b181868aba
x86/h26x/h2656dsp: add missing preprocessor wrappers
...
Fixes compilation on x86_32 targets.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
6b6eb7d74e
x86/Makefile: fix hevc and vvc dependency of h2656dsp.o
...
And remove tabs while at it.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
2dc8221e66
x86/hevcdsp_init.c: fix preprocessor check
...
HAVE_AVX2_EXTERNAL has a value, so check for it.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
78a7927df7
x86/vvc/vvc_mc: wrap the entire file in x86_64 and AVX2 checks
...
Fixes compilation with old yasm.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
bf62ddc7bf
x86/vvc/vvc_mc: set the correct number of used registers in vvc_w_avg functions
...
Fixes crashes when running fate-vvc-conformance-WP_A_3 on Win64 targets
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
326cc01de9
avcodec/x86/vvc: add avg and avg_w AVX2 optimizations
...
The avg/avg_w is based on dav1d.
See https://code.videolan.org/videolan/dav1d/-/blob/master/src/x86/mc_avx2.asm
vvc_avg_8_2x2_c: 71.6
vvc_avg_8_2x2_avx2: 26.8
vvc_avg_8_2x4_c: 140.8
vvc_avg_8_2x4_avx2: 34.6
vvc_avg_8_2x8_c: 410.3
vvc_avg_8_2x8_avx2: 41.3
vvc_avg_8_2x16_c: 769.3
vvc_avg_8_2x16_avx2: 60.3
vvc_avg_8_2x32_c: 1669.6
vvc_avg_8_2x32_avx2: 105.1
vvc_avg_8_2x64_c: 1978.3
vvc_avg_8_2x64_avx2: 425.8
vvc_avg_8_2x128_c: 6536.8
vvc_avg_8_2x128_avx2: 1315.1
vvc_avg_8_4x2_c: 155.6
vvc_avg_8_4x2_avx2: 26.1
vvc_avg_8_4x4_c: 250.3
vvc_avg_8_4x4_avx2: 31.3
vvc_avg_8_4x8_c: 831.8
vvc_avg_8_4x8_avx2: 41.3
vvc_avg_8_4x16_c: 1461.1
vvc_avg_8_4x16_avx2: 57.1
vvc_avg_8_4x32_c: 2821.6
vvc_avg_8_4x32_avx2: 105.1
vvc_avg_8_4x64_c: 3615.8
vvc_avg_8_4x64_avx2: 412.6
vvc_avg_8_4x128_c: 11962.6
vvc_avg_8_4x128_avx2: 1274.3
vvc_avg_8_8x2_c: 215.8
vvc_avg_8_8x2_avx2: 29.1
vvc_avg_8_8x4_c: 430.6
vvc_avg_8_8x4_avx2: 37.6
vvc_avg_8_8x8_c: 1463.3
vvc_avg_8_8x8_avx2: 51.8
vvc_avg_8_8x16_c: 2630.1
vvc_avg_8_8x16_avx2: 97.6
vvc_avg_8_8x32_c: 5813.8
vvc_avg_8_8x32_avx2: 196.6
vvc_avg_8_8x64_c: 6687.3
vvc_avg_8_8x64_avx2: 487.8
vvc_avg_8_8x128_c: 13178.6
vvc_avg_8_8x128_avx2: 1290.6
vvc_avg_8_16x2_c: 443.8
vvc_avg_8_16x2_avx2: 28.3
vvc_avg_8_16x4_c: 1253.3
vvc_avg_8_16x4_avx2: 32.1
vvc_avg_8_16x8_c: 2236.3
vvc_avg_8_16x8_avx2: 44.3
vvc_avg_8_16x16_c: 5127.8
vvc_avg_8_16x16_avx2: 63.3
vvc_avg_8_16x32_c: 6573.3
vvc_avg_8_16x32_avx2: 223.6
vvc_avg_8_16x64_c: 30311.8
vvc_avg_8_16x64_avx2: 437.8
vvc_avg_8_16x128_c: 25693.3
vvc_avg_8_16x128_avx2: 1266.8
vvc_avg_8_32x2_c: 954.6
vvc_avg_8_32x2_avx2: 32.1
vvc_avg_8_32x4_c: 2359.6
vvc_avg_8_32x4_avx2: 39.6
vvc_avg_8_32x8_c: 5703.6
vvc_avg_8_32x8_avx2: 57.1
vvc_avg_8_32x16_c: 9967.6
vvc_avg_8_32x16_avx2: 107.1
vvc_avg_8_32x32_c: 21327.6
vvc_avg_8_32x32_avx2: 272.6
vvc_avg_8_32x64_c: 39240.8
vvc_avg_8_32x64_avx2: 529.6
vvc_avg_8_32x128_c: 52580.8
vvc_avg_8_32x128_avx2: 1338.8
vvc_avg_8_64x2_c: 1647.3
vvc_avg_8_64x2_avx2: 38.8
vvc_avg_8_64x4_c: 5130.1
vvc_avg_8_64x4_avx2: 58.8
