std/jpeg: optimize IDCT when AC terms are zero

name                                             old speed     new speed     delta

wuffs_jpeg_decode_19k_8bpp/clang11               199MB/s ± 0%  211MB/s ± 0%   +5.69%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_progressive/clang11  169MB/s ± 1%  176MB/s ± 0%   +4.32%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_sequential/clang11   269MB/s ± 0%  291MB/s ± 0%   +7.89%  (p=0.008 n=5+5)
wuffs_jpeg_decode_77k_24bpp/clang11              250MB/s ± 0%  254MB/s ± 0%   +1.90%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_420/clang11         258MB/s ± 0%  262MB/s ± 0%   +1.51%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_444/clang11         275MB/s ± 0%  356MB/s ± 0%  +29.59%  (p=0.016 n=5+4)
wuffs_jpeg_decode_4002k_24bpp/clang11            261MB/s ± 0%  263MB/s ± 1%   +0.61%  (p=0.008 n=5+5)

wuffs_jpeg_decode_19k_8bpp/gcc10                 153MB/s ± 0%  165MB/s ± 0%   +8.09%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_progressive/gcc10    180MB/s ± 0%  197MB/s ± 0%   +9.56%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_sequential/gcc10     288MB/s ± 0%  337MB/s ± 0%  +16.81%  (p=0.008 n=5+5)
wuffs_jpeg_decode_77k_24bpp/gcc10                237MB/s ± 0%  245MB/s ± 0%   +3.47%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_420/gcc10           247MB/s ± 0%  256MB/s ± 0%   +3.71%  (p=0.016 n=4+5)
wuffs_jpeg_decode_552k_24bpp_444/gcc10           249MB/s ± 0%  348MB/s ± 0%  +39.88%  (p=0.008 n=5+5)
wuffs_jpeg_decode_4002k_24bpp/gcc10              247MB/s ± 0%  254MB/s ± 0%   +2.81%  (p=0.008 n=5+5)
This commit is contained in:
Nigel Tao committed 2023-06-08 22:03:46 +10:00
1 parent ca2909b6d7
commit d6cc525059
3 files changed
+2281 -1524

No files matched your search

+1030 -727
View File
@@ -37385,741 +37385,1044 @@ wuffs_jpeg__decoder__decode_idct(
if (8u > a_dst_stride) {
return wuffs_base__make_empty_struct();
}
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][16u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][16u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][48u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][48u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][0u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][0u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][32u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][32u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][8u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][8u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][24u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][24u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][40u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][40u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][56u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][56u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[0u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[56u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[8u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[48u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[16u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[40u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[24u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[32u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][17u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][17u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][49u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][49u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][1u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][1u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][33u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][33u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][9u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][9u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][25u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][25u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][41u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][41u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][57u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][57u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[1u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[57u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[9u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[49u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[17u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[41u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[25u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[33u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][18u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][18u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][50u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][50u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][2u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][2u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][34u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][34u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][10u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][10u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][26u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][26u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][42u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][42u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][58u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][58u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[2u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[58u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[10u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[50u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[18u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[42u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[26u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[34u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][19u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][19u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][51u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][51u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][3u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][3u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][35u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][35u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][11u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][11u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][27u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][27u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][43u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][43u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][59u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][59u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[3u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[59u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[11u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[51u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[19u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[43u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[27u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[35u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][20u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][20u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][52u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][52u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][4u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][4u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][36u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][36u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][12u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][12u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][28u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][28u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][44u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][44u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][60u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][60u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[4u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[60u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[12u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[52u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[20u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[44u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[28u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[36u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][21u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][21u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][53u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][53u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][5u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][5u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][37u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][37u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][13u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][13u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][29u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][29u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][45u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][45u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][61u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][61u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[5u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[61u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[13u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[53u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[21u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[45u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[29u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[37u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][22u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][22u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][54u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][54u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][6u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][6u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][38u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][38u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][14u