From 8cfe304f3c0ddcb8efbec2653f335382f5236849 Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Sun, 7 May 2023 15:48:43 +1000 Subject: [PATCH] std/jpeg: implement Inverse DCT Updates #42 --- internal/cgen/base/fundamental-private.h | 8 + lang/builtin/builtin.go | 2 + release/c/wuffs-unsupported-snapshot.c | 956 +++++++++++++++++ script/print-jpeg-idct-code.go | 323 ++++++ std/jpeg/common_consts.wuffs | 76 ++ std/jpeg/decode_idct_default.wuffs | 1230 ++++++++++++++++++++++ test/c/std/jpeg.c | 67 ++ 7 files changed, 2662 insertions(+) create mode 100644 script/print-jpeg-idct-code.go create mode 100644 std/jpeg/decode_idct_default.wuffs diff --git a/internal/cgen/base/fundamental-private.h b/internal/cgen/base/fundamental-private.h index 71d33f23..316720f6 100644 --- a/internal/cgen/base/fundamental-private.h +++ b/internal/cgen/base/fundamental-private.h @@ -170,6 +170,14 @@ wuffs_base__u64__sat_sub_indirect(uint64_t* x, uint64_t y) { *x = wuffs_base__u64__sat_sub(*x, y); } +// ---------------- Numeric Types (Utility) + +#define wuffs_base__utility__sign_extend_convert_u16_u32(a) \ + ((uint32_t)(int32_t)(int16_t)(a)) + +#define wuffs_base__utility__sign_extend_rshift_u32(a, n) \ + ((uint32_t)(((int32_t)(a)) >> (n))) + // ---------------- Slices and Tables // wuffs_base__slice_u8__prefix returns up to the first up_to bytes of s. diff --git a/lang/builtin/builtin.go b/lang/builtin/builtin.go index 77a6ccd2..e3a172a7 100644 --- a/lang/builtin/builtin.go +++ b/lang/builtin/builtin.go @@ -386,6 +386,8 @@ var funcsOther = [...]string{ "min_incl_x: u32, min_incl_y: u32, max_incl_x: u32, max_incl_y: u32) rect_ii_u32", "utility.make_rect_ie_u32(" + "min_incl_x: u32, min_incl_y: u32, max_excl_x: u32, max_excl_y: u32) rect_ie_u32", + "utility.sign_extend_convert_u16_u32(a: u16) u32", + "utility.sign_extend_rshift_u32(a: u32, n: u32[..= 31]) u32", // ---- ranges diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index 03c472ed..bf4c1650 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -11841,6 +11841,14 @@ wuffs_base__u64__sat_sub_indirect(uint64_t* x, uint64_t y) { *x = wuffs_base__u64__sat_sub(*x, y); } +// ---------------- Numeric Types (Utility) + +#define wuffs_base__utility__sign_extend_convert_u16_u32(a) \ + ((uint32_t)(int32_t)(int16_t)(a)) + +#define wuffs_base__utility__sign_extend_rshift_u32(a, n) \ + ((uint32_t)(((int32_t)(a)) >> (n))) + // ---------------- Slices and Tables // wuffs_base__slice_u8__prefix returns up to the first up_to bytes of s. @@ -35399,10 +35407,150 @@ WUFFS_JPEG__UNZIG[80] WUFFS_BASE__POTENTIALLY_UNUSED = { 63, 63, 63, 63, 63, 63, 63, 63, }; +static const uint8_t +WUFFS_JPEG__BIAS_AND_CLAMP[1024] WUFFS_BASE__POTENTIALLY_UNUSED = { + 128, 129, 130, 131, 132, 133, 134, 135, + 136, 137, 138, 139, 140, 141, 142, 143, + 144, 145, 146, 147, 148, 149, 150, 151, + 152, 153, 154, 155, 156, 157, 158, 159, + 160, 161, 162, 163, 164, 165, 166, 167, + 168, 169, 170, 171, 172, 173, 174, 175, + 176, 177, 178, 179, 180, 181, 182, 183, + 184, 185, 186, 187, 188, 189, 190, 191, + 192, 193, 194, 195, 196, 197, 198, 199, + 200, 201, 202, 203, 204, 205, 206, 207, + 208, 209, 210, 211, 212, 213, 214, 215, + 216, 217, 218, 219, 220, 221, 222, 223, + 224, 225, 226, 227, 228, 229, 230, 231, + 232, 233, 234, 235, 236, 237, 238, 239, + 240, 241, 242, 243, 244, 245, 246, 247, + 248, 249, 250, 251, 252, 253, 254, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 1, 2, 3, 4, 5, 6, 7, + 8, 9, 10, 11, 12, 13, 14, 15, + 16, 17, 18, 19, 20, 21, 22, 23, + 24, 25, 26, 27, 28, 29, 30, 31, + 32, 33, 34, 35, 36, 37, 38, 39, + 40, 41, 42, 43, 44, 45, 46, 47, + 48, 49, 50, 51, 52, 53, 54, 55, + 56, 57, 58, 59, 60, 61, 62, 63, + 64, 65, 66, 67, 68, 69, 70, 71, + 72, 73, 74, 75, 76, 77, 78, 79, + 80, 81, 82, 83, 84, 85, 86, 87, + 88, 89, 90, 91, 92, 93, 94, 95, + 96, 97, 98, 99, 100, 101, 102, 103, + 104, 105, 106, 107, 108, 109, 110, 111, + 112, 113, 114, 115, 116, 117, 118, 119, + 120, 121, 122, 123, 124, 125, 126, 127, +}; + // ---------------- Private Initializer Prototypes // ---------------- Private Function Prototypes +static wuffs_base__empty_struct +wuffs_jpeg__decoder__decode_idct( + wuffs_jpeg__decoder* self, + wuffs_base__slice_u8 a_dst_buffer, + uint64_t a_dst_stride, + uint32_t a_b, + uint32_t a_q); + static wuffs_base__status wuffs_jpeg__decoder__do_decode_image_config( wuffs_jpeg__decoder* self, @@ -35588,6 +35736,814 @@ sizeof__wuffs_jpeg__decoder() { // ---------------- Function Implementations +// -------- func jpeg.decoder.decode_idct + +static wuffs_base__empty_struct +wuffs_jpeg__decoder__decode_idct( + wuffs_jpeg__decoder* self, + wuffs_base__slice_u8 a_dst_buffer, + uint64_t a_dst_stride, + uint32_t a_b, + uint32_t a_q) { + uint32_t v_bq0 = 0; + uint32_t v_bq2 = 0; + uint32_t v_bq4 = 0; + uint32_t v_bq6 = 0; + uint32_t v_ca = 0; + uint32_t v_cb2 = 0; + uint32_t v_cb6 = 0; + uint32_t v_ccp = 0; + uint32_t v_ccm = 0; + uint32_t v_cd0 = 0; + uint32_t v_cd1 = 0; + uint32_t v_cd2 = 0; + uint32_t v_cd3 = 0; + uint32_t v_bq1 = 0; + uint32_t v_bq3 = 0; + uint32_t v_bq5 = 0; + uint32_t v_bq7 = 0; + uint32_t v_ci51 = 0; + uint32_t v_ci53 = 0; + uint32_t v_ci71 = 0; + uint32_t v_ci73 = 0; + uint32_t v_cj = 0; + uint32_t v_ck1 = 0; + uint32_t v_ck3 = 0; + uint32_t v_ck5 = 0; + uint32_t v_ck7 = 0; + uint32_t v_cl51 = 0; + uint32_t v_cl73 = 0; + uint32_t v_in0 = 0; + uint32_t v_in2 = 0; + uint32_t v_in4 = 0; + uint32_t v_in6 = 0; + uint32_t v_ra = 0; + uint32_t v_rb2 = 0; + uint32_t v_rb6 = 0; + uint32_t v_rcp = 0; + uint32_t v_rcm = 0; + uint32_t v_rd0 = 0; + uint32_t v_rd1 = 0; + uint32_t v_rd2 = 0; + uint32_t v_rd3 = 0; + uint32_t v_in1 = 0; + uint32_t v_in3 = 