vvc_avg_8_64x8_c: 6529.3
vvc_avg_8_64x8_avx2: 88.3
vvc_avg_8_64x16_c: 19913.6
vvc_avg_8_64x16_avx2: 162.3
vvc_avg_8_64x32_c: 39360.8
vvc_avg_8_64x32_avx2: 295.8
vvc_avg_8_64x64_c: 49658.3
vvc_avg_8_64x64_avx2: 784.1
vvc_avg_8_64x128_c: 108513.1
vvc_avg_8_64x128_avx2: 1977.1
vvc_avg_8_128x2_c: 3226.1
vvc_avg_8_128x2_avx2: 61.1
vvc_avg_8_128x4_c: 10280.3
vvc_avg_8_128x4_avx2: 94.6
vvc_avg_8_128x8_c: 18079.3
vvc_avg_8_128x8_avx2: 155.3
vvc_avg_8_128x16_c: 45121.8
vvc_avg_8_128x16_avx2: 285.3
vvc_avg_8_128x32_c: 48651.8
vvc_avg_8_128x32_avx2: 581.6
vvc_avg_8_128x64_c: 165078.6
vvc_avg_8_128x64_avx2: 1942.8
vvc_avg_8_128x128_c: 339103.1
vvc_avg_8_128x128_avx2: 4332.6
vvc_avg_10_2x2_c: 144.3
vvc_avg_10_2x2_avx2: 26.8
vvc_avg_10_2x4_c: 142.6
vvc_avg_10_2x4_avx2: 45.3
vvc_avg_10_2x8_c: 478.1
vvc_avg_10_2x8_avx2: 38.1
vvc_avg_10_2x16_c: 518.3
vvc_avg_10_2x16_avx2: 58.1
vvc_avg_10_2x32_c: 2059.8
vvc_avg_10_2x32_avx2: 93.1
vvc_avg_10_2x64_c: 2383.8
vvc_avg_10_2x64_avx2: 714.8
vvc_avg_10_2x128_c: 4498.3
vvc_avg_10_2x128_avx2: 1466.3
vvc_avg_10_4x2_c: 228.6
vvc_avg_10_4x2_avx2: 26.8
vvc_avg_10_4x4_c: 378.3
vvc_avg_10_4x4_avx2: 30.6
vvc_avg_10_4x8_c: 866.8
vvc_avg_10_4x8_avx2: 44.6
vvc_avg_10_4x16_c: 1018.1
vvc_avg_10_4x16_avx2: 58.1
vvc_avg_10_4x32_c: 3590.8
vvc_avg_10_4x32_avx2: 128.8
vvc_avg_10_4x64_c: 4200.8
vvc_avg_10_4x64_avx2: 663.6
vvc_avg_10_4x128_c: 8450.8
vvc_avg_10_4x128_avx2: 1531.8
vvc_avg_10_8x2_c: 369.3
vvc_avg_10_8x2_avx2: 28.3
vvc_avg_10_8x4_c: 513.8
vvc_avg_10_8x4_avx2: 32.1
vvc_avg_10_8x8_c: 1720.3
vvc_avg_10_8x8_avx2: 49.1
vvc_avg_10_8x16_c: 1894.8
vvc_avg_10_8x16_avx2: 71.6
vvc_avg_10_8x32_c: 3931.3
vvc_avg_10_8x32_avx2: 148.1
vvc_avg_10_8x64_c: 7964.3
vvc_avg_10_8x64_avx2: 613.1
vvc_avg_10_8x128_c: 15540.1
vvc_avg_10_8x128_avx2: 1585.1
vvc_avg_10_16x2_c: 877.3
vvc_avg_10_16x2_avx2: 27.6
vvc_avg_10_16x4_c: 955.8
vvc_avg_10_16x4_avx2: 29.8
vvc_avg_10_16x8_c: 3419.6
vvc_avg_10_16x8_avx2: 62.6
vvc_avg_10_16x16_c: 3826.8
vvc_avg_10_16x16_avx2: 54.3
vvc_avg_10_16x32_c: 7655.3
vvc_avg_10_16x32_avx2: 86.3
vvc_avg_10_16x64_c: 30011.1
vvc_avg_10_16x64_avx2: 692.6
vvc_avg_10_16x128_c: 47894.8
vvc_avg_10_16x128_avx2: 1580.3
vvc_avg_10_32x2_c: 944.3
vvc_avg_10_32x2_avx2: 29.8
vvc_avg_10_32x4_c: 2022.6
vvc_avg_10_32x4_avx2: 35.1
vvc_avg_10_32x8_c: 6148.8
vvc_avg_10_32x8_avx2: 51.3
vvc_avg_10_32x16_c: 12601.6
vvc_avg_10_32x16_avx2: 70.8
vvc_avg_10_32x32_c: 15958.6
vvc_avg_10_32x32_avx2: 124.3
vvc_avg_10_32x64_c: 31784.6
vvc_avg_10_32x64_avx2: 757.3
vvc_avg_10_32x128_c: 63892.8
vvc_avg_10_32x128_avx2: 1711.3
vvc_avg_10_64x2_c: 1890.8
vvc_avg_10_64x2_avx2: 34.3
vvc_avg_10_64x4_c: 6267.3
vvc_avg_10_64x4_avx2: 42.6
vvc_avg_10_64x8_c: 12778.1
vvc_avg_10_64x8_avx2: 67.8
vvc_avg_10_64x16_c: 22304.3
vvc_avg_10_64x16_avx2: 116.8
vvc_avg_10_64x32_c: 30777.1
vvc_avg_10_64x32_avx2: 201.1
vvc_avg_10_64x64_c: 60169.1
vvc_avg_10_64x64_avx2: 1454.3
vvc_avg_10_64x128_c: 124392.8
vvc_avg_10_64x128_avx2: 3648.6
vvc_avg_10_128x2_c: 3650.1