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][14u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][30u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][30u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][46u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][46u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][62u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][62u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[6u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[62u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[14u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[54u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[22u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[46u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[30u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[38u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][23u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][23u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][55u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][55u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][7u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][7u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][39u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][39u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][15u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][15u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][31u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][31u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][47u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][47u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][63u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][63u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[7u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[63u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[15u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[55u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[23u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[47u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[31u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[39u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
v_in2 = v_intermediate[2u];
v_in6 = v_intermediate[6u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[0u];
v_in4 = v_intermediate[4u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[1u];
v_in3 = v_intermediate[3u];
v_in5 = v_intermediate[5u];
v_in7 = v_intermediate[7u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][8u] |
self->private_data.f_mcu_blocks[a_b][16u] |
self->private_data.f_mcu_blocks[a_b][24u] |
self->private_data.f_mcu_blocks[a_b][32u] |
self->private_data.f_mcu_blocks[a_b][40u] |
self->private_data.f_mcu_blocks[a_b][48u] |
self->private_data.f_mcu_blocks[a_b][56u])) {
v_intermediate[0u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][0u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][0u])))) << 2u));
v_intermediate[8u] = v_intermediate[0u];
v_intermediate[16u] = v_intermediate[0u];
v_intermediate[24u] = v_intermediate[0u];
v_intermediate[32u] = v_intermediate[0u];
v_intermediate[40u] = v_intermediate[0u];
v_intermediate[48u] = v_intermediate[0u];
v_intermediate[56u] = v_intermediate[0u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][16u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][16u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][48u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][48u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][0u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][0u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][32u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][32u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][8u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][8u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][24u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][24u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][40u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][40u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][56u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][56u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[0u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[56u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[8u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[48u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[16u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[40u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[24u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[32u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[10u];
v_in6 = v_intermediate[14u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[8u];
v_in4 = v_intermediate[12u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[9u];
v_in3 = v_intermediate[11u];
v_in5 = v_intermediate[13u];
v_in7 = v_intermediate[15u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][9u] |
self->private_data.f_mcu_blocks[a_b][17u] |
self->private_data.f_mcu_blocks[a_b][25u] |
self->private_data.f_mcu_blocks[a_b][33u] |
self->private_data.f_mcu_blocks[a_b][41u] |
self->private_data.f_mcu_blocks[a_b][49u] |
self->private_data.f_mcu_blocks[a_b][57u])) {
v_intermediate[1u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][1u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][1u])))) << 2u));
v_intermediate[9u] = v_intermediate[1u];
v_intermediate[17u] = v_intermediate[1u];
v_intermediate[25u] = v_intermediate[1u];
v_intermediate[33u] = v_intermediate[1u];
v_intermediate[41u] = v_intermediate[1u];
v_intermediate[49u] = v_intermediate[1u];
v_intermediate[57u] = v_intermediate[1u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][17u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][17u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][49u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][49u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][1u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][1u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][33u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][33u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][9u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][9u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][25u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][25u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][41u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][41u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][57u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][57u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[1u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[57u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[9u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[49u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[17u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[41u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[25u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[33u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[18u];
v_in6 = v_intermediate[22u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[16u];
v_in4 = v_intermediate[20u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[17u];
v_in3 = v_intermediate[19u];
v_in5 = v_intermediate[21u];
v_in7 = v_intermediate[23u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][10u] |
self->private_data.f_mcu_blocks[a_b][18u] |
self->private_data.f_mcu_blocks[a_b][26u] |
self->private_data.f_mcu_blocks[a_b][34u] |
self->private_data.f_mcu_blocks[a_b][42u] |
self->private_data.f_mcu_blocks[a_b][50u] |
self->private_data.f_mcu_blocks[a_b][58u])) {
v_intermediate[2u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][2u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][2u])))) << 2u));
v_intermediate[10u] = v_intermediate[2u];
v_intermediate[18u] = v_intermediate[2u];
v_intermediate[26u] = v_intermediate[2u];
v_intermediate[34u] = v_intermediate[2u];
v_intermediate[42u] = v_intermediate[2u];
v_intermediate[50u] = v_intermediate[2u];
v_intermediate[58u] = v_intermediate[2u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][18u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][18u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][50u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][50u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][2u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][2u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][34u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][34u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][10u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][10u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][26u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][26u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][42u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][42u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][58u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][58u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[2u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[58u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[10u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[50u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[18u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[42u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[26u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[34u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[26u];
v_in6 = v_intermediate[30u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[24u];
v_in4 = v_intermediate[28u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[25u];
v_in3 = v_intermediate[27u];
v_in5 = v_intermediate[29u];
v_in7 = v_intermediate[31u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][11u] |
self->private_data.f_mcu_blocks[a_b][19u] |
self->private_data.f_mcu_blocks[a_b][27u] |
self->private_data.f_mcu_blocks[a_b][35u] |
self->private_data.f_mcu_blocks[a_b][43u] |
self->private_data.f_mcu_blocks[a_b][51u] |
self->private_data.f_mcu_blocks[a_b][59u])) {
v_intermediate[3u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][3u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][3u])))) << 2u));
v_intermediate[11u] = v_intermediate[3u];
v_intermediate[19u] = v_intermediate[3u];
v_intermediate[27u] = v_intermediate[3u];
v_intermediate[35u] = v_intermediate[3u];
v_intermediate[43u] = v_intermediate[3u];
v_intermediate[51u] = v_intermediate[3u];
