原载于：
http://vx.moe-sos.com/?action=show&id=20==================================================================================
这两天x264又开始发疯似的更新…………
今天的767版changelog里这么写到
skip intra pred+dct+quant in cases where it's redundant (analyse vs encode)
large speedup with trellis=2, small speedup with trellis=0 and/or subme>=6
到底这个large speedup是多么的large呢？
不禁激起我测试一下的欲望
测试方法如下
从DVD里截取一段1m11s的视频，然后Resize到864x480，并输出无压缩YV12的AVI，用AVS载入AVI，在CMD下送给x264压制。
我的CPU是Core2Duo E6550，超频到3.2G
首先用MEGUI自动更新的763 patched版。
命令行如下
--qp 18 --ref 3 --aq-strength 0.0 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 6 --trellis 1 --analyse all --8x8dct --me umh --threads 3 --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --output "test.mp4" "test.avs"
测试结果
avis [info]: 864x480 @ 23.98 fps (1724 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSE3 SSSE3 Cache64
mp4 [info]: initial delay 2002 (scale 24000)
x264 [info]: slice I:19Avg QP:15.00 size: 47242:00:00
x264 [info]: slice P:659 Avg QP:18.00 size: 9882
x264 [info]: slice B:1046 Avg QP:19.71 size: 1509
x264 [info]: mb I I16..4: 39.4% 23.1% 37.6%
x264 [info]: mb P I16..4: 8.1% 3.3% 3.0% P16..4: 47.2% 8.6% 5.1% 0.6% 0.3%skip:23.7%
x264 [info]: mb B I16..4: 0.1% 0.0% 0.1% B16..8: 20.7% 0.7% 1.3% direct: 0.9% skip:76.1%
x264 [info]: 8x8 transform intra:22.8% inter:59.3%
x264 [info]: ref P 73.0% 16.3% 10.7%
x264 [info]: ref B 88.9% 11.1%
x264 [info]: kb/s:1000.0
encoded 1724 frames, 18.04 fps, 1000.10 kb/s
再来测试x264.nl上的767版，参数一样。
测试结果
avis [info]: 864x480 @ 23.98 fps (1724 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSE3 SSSE3 Cache64
mp4 [info]: initial delay 2002 (scale 24000)
x264 [info]: slice I:19Avg QP:15.00 size: 47242:00:00
x264 [info]: slice P:660 Avg QP:18.00 size: 9868
x264 [info]: slice B:1045 Avg QP:19.71 size: 1511
x264 [info]: mb I I16..4: 39.4% 23.1% 37.6%
x264 [info]: mb P I16..4: 8.1% 3.3% 3.1% P16..4: 47.2% 8.6% 5.1% 0.6% 0.3%skip:23.7%
x264 [info]: mb B I16..4: 0.1% 0.0% 0.1% B16..8: 20.7% 0.7% 1.3% direct: 0.9% skip:76.1%
x264 [info]: 8x8 transform intra:22.8% inter:59.2%
x264 [info]: ref P 73.0% 16.3% 10.6%
x264 [info]: ref B 88.9% 11.1%
x264 [info]: kb/s:1000.1
encoded 1724 frames, 22.53 fps, 1000.24 kb/s
下面使用设置高一些的参数再压一次
--qp 18 --ref 3 --aq-strength 0.0 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 7 --trellis 2 --analyse all --8x8dct --me umh --threads 3 --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --output "test.mp4" "test.avs"
现在用MEGUI自动更新的763压制，结果如下
avis [info]: 864x480 @ 23.98 fps (1724 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSE3 SSSE3 Cache64
mp4 [info]: initial delay 2002 (scale 24000)
x264 [info]: slice I:19Avg QP:15.00 size: 45907:00:00
x264 [info]: slice P:659 Avg QP:18.00 size: 9229
x264 [info]: slice B:1046 Avg QP:19.71 size: 1420
x264 [info]: mb I I16..4: 34.9% 27.7% 37.4%
x264 [info]: mb P I16..4: 8.6% 3.1% 3.1% P16..4: 43.4% 8.6% 4.1% 0.5% 0.3%skip:28.1%
x264 [info]: mb B I16..4: 0.1% 0.0% 0.1% B16..8: 20.8% 0.7% 1.2% direct: 1.2% skip:75.8%
x264 [info]: 8x8 transform intra:22.1% inter:52.4%
x264 [info]: ref P 74.1% 15.7% 10.1%
x264 [info]: ref B 90.2% 9.8%
x264 [info]: kb/s:938.9
encoded 1724 frames, 13.33 fps, 939.02 kb/s
使用767压制，结果如下
avis [info]: 864x480 @ 23.98 fps (1724 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSE3 SSSE3 Cache64
mp4 [info]: initial delay 2002 (scale 24000)
x264 [info]: slice I:19Avg QP:15.00 size: 45907:00:00
x264 [info]: slice P:660 Avg QP:18.00 size: 9214
x264 [info]: slice B:1045 Avg QP:19.71 size: 1420
x264 [info]: mb I I16..4: 34.9% 27.7% 37.4%
x264 [info]: mb P I16..4: 8.5% 3.1% 3.1% P16..4: 43.4% 8.7% 4.1% 0.5% 0.3%skip:28.2%
x264 [info]: mb B I16..4: 0.1% 0.0% 0.1% B16..8: 20.8% 0.7% 1.2% direct: 1.2% skip:75.9%
x264 [info]: 8x8 transform intra:22.2% inter:52.4%
x264 [info]: ref P 74.1% 15.7% 10.2%
x264 [info]: ref B 90.1% 9.9%
x264 [info]: kb/s:938.8
encoded 1724 frames, 13.51 fps, 938.90 kb/s
变化对比
结果好像和changelog不一样哦~
在me6和tre1的情况下，767的速度确实大大的提高，提高幅度较763达到25%，很惊人！但是在me7和tre2的情况下，这个速度的提高基本可以忽略了。
再来看两个参数对码率的影响。同样是CQ18，第二套参数的码率要比第一套小6%左右，高参数对体积的影响还是可观的。不过缺点是会牺牲速度。到底是要速度还是要码率，就看各位Riper的意思了~