vvc_avg_10_128x2_avx2: 41.1
vvc_avg_10_128x4_c: 22887.8
vvc_avg_10_128x4_avx2: 64.1
vvc_avg_10_128x8_c: 14622.6
vvc_avg_10_128x8_avx2: 111.6
vvc_avg_10_128x16_c: 62207.6
vvc_avg_10_128x16_avx2: 186.3
vvc_avg_10_128x32_c: 59761.3
vvc_avg_10_128x32_avx2: 374.6
vvc_avg_10_128x64_c: 117504.3
vvc_avg_10_128x64_avx2: 2684.6
vvc_avg_10_128x128_c: 236767.6
vvc_avg_10_128x128_avx2: 15278.1
vvc_avg_12_2x2_c: 78.6
vvc_avg_12_2x2_avx2: 26.1
vvc_avg_12_2x4_c: 254.1
vvc_avg_12_2x4_avx2: 30.6
vvc_avg_12_2x8_c: 261.8
vvc_avg_12_2x8_avx2: 39.1
vvc_avg_12_2x16_c: 527.6
vvc_avg_12_2x16_avx2: 57.3
vvc_avg_12_2x32_c: 1089.1
vvc_avg_12_2x32_avx2: 93.8
vvc_avg_12_2x64_c: 2337.6
vvc_avg_12_2x64_avx2: 707.1
vvc_avg_12_2x128_c: 4582.1
vvc_avg_12_2x128_avx2: 1414.6
vvc_avg_12_4x2_c: 129.6
vvc_avg_12_4x2_avx2: 26.8
vvc_avg_12_4x4_c: 427.3
vvc_avg_12_4x4_avx2: 30.6
vvc_avg_12_4x8_c: 529.6
vvc_avg_12_4x8_avx2: 36.6
vvc_avg_12_4x16_c: 1022.1
vvc_avg_12_4x16_avx2: 57.3
vvc_avg_12_4x32_c: 1987.6
vvc_avg_12_4x32_avx2: 84.3
vvc_avg_12_4x64_c: 4147.6
vvc_avg_12_4x64_avx2: 706.3
vvc_avg_12_4x128_c: 8469.3
vvc_avg_12_4x128_avx2: 1448.3
vvc_avg_12_8x2_c: 253.6
vvc_avg_12_8x2_avx2: 27.6
vvc_avg_12_8x4_c: 836.3
vvc_avg_12_8x4_avx2: 32.1
vvc_avg_12_8x8_c: 1074.6
vvc_avg_12_8x8_avx2: 45.1
vvc_avg_12_8x16_c: 3616.8
vvc_avg_12_8x16_avx2: 71.6
vvc_avg_12_8x32_c: 3823.6
vvc_avg_12_8x32_avx2: 140.1
vvc_avg_12_8x64_c: 7764.8
vvc_avg_12_8x64_avx2: 656.1
vvc_avg_12_8x128_c: 15896.1
vvc_avg_12_8x128_avx2: 1232.8
vvc_avg_12_16x2_c: 462.1
vvc_avg_12_16x2_avx2: 26.8
vvc_avg_12_16x4_c: 1732.1
vvc_avg_12_16x4_avx2: 29.1
vvc_avg_12_16x8_c: 2097.6
vvc_avg_12_16x8_avx2: 62.6
vvc_avg_12_16x16_c: 6753.1
vvc_avg_12_16x16_avx2: 47.8
vvc_avg_12_16x32_c: 7373.1
vvc_avg_12_16x32_avx2: 80.8
vvc_avg_12_16x64_c: 15046.3
vvc_avg_12_16x64_avx2: 621.1
vvc_avg_12_16x128_c: 52574.6
vvc_avg_12_16x128_avx2: 1417.1
vvc_avg_12_32x2_c: 1712.1
vvc_avg_12_32x2_avx2: 29.8
vvc_avg_12_32x4_c: 2036.8
vvc_avg_12_32x4_avx2: 37.6
vvc_avg_12_32x8_c: 4017.6
vvc_avg_12_32x8_avx2: 44.1
vvc_avg_12_32x16_c: 8018.6
vvc_avg_12_32x16_avx2: 70.8
vvc_avg_12_32x32_c: 15637.6
vvc_avg_12_32x32_avx2: 124.3
vvc_avg_12_32x64_c: 31143.3
vvc_avg_12_32x64_avx2: 830.3
vvc_avg_12_32x128_c: 75706.8
vvc_avg_12_32x128_avx2: 1604.8
vvc_avg_12_64x2_c: 3230.3
vvc_avg_12_64x2_avx2: 33.6
vvc_avg_12_64x4_c: 4139.6
vvc_avg_12_64x4_avx2: 45.1
vvc_avg_12_64x8_c: 8201.6
vvc_avg_12_64x8_avx2: 67.1
vvc_avg_12_64x16_c: 25632.3
vvc_avg_12_64x16_avx2: 110.3
vvc_avg_12_64x32_c: 30744.3
vvc_avg_12_64x32_avx2: 200.3
vvc_avg_12_64x64_c: 105554.8
vvc_avg_12_64x64_avx2: 1325.6
vvc_avg_12_64x128_c: 235254.3
vvc_avg_12_64x128_avx2: 3132.6
vvc_avg_12_128x2_c: 6194.3
vvc_avg_12_128x2_avx2: 55.1
vvc_avg_12_128x4_c: 7583.8
vvc_avg_12_128x4_avx2: 79.3
vvc_avg_12_128x8_c: 14635.6
vvc_avg_12_128x8_avx2: 104.3
vvc_avg_12_128x16_c: 29270.8
vvc_avg_12_128x16_avx2: 194.3
vvc_avg_12_128x32_c: 60113.6
vvc_avg_12_128x32_avx2: 346.3