v_intermediate[59u] = v_intermediate[3u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][19u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][19u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][51u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][51u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][3u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][3u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][35u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][35u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][11u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][11u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][27u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][27u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][43u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][43u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][59u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][59u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[3u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[59u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[11u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[51u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[19u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[43u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[27u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[35u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[34u];
v_in6 = v_intermediate[38u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[32u];
v_in4 = v_intermediate[36u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[33u];
v_in3 = v_intermediate[35u];
v_in5 = v_intermediate[37u];
v_in7 = v_intermediate[39u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][12u] |
self->private_data.f_mcu_blocks[a_b][20u] |
self->private_data.f_mcu_blocks[a_b][28u] |
self->private_data.f_mcu_blocks[a_b][36u] |
self->private_data.f_mcu_blocks[a_b][44u] |
self->private_data.f_mcu_blocks[a_b][52u] |
self->private_data.f_mcu_blocks[a_b][60u])) {
v_intermediate[4u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][4u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][4u])))) << 2u));
v_intermediate[12u] = v_intermediate[4u];
v_intermediate[20u] = v_intermediate[4u];
v_intermediate[28u] = v_intermediate[4u];
v_intermediate[36u] = v_intermediate[4u];
v_intermediate[44u] = v_intermediate[4u];
v_intermediate[52u] = v_intermediate[4u];
v_intermediate[60u] = v_intermediate[4u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][20u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][20u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][52u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][52u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][4u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][4u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][36u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][36u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][12u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][12u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][28u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][28u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][44u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][44u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][60u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][60u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[4u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[60u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[12u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[52u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[20u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[44u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[28u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[36u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[42u];
v_in6 = v_intermediate[46u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[40u];
v_in4 = v_intermediate[44u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[41u];
v_in3 = v_intermediate[43u];
v_in5 = v_intermediate[45u];
v_in7 = v_intermediate[47u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][13u] |
self->private_data.f_mcu_blocks[a_b][21u] |
self->private_data.f_mcu_blocks[a_b][29u] |
self->private_data.f_mcu_blocks[a_b][37u] |
self->private_data.f_mcu_blocks[a_b][45u] |
self->private_data.f_mcu_blocks[a_b][53u] |
self->private_data.f_mcu_blocks[a_b][61u])) {
v_intermediate[5u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][5u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][5u])))) << 2u));
v_intermediate[13u] = v_intermediate[5u];
v_intermediate[21u] = v_intermediate[5u];
v_intermediate[29u] = v_intermediate[5u];
v_intermediate[37u] = v_intermediate[5u];
v_intermediate[45u] = v_intermediate[5u];
v_intermediate[53u] = v_intermediate[5u];
v_intermediate[61u] = v_intermediate[5u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][21u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][21u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][53u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][53u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][5u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][5u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][37u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][37u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][13u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][13u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][29u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][29u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][45u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][45u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][61u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][61u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[5u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[61u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[13u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[53u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[21u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[45u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[29u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[37u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[50u];
v_in6 = v_intermediate[54u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[48u];
v_in4 = v_intermediate[52u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[49u];
v_in3 = v_intermediate[51u];
v_in5 = v_intermediate[53u];
v_in7 = v_intermediate[55u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][14u] |
self->private_data.f_mcu_blocks[a_b][22u] |
self->private_data.f_mcu_blocks[a_b][30u] |
self->private_data.f_mcu_blocks[a_b][38u] |
self->private_data.f_mcu_blocks[a_b][46u] |
self->private_data.f_mcu_blocks[a_b][54u] |
self->private_data.f_mcu_blocks[a_b][62u])) {
v_intermediate[6u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][6u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][6u])))) << 2u));
v_intermediate[14u] = v_intermediate[6u];
v_intermediate[22u] = v_intermediate[6u];
v_intermediate[30u] = v_intermediate[6u];
v_intermediate[38u] = v_intermediate[6u];
v_intermediate[46u] = v_intermediate[6u];
v_intermediate[54u] = v_intermediate[6u];
v_intermediate[62u] = v_intermediate[6u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][22u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][22u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][54u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][54u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][6u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][6u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][38u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][38u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][14u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][14u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][30u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][30u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][46u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][46u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][62u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][62u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[6u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[62u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[14u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[54u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[22u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[46u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[30u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[38u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
v_in2 = v_intermediate[58u];
v_in6 = v_intermediate[62u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[56u];
v_in4 = v_intermediate[60u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[57u];
v_in3 = v_intermediate[59u];
v_in5 = v_intermediate[61u];
v_in7 = v_intermediate[63u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (8u > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
if (0u == (self->private_data.f_mcu_blocks[a_b][15u] |
self->private_data.f_mcu_blocks[a_b][23u] |
self->private_data.f_mcu_blocks[a_b][31u] |
self->private_data.f_mcu_blocks[a_b][39u] |
self->private_data.f_mcu_blocks[a_b][47u] |
self->private_data.f_mcu_blocks[a_b][55u] |
self->private_data.f_mcu_blocks[a_b][63u])) {
v_intermediate[7u] = ((uint32_t)(((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][7u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][7u])))) << 2u));
v_intermediate[15u] = v_intermediate[7u];
v_intermediate[23u] = v_intermediate[7u];
v_intermediate[31u] = v_intermediate[7u];
v_intermediate[39u] = v_intermediate[7u];
v_intermediate[47u] = v_intermediate[7u];
v_intermediate[55u] = v_intermediate[7u];
v_intermediate[63u] = v_intermediate[7u];
} else {
v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][23u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][23u]))));
v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][55u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][55u]))));
v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433u));
v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270u))));
v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137u))));
v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][7u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][7u]))));
v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][39u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][39u]))));
v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13u));
v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13u));