0; + uint32_t v_in5 = 0; + uint32_t v_in7 = 0; + uint32_t v_ri51 = 0; + uint32_t v_ri53 = 0; + uint32_t v_ri71 = 0; + uint32_t v_ri73 = 0; + uint32_t v_rj = 0; + uint32_t v_rk1 = 0; + uint32_t v_rk3 = 0; + uint32_t v_rk5 = 0; + uint32_t v_rk7 = 0; + uint32_t v_rl51 = 0; + uint32_t v_rl73 = 0; + uint32_t v_intermediate[64] = {0}; + + if (8 > 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][16]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][16])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][48]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][48])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][0]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][0])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][32]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][32])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][8]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][8])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][24]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][24])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][40]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][40])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][56]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][56])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[0] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[56] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[8] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[48] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[16] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[40] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[24] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[32] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][17]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][17])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][49]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][49])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][1]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][1])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][33]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][33])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][9]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][9])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][25]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][25])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][41]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][41])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][57]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][57])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[1] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[57] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[9] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[49] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[17] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[41] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[25] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[33] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][18]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][18])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][50]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][50])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][2]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][2])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][34]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][34])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][10]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][10])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][26]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][26])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][42]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][42])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][58]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][58])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[2] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[58] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[10] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[50] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[18] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[42] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[26] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[34] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][19]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][19])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][51]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][51])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][3]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][3])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][35]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][35])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][11]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][11])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][27]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][27])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][43]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][43])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][59]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][59])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[3] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[59] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[11] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[51] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[19] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[43] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[27] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[35] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][20]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][20])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][52]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][52])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][4]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][4])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][36]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][36])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][12]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][12])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][28]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][28])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][44]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][44])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][60]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][60])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[4] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[60] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[12] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[52] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[20] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[44] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[28] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[36] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][21]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][21])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][53]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][53])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][5]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][5])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][37]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][37])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][13]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][13])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][29]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][29])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][45]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][45])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][61]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][61])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[5] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[61] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[13] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[53] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[21] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[45] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[29] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[37] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][22]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][22])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][54]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][54])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][6]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][6])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][38]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][38])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][14]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][14])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][30]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][30])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][46]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][46])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][62]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][62])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[6] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[62] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[14] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[54] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[22] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[46] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[30] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[38] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_bq2 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][23]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][23])))); + v_bq6 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][55]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][55])))); + v_ca = ((uint32_t)(((uint32_t)(v_bq2 + v_bq6)) * 4433)); + v_cb2 = ((uint32_t)(v_ca + ((uint32_t)(v_bq2 * 6270)))); + v_cb6 = ((uint32_t)(v_ca - ((uint32_t)(v_bq6 * 15137)))); + v_bq0 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][7]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][7])))); + v_bq4 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][39]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][39])))); + v_ccp = ((uint32_t)(((uint32_t)(v_bq0 + v_bq4)) << 13)); + v_ccm = ((uint32_t)(((uint32_t)(v_bq0 - v_bq4)) << 13)); + 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][15]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][15])))); + v_bq3 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][31]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][31])))); + v_bq5 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][47]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][47])))); + v_bq7 = ((uint32_t)(wuffs_base__utility__sign_extend_convert_u16_u32(self->private_data.f_mcu_blocks[a_b][63]) * ((uint32_t)(self->private_impl.f_quant_tables[a_q][63])))); + 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)) * 9633)); + v_ck1 = ((uint32_t)(v_bq1 * 12299)); + v_ck3 = ((uint32_t)(v_bq3 * 25172)); + v_ck5 = ((uint32_t)(v_bq5 * 16819)); + v_ck7 = ((uint32_t)(v_bq7 * 2446)); + v_ci51 *= 4294964100; + v_ci53 *= 4294946301; + v_ci71 *= 4294959923; + v_ci73 *= 4294951227; + 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[7] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 + v_ck1)) + 1024)), 11); + v_intermediate[63] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd0 - v_ck1)) + 1024)), 11); + v_intermediate[15] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 + v_ck3)) + 1024)), 11); + v_intermediate[55] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd1 - v_ck3)) + 1024)), 11); + v_intermediate[23] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 + v_ck5)) + 1024)), 11); + v_intermediate[47] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd2 - v_ck5)) + 1024)), 11); + v_intermediate[31] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 + v_ck7)) + 1024)), 11); + v_intermediate[39] = wuffs_base__utility__sign_extend_rshift_u32(((uint32_t)(((uint32_t)(v_cd3 - v_ck7)) + 1024)), 11); + v_in2 = v_intermediate[2]; + v_in6 = v_intermediate[6]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[0]; + v_in4 = v_intermediate[4]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[1]; + v_in3 = v_intermediate[3]; + v_in5 = v_intermediate[5]; + v_in7 = v_intermediate[7]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[10]; + v_in6 = v_intermediate[14]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[8]; + v_in4 = v_intermediate[12]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[9]; + v_in3 = v_intermediate[11]; + v_in5 = v_intermediate[13]; + v_in7 = v_intermediate[15]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[18]; + v_in6 = v_intermediate[22]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[16]; + v_in4 = v_intermediate[20]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[17]; + v_in3 = v_intermediate[19]; + v_in5 = v_intermediate[21]; + v_in7 = v_intermediate[23]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[26]; + v_in6 = v_intermediate[30]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[24]; + v_in4 = v_intermediate[28]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[25]; + v_in3 = v_intermediate[27]; + v_in5 = v_intermediate[29]; + v_in7 = v_intermediate[31]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[34]; + v_in6 = v_intermediate[38]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[32]; + v_in4 = v_intermediate[36]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[33]; + v_in3 = v_intermediate[35]; + v_in5 = v_intermediate[37]; + v_in7 = v_intermediate[39]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[42]; + v_in6 = v_intermediate[46]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[40]; + v_in4 = v_intermediate[44]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[41]; + v_in3 = v_intermediate[43]; + v_in5 = v_intermediate[45]; + v_in7 = v_intermediate[47]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[50]; + v_in6 = v_intermediate[54]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[48]; + v_in4 = v_intermediate[52]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[49]; + v_in3 = v_intermediate[51]; + v_in5 = v_intermediate[53]; + v_in7 = v_intermediate[55]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer = wuffs_base__slice_u8__subslice_i(a_dst_buffer, a_dst_stride); + v_in2 = v_intermediate[58]; + v_in6 = v_intermediate[62]; + v_ra = ((uint32_t)(((uint32_t)(v_in2 + v_in6)) * 4433)); + v_rb2 = ((uint32_t)(v_ra + ((uint32_t)(v_in2 * 6270)))); + v_rb6 = ((uint32_t)(v_ra - ((uint32_t)(v_in6 * 15137)))); + v_in0 = v_intermediate[56]; + v_in4 = v_intermediate[60]; + v_rcp = ((uint32_t)(((uint32_t)(v_in0 + v_in4)) << 13)); + v_rcm = ((uint32_t)(((uint32_t)(v_in0 - v_in4)) << 13)); + 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[57]; + v_in3 = v_intermediate[59]; + v_in5 = v_intermediate[61]; + v_in7 = v_intermediate[63]; + 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)) * 9633)); + v_rk1 = ((uint32_t)(v_in1 * 12299)); + v_rk3 = ((uint32_t)(v_in3 * 25172)); + v_rk5 = ((uint32_t)(v_in5 * 16819)); + v_rk7 = ((uint32_t)(v_in7 * 2446)); + v_ri51 *= 4294964100; + v_ri53 *= 4294946301; + v_ri71 *= 4294959923; + v_ri73 *= 4294951227; + 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 (8 > ((uint64_t)(a_dst_buffer.len))) { + return wuffs_base__make_empty_struct(); + } + a_dst_buffer.ptr[0] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 + v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[7] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd0 - v_rk1)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[1] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 + v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[6] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd1 - v_rk3)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[2] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 + v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[5] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd2 - v_rk5)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[3] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 + v_rk7)) + 131072)) >> 18) & 1023)]; + a_dst_buffer.ptr[4] = WUFFS_JPEG__BIAS_AND_CLAMP[((((uint32_t)(((uint32_t)(v_rd3 - v_rk7)) + 131072)) >> 18) & 1023)]; + return wuffs_base__make_empty_struct(); +} + // -------- func