vvc_avg_12_128x64_c: 197030.3
vvc_avg_12_128x64_avx2: 2779.6
vvc_avg_12_128x128_c: 432809.6
vvc_avg_12_128x128_avx2: 5513.3
vvc_w_avg_8_2x2_c: 84.3
vvc_w_avg_8_2x2_avx2: 42.6
vvc_w_avg_8_2x4_c: 156.3
vvc_w_avg_8_2x4_avx2: 58.8
vvc_w_avg_8_2x8_c: 310.6
vvc_w_avg_8_2x8_avx2: 73.1
vvc_w_avg_8_2x16_c: 942.1
vvc_w_avg_8_2x16_avx2: 113.3
vvc_w_avg_8_2x32_c: 1098.8
vvc_w_avg_8_2x32_avx2: 202.6
vvc_w_avg_8_2x64_c: 2414.3
vvc_w_avg_8_2x64_avx2: 467.6
vvc_w_avg_8_2x128_c: 4763.8
vvc_w_avg_8_2x128_avx2: 1333.1
vvc_w_avg_8_4x2_c: 140.1
vvc_w_avg_8_4x2_avx2: 49.8
vvc_w_avg_8_4x4_c: 276.3
vvc_w_avg_8_4x4_avx2: 58.1
vvc_w_avg_8_4x8_c: 524.3
vvc_w_avg_8_4x8_avx2: 72.3
vvc_w_avg_8_4x16_c: 1108.1
vvc_w_avg_8_4x16_avx2: 111.8
vvc_w_avg_8_4x32_c: 2149.8
vvc_w_avg_8_4x32_avx2: 199.6
vvc_w_avg_8_4x64_c: 12288.1
vvc_w_avg_8_4x64_avx2: 509.3
vvc_w_avg_8_4x128_c: 8398.6
vvc_w_avg_8_4x128_avx2: 1319.6
vvc_w_avg_8_8x2_c: 271.1
vvc_w_avg_8_8x2_avx2: 44.1
vvc_w_avg_8_8x4_c: 503.3
vvc_w_avg_8_8x4_avx2: 61.8
vvc_w_avg_8_8x8_c: 1031.1
vvc_w_avg_8_8x8_avx2: 93.8
vvc_w_avg_8_8x16_c: 2009.8
vvc_w_avg_8_8x16_avx2: 163.1
vvc_w_avg_8_8x32_c: 4161.3
vvc_w_avg_8_8x32_avx2: 292.1
vvc_w_avg_8_8x64_c: 7940.6
vvc_w_avg_8_8x64_avx2: 592.1
vvc_w_avg_8_8x128_c: 16802.3
vvc_w_avg_8_8x128_avx2: 1287.6
vvc_w_avg_8_16x2_c: 762.6
vvc_w_avg_8_16x2_avx2: 53.6
vvc_w_avg_8_16x4_c: 1486.3
vvc_w_avg_8_16x4_avx2: 67.1
vvc_w_avg_8_16x8_c: 1907.8
vvc_w_avg_8_16x8_avx2: 96.8
vvc_w_avg_8_16x16_c: 3883.6
vvc_w_avg_8_16x16_avx2: 151.3
vvc_w_avg_8_16x32_c: 7974.8
vvc_w_avg_8_16x32_avx2: 285.8
vvc_w_avg_8_16x64_c: 25160.6
vvc_w_avg_8_16x64_avx2: 589.8
vvc_w_avg_8_16x128_c: 58328.1
vvc_w_avg_8_16x128_avx2: 1169.8
vvc_w_avg_8_32x2_c: 1009.1
vvc_w_avg_8_32x2_avx2: 65.6
vvc_w_avg_8_32x4_c: 2091.1
vvc_w_avg_8_32x4_avx2: 96.8
vvc_w_avg_8_32x8_c: 3997.8
vvc_w_avg_8_32x8_avx2: 156.3
vvc_w_avg_8_32x16_c: 8216.8
vvc_w_avg_8_32x16_avx2: 269.6
vvc_w_avg_8_32x32_c: 21746.1
vvc_w_avg_8_32x32_avx2: 635.3
vvc_w_avg_8_32x64_c: 31564.8
vvc_w_avg_8_32x64_avx2: 1010.6
vvc_w_avg_8_32x128_c: 114373.3
vvc_w_avg_8_32x128_avx2: 2013.6
vvc_w_avg_8_64x2_c: 2067.3
vvc_w_avg_8_64x2_avx2: 97.6
vvc_w_avg_8_64x4_c: 3901.1
vvc_w_avg_8_64x4_avx2: 154.8
vvc_w_avg_8_64x8_c: 7911.6
vvc_w_avg_8_64x8_avx2: 268.8
vvc_w_avg_8_64x16_c: 16508.8
vvc_w_avg_8_64x16_avx2: 501.8
vvc_w_avg_8_64x32_c: 38770.3
vvc_w_avg_8_64x32_avx2: 1287.6
vvc_w_avg_8_64x64_c: 110350.6
vvc_w_avg_8_64x64_avx2: 1890.8
vvc_w_avg_8_64x128_c: 141354.6
vvc_w_avg_8_64x128_avx2: 3839.6
vvc_w_avg_8_128x2_c: 7012.1
vvc_w_avg_8_128x2_avx2: 159.3
vvc_w_avg_8_128x4_c: 8146.8
vvc_w_avg_8_128x4_avx2: 272.6
vvc_w_avg_8_128x8_c: 24596.8
vvc_w_avg_8_128x8_avx2: 501.1
vvc_w_avg_8_128x16_c: 35918.1
vvc_w_avg_8_128x16_avx2: 948.8
vvc_w_avg_8_128x32_c: 68799.6
vvc_w_avg_8_128x32_avx2: 1963.1
vvc_w_avg_8_128x64_c: 133862.1
vvc_w_avg_8_128x64_avx2: 3833.6
vvc_w_avg_8_128x128_c: 348427.8
vvc_w_avg_8_128x128_avx2: 7682.8
vvc_w_avg_10_2x2_c: 118.6
vvc_w_avg_10_2x2_avx2: 73.1