v_cd0 = ((uint32_t)(v_ccp + v_cb2));
v_cd1 = ((uint32_t)(v_ccm + v_cb6));
v_cd2 = ((uint32_t)(v_ccm - v_cb6));
v_cd3 = ((uint32_t)(v_ccp - v_cb2));
v_bq1 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][15u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][15u]))));
v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][31u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][31u]))));
v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][47u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][47u]))));
v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][63u]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][63u]))));
v_ci51 = ((uint32_t)(v_bq5 + v_bq1));
v_ci53 = ((uint32_t)(v_bq5 + v_bq3));
v_ci71 = ((uint32_t)(v_bq7 + v_bq1));
v_ci73 = ((uint32_t)(v_bq7 + v_bq3));
v_cj = ((uint32_t)(((uint32_t)(v_ci73 + v_ci51)) * 9633u));
v_ck1 = ((uint32_t)(v_bq1 * 12299u));
v_ck3 = ((uint32_t)(v_bq3 * 25172u));
v_ck5 = ((uint32_t)(v_bq5 * 16819u));
v_ck7 = ((uint32_t)(v_bq7 * 2446u));
v_ci51 *= 4294964100u;
v_ci53 *= 4294946301u;
v_ci71 *= 4294959923u;
v_ci73 *= 4294951227u;
v_cl51 = ((uint32_t)(v_ci51 + v_cj));
v_cl73 = ((uint32_t)(v_ci73 + v_cj));
v_ck1 += ((uint32_t)(v_ci71 + v_cl51));
v_ck3 += ((uint32_t)(v_ci53 + v_cl73));
v_ck5 += ((uint32_t)(v_ci53 + v_cl51));
v_ck7 += ((uint32_t)(v_ci71 + v_cl73));
v_intermediate[7u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024u)), 11u);
v_intermediate[63u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024u)), 11u);
v_intermediate[15u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024u)), 11u);
v_intermediate[55u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024u)), 11u);
v_intermediate[23u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024u)), 11u);
v_intermediate[47u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024u)), 11u);
v_intermediate[31u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024u)), 11u);
v_intermediate[39u] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024u)), 11u);
}
if (0u == (v_intermediate[1u] |
v_intermediate[2u] |
v_intermediate[3u] |
v_intermediate[4u] |
v_intermediate[5u] |
v_intermediate[6u] |
v_intermediate[7u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[0u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[2u];
v_in6 = v_intermediate[6u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[0u];
v_in4 = v_intermediate[4u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[1u];
v_in3 = v_intermediate[3u];
v_in5 = v_intermediate[5u];
v_in7 = v_intermediate[7u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[9u] |
v_intermediate[10u] |
v_intermediate[11u] |
v_intermediate[12u] |
v_intermediate[13u] |
v_intermediate[14u] |
v_intermediate[15u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[8u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[10u];
v_in6 = v_intermediate[14u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[8u];
v_in4 = v_intermediate[12u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[9u];
v_in3 = v_intermediate[11u];
v_in5 = v_intermediate[13u];
v_in7 = v_intermediate[15u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[17u] |
v_intermediate[18u] |
v_intermediate[19u] |
v_intermediate[20u] |
v_intermediate[21u] |
v_intermediate[22u] |
v_intermediate[23u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[16u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[18u];
v_in6 = v_intermediate[22u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[16u];
v_in4 = v_intermediate[20u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[17u];
v_in3 = v_intermediate[19u];
v_in5 = v_intermediate[21u];
v_in7 = v_intermediate[23u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[25u] |
v_intermediate[26u] |
v_intermediate[27u] |
v_intermediate[28u] |
v_intermediate[29u] |
v_intermediate[30u] |
v_intermediate[31u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[24u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[26u];
v_in6 = v_intermediate[30u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[24u];
v_in4 = v_intermediate[28u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[25u];
v_in3 = v_intermediate[27u];
v_in5 = v_intermediate[29u];
v_in7 = v_intermediate[31u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[33u] |
v_intermediate[34u] |
v_intermediate[35u] |
v_intermediate[36u] |
v_intermediate[37u] |
v_intermediate[38u] |
v_intermediate[39u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[32u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[34u];
v_in6 = v_intermediate[38u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[32u];
v_in4 = v_intermediate[36u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[33u];
v_in3 = v_intermediate[35u];
v_in5 = v_intermediate[37u];
v_in7 = v_intermediate[39u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[41u] |
v_intermediate[42u] |
v_intermediate[43u] |
v_intermediate[44u] |
v_intermediate[45u] |
v_intermediate[46u] |
v_intermediate[47u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[40u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[42u];
v_in6 = v_intermediate[46u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[40u];
v_in4 = v_intermediate[44u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[41u];
v_in3 = v_intermediate[43u];
v_in5 = v_intermediate[45u];
v_in7 = v_intermediate[47u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[49u] |
v_intermediate[50u] |
v_intermediate[51u] |
v_intermediate[52u] |
v_intermediate[53u] |
v_intermediate[54u] |
v_intermediate[55u])) {
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[48u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
} else {
v_in2 = v_intermediate[50u];
v_in6 = v_intermediate[54u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[48u];
v_in4 = v_intermediate[52u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[49u];
v_in3 = v_intermediate[51u];
v_in5 = v_intermediate[53u];
v_in7 = v_intermediate[55u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (a_dst_stride > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride);
}
if (0u == (v_intermediate[57u] |
v_intermediate[58u] |
v_intermediate[59u] |
v_intermediate[60u] |
v_intermediate[61u] |
v_intermediate[62u] |
v_intermediate[63u])) {
if (8u > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(v_intermediate[56u] + 16u)) >> 5u) & 1023u)];
a_dst_buffer.ptr[1u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[2u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[3u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[4u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[5u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[6u] = a_dst_buffer.ptr[0u];
a_dst_buffer.ptr[7u] = a_dst_buffer.ptr[0u];
} else {
v_in2 = v_intermediate[58u];
v_in6 = v_intermediate[62u];
v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433u));
v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270u))));
v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137u))));
v_in0 = v_intermediate[56u];
v_in4 = v_intermediate[60u];
v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13u));
v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13u));
v_rd0 = ((uint32_t)(v_rcp + v_rb2));
v_rd1 = ((uint32_t)(v_rcm + v_rb6));
v_rd2 = ((uint32_t)(v_rcm - v_rb6));
v_rd3 = ((uint32_t)(v_rcp - v_rb2));
v_in1 = v_intermediate[57u];
v_in3 = v_intermediate[59u];
v_in5 = v_intermediate[61u];
v_in7 = v_intermediate[63u];
v_ri51 = ((uint32_t)(v_in5 + v_in1));
v_ri53 = ((uint32_t)(v_in5 + v_in3));
v_ri71 = ((uint32_t)(v_in7 + v_in1));
v_ri73 = ((uint32_t)(v_in7 + v_in3));
v_rj = ((uint32_t)(((uint32_t)(v_ri73 + v_ri51)) * 9633u));
v_rk1 = ((uint32_t)(v_in1 * 12299u));
v_rk3 = ((uint32_t)(v_in3 * 25172u));
v_rk5 = ((uint32_t)(v_in5 * 16819u));
v_rk7 = ((uint32_t)(v_in7 * 2446u));
v_ri51 *= 4294964100u;
v_ri53 *= 4294946301u;
v_ri71 *= 4294959923u;
v_ri73 *= 4294951227u;
v_rl51 = ((uint32_t)(v_ri51 + v_rj));
v_rl73 = ((uint32_t)(v_ri73 + v_rj));
v_rk1 += ((uint32_t)(v_ri71 + v_rl51));
v_rk3 += ((uint32_t)(v_ri53 + v_rl73));
v_rk5 += ((uint32_t)(v_ri53 + v_rl51));
v_rk7 += ((uint32_t)(v_ri71 + v_rl73));
if (8u > ((uint64_t)(a_dst_buffer.len))) {
return wuffs_base__make_empty_struct();
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
}
a_dst_buffer.ptr[0u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[7u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[1u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[6u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[2u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[5u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[3u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072u)) >> 18u) & 1023u)];
a_dst_buffer.ptr[4u] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072u)) >> 18u) & 1023u)];
return wuffs_base__make_empty_struct();
}
+48
View File
@@ -186,6 +186,28 @@ const (
const pass0 = `// ==== First pass, column $colX$.
if (0 == (
this.mcu_blocks[args.b][$row1colX$] |
this.mcu_blocks[args.b][$row2colX$] |
this.mcu_blocks[args.b][$row3colX$] |
this.mcu_blocks[args.b][$row4colX$] |
this.mcu_blocks[args.b][$row5colX$] |
this.mcu_blocks[args.b][$row6colX$] |
this.mcu_blocks[args.b][$row7colX$])) {
// Fast path when the 1-dimensional AC terms are all zero.
intermediate[$row0colX$] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row0colX$]) ~mod*
(this.quant_tables[args.q][$row0colX$] as base.u32)) ~mod<< 2
intermediate[$row1colX$] = intermediate[$row0colX$]
intermediate[$row2colX$] = intermediate[$row0colX$]
intermediate[$row3colX$] = intermediate[$row0colX$]
intermediate[$row4colX$] = intermediate[$row0colX$]
intermediate[$row5colX$] = intermediate[$row0colX$]
intermediate[$row6colX$] = intermediate[$row0colX$]
intermediate[$row7colX$] = intermediate[$row0colX$]
} else {
// Even rows.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row2colX$]) ~mod* (this.quant_tables[args.q][$row2colX$] as base.u32)
@@ -249,10 +271,35 @@ intermediate[$row2colX$] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~
intermediate[$row5colX$] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[$row3colX$] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[$row4colX$] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
`
const pass1 = `// ==== Second pass, row $rowY$.
if (0 == (
intermediate[$rowYcol1$] |
intermediate[$rowYcol2$] |
intermediate[$rowYcol3$] |
intermediate[$rowYcol4$] |
intermediate[$rowYcol5$] |
intermediate[$rowYcol6$] |
intermediate[$rowYcol7$])) {
// Fast path when the 1-dimensional AC terms are all zero.
$bounds_check$
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[$rowYcol0$] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
$advance$
} else {
// Even columns.