jpeg.decoder.set_quirk WUFFS_BASE__MAYBE_STATIC wuffs_base__status diff --git a/script/print-jpeg-idct-code.go b/script/print-jpeg-idct-code.go new file mode 100644 index 00000000..76a16d17 --- /dev/null +++ b/script/print-jpeg-idct-code.go @@ -0,0 +1,323 @@ +// Copyright 2023 The Wuffs Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +//go:build ignore +// +build ignore + +package main + +// print-jpeg-idct-code.go prints the "func jpeg.decoder.decode_idct" code. +// +// Usage: go run print-jpeg-idct-code.go + +import ( + "fmt" + "os" + "strings" +) + +func main() { + if err := main1(); err != nil { + os.Stderr.WriteString(err.Error() + "\n") + os.Exit(1) + } +} + +func main1() error { + fmt.Printf("// -------- BEGIN generated by script/print-jpeg-idct-code.go\n") + fmt.Println() + + fmt.Printf("// p0_298631336 = 0x%04X = %5d\n", round(p0_298631336), round(p0_298631336)) + fmt.Printf("// p0_390180644 = 0x%04X = %5d\n", round(p0_390180644), round(p0_390180644)) + fmt.Printf("// p0_541196100 = 0x%04X = %5d\n", round(p0_541196100), round(p0_541196100)) + fmt.Printf("// p0_765366865 = 0x%04X = %5d\n", round(p0_765366865), round(p0_765366865)) + fmt.Printf("// p0_899976223 = 0x%04X = %5d\n", round(p0_899976223), round(p0_899976223)) + fmt.Printf("// p1_175875602 = 0x%04X = %5d\n", round(p1_175875602), round(p1_175875602)) + fmt.Printf("// p1_501321110 = 0x%04X = %5d\n", round(p1_501321110), round(p1_501321110)) + fmt.Printf("// p1_847759065 = 0x%04X = %5d\n", round(p1_847759065), round(p1_847759065)) + fmt.Printf("// p1_961570560 = 0x%04X = %5d\n", round(p1_961570560), round(p1_961570560)) + fmt.Printf("// p2_053119869 = 0x%04X = %5d\n", round(p2_053119869), round(p2_053119869)) + fmt.Printf("// p2_562915447 = 0x%04X = %5d\n", round(p2_562915447), round(p2_562915447)) + fmt.Printf("// p3_072711026 = 0x%04X = %5d\n", round(p3_072711026), round(p3_072711026)) + fmt.Printf("//\n") + fmt.Printf("// m0_390180644 = 0x%08X = %10d\n", round(m0_390180644), round(m0_390180644)) + fmt.Printf("// m0_899976223 = 0x%08X = %10d\n", round(m0_899976223), round(m0_899976223)) + fmt.Printf("// m1_961570560 = 0x%08X = %10d\n", round(m1_961570560), round(m1_961570560)) + fmt.Printf("// m2_562915447 = 0x%08X = %10d\n", round(m2_562915447), round(m2_562915447)) + fmt.Println() + + for x := 0; x < 8; x++ { + fmt.Println(strings.TrimSpace(replace(pass0, map[string]string{ + "$colX$": fmt.Sprint(x), + "$row0colX$": fmt.Sprintf("0x%02X", (8*0)|x), + "$row1colX$": fmt.Sprintf("0x%02X", (8*1)|x), + "$row2colX$": fmt.Sprintf("0x%02X", (8*2)|x), + "$row3colX$": fmt.Sprintf("0x%02X", (8*3)|x), + "$row4colX$": fmt.Sprintf("0x%02X", (8*4)|x), + "$row5colX$": fmt.Sprintf("0x%02X", (8*5)|x), + "$row6colX$": fmt.Sprintf("0x%02X", (8*6)|x), + "$row7colX$": fmt.Sprintf("0x%02X", (8*7)|x), + "$p0_298631336$": fmt.Sprintf("0x%04X", round(p0_298631336)), + "$p0_541196100$": fmt.Sprintf("0x%04X", round(p0_541196100)), + "$p0_765366865$": fmt.Sprintf("0x%04X", round(p0_765366865)), + "$p1_175875602$": fmt.Sprintf("0x%04X", round(p1_175875602)), + "$p1_501321110$": fmt.Sprintf("0x%04X", round(p1_501321110)), + "$p1_847759065$": fmt.Sprintf("0x%04X", round(p1_847759065)), + "$p2_053119869$": fmt.Sprintf("0x%04X", round(p2_053119869)), + "$p3_072711026$": fmt.Sprintf("0x%04X", round(p3_072711026)), + "$m0_390180644$": fmt.Sprintf("0x%08X", round(m0_390180644)), + "$m0_899976223$": fmt.Sprintf("0x%08X", round(m0_899976223)), + "$m1_961570560$": fmt.Sprintf("0x%08X", round(m1_961570560)), + "$m2_562915447$": fmt.Sprintf("0x%08X", round(m2_562915447)), + }))) + fmt.Println() + } + + for y := 0; y < 8; y++ { + fmt.Println(strings.TrimSpace(replace(pass1, map[string]string{ + "$rowY$": fmt.Sprint(y), + "$rowYcol0$": fmt.Sprintf("0x%02X", (8*y)|0), + "$rowYcol1$": fmt.Sprintf("0x%02X", (8*y)|1), + "$rowYcol2$": fmt.Sprintf("0x%02X", (8*y)|2), + "$rowYcol3$": fmt.Sprintf("0x%02X", (8*y)|3), + "$rowYcol4$": fmt.Sprintf("0x%02X", (8*y)|4), + "$rowYcol5$": fmt.Sprintf("0x%02X", (8*y)|5), + "$rowYcol6$": fmt.Sprintf("0x%02X", (8*y)|6), + "$rowYcol7$": fmt.Sprintf("0x%02X", (8*y)|7), + "$p0_298631336$": fmt.Sprintf("0x%04X", round(p0_298631336)), + "$p0_541196100$": fmt.Sprintf("0x%04X", round(p0_541196100)), + "$p0_765366865$": fmt.Sprintf("0x%04X", round(p0_765366865)), + "$p1_175875602$": fmt.Sprintf("0x%04X", round(p1_175875602)), + "$p1_501321110$": fmt.Sprintf("0x%04X", round(p1_501321110)), + "$p1_847759065$": fmt.Sprintf("0x%04X", round(p1_847759065)), + "$p2_053119869$": fmt.Sprintf("0x%04X", round(p2_053119869)), + "$p3_072711026$": fmt.Sprintf("0x%04X", round(p3_072711026)), + "$m0_390180644$": fmt.Sprintf("0x%08X", round(m0_390180644)), + "$m0_899976223$": fmt.Sprintf("0x%08X", round(m0_899976223)), + "$m1_961570560$": fmt.Sprintf("0x%08X", round(m1_961570560)), + "$m2_562915447$": fmt.Sprintf("0x%08X", round(m2_562915447)), + "$bounds_check$": boundsCheck(y == 7), + "$advance$": advance(y == 7), + }))) + fmt.Println() + } + + fmt.Printf("// -------- END generated by script/print-jpeg-idct-code.go\n") + return nil +} + +func replace(s string, m map[string]string) string { + for k, v := range m { + s = strings.ReplaceAll(s, k, v) + } + return s +} + +func round(x float64) uint32 { + return uint32(0.5 + (x * (1 << 13))) +} + +func boundsCheck(final bool) string { + if final { + return "" + + "if 8 > args.dst_buffer.length() {\n" + + " return nothing\n" + + "}" + } + return "" + + "if args.dst_stride > args.dst_buffer.length() {\n" + + " return nothing\n" + + "}\n" + + `assert 8 <= args.dst_buffer.length() via "a <= b: a <= c; c <= b"(c: args.dst_stride)` +} + +func advance(final bool) string { + if final { + return "" + } + return "args.dst_buffer = args.dst_buffer[args.dst_stride ..]" +} + +const ( + sqrt2 = 1.4142135623730950488016887242096980785696718753769480731766797379 + + // cosNpi16 ≈ cos(N * (pi / 16)). + cos1pi16 = 0.9807852804032304491261822361342390369739337308933360950029160885 + cos2pi16 = 0.9238795325112867561281831893967882868224166258636424861150977312 + cos3pi16 = 0.8314696123025452370787883776179057567385608119872499634461245902 + cos5pi16 = 0.5555702330196022247428308139485328743749371907548040459241535282 + cos6pi16 = 0.3826834323650897717284599840303988667613445624856270414338006356 + cos7pi16 = 0.1950903220161282678482848684770222409276916177519548077545020894 +) + +const ( + p0_541196100 = sqrt2 * (+cos6pi16) + p1_175875602 = sqrt2 * (+cos3pi16) + + p0_765366865 = sqrt2 * (+cos2pi16 - cos6pi16) + p1_847759065 = sqrt2 * (+cos2pi16 + cos6pi16) + + p0_390180644 = sqrt2 * (+cos3pi16 - cos5pi16) + p0_899976223 = sqrt2 * (+cos3pi16 - cos7pi16) + p1_961570560 = sqrt2 * (+cos3pi16 + cos5pi16) + p2_562915447 = sqrt2 * (+cos3pi16 + cos1pi16) + + p0_298631336 = sqrt2 * (-cos1pi16 + cos3pi16 + cos5pi16 - cos7pi16) + p1_501321110 = sqrt2 * (+cos1pi16 + cos3pi16 - cos5pi16 - cos7pi16) + p2_053119869 = sqrt2 * (+cos1pi16 + cos3pi16 - cos5pi16 + cos7pi16) + p3_072711026 = sqrt2 * (+cos1pi16 + cos3pi16 + cos5pi16 - cos7pi16) + + m0_390180644 = (1 << (32 - 13)) - p0_390180644 + m0_899976223 = (1 << (32 - 13)) - p0_899976223 + m1_961570560 = (1 << (32 - 13)) - p1_961570560 + m2_562915447 = (1 << (32 - 13)) - p2_562915447 +) + +const pass0 = `// ==== First pass, column $colX$. + +// 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) +bq6 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row6colX$]) ~mod* (this.quant_tables[args.q][$row6colX$] as base.u32) + +ca = (bq2 ~mod+ bq6) ~mod* $p0_541196100$ + +cb2 = ca ~mod+ (bq2 ~mod* $p0_765366865$) +cb6 = ca ~mod- (bq6 ~mod* $p1_847759065$) + +bq0 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row0colX$]) ~mod* (this.quant_tables[args.q][$row0colX$] as base.u32) +bq4 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row4colX$]) ~mod* (this.quant_tables[args.q][$row4colX$] as base.u32) + +ccp = (bq0 ~mod+ bq4) ~mod<< 13 +ccm = (bq0 ~mod- bq4) ~mod<< 13 + +cd0 = ccp ~mod+ cb2 +cd1 = ccm ~mod+ cb6 +cd2 = ccm ~mod- cb6 +cd3 = ccp ~mod- cb2 + +// Odd rows. + +bq1 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row1colX$]) ~mod* (this.quant_tables[args.q][$row1colX$] as base.u32) +bq3 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row3colX$]) ~mod* (this.quant_tables[args.q][$row3colX$] as base.u32) +bq5 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row5colX$]) ~mod* (this.quant_tables[args.q][$row5colX$] as base.u32) +bq7 = this.util.sign_extend_convert_u16_u32(a: this.mcu_blocks[args.b][$row7colX$]) ~mod* (this.quant_tables[args.q][$row7colX$] as base.u32) + +ci51 = bq5 ~mod+ bq1 +ci53 = bq5 ~mod+ bq3 +ci71 = bq7 ~mod+ bq1 +ci73 = bq7 ~mod+ bq3 + +cj = (ci73 ~mod+ ci51) ~mod* $p1_175875602$ + +ck1 = bq1 ~mod* $p1_501321110$ +ck3 = bq3 ~mod* $p3_072711026$ +ck5 = bq5 ~mod* $p2_053119869$ +ck7 = bq7 ~mod* $p0_298631336$ + +ci51 ~mod*= $m0_390180644$ +ci53 ~mod*= $m2_562915447$ +ci71 ~mod*= $m0_899976223$ +ci73 ~mod*= $m1_961570560$ + +cl51 = ci51 ~mod+ cj +cl73 = ci73 ~mod+ cj + +ck1 ~mod+= ci71 ~mod+ cl51 +ck3 ~mod+= ci53 ~mod+ cl73 +ck5 ~mod+= ci53 ~mod+ cl51 +ck7 ~mod+= ci71 ~mod+ cl73 + +// Combine rows. + +intermediate[$row0colX$] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod+ ck1) ~mod+ (1 << 10), n: 11) +intermediate[$row7colX$] = this.util.sign_extend_rshift_u32(a: (cd0 ~mod- ck1) ~mod+ (1 << 10), n: 11) +intermediate[$row1colX$] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod+ ck3) ~mod+ (1 << 10), n: 11) +intermediate[$row6colX$] = this.util.sign_extend_rshift_u32(a: (cd1 ~mod- ck3) ~mod+ (1 << 10), n: 11) +intermediate[$row2colX$] = this.util.sign_extend_rshift_u32(a: (cd2 ~mod+ ck5) ~mod+ (1 << 10), n: 11) +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 $rowY$. + +// Even columns. + +in2 = intermediate[$rowYcol2$] +in6 = intermediate[$rowYcol6$] + +ra = (in2 ~mod+ in6) ~mod* $p0_541196100$ + +rb2 = ra ~mod+ (in2 ~mod* $p0_765366865$) +rb6 = ra ~mod- (in6 ~mod* $p1_847759065$) + +in0 = intermediate[$rowYcol0$] +in4 = intermediate[$rowYcol4$] + +rcp = (in0 ~mod+ in4) ~mod<< 13 +rcm = (in0 ~mod- in4) ~mod<< 13 + +rd0 = rcp ~mod+ rb2 +rd1 = rcm ~mod+ rb6 +rd2 = rcm ~mod- rb6 +rd3 = rcp ~mod- rb2 + +// Odd columns. + +in1 = intermediate[$rowYcol1$] +in3 = intermediate[$rowYcol3$] +in5 = intermediate[$rowYcol5$] +in7 = intermediate[$rowYcol7$] + +ri51 = in5 ~mod+ in1 +ri53 = in5 ~mod+ in3 +ri71 = in7 ~mod+ in1 +ri73 = in7 ~mod+ in3 + +rj = (ri73 ~mod+ ri51) ~mod* $p1_175875602$ + +rk1 = in1 ~mod* $p1_501321110$ +rk3 = in3 ~mod* $p3_072711026$ +rk5 = in5 ~mod* $p2_053119869$ +rk7 = in7 ~mod* $p0_298631336$ + +ri51 ~mod*= $m0_390180644$ +ri53 ~mod*= $m2_562915447$ +ri71 ~mod*= $m0_899976223$ +ri73 ~mod*= $m1_961570560$ + +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. + +$bounds_check$ + +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] + +$advance$ +` diff --git a/std/jpeg/common_consts.wuffs b/std/jpeg/common_consts.wuffs index e507bb21..fe007ac2 100644 --- a/std/jpeg/common_consts.wuffs +++ b/std/jpeg/common_consts.wuffs @@ -30,3 +30,79 @@ pri const UNZIG : roarray[80] base.u8[..= 63] = [ 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, 0x3F, ] + +// BIAS_AND_CLAMP maps 10-bit signed values (0 centered, in the range -512 .. +// +512) to 8-bit unsigned values (128 centered, in the range 0 .. 