vvc_w_avg_10_2x4_c: 189.1
vvc_w_avg_10_2x4_avx2: 89.3
vvc_w_avg_10_2x8_c: 382.8
vvc_w_avg_10_2x8_avx2: 179.8
vvc_w_avg_10_2x16_c: 658.3
vvc_w_avg_10_2x16_avx2: 185.1
vvc_w_avg_10_2x32_c: 1409.3
vvc_w_avg_10_2x32_avx2: 290.8
vvc_w_avg_10_2x64_c: 2906.8
vvc_w_avg_10_2x64_avx2: 793.1
vvc_w_avg_10_2x128_c: 6292.6
vvc_w_avg_10_2x128_avx2: 1696.8
vvc_w_avg_10_4x2_c: 178.8
vvc_w_avg_10_4x2_avx2: 80.1
vvc_w_avg_10_4x4_c: 581.6
vvc_w_avg_10_4x4_avx2: 97.6
vvc_w_avg_10_4x8_c: 693.3
vvc_w_avg_10_4x8_avx2: 128.1
vvc_w_avg_10_4x16_c: 1436.6
vvc_w_avg_10_4x16_avx2: 179.8
vvc_w_avg_10_4x32_c: 2409.1
vvc_w_avg_10_4x32_avx2: 292.3
vvc_w_avg_10_4x64_c: 4925.3
vvc_w_avg_10_4x64_avx2: 746.1
vvc_w_avg_10_4x128_c: 10664.6
vvc_w_avg_10_4x128_avx2: 1647.6
vvc_w_avg_10_8x2_c: 359.3
vvc_w_avg_10_8x2_avx2: 80.1
vvc_w_avg_10_8x4_c: 925.6
vvc_w_avg_10_8x4_avx2: 97.6
vvc_w_avg_10_8x8_c: 1360.6
vvc_w_avg_10_8x8_avx2: 121.8
vvc_w_avg_10_8x16_c: 3490.3
vvc_w_avg_10_8x16_avx2: 203.3
vvc_w_avg_10_8x32_c: 5266.1
vvc_w_avg_10_8x32_avx2: 325.8
vvc_w_avg_10_8x64_c: 11127.1
vvc_w_avg_10_8x64_avx2: 747.8
vvc_w_avg_10_8x128_c: 31058.3
vvc_w_avg_10_8x128_avx2: 1424.6
vvc_w_avg_10_16x2_c: 624.8
vvc_w_avg_10_16x2_avx2: 84.6
vvc_w_avg_10_16x4_c: 1389.6
vvc_w_avg_10_16x4_avx2: 109.1
vvc_w_avg_10_16x8_c: 2688.3
vvc_w_avg_10_16x8_avx2: 137.1
vvc_w_avg_10_16x16_c: 5387.1
vvc_w_avg_10_16x16_avx2: 224.6
vvc_w_avg_10_16x32_c: 10776.3
vvc_w_avg_10_16x32_avx2: 312.1
vvc_w_avg_10_16x64_c: 18069.1
vvc_w_avg_10_16x64_avx2: 858.6
vvc_w_avg_10_16x128_c: 43460.3
vvc_w_avg_10_16x128_avx2: 1411.6
vvc_w_avg_10_32x2_c: 1232.8
vvc_w_avg_10_32x2_avx2: 99.1
vvc_w_avg_10_32x4_c: 4017.6
vvc_w_avg_10_32x4_avx2: 134.1
vvc_w_avg_10_32x8_c: 9306.3
vvc_w_avg_10_32x8_avx2: 208.1
vvc_w_avg_10_32x16_c: 8424.6
vvc_w_avg_10_32x16_avx2: 349.3
vvc_w_avg_10_32x32_c: 20787.8
vvc_w_avg_10_32x32_avx2: 655.3
vvc_w_avg_10_32x64_c: 40972.1
vvc_w_avg_10_32x64_avx2: 904.8
vvc_w_avg_10_32x128_c: 85670.3
vvc_w_avg_10_32x128_avx2: 1751.6
vvc_w_avg_10_64x2_c: 2454.1
vvc_w_avg_10_64x2_avx2: 132.6
vvc_w_avg_10_64x4_c: 5012.6
vvc_w_avg_10_64x4_avx2: 215.6
vvc_w_avg_10_64x8_c: 10811.3
vvc_w_avg_10_64x8_avx2: 361.1
vvc_w_avg_10_64x16_c: 33349.1
vvc_w_avg_10_64x16_avx2: 904.1
vvc_w_avg_10_64x32_c: 41892.3
vvc_w_avg_10_64x32_avx2: 1220.6
vvc_w_avg_10_64x64_c: 66983.3
vvc_w_avg_10_64x64_avx2: 2622.1
vvc_w_avg_10_64x128_c: 246508.8
vvc_w_avg_10_64x128_avx2: 3316.8
vvc_w_avg_10_128x2_c: 7791.6
vvc_w_avg_10_128x2_avx2: 198.8
vvc_w_avg_10_128x4_c: 10534.3
vvc_w_avg_10_128x4_avx2: 337.3
vvc_w_avg_10_128x8_c: 21142.3
vvc_w_avg_10_128x8_avx2: 614.8
vvc_w_avg_10_128x16_c: 40968.6
vvc_w_avg_10_128x16_avx2: 1160.6
vvc_w_avg_10_128x32_c: 113043.3
vvc_w_avg_10_128x32_avx2: 1644.6
vvc_w_avg_10_128x64_c: 230658.3
vvc_w_avg_10_128x64_avx2: 5065.3
vvc_w_avg_10_128x128_c: 335236.3
vvc_w_avg_10_128x128_avx2: 6450.3
vvc_w_avg_12_2x2_c: 185.3
vvc_w_avg_12_2x2_avx2: 43.6
vvc_w_avg_12_2x4_c: 340.3