in2 = intermediate[$rowYcol2$]
@@ -320,4 +367,5 @@ args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) &
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
$advance$
}
`
+1203 -797
View File
@@ -119,1112 +119,1518 @@ pri func decoder.decode_idct!(dst_buffer: slice base.u8, dst_stride: base.u64, b
// ==== First pass, column 0.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x08] |
this.mcu_blocks[args.b][0x10] |
this.mcu_blocks[args.b][0x18] |
this.mcu_blocks[args.b][0x20] |
this.mcu_blocks[args.b][0x28] |
this.mcu_blocks[args.b][0x30] |
this.mcu_blocks[args.b][0x38])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x10]) ~mod* (this.quant_tables[args.q][0x10] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x30]) ~mod* (this.quant_tables[args.q][0x30] as base.u32)
intermediate[0x00] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x00]) ~mod*
(this.quant_tables[args.q][0x00] as base.u32)) ~mod<< 2
intermediate[0x08] = intermediate[0x00]
intermediate[0x10] = intermediate[0x00]
intermediate[0x18] = intermediate[0x00]
intermediate[0x20] = intermediate[0x00]
intermediate[0x28] = intermediate[0x00]
intermediate[0x30] = intermediate[0x00]
intermediate[0x38] = intermediate[0x00]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x10]) ~mod* (this.quant_tables[args.q][0x10] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x30]) ~mod* (this.quant_tables[args.q][0x30] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x00]) ~mod* (this.quant_tables[args.q][0x00] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x20]) ~mod* (this.quant_tables[args.q][0x20] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x00]) ~mod* (this.quant_tables[args.q][0x00] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x20]) ~mod* (this.quant_tables[args.q][0x20] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x08]) ~mod* (this.quant_tables[args.q][0x08] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x18]) ~mod* (this.quant_tables[args.q][0x18] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x28]) ~mod* (this.quant_tables[args.q][0x28] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x38]) ~mod* (this.quant_tables[args.q][0x38] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x08]) ~mod* (this.quant_tables[args.q][0x08] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x18]) ~mod* (this.quant_tables[args.q][0x18] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x28]) ~mod* (this.quant_tables[args.q][0x28] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x38]) ~mod* (this.quant_tables[args.q][0x38] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x00] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x38] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x08] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x30] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x10] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x28] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x18] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x20] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x00] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x38] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x08] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x30] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x10] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x28] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x18] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x20] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 1.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x09] |
this.mcu_blocks[args.b][0x11] |
this.mcu_blocks[args.b][0x19] |
this.mcu_blocks[args.b][0x21] |
this.mcu_blocks[args.b][0x29] |
this.mcu_blocks[args.b][0x31] |
this.mcu_blocks[args.b][0x39])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x11]) ~mod* (this.quant_tables[args.q][0x11] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x31]) ~mod* (this.quant_tables[args.q][0x31] as base.u32)
intermediate[0x01] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x01]) ~mod*
(this.quant_tables[args.q][0x01] as base.u32)) ~mod<< 2
intermediate[0x09] = intermediate[0x01]
intermediate[0x11] = intermediate[0x01]
intermediate[0x19] = intermediate[0x01]
intermediate[0x21] = intermediate[0x01]
intermediate[0x29] = intermediate[0x01]
intermediate[0x31] = intermediate[0x01]
intermediate[0x39] = intermediate[0x01]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x11]) ~mod* (this.quant_tables[args.q][0x11] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x31]) ~mod* (this.quant_tables[args.q][0x31] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x01]) ~mod* (this.quant_tables[args.q][0x01] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x21]) ~mod* (this.quant_tables[args.q][0x21] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x01]) ~mod* (this.quant_tables[args.q][0x01] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x21]) ~mod* (this.quant_tables[args.q][0x21] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x09]) ~mod* (this.quant_tables[args.q][0x09] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x19]) ~mod* (this.quant_tables[args.q][0x19] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x29]) ~mod* (this.quant_tables[args.q][0x29] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x39]) ~mod* (this.quant_tables[args.q][0x39] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x09]) ~mod* (this.quant_tables[args.q][0x09] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x19]) ~mod* (this.quant_tables[args.q][0x19] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x29]) ~mod* (this.quant_tables[args.q][0x29] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x39]) ~mod* (this.quant_tables[args.q][0x39] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x01] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x39] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x09] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x31] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x11] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x29] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x19] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x21] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x01] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x39] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x09] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x31] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x11] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x29] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x19] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x21] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 2.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x0A] |
this.mcu_blocks[args.b][0x12] |
this.mcu_blocks[args.b][0x1A] |
this.mcu_blocks[args.b][0x22] |
this.mcu_blocks[args.b][0x2A] |
this.mcu_blocks[args.b][0x32] |
this.mcu_blocks[args.b][0x3A])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x12]) ~mod* (this.quant_tables[args.q][0x12] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x32]) ~mod* (this.quant_tables[args.q][0x32] as base.u32)
intermediate[0x02] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x02]) ~mod*
(this.quant_tables[args.q][0x02] as base.u32)) ~mod<< 2
intermediate[0x0A] = intermediate[0x02]
intermediate[0x12] = intermediate[0x02]
intermediate[0x1A] = intermediate[0x02]
intermediate[0x22] = intermediate[0x02]
intermediate[0x2A] = intermediate[0x02]
intermediate[0x32] = intermediate[0x02]
intermediate[0x3A] = intermediate[0x02]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x12]) ~mod* (this.quant_tables[args.q][0x12] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x32]) ~mod* (this.quant_tables[args.q][0x32] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x02]) ~mod* (this.quant_tables[args.q][0x02] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x22]) ~mod* (this.quant_tables[args.q][0x22] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x02]) ~mod* (this.quant_tables[args.q][0x02] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x22]) ~mod* (this.quant_tables[args.q][0x22] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0A]) ~mod* (this.quant_tables[args.q][0x0A] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1A]) ~mod* (this.quant_tables[args.q][0x1A] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2A]) ~mod* (this.quant_tables[args.q][0x2A] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3A]) ~mod* (this.quant_tables[args.q][0x3A] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0A]) ~mod* (this.quant_tables[args.q][0x0A] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1A]) ~mod* (this.quant_tables[args.q][0x1A] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2A]) ~mod* (this.quant_tables[args.q][0x2A] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3A]) ~mod* (this.quant_tables[args.q][0x3A] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x02] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3A] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0A] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x32] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x12] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2A] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1A] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x22] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x02] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3A] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0A] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x32] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x12] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2A] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1A] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x22] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 3.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x0B] |
this.mcu_blocks[args.b][0x13] |
this.mcu_blocks[args.b][0x1B] |
this.mcu_blocks[args.b][0x23] |
this.mcu_blocks[args.b][0x2B] |
this.mcu_blocks[args.b][0x33] |
this.mcu_blocks[args.b][0x3B])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x13]) ~mod* (this.quant_tables[args.q][0x13] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x33]) ~mod* (this.quant_tables[args.q][0x33] as base.u32)
intermediate[0x03] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x03]) ~mod*
(this.quant_tables[args.q][0x03] as base.u32)) ~mod<< 2
intermediate[0x0B] = intermediate[0x03]
intermediate[0x13] = intermediate[0x03]
intermediate[0x1B] = intermediate[0x03]
intermediate[0x23] = intermediate[0x03]
intermediate[0x2B] = intermediate[0x03]
intermediate[0x33] = intermediate[0x03]