256). +pri const BIAS_AND_CLAMP : roarray[1024] base.u8 = [ + 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F, + 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9A, 0x9B, 0x9C, 0x9D, 0x9E, 0x9F, + 0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7, 0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF, + 0xB0, 0xB1, 0xB2, 0xB3, 0xB4, 0xB5, 0xB6, 0xB7, 0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF, + 0xC0, 0xC1, 0xC2, 0xC3, 0xC4, 0xC5, 0xC6, 0xC7, 0xC8, 0xC9, 0xCA, 0xCB, 0xCC, 0xCD, 0xCE, 0xCF, + 0xD0, 0xD1, 0xD2, 0xD3, 0xD4, 0xD5, 0xD6, 0xD7, 0xD8, 0xD9, 0xDA, 0xDB, 0xDC, 0xDD, 0xDE, 0xDF, + 0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF, + 0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xFF, + + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, + 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, + 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F, + 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B, 0x3C, 0x3D, 0x3E, 0x3F, + 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4A, 0x4B, 0x4C, 0x4D, 0x4E, 0x4F, + 0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F, + 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F, + 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F, +] diff --git a/std/jpeg/decode_idct_default.wuffs b/std/jpeg/decode_idct_default.wuffs new file mode 100644 index 00000000..3cc58108 --- /dev/null +++ b/std/jpeg/decode_idct_default.wuffs @@ -0,0 +1,1230 @@ +// Copyright 2023 The Wuffs Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +pri func decoder.decode_idct!(dst_buffer: slice base.u8, dst_stride: base.u64, b: base.u32[..= 9], q: base.u32[..= 3]) { + // This method implements the same algorithm as libjpeg-turbo's jidctint.c. + + var bq0 : base.u32 + var bq2 : base.u32 + var bq4 : base.u32 + var bq6 : base.u32 + + var ca : base.u32 + + var cb2 : base.u32 + var cb6 : base.u32 + + var ccp : base.u32 + var ccm : base.u32 + + var cd0 : base.u32 + var cd1 : base.u32 + var cd2 : base.u32 + var cd3 : base.u32 + + var bq1 : base.u32 + var bq3 : base.u32 + var bq5 : base.u32 + var bq7 : base.u32 + + var ci51 : base.u32 + var ci53 : base.u32 + var ci71 : base.u32 + var ci73 : base.u32 + + var cj : base.u32 + + var ck1 : base.u32 + var ck3 : base.u32 + var ck5 : base.u32 + var ck7 : base.u32 + + var cl51 : base.u32 + var cl73 : base.u32 + + var in0 : base.u32 + var in2 : base.u32 + var in4 : base.u32 + var in6 : base.u32 + + var ra : base.u32 + + var rb2 : base.u32 + var rb6 : base.u32 + + var rcp : base.u32 + var rcm : base.u32 + + var rd0 : base.u32 + var rd1 : base.u32 + var rd2 : base.u32 + var rd3 : base.u32 + + var in1 : base.u32 + var in3 : base.u32 + var in5 : base.u32 + var in7 : base.u32 + + var ri51 : base.u32 + var ri53 : base.u32 + var ri71 : base.u32 + var ri73 : base.u32 + + var rj : base.u32 + + var rk1 : base.u32 + var rk3 : base.u32 + var rk5 : base.u32 + var rk7 : base.u32 + + var rl51 : base.u32 + var rl73 : base.u32 + + var intermediate : array[64] base.u32 + + if 8 > args.dst_stride { + return nothing + } + + // -------- BEGIN generated by script/print-jpeg-idct-code.go + + // p0_298631336 = 0x098E = 2446 + // p0_390180644 = 0x0C7C = 3196 + // p0_541196100 = 0x1151 = 4433 + // p0_765366865 = 0x187E = 6270 + // p0_899976223 = 0x1CCD = 7373 + // p1_175875602 = 0x25A1 = 9633 + // p1_501321110 = 0x300B = 12299 + // p1_847759065 = 0x3B21 = 15137 + // p1_961570560 = 0x3EC5 = 16069 + // p2_053119869 = 0x41B3 = 16819 + // p2_562915447 = 0x5203 = 20995 + // p3_072711026 = 0x6254 = 25172 + // + // m0_390180644 = 0xFFFFF384 = 4294964100 + // m0_899976223 = 0xFFFFE333 = 4294959923 + // m1_961570560 = 0xFFFFC13B = 4294951227 + // m2_562915447 = 0xFFFFADFD = 4294946301 + + // ==== First pass, column 0. + + // Even rows. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + 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) + + ca = (bq2 ~mod+ bq6) ~mod* 0x1151 + + cb2 = ca ~mod+ (bq2 ~mod* 0x187E) + cb6 = ca ~mod- (bq6 ~mod* 0x3B21) + + 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) + + ccp = (bq0 ~mod+ bq4) ~mod<< 13 + ccm = (bq0 ~mod- bq4) ~mod<< 13 + + cd0 = ccp ~mod+ cb2 + cd1 = ccm ~mod+ cb6 + cd2 = ccm ~mod- cb6 + cd3 = ccp ~mod- cb2 + + // Odd rows. + + 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) + + ci51 = bq5 ~mod+ bq1 + ci53 = bq5 ~mod+ bq3 + ci71 = bq7 ~mod+ bq1 + ci73 = bq7 ~mod+ bq3 + + cj = (ci73 ~mod+ ci51) ~mod* 0x25A1 + + ck1 = bq1 ~mod* 0x300B + ck3 = bq3 ~mod* 0x6254 + ck5 = bq5 ~mod* 0x41B3 + ck7 = bq7 ~mod* 0x098E + + ci51 ~mod*= 0xFFFF_F384 + ci53 ~mod*= 0xFFFF_ADFD + ci71 ~mod*= 0xFFFF_E333 + ci73 ~mod*= 0xFFFF_C13B + + cl51 = ci51 ~mod+ cj + cl73 = ci73 ~mod+ cj + + 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. + + in2 = intermediate[0x02] + in6 = intermediate[0x06] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x00] + in4 = intermediate[0x04] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x01] + in3 = intermediate[0x03] + in5 = intermediate[0x05] + in7 = intermediate[0x07] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 1. + + // Even columns. + + in2 = intermediate[0x0A] + in6 = intermediate[0x0E] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x08] + in4 = intermediate[0x0C] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x09] + in3 = intermediate[0x0B] + in5 = intermediate[0x0D] + in7 = intermediate[0x0F] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 2. + + // Even columns. + + in2 = intermediate[0x12] + in6 = intermediate[0x16] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x10] + in4 = intermediate[0x14] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x11] + in3 = intermediate[0x13] + in5 = intermediate[0x15] + in7 = intermediate[0x17] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 3. + + // Even columns. + + in2 = intermediate[0x1A] + in6 = intermediate[0x1E] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x18] + in4 = intermediate[0x1C] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x19] + in3 = intermediate[0x1B] + in5 = intermediate[0x1D] + in7 = intermediate[0x1F] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 4. + + // Even columns. + + in2 = intermediate[0x22] + in6 = intermediate[0x26] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x20] + in4 = intermediate[0x24] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x21] + in3 = intermediate[0x23] + in5 = intermediate[0x25] + in7 = intermediate[0x27] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 5. + + // Even columns. + + in2 = intermediate[0x2A] + in6 = intermediate[0x2E] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x28] + in4 = intermediate[0x2C] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x29] + in3 = intermediate[0x2B] + in5 = intermediate[0x2D] + in7 = intermediate[0x2F] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 6. + + // Even columns. + + in2 = intermediate[0x32] + in6 = intermediate[0x36] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x30] + in4 = intermediate[0x34] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x31] + in3 = intermediate[0x33] + in5 = intermediate[0x35] + in7 = intermediate[0x37] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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 ..] + + // ==== Second pass, row 7. + + // Even columns. + + in2 = intermediate[0x3A] + in6 = intermediate[0x3E] + + ra = (in2 ~mod+ in6) ~mod* 0x1151 + + rb2 = ra ~mod+ (in2 ~mod* 0x187E) + rb6 = ra ~mod- (in6 ~mod* 0x3B21) + + in0 = intermediate[0x38] + in4 = intermediate[0x3C] + + rcp = (in0 ~mod+ in4) ~mod<< 13 + rcm = (in0 ~mod- in4) ~mod<< 13 + + rd0 = rcp ~mod+ rb2 + rd1 = rcm ~mod+ rb6 + rd2 = rcm ~mod- rb6 + rd3 = rcp ~mod- rb2 + + // Odd columns. + + in1 = intermediate[0x39] + in3 = intermediate[0x3B] + in5 = intermediate[0x3D] + in7 = intermediate[0x3F] + + ri51 = in5 ~mod+ in1 + ri53 = in5 ~mod+ in3 + ri71 = in7 ~mod+ in1 + ri73 = in7 ~mod+ in3 + + rj = (ri73 ~mod+ ri51) ~mod* 0x25A1 + + rk1 = in1 ~mod* 0x300B + rk3 = in3 ~mod* 0x6254 + rk5 = in5 ~mod* 0x41B3 + rk7 = in7 ~mod* 0x098E + + 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] + + // -------- END generated by script/print-jpeg-idct-code.go +} diff --git a/test/c/std/jpeg.c b/test/c/std/jpeg.c index cffac60b..81696bb4 100644 --- a/test/c/std/jpeg.c +++ b/test/c/std/jpeg.c @@ -531,6 +531,72 @@ test_wuffs_jpeg_decode_dht_hard() { &want_fast); } +const char* // +test_wuffs_jpeg_decode_idct() { + CHECK_FOCUS(__func__); + + // This is "test/data/bricks-color.jpeg"'s first MCU's first block, in + // natural (not zig-zag) order. + const uint16_t mcu_block[64] = { + 0xFFC9, 0xFFD8, 0x0014, 0xFFF7, 0x0002, 0x0000, 0x0000, 0x0000, // + 0x006A, 0xFFE3, 0x001C, 0xFFF9, 0x0002, 0x0000, 0x0000, 0x0000, // + 0x0015, 0x0002, 0x0002, 0xFFFE, 0x0001, 0x0000, 0x0000, 0x0001, // + 0x000D, 0xFFEC, 0x0005, 0xFFFE, 0x0000, 0x0000, 0x0000, 0x0000, // + 0xFFFA, 0xFFFA, 0x0002, 0xFFFF, 0x0000, 0x0000, 0x0000, 0x0000, // + 0x0001, 0xFFFD, 0x0001, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, // + 0x0000, 0x0001, 0x0001, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, // + 0x0001, 0x0001, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, // + }; + + // This is "test/data/bricks-color.jpeg"'s first quantization table, in + // natural (not zig-zag) order. + const uint8_t quant_table[64] = { + 0x03, 0x02, 0x02, 0x03, 0x04, 0x06, 0x08, 0x0A, // + 0x02, 0x02, 0x02, 0x03, 0x04, 0x09, 0x0A, 0x09, // + 0x02, 0x02, 0x03, 0x04, 0x06, 0x09, 0x0B, 0x09, // + 0x02, 0x03, 0x04, 0x05, 0x08, 0x0E, 0x0D, 0x0A, // + 0x03, 0x04, 0x06, 0x09, 0x0B, 0x11, 0x10, 0x0C, // + 0x04, 0x06, 0x09, 0x0A, 0x0D, 0x11, 0x12, 0x0F, // + 0x08, 0x0A, 0x0C, 0x0E, 0x10, 0x13, 0x13, 0x10, // + 0x0C, 0x0F, 0x0F, 0x10, 0x12, 0x10, 0x10, 0x10, // + }; + + // This is the IDCT's expected result (including dequantization), again in + // natural (not zig-zag) order. + const uint8_t want_array[64] = { + 0x81, 0x7E, 0x82, 0x7E, 0x82, 0x92, 0xC5, 0xF2, // + 0x81, 0x80, 0x84, 0x85, 0x85, 0x88, 0x9D, 0xB2, // + 0x86, 0x81, 0x7A, 0x77, 0x72, 0x75, 0x7E, 0x8A, // + 0x54, 0x58, 0x58, 0x5E, 0x5E, 0x6C, 0x79, 0x87, // + 0x4D, 0x54, 0x56, 0x5B, 0x59, 0x65, 0x6E, 0x7A, // + 0x4A, 0x4D, 0x4F, 0x53, 0x56, 0x5F, 0x67, 0x6E, // + 0x4A, 0x4D, 0x54, 0x58, 0x5B, 0x58, 0x56, 0x54, // + 0x4C, 0x4C, 0x52, 0x4F, 0x4D, 0x40, 0x3A, 0x35, // + }; + + wuffs_jpeg__decoder dec; + CHECK_STATUS("initialize", wuffs_jpeg__decoder__initialize( + &dec, sizeof dec, WUFFS_VERSION, + WUFFS_INITIALIZE__DEFAULT_OPTIONS)); + + const uint32_t b = 0; + memcpy(&dec.private_data.f_mcu_blocks[b], mcu_block, sizeof(mcu_block)); + + const uint32_t q = 0; + memcpy(&dec.private_impl.f_quant_tables[q], quant_table, sizeof(quant_table)); + + uint8_t dst_array[64] = {0}; + wuffs_jpeg__decoder__decode_idct( + &dec, wuffs_base__make_slice_u8(&dst_array[0], 64), 8, b, q); + + wuffs_base__io_buffer have = + wuffs_base__ptr_u8__reader(&dst_array[0], 64, true); + wuffs_base__io_buffer want = + wuffs_base__ptr_u8__reader((void*)(&want_array[0]), 64, true); + + return check_io_buffers_equal("", &have, &want); +} + const char* // test_wuffs_jpeg_decode_mcu() { CHECK_FOCUS(__func__); @@ -705,6 +771,7 @@ proc g_tests[] = { test_wuffs_jpeg_decode_dht_easy, test_wuffs_jpeg_decode_dht_hard, + test_wuffs_jpeg_decode_idct, test_wuffs_jpeg_decode_mcu, test_wuffs_jpeg_decode_interface, test_wuffs_jpeg_decode_truncated_input,