vvc_w_avg_12_2x4_avx2: 55.8
vvc_w_avg_12_2x8_c: 632.3
vvc_w_avg_12_2x8_avx2: 70.1
vvc_w_avg_12_2x16_c: 728.3
vvc_w_avg_12_2x16_avx2: 108.1
vvc_w_avg_12_2x32_c: 1392.6
vvc_w_avg_12_2x32_avx2: 176.8
vvc_w_avg_12_2x64_c: 2618.3
vvc_w_avg_12_2x64_avx2: 757.3
vvc_w_avg_12_2x128_c: 6408.8
vvc_w_avg_12_2x128_avx2: 1435.1
vvc_w_avg_12_4x2_c: 349.3
vvc_w_avg_12_4x2_avx2: 44.3
vvc_w_avg_12_4x4_c: 607.1
vvc_w_avg_12_4x4_avx2: 52.6
vvc_w_avg_12_4x8_c: 1134.8
vvc_w_avg_12_4x8_avx2: 70.1
vvc_w_avg_12_4x16_c: 1378.1
vvc_w_avg_12_4x16_avx2: 115.3
vvc_w_avg_12_4x32_c: 2599.3
vvc_w_avg_12_4x32_avx2: 174.3
vvc_w_avg_12_4x64_c: 4474.8
vvc_w_avg_12_4x64_avx2: 656.1
vvc_w_avg_12_4x128_c: 11319.6
vvc_w_avg_12_4x128_avx2: 1373.1
vvc_w_avg_12_8x2_c: 595.8
vvc_w_avg_12_8x2_avx2: 44.3
vvc_w_avg_12_8x4_c: 1164.3
vvc_w_avg_12_8x4_avx2: 56.6
vvc_w_avg_12_8x8_c: 2019.6
vvc_w_avg_12_8x8_avx2: 80.1
vvc_w_avg_12_8x16_c: 4071.6
vvc_w_avg_12_8x16_avx2: 139.3
vvc_w_avg_12_8x32_c: 4485.1
vvc_w_avg_12_8x32_avx2: 250.6
vvc_w_avg_12_8x64_c: 8404.8
vvc_w_avg_12_8x64_avx2: 735.8
vvc_w_avg_12_8x128_c: 35679.8
vvc_w_avg_12_8x128_avx2: 1252.6
vvc_w_avg_12_16x2_c: 1114.8
vvc_w_avg_12_16x2_avx2: 46.6
vvc_w_avg_12_16x4_c: 2240.1
vvc_w_avg_12_16x4_avx2: 62.6
vvc_w_avg_12_16x8_c: 13174.6
vvc_w_avg_12_16x8_avx2: 88.6
vvc_w_avg_12_16x16_c: 5334.6
vvc_w_avg_12_16x16_avx2: 144.3
vvc_w_avg_12_16x32_c: 8378.1
vvc_w_avg_12_16x32_avx2: 234.6
vvc_w_avg_12_16x64_c: 21300.8
vvc_w_avg_12_16x64_avx2: 761.8
vvc_w_avg_12_16x128_c: 32786.8
vvc_w_avg_12_16x128_avx2: 1432.8
vvc_w_avg_12_32x2_c: 2154.3
vvc_w_avg_12_32x2_avx2: 61.1
vvc_w_avg_12_32x4_c: 4299.8
vvc_w_avg_12_32x4_avx2: 83.1
vvc_w_avg_12_32x8_c: 7964.8
vvc_w_avg_12_32x8_avx2: 132.6
vvc_w_avg_12_32x16_c: 13321.6
vvc_w_avg_12_32x16_avx2: 234.6
vvc_w_avg_12_32x32_c: 21149.3
vvc_w_avg_12_32x32_avx2: 433.3
vvc_w_avg_12_32x64_c: 43666.6
vvc_w_avg_12_32x64_avx2: 876.6
vvc_w_avg_12_32x128_c: 83189.8
vvc_w_avg_12_32x128_avx2: 1756.6
vvc_w_avg_12_64x2_c: 3829.8
vvc_w_avg_12_64x2_avx2: 83.1
vvc_w_avg_12_64x4_c: 8588.1
vvc_w_avg_12_64x4_avx2: 127.1
vvc_w_avg_12_64x8_c: 17027.6
vvc_w_avg_12_64x8_avx2: 310.6
vvc_w_avg_12_64x16_c: 29797.8
vvc_w_avg_12_64x16_avx2: 415.6
vvc_w_avg_12_64x32_c: 43854.3
vvc_w_avg_12_64x32_avx2: 773.3
vvc_w_avg_12_64x64_c: 137767.3
vvc_w_avg_12_64x64_avx2: 1608.6
vvc_w_avg_12_64x128_c: 316428.3
vvc_w_avg_12_64x128_avx2: 3249.8
vvc_w_avg_12_128x2_c: 8824.6
vvc_w_avg_12_128x2_avx2: 130.3
vvc_w_avg_12_128x4_c: 17173.6
vvc_w_avg_12_128x4_avx2: 219.3
vvc_w_avg_12_128x8_c: 21997.8
vvc_w_avg_12_128x8_avx2: 397.3
vvc_w_avg_12_128x16_c: 43553.8
vvc_w_avg_12_128x16_avx2: 790.1
vvc_w_avg_12_128x32_c: 89792.1
vvc_w_avg_12_128x32_avx2: 1497.6
vvc_w_avg_12_128x64_c: 226573.3
vvc_w_avg_12_128x64_avx2: 3153.1
vvc_w_avg_12_128x128_c: 332090.1
vvc_w_avg_12_128x128_avx2: 6499.6
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
1 year ago
70889620f2
avcodec/vvcdec: reuse h26x/2656_inter.asm to enable x86 optimizations
...