intermediate[0x3B] = intermediate[0x03]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x13]) ~mod* (this.quant_tables[args.q][0x13] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x33]) ~mod* (this.quant_tables[args.q][0x33] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x03]) ~mod* (this.quant_tables[args.q][0x03] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x23]) ~mod* (this.quant_tables[args.q][0x23] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x03]) ~mod* (this.quant_tables[args.q][0x03] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x23]) ~mod* (this.quant_tables[args.q][0x23] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0B]) ~mod* (this.quant_tables[args.q][0x0B] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1B]) ~mod* (this.quant_tables[args.q][0x1B] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2B]) ~mod* (this.quant_tables[args.q][0x2B] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3B]) ~mod* (this.quant_tables[args.q][0x3B] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0B]) ~mod* (this.quant_tables[args.q][0x0B] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1B]) ~mod* (this.quant_tables[args.q][0x1B] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2B]) ~mod* (this.quant_tables[args.q][0x2B] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3B]) ~mod* (this.quant_tables[args.q][0x3B] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x03] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3B] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0B] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x33] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x13] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2B] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1B] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x23] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x03] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3B] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0B] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x33] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x13] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2B] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1B] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x23] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 4.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x0C] |
this.mcu_blocks[args.b][0x14] |
this.mcu_blocks[args.b][0x1C] |
this.mcu_blocks[args.b][0x24] |
this.mcu_blocks[args.b][0x2C] |
this.mcu_blocks[args.b][0x34] |
this.mcu_blocks[args.b][0x3C])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x14]) ~mod* (this.quant_tables[args.q][0x14] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x34]) ~mod* (this.quant_tables[args.q][0x34] as base.u32)
intermediate[0x04] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x04]) ~mod*
(this.quant_tables[args.q][0x04] as base.u32)) ~mod<< 2
intermediate[0x0C] = intermediate[0x04]
intermediate[0x14] = intermediate[0x04]
intermediate[0x1C] = intermediate[0x04]
intermediate[0x24] = intermediate[0x04]
intermediate[0x2C] = intermediate[0x04]
intermediate[0x34] = intermediate[0x04]
intermediate[0x3C] = intermediate[0x04]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x14]) ~mod* (this.quant_tables[args.q][0x14] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x34]) ~mod* (this.quant_tables[args.q][0x34] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x04]) ~mod* (this.quant_tables[args.q][0x04] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x24]) ~mod* (this.quant_tables[args.q][0x24] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x04]) ~mod* (this.quant_tables[args.q][0x04] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x24]) ~mod* (this.quant_tables[args.q][0x24] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0C]) ~mod* (this.quant_tables[args.q][0x0C] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1C]) ~mod* (this.quant_tables[args.q][0x1C] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2C]) ~mod* (this.quant_tables[args.q][0x2C] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3C]) ~mod* (this.quant_tables[args.q][0x3C] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0C]) ~mod* (this.quant_tables[args.q][0x0C] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1C]) ~mod* (this.quant_tables[args.q][0x1C] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2C]) ~mod* (this.quant_tables[args.q][0x2C] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3C]) ~mod* (this.quant_tables[args.q][0x3C] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x04] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3C] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0C] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x34] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x14] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2C] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1C] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x24] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x04] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3C] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0C] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x34] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x14] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2C] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1C] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x24] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 5.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x0D] |
this.mcu_blocks[args.b][0x15] |
this.mcu_blocks[args.b][0x1D] |
this.mcu_blocks[args.b][0x25] |
this.mcu_blocks[args.b][0x2D] |
this.mcu_blocks[args.b][0x35] |
this.mcu_blocks[args.b][0x3D])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x15]) ~mod* (this.quant_tables[args.q][0x15] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x35]) ~mod* (this.quant_tables[args.q][0x35] as base.u32)
intermediate[0x05] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x05]) ~mod*
(this.quant_tables[args.q][0x05] as base.u32)) ~mod<< 2
intermediate[0x0D] = intermediate[0x05]
intermediate[0x15] = intermediate[0x05]
intermediate[0x1D] = intermediate[0x05]
intermediate[0x25] = intermediate[0x05]
intermediate[0x2D] = intermediate[0x05]
intermediate[0x35] = intermediate[0x05]
intermediate[0x3D] = intermediate[0x05]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x15]) ~mod* (this.quant_tables[args.q][0x15] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x35]) ~mod* (this.quant_tables[args.q][0x35] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x05]) ~mod* (this.quant_tables[args.q][0x05] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x25]) ~mod* (this.quant_tables[args.q][0x25] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x05]) ~mod* (this.quant_tables[args.q][0x05] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x25]) ~mod* (this.quant_tables[args.q][0x25] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0D]) ~mod* (this.quant_tables[args.q][0x0D] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1D]) ~mod* (this.quant_tables[args.q][0x1D] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2D]) ~mod* (this.quant_tables[args.q][0x2D] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3D]) ~mod* (this.quant_tables[args.q][0x3D] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0D]) ~mod* (this.quant_tables[args.q][0x0D] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1D]) ~mod* (this.quant_tables[args.q][0x1D] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2D]) ~mod* (this.quant_tables[args.q][0x2D] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3D]) ~mod* (this.quant_tables[args.q][0x3D] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x05] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3D] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0D] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x35] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x15] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2D] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1D] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x25] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x05] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3D] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0D] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x35] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x15] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2D] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1D] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x25] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 6.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x0E] |
this.mcu_blocks[args.b][0x16] |
this.mcu_blocks[args.b][0x1E] |
this.mcu_blocks[args.b][0x26] |
this.mcu_blocks[args.b][0x2E] |
this.mcu_blocks[args.b][0x36] |
this.mcu_blocks[args.b][0x3E])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x16]) ~mod* (this.quant_tables[args.q][0x16] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x36]) ~mod* (this.quant_tables[args.q][0x36] as base.u32)
intermediate[0x06] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x06]) ~mod*
(this.quant_tables[args.q][0x06] as base.u32)) ~mod<< 2
intermediate[0x0E] = intermediate[0x06]
intermediate[0x16] = intermediate[0x06]
intermediate[0x1E] = intermediate[0x06]
intermediate[0x26] = intermediate[0x06]
intermediate[0x2E] = intermediate[0x06]
intermediate[0x36] = intermediate[0x06]
intermediate[0x3E] = intermediate[0x06]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x16]) ~mod* (this.quant_tables[args.q][0x16] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x36]) ~mod* (this.quant_tables[args.q][0x36] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x06]) ~mod* (this.quant_tables[args.q][0x06] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x26]) ~mod* (this.quant_tables[args.q][0x26] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x06]) ~mod* (this.quant_tables[args.q][0x06] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x26]) ~mod* (this.quant_tables[args.q][0x26] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0E]) ~mod* (this.quant_tables[args.q][0x0E] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1E]) ~mod* (this.quant_tables[args.q][0x1E] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2E]) ~mod* (this.quant_tables[args.q][0x2E] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3E]) ~mod* (this.quant_tables[args.q][0x3E] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0E]) ~mod* (this.quant_tables[args.q][0x0E] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1E]) ~mod* (this.quant_tables[args.q][0x1E] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2E]) ~mod* (this.quant_tables[args.q][0x2E] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3E]) ~mod* (this.quant_tables[args.q][0x3E] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x06] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3E] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0E] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x36] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x16] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2E] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1E] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x26] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x06] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3E] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0E] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x36] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x16] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2E] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1E] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x26] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== First pass, column 7.
// Even rows.
if (0 == (
this.mcu_blocks[args.b][0x0F] |
this.mcu_blocks[args.b][0x17] |
this.mcu_blocks[args.b][0x1F] |
this.mcu_blocks[args.b][0x27] |
this.mcu_blocks[args.b][0x2F] |
this.mcu_blocks[args.b][0x37] |
this.mcu_blocks[args.b][0x3F])) {
// Fast path when the 1-dimensional AC terms are all zero.
