_AV1_ENCODER") eq "yes") || (aom_config("CONFIG_PVQ") eq add_proto qw/void aom_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct8x8 sse2/, "$ssse3_x86_64"; - add_proto qw/void aom_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride"; - specialize qw/aom_fdct8x8_1 sse2/; - add_proto qw/void aom_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct16x16 sse2/; - add_proto qw/void aom_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride"; - specialize qw/aom_fdct16x16_1 sse2 avx2/; - add_proto qw/void aom_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct32x32 sse2 avx2/; add_proto qw/void aom_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct32x32_rd sse2 avx2/; - add_proto qw/void aom_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride"; - specialize qw/aom_fdct32x32_1 sse2 avx2/; - # High bit depth add_proto qw/void aom_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_highbd_fdct4x4 sse2/; @@ -359,20 +370,15 @@ if ((aom_config("CONFIG_AV1_ENCODER") eq "yes") || (aom_config("CONFIG_PVQ") eq add_proto qw/void aom_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_highbd_fdct8x8 sse2/; - add_proto qw/void aom_highbd_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride"; - add_proto qw/void aom_highbd_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_highbd_fdct16x16 sse2/; - add_proto qw/void aom_highbd_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride"; - add_proto qw/void aom_highbd_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_highbd_fdct32x32 sse2/; add_proto qw/void aom_highbd_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_highbd_fdct32x32_rd sse2/; - add_proto qw/void aom_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride"; } else { add_proto qw/void aom_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct4x4 sse2 msa/; @@ -383,47 +389,25 @@ if ((aom_config("CONFIG_AV1_ENCODER") eq "yes") || (aom_config("CONFIG_PVQ") eq add_proto qw/void aom_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct8x8 sse2 neon msa/, "$ssse3_x86_64"; - add_proto qw/void aom_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride"; - specialize qw/aom_fdct8x8_1 sse2 neon msa/; - add_proto qw/void aom_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct16x16 sse2 msa/; - add_proto qw/void aom_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride"; - specialize qw/aom_fdct16x16_1 sse2 avx2 msa/; - add_proto qw/void aom_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct32x32 sse2 avx2 msa/; add_proto qw/void aom_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride"; specialize qw/aom_fdct32x32_rd sse2 avx2 msa/; - - add_proto qw/void aom_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride"; - specialize qw/aom_fdct32x32_1 sse2 avx2 msa/; } # CONFIG_HIGHBITDEPTH } # CONFIG_AV1_ENCODER # # Inverse transform if (aom_config("CONFIG_AV1") eq "yes") { -if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { add_proto qw/void aom_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride"; add_proto qw/void aom_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride"; specialize qw/aom_iwht4x4_16_add sse2/; - add_proto qw/void aom_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - - add_proto qw/void aom_highbd_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - - add_proto qw/void aom_highbd_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - - add_proto qw/void aom_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - - add_proto qw/void aom_highbd_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - - add_proto qw/void aom_highbd_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - add_proto qw/void aom_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; add_proto qw/void aom_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; @@ -469,10 +453,8 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { add_proto qw/void aom_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride"; specialize qw/aom_idct32x32_1_add sse2 avx2/; - - add_proto qw/void aom_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd"; - specialize qw/aom_highbd_idct4x4_16_add sse2/; } +if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { } else { { add_proto qw/void aom_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride"; @@ -541,13 +523,12 @@ if (aom_config("CONFIG_AOM_QM") eq "yes") { add_proto qw/void aom_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; - if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { - add_proto qw/void aom_highbd_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; + add_proto qw/void aom_highbd_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; - add_proto qw/void aom_highbd_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; + add_proto qw/void aom_highbd_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; + + add_proto qw/void aom_highbd_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; - add_proto qw/void aom_highbd_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr"; - } # CONFIG_HIGHBITDEPTH } # CONFIG_AV1_ENCODER } else { if (aom_config("CONFIG_AV1_ENCODER") eq "yes") { @@ -559,21 +540,23 @@ if (aom_config("CONFIG_AOM_QM") eq "yes") { add_proto qw/void aom_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; - if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { - add_proto qw/void aom_highbd_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; - specialize qw/aom_highbd_quantize_b sse2/; + add_proto qw/void aom_highbd_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; + specialize qw/aom_highbd_quantize_b sse2 avx2/; - add_proto qw/void aom_highbd_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; - specialize qw/aom_highbd_quantize_b_32x32 sse2/; + add_proto qw/void aom_highbd_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; + specialize qw/aom_highbd_quantize_b_32x32 sse2/; + + add_proto qw/void aom_highbd_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; - add_proto qw/void aom_highbd_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; - } # CONFIG_HIGHBITDEPTH } # CONFIG_AV1_ENCODER } # CONFIG_AOM_QM if (aom_config("CONFIG_AV1") eq "yes") { # # Alpha blending with mask # + if (aom_config("CONFIG_CONVOLVE_ROUND") eq "yes") { + add_proto qw/void aom_blend_a64_d32_mask/, "int32_t *dst, uint32_t dst_stride, const int32_t *src0, uint32_t src0_stride, const int32_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx"; + } add_proto qw/void aom_blend_a64_mask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx"; add_proto qw/void aom_blend_a64_hmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w"; add_proto qw/void aom_blend_a64_vmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w"; @@ -927,15 +910,15 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { # # ... # -add_proto qw/void aom_upsampled_pred/, "uint8_t *comp_pred, int width, int height, const uint8_t *ref, int ref_stride"; +add_proto qw/void aom_upsampled_pred/, "uint8_t *comp_pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref, int ref_stride"; specialize qw/aom_upsampled_pred sse2/; -add_proto qw/void aom_comp_avg_upsampled_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride"; +add_proto qw/void aom_comp_avg_upsampled_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref, int ref_stride"; specialize qw/aom_comp_avg_upsampled_pred sse2/; if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { - add_proto qw/void aom_highbd_upsampled_pred/, "uint16_t *comp_pred, int width, int height, const uint8_t *ref8, int ref_stride"; + add_proto qw/void aom_highbd_upsampled_pred/, "uint16_t *comp_pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, int bd"; specialize qw/aom_highbd_upsampled_pred sse2/; - add_proto qw/void aom_highbd_comp_avg_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride"; + add_proto qw/void aom_highbd_comp_avg_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, int bd"; specialize qw/aom_highbd_comp_avg_upsampled_pred sse2/; } @@ -1005,6 +988,22 @@ specialize qw/aom_sub_pixel_avg_variance8x8 msa sse2 ssse3/; specialize qw/aom_sub_pixel_avg_variance8x4 msa sse2 ssse3/; specialize qw/aom_sub_pixel_avg_variance4x8 msa sse2 ssse3/; specialize qw/aom_sub_pixel_avg_variance4x4 msa sse2 ssse3/; + +if (aom_config("CONFIG_EXT_PARTITION_TYPES")) { + specialize qw/aom_variance4x16 sse2/; + specialize qw/aom_variance16x4 sse2/; + specialize qw/aom_variance8x32 sse2/; + specialize qw/aom_variance32x8 sse2/; + specialize qw/aom_sub_pixel_variance4x16 sse2 ssse3/; + specialize qw/aom_sub_pixel_variance16x4 sse2 ssse3/; + specialize qw/aom_sub_pixel_variance8x32 sse2 ssse3/; + specialize qw/aom_sub_pixel_variance32x8 sse2 ssse3/; + specialize qw/aom_sub_pixel_avg_variance4x16 sse2 ssse3/; + specialize qw/aom_sub_pixel_avg_variance16x4 sse2 ssse3/; + specialize qw/aom_sub_pixel_avg_variance8x32 sse2 ssse3/; + specialize qw/aom_sub_pixel_avg_variance32x8 sse2 ssse3/; +} + if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { foreach $bd (8, 10, 12) { add_proto qw/unsigned int/, "aom_highbd_${bd}_variance2x2", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"; @@ -1021,6 +1020,8 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { if ($w != 128 && $h != 128 && $w != 4 && $h != 4) { specialize "aom_highbd_${bd}_variance${w}x${h}", "sse2"; } + # TODO(david.barker): When ext-partition-types is enabled, we currenly + # don't have vectorized 4x16 highbd variance functions if ($w == 4 && $h == 4) { specialize "aom_highbd_${bd}_variance${w}x${h}", "sse4_1"; } @@ -1496,10 +1497,10 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { if (aom_config("CONFIG_EXT_INTER") eq "yes") { add_proto qw/void aom_comp_mask_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask"; - add_proto qw/void aom_comp_mask_upsampled_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask"; + add_proto qw/void aom_comp_mask_upsampled_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask"; if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") { add_proto qw/void aom_highbd_comp_mask_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask"; - add_proto qw/void aom_highbd_comp_mask_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask"; + add_proto qw/void aom_highbd_comp_mask_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask, int bd"; } } |