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
1 year ago
fc5ff6b0b8
avcodec/x86/h26x/h2656_inter: add dststride to put
...
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
1 year ago
7d9f1f5485
avcodec/x86/hevc_mc: move put/put_uni to h26x/h2656_inter.asm
...
This enable that the asm optimization can be reused by VVC
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
1 year ago
04c2e246a3
avcodec/hevcdsp_template: reuse put/put_luma/put_chroma from h2656_inter_template
...
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
1 year ago
639b1820ce
avcodec/vvc/vvc_inter_template: move put/put_luma/put_chroma template to h2656_inter_template.c
...
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
1 year ago
fa469545ba
avcodec: move leb reading functions to its own header
...
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
7252e4f8ee
avcodec/aac_defines: Remove unused AAC_RENAME_32
...
Unused since fbe6a51b11
1 year ago
36a986d9a1
lavc/vvc: Add check to num_multi_layer_olss
...
Check that vps_each_layer_is_an_ols_flag, which indicates that "at
least one OLS specified by the VPS contains more than one layer," is
set if num_multi_layer_olss is non-zero.
Fixes: 65160/clusterfuzz-testcase-minimized-ffmpeg_BSF_VVC_METADATA_fuzzer-4665241535119360
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Frank Plowman <post@frankplowman.com>
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
bc944168db
avcodec/mpeg4videodec: Remove write-only variable
...
Fix warning: variable 'time_incr' set but not used.
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
1 year ago
66f028accb
avcodec/cbs_h266: fix logic setting num_layers_in_ols when vps_ols_mode_idc is 2
...
The old code did not follow the syntax from the spec.
Reviewed-by: Frank Plowman <post@frankplowman.com>
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
85e031d5bf
lavc/vvc: Increase IntraEdgeParams buffer size
...
The reference line buffers are used with indices in the range
-MAX_TB_SIZE - 3 to refw + FFMAX(1, w/h) * ref_idx + 1, which is
at most 5*MAX_TB_SIZE + 1.
Fixes buffer overflows.
http://fate.ffmpeg.org/report.cgi?slot=armv7-linux-gcc-9&time=20240124051736
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
887a7817b6
lavc: move bitstream filters into bsf/ subdir
1 year ago
341d0419e1
avcodec/texturedsp: Factor common code out
...
Namely calling avctx->execute2(avctx,...).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
4f58372c10
avcodec/hap: Avoid unnecessary opt.h inclusion
...
It presumably exists because HapContext contains an AVClass*.
Yet AVClass is actually defined in log.h and even this inclusion
can be avoided by struct AVClass*. This avoids opt.h inclusions
in hap.c and hapdec.c.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
a3fc9fb9fb
avcodec/texturedsp: Add separate TextureDSPEncContext
...
ff_texturedspenc_init() doesn't support most of the function types
supported for decoding; add a separate context containing only pointers
for the actually supported types.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
8f791304f2
avcodec/dxvenc, hap(dec|enc): Move TextureDSPContext to stack
...
Only used during init.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
e1d1304b4b
avcodec/texturedspenc: Remove unused rgtc1_u_alpha encoding func
...
Effectively reverts 50a20de6b9
1 year ago
916f016741
avcodec/dxvenc: Fix data races with slice threading
...
The old code set a common struct from each thread;
this only "worked" (but is still UB) because the values
written are the same for each thread.
Fix this by moving the assignments to the main thread.
(This also avoids casting const away from a const AVFrame*.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
555879ca7c
avcodec/dxvenc: Don't cast const away
...
Reviewed-by: Connor Worley <connorbworley@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
0c517fcbe8
lavc/vvc: Fix emulation prevention byte handling
...
nal->skipped_bytes_pos contains the positions of errors relative to the
start of the slice header, whereas the position they were tested against
is relative to the start of the slice data, i.e. one byte after the end
of the slice header.
Patch fixes this by storing the size of the slice header in H266RawSlice
and adding it to the position given by the GetBitContext before
comparing to skipped_bytes_pos. This fixes AVERROR_INVALIDDATAs for
various valid bitstreams, such as the LMCS_B_Dolby_2 conformance
bitstream.
Signed-off-by: Frank Plowman <post@frankplowman.com>
1 year ago
eb4584f994
avcodec/vvc_ps: remove duplicated enum
...
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
763e31a8d3
lavc/vvc: Clamp shift RHS
...
Resolves the following undefined behavior sanitiser error:
runtime error: shift exponent 32 is too large for 32-bit type 'int'
Signed-off-by: Frank Plowman <post@frankplowman.com>
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
cb7b4ee024
lavc/vvc: Use av_log2 when destination is integer
...