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x17]) ~mod* (this.quant_tables[args.q][0x17] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x37]) ~mod* (this.quant_tables[args.q][0x37] as base.u32)
intermediate[0x07] =
(this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x07]) ~mod*
(this.quant_tables[args.q][0x07] as base.u32)) ~mod<< 2
intermediate[0x0F] = intermediate[0x07]
intermediate[0x17] = intermediate[0x07]
intermediate[0x1F] = intermediate[0x07]
intermediate[0x27] = intermediate[0x07]
intermediate[0x2F] = intermediate[0x07]
intermediate[0x37] = intermediate[0x07]
intermediate[0x3F] = intermediate[0x07]
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
} else {
// Even rows.
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
bq2 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x17]) ~mod* (this.quant_tables[args.q][0x17] as base.u32)
bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x37]) ~mod* (this.quant_tables[args.q][0x37] as base.u32)
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x07]) ~mod* (this.quant_tables[args.q][0x07] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x27]) ~mod* (this.quant_tables[args.q][0x27] as base.u32)
ca = (bq2 ~mod+ bq6) ~mod* 0x1151
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
cb2 = ca ~mod+ (bq2 ~mod* 0x187E)
cb6 = ca ~mod- (bq6 ~mod* 0x3B21)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x07]) ~mod* (this.quant_tables[args.q][0x07] as base.u32)
bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x27]) ~mod* (this.quant_tables[args.q][0x27] as base.u32)
// Odd rows.
ccp = (bq0 ~mod+ bq4) ~mod<< 13
ccm = (bq0 ~mod- bq4) ~mod<< 13
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0F]) ~mod* (this.quant_tables[args.q][0x0F] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1F]) ~mod* (this.quant_tables[args.q][0x1F] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2F]) ~mod* (this.quant_tables[args.q][0x2F] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3F]) ~mod* (this.quant_tables[args.q][0x3F] as base.u32)
cd0 = ccp ~mod+ cb2
cd1 = ccm ~mod+ cb6
cd2 = ccm ~mod- cb6
cd3 = ccp ~mod- cb2
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
// Odd rows.
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x0F]) ~mod* (this.quant_tables[args.q][0x0F] as base.u32)
bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x1F]) ~mod* (this.quant_tables[args.q][0x1F] as base.u32)
bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x2F]) ~mod* (this.quant_tables[args.q][0x2F] as base.u32)
bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][0x3F]) ~mod* (this.quant_tables[args.q][0x3F] as base.u32)
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ci51 = bq5 ~mod+ bq1
ci53 = bq5 ~mod+ bq3
ci71 = bq7 ~mod+ bq1
ci73 = bq7 ~mod+ bq3
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
cj = (ci73 ~mod+ ci51) ~mod* 0x25A1
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
ck1 = bq1 ~mod* 0x300B
ck3 = bq3 ~mod* 0x6254
ck5 = bq5 ~mod* 0x41B3
ck7 = bq7 ~mod* 0x098E
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
ci51 ~mod*= 0xFFFF_F384
ci53 ~mod*= 0xFFFF_ADFD
ci71 ~mod*= 0xFFFF_E333
ci73 ~mod*= 0xFFFF_C13B
// Combine rows.
cl51 = ci51 ~mod+ cj
cl73 = ci73 ~mod+ cj
intermediate[0x07] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3F] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0F] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x37] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x17] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2F] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1F] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x27] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
ck1 ~mod+= ci71 ~mod+ cl51
ck3 ~mod+= ci53 ~mod+ cl73
ck5 ~mod+= ci53 ~mod+ cl51
ck7 ~mod+= ci71 ~mod+ cl73
// Combine rows.
intermediate[0x07] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x3F] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11)
intermediate[0x0F] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x37] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11)
intermediate[0x17] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x2F] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod- ck5) ~mod+ (1 << 10), n: 11)
intermediate[0x1F] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod+ ck7) ~mod+ (1 << 10), n: 11)
intermediate[0x27] = this.util.sign_extend_rshift_u32(a: (cd3 ~mod- ck7) ~mod+ (1 << 10), n: 11)
}
// ==== Second pass, row 0.
// Even columns.
if (0 == (
intermediate[0x01] |
intermediate[0x02] |
intermediate[0x03] |
intermediate[0x04] |
intermediate[0x05] |
intermediate[0x06] |
intermediate[0x07])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x02]
in6 = intermediate[0x06]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x00] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x00]
in4 = intermediate[0x04]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x02]
in6 = intermediate[0x06]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x01]
in3 = intermediate[0x03]
in5 = intermediate[0x05]
in7 = intermediate[0x07]
in0 = intermediate[0x00]
in4 = intermediate[0x04]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x01]
in3 = intermediate[0x03]
in5 = intermediate[0x05]
in7 = intermediate[0x07]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 1.
// Even columns.
if (0 == (
intermediate[0x09] |
intermediate[0x0A] |
intermediate[0x0B] |
intermediate[0x0C] |
intermediate[0x0D] |
intermediate[0x0E] |
intermediate[0x0F])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x0A]
in6 = intermediate[0x0E]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x08] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x08]
in4 = intermediate[0x0C]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x0A]
in6 = intermediate[0x0E]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x09]
in3 = intermediate[0x0B]
in5 = intermediate[0x0D]
in7 = intermediate[0x0F]
in0 = intermediate[0x08]
in4 = intermediate[0x0C]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x09]
in3 = intermediate[0x0B]
in5 = intermediate[0x0D]
in7 = intermediate[0x0F]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 2.
// Even columns.
if (0 == (
intermediate[0x11] |
intermediate[0x12] |
intermediate[0x13] |
intermediate[0x14] |
intermediate[0x15] |
intermediate[0x16] |
intermediate[0x17])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x12]
in6 = intermediate[0x16]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x10] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x10]
in4 = intermediate[0x14]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x12]
in6 = intermediate[0x16]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x11]
in3 = intermediate[0x13]
in5 = intermediate[0x15]
in7 = intermediate[0x17]
in0 = intermediate[0x10]
in4 = intermediate[0x14]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x11]
in3 = intermediate[0x13]
in5 = intermediate[0x15]
in7 = intermediate[0x17]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 3.
// Even columns.
if (0 == (
intermediate[0x19] |
intermediate[0x1A] |
intermediate[0x1B] |
intermediate[0x1C] |
intermediate[0x1D] |
intermediate[0x1E] |
intermediate[0x1F])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x1A]
in6 = intermediate[0x1E]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x18] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x18]
in4 = intermediate[0x1C]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x1A]
in6 = intermediate[0x1E]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x19]
in3 = intermediate[0x1B]
in5 = intermediate[0x1D]
in7 = intermediate[0x1F]
in0 = intermediate[0x18]
in4 = intermediate[0x1C]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x19]
in3 = intermediate[0x1B]
in5 = intermediate[0x1D]
in7 = intermediate[0x1F]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 4.
// Even columns.
if (0 == (
intermediate[0x21] |
intermediate[0x22] |
intermediate[0x23] |
intermediate[0x24] |
intermediate[0x25] |
intermediate[0x26] |
intermediate[0x27])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x22]
in6 = intermediate[0x26]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x20] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x20]
in4 = intermediate[0x24]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x22]
in6 = intermediate[0x26]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x21]
in3 = intermediate[0x23]
in5 = intermediate[0x25]
in7 = intermediate[0x27]
in0 = intermediate[0x20]
in4 = intermediate[0x24]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x21]
in3 = intermediate[0x23]
in5 = intermediate[0x25]
in7 = intermediate[0x27]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 5.
// Even columns.
if (0 == (
intermediate[0x29] |
intermediate[0x2A] |
intermediate[0x2B] |
intermediate[0x2C] |
intermediate[0x2D] |
intermediate[0x2E] |
intermediate[0x2F])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x2A]
in6 = intermediate[0x2E]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x28] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x28]
in4 = intermediate[0x2C]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x2A]
in6 = intermediate[0x2E]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x29]
in3 = intermediate[0x2B]
in5 = intermediate[0x2D]
in7 = intermediate[0x2F]
in0 = intermediate[0x28]
in4 = intermediate[0x2C]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x29]
in3 = intermediate[0x2B]
in5 = intermediate[0x2D]
in7 = intermediate[0x2F]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 6.
// Even columns.
if (0 == (
intermediate[0x31] |
intermediate[0x32] |
intermediate[0x33] |
intermediate[0x34] |
intermediate[0x35] |
intermediate[0x36] |
intermediate[0x37])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x32]
in6 = intermediate[0x36]
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x30] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
in0 = intermediate[0x30]
in4 = intermediate[0x34]
} else {
// Even columns.
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
in2 = intermediate[0x32]
in6 = intermediate[0x36]
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
ra = (in2 ~mod+ in6) ~mod* 0x1151
// Odd columns.
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
in1 = intermediate[0x31]
in3 = intermediate[0x33]
in5 = intermediate[0x35]
in7 = intermediate[0x37]
in0 = intermediate[0x30]
in4 = intermediate[0x34]
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Odd columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
in1 = intermediate[0x31]
in3 = intermediate[0x33]
in5 = intermediate[0x35]
in7 = intermediate[0x37]
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Combine columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
if args.dst_stride > args.dst_buffer.length() {
return nothing
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if args.dst_stride > args.dst_buffer.length() {
return nothing
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
}
assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer = args.dst_buffer[args.dst_stride ..]
// ==== Second pass, row 7.
// Even columns.
if (0 == (
intermediate[0x39] |
intermediate[0x3A] |
intermediate[0x3B] |
intermediate[0x3C] |
intermediate[0x3D] |
intermediate[0x3E] |
intermediate[0x3F])) {
// Fast path when the 1-dimensional AC terms are all zero.
in2 = intermediate[0x3A]
in6 = intermediate[0x3E]
if 8 > args.dst_buffer.length() {
return nothing
}
ra = (in2 ~mod+ in6) ~mod* 0x1151
args.dst_buffer[0] = BIAS_AND_CLAMP[((intermediate[0x38] ~mod+ (1 << 4)) >> 5) & 1023]
args.dst_buffer[1] = args.dst_buffer[0]
args.dst_buffer[2] = args.dst_buffer[0]
args.dst_buffer[3] = args.dst_buffer[0]
args.dst_buffer[4] = args.dst_buffer[0]
args.dst_buffer[5] = args.dst_buffer[0]
args.dst_buffer[6] = args.dst_buffer[0]
args.dst_buffer[7] = args.dst_buffer[0]
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
} else {
// Even columns.
in0 = intermediate[0x38]
in4 = intermediate[0x3C]
in2 = intermediate[0x3A]
in6 = intermediate[0x3E]
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
ra = (in2 ~mod+ in6) ~mod* 0x1151
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rb2 = ra ~mod+ (in2 ~mod* 0x187E)
rb6 = ra ~mod- (in6 ~mod* 0x3B21)
// Odd columns.
in0 = intermediate[0x38]
in4 = intermediate[0x3C]
in1 = intermediate[0x39]
in3 = intermediate[0x3B]
in5 = intermediate[0x3D]
in7 = intermediate[0x3F]
rcp = (in0 ~mod+ in4) ~mod<< 13
rcm = (in0 ~mod- in4) ~mod<< 13
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rd0 = rcp ~mod+ rb2
rd1 = rcm ~mod+ rb6
rd2 = rcm ~mod- rb6
rd3 = rcp ~mod- rb2
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
// Odd columns.
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
in1 = intermediate[0x39]
in3 = intermediate[0x3B]
in5 = intermediate[0x3D]
in7 = intermediate[0x3F]
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
ri51 = in5 ~mod+ in1
ri53 = in5 ~mod+ in3
ri71 = in7 ~mod+ in1
ri73 = in7 ~mod+ in3
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rj = (ri73 ~mod+ ri51) ~mod* 0x25A1
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
rk1 = in1 ~mod* 0x300B
rk3 = in3 ~mod* 0x6254
rk5 = in5 ~mod* 0x41B3
rk7 = in7 ~mod* 0x098E
// Combine columns.
ri51 ~mod*= 0xFFFF_F384
ri53 ~mod*= 0xFFFF_ADFD
ri71 ~mod*= 0xFFFF_E333
ri73 ~mod*= 0xFFFF_C13B
rl51 = ri51 ~mod+ rj
rl73 = ri73 ~mod+ rj
rk1 ~mod+= ri71 ~mod+ rl51
rk3 ~mod+= ri53 ~mod+ rl73
rk5 ~mod+= ri53 ~mod+ rl51
rk7 ~mod+= ri71 ~mod+ rl73
// Combine columns.
if 8 > args.dst_buffer.length() {
return nothing
}
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
if 8 > args.dst_buffer.length() {
return nothing
}
args.dst_buffer[0] = BIAS_AND_CLAMP[(((rd0 ~mod+ rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[7] = BIAS_AND_CLAMP[(((rd0 ~mod- rk1) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[1] = BIAS_AND_CLAMP[(((rd1 ~mod+ rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[6] = BIAS_AND_CLAMP[(((rd1 ~mod- rk3) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[2] = BIAS_AND_CLAMP[(((rd2 ~mod+ rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[5] = BIAS_AND_CLAMP[(((rd2 ~mod- rk5) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[3] = BIAS_AND_CLAMP[(((rd3 ~mod+ rk7) ~mod+ (1 << 17)) >> 18) & 1023]
args.dst_buffer[4] = BIAS_AND_CLAMP[(((rd3 ~mod- rk7) ~mod+ (1 << 17)) >> 18) & 1023]
// -------- END generated by script/print-jpeg-idct-code.go
}