Signed-off-by: Frank Plowman <post@frankplowman.com>
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
8c99a1429a
avcodec/d3d12va_mpeg2|vc1: remove the unused macros
...
These macros are no longer used. Remove them.
Signed-off-by: Tong Wu <tong1.wu@intel.com>
1 year ago
8b41e9cfbe
avcodec/d3d12va_decode: check existance before assigning a new index
...
Fixes #10759 .
It can happen in H.264, MPEG2, VC1 that the current frame resource
memory is already in ref_resource. For example, for a interlaced frame,
the same curr memory is passed twice. For the second time it could possibly
reference itself. When this happens the curr is already given an index and
in ref_resources. When the reference frame index is required, we should check
the existance in the ref_resources first before assigning a new index for it.
Signed-off-by: Tong Wu <tong1.wu@intel.com>
1 year ago
ac06190a5a
avcodec/libjxl.h: include version.h
...
This file has been exported since our minimum required version (0.7.0),
but it wasn't documented. Instead it was transitively included by
<jxl/decode.h> (but not jxl/encode.h), which ffmpeg relied on.
libjxl broke its API in libjxl/libjxl@66b959239355aef5255 by removing
the transitive include of version.h, and they do not plan on adding
it back. Instead they are choosing to leave the API backwards-
incompatible with downstream callers written for some fairly recent
versions of their API.
As a result, we include <jxl/version.h> to continue to build against
more recent versions of libjxl. The version macros removed are also
present in that file, so we no longer need to redefine them.
Signed-off-by: Leo Izen <leo.izen@gmail.com>
1 year ago
45a2f2635d
avcodec/d3d12va: remove unused variables
...
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
49a7fe86fe
avcodec/d3d12va_vc1: cast mapped_data to void* before passing it to mapped_data()
...
Fixes -Wincompatible-pointer-types warnings.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
9aa388e758
avcodec/d3d12va_hevc: cast mapped_data to void* before passing it to mapped_data()
...
Fixes -Wincompatible-pointer-types warnings.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
342cc1792f
avcodec/d3d12va_h264: cast mapped_data to void* before passing it to mapped_data()
...
Fixes -Wincompatible-pointer-types warnings.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
04332ca35e
avcodec/d3d12va_vp9.c: change the type for the ID3D12Resource_Map mapped_data argument
...
Fixes -Wincompatible-pointer-types warnings.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
b4d871fdc8
avcodec/d3d12va_av1.c: change the type for the ID3D12Resource_Map mapped_data argument
...
Fixes -Wincompatible-pointer-types warnings.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
cb6a488fba
avcodec/vvc_mvs: remove an unnecessary AV_ZERO64() call
...
Should fix "member access within misaligned address 0xf00 for type 'const union
av_alias64', which requires 8 byte alignment" errors as reported by GCC ubsan.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
bc1d8a9b76
avcodec/vvc_mvs: align local motion vector fields
...
Should fix "member access within misaligned address 0xf00 for type 'const union
av_alias64', which requires 8 byte alignment" errors as reported by GCC ubsan.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
3435565e26
avcodec/d3d12va_(av1|hevc|vp9): Don't use deprecated FF_PROFILE_*
...
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
dfbbd11a4b
lavc/dxvenc: add DXV encoder with support for DXT1 texture format
...
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
1 year ago
1496ce8f6b
avcodec/vvc_ctu: align motion vector fields
...
Should fix "member access within misaligned address 0xf00 for type 'const union
av_alias64', which requires 8 byte alignment" errors as reported by GCC ubsan.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
8157b5d405
lavc/vvc: Remove left shifts of negative values
...
VVC specifies << as arithmetic left shift, i.e. x << y is equivalent to
x * pow2(y). C's << on the other hand has UB if x is negative. This
patch removes all UB resulting from this, mostly by replacing x << y
with x * (1 << y), but there are also a couple places where the OOP was
changed instead.
Signed-off-by: Frank Plowman <post@frankplowman.com>
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
ab39cc36c7
avcodec/speexdec: fix setting frame_size from extradata
...
Finishes fixing vp5/potter512-400-partial.avi
The fate-matroska-ms-mode test ref is updated to reflect that the Speex decoder
can now read the stream.
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
cad35f0a77
avcodec/speexdec: relax the extradata check for the speex string
...
There could be bogus bytes at the start, as is the case of
vp5/potter512-400-partial.avi from the FATE suite, which could be a case of bad
remuxing from an OGG source.
Partially fixes decoding of vp5/potter512-400-partial.avi
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
08bebeb1be
Revert "all: Don't set AVClass.item_name to its default value"
...
Some callers assume that item_name is always set, so this may be
considered an API break.
This reverts commit 0c6203c97a
1 year ago
0a5813fc68
avcodec/vvcdec: allocate and store structs on their own within the table list
...
Fixes "runtime error: member access within misaligned address 0xf00 for type
'struct bar', which requires # byte alignment" errors under GCC ubsan.
Reviewed-by: Nuo Mi <nuomi2021@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
8e23ebe6f9
lavc/svq1enc: R-V V ssd_int8_vs_int16
...
C908
ssd_int8_vs_int16_c: 207.7
ssd_int8_vs_int16_rvv_i32: 14.2
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>
1 year ago
d595e0a0b6
avcodec/vvcdec: misc, constify hor_ctu_edge
...
Signed-off-by: James Almer <jamrial@gmail.com>
1 year ago
James Almer
Wu Jianhua
Andreas Rheinhardt
Frank Plowman
Zhao Zhili
Anton Khirnov
Tong Wu
Leo Izen
Connor Worley
sunyuechi
Nuo Mi