x264 1360 gcc 3.4.6 vs gcc 4.4.2

[bob0r]

http://x264.nl/x264.1360.gcc.3.4.6.make.exe
http://x264.nl/x264.1360.gcc.3.4.6.make.fprofiled.exe
http://x264.nl/x264.1360.gcc.4.4.2.make.exe
http://x264.nl/x264.1360.gcc.4.4.2.make.fprofiled.exe

ran speed.bat twice:

x264.1360.gcc.3.4.6.make.exe -b 2000 ../1280x720p50_parkrun_ter.yuv -o NUL
x264.1360.gcc.3.4.6.make.fprofiled.exe -b 2000 ../1280x720p50_parkrun_ter.yuv -o NUL
x264.1360.gcc.4.4.2.make.exe -b 2000 ../1280x720p50_parkrun_ter.yuv -o NUL
x264.1360.gcc.4.4.2.make.fprofiled.exe -b 2000 ../1280x720p50_parkrun_ter.yuv -o NUL


my results: http://x264.nl/x264.1360.gcc.3.4.6.vs.gcc.4.4.2.txt

gcc 4.4.2 make seems fastest here (*), how is it for you?



* cpu: Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz (x86) at 2405MHz
* ram: 894/2047MB
* os: Microsoft Windows XP Professional - Service Pack 3 (5.1.2600)

[XhmikosR]

Video: 2000 frames Big Buck Bunny 1280x720 lagarith
CPU: Q6600 @ 3Ghz
RAM: 2GB-540MB used @ idle
OS: Windows 7 Pro

x264.1360.gcc.3.4.6.make.exe -B 2000 --quiet -o NUL "test.avs"
x264.1360.gcc.3.4.6.make.fprofiled.exe -B 2000 --quiet -o NUL "test.avs"
x264.1360.gcc.4.2.2.make.exe -B 2000 --quiet -o NUL "test.avs"
x264.1360.gcc.4.2.2.make.fprofiled.exe -B 2000 --quiet -o NUL "test.avs"

x264.1360.gcc.3.4.6.make
encoded 2000 frames, 23.40 fps, 2022.51 kb/s
encoded 2000 frames, 23.33 fps, 2022.51 kb/s

x264.1360.gcc.3.4.6.make.fprofiled
encoded 2000 frames, 23.74 fps, 2022.51 kb/s
encoded 2000 frames, 23.56 fps, 2022.51 kb/s

x264.1360.gcc.4.4.2.make
encoded 2000 frames, 23.44 fps, 2022.44 kb/s
encoded 2000 frames, 23.74 fps, 2022.44 kb/s

x264.1360.gcc.4.4.2.make.fprofiled
encoded 2000 frames, 23.45 fps, 2022.44 kb/s
encoded 2000 frames, 23.77 fps, 2022.44 kb/s

I don't think two runs are enough to make a conclusion.

[VempX]

I remember darkshikari said that if you use gcc 4.* you will be get a mistake.
Your weightp may be miscomplied.:confused:

[wyti]

yeah... and output seems different between the two builds regarding XhmikosR post.

Is there a solutionl, because x64 build needs gcc 4.*

Oh and a little typo in your texte bobor, one time it's 4.4.2 and one time it's 4.2.2

[XhmikosR]

Video: 2000 frames ParkJoy 1280x720 YUV
CPU: Q6600 @ 3Ghz
RAM: 2GB-540MB used @ idle
OS: Windows 7 Pro

x264.1360.gcc.3.4.6.make.exe -B 2000 --quiet -o NUL "ParkJoy_1280x720_50.avi"
x264.1360.gcc.3.4.6.make.fprofiled.exe -B 2000 --quiet -o NUL "ParkJoy_1280x720_50.avi"
x264.1360.gcc.4.2.2.make.exe -B 2000 --quiet -o NUL "ParkJoy_1280x720_50.avi"
x264.1360.gcc.4.2.2.make.fprofiled.exe -B 2000 --quiet -o NUL "ParkJoy_1280x720_50.avi"


x264.1360.gcc.3.4.6.make
encoded 2000 frames, 35.27 fps, 1983.57 kb/s
encoded 2000 frames, 35.07 fps, 1983.57 kb/s
encoded 2000 frames, 35.14 fps, 1983.57 kb/s
encoded 2000 frames, 35.11 fps, 1983.57 kb/s
average: 35.15 fps

x264.1360.gcc.3.4.6.make.fprofiled
encoded 2000 frames, 35.69 fps, 1983.57 kb/s
encoded 2000 frames, 35.29 fps, 1983.57 kb/s
encoded 2000 frames, 35.62 fps, 1983.57 kb/s
encoded 2000 frames, 35.53 fps, 1983.57 kb/s
average: 35.53 fps

x264.1360.gcc.4.4.2.make
encoded 2000 frames, 35.70 fps, 1983.92 kb/s
encoded 2000 frames, 35.60 fps, 1983.92 kb/s
encoded 2000 frames, 35.73 fps, 1983.92 kb/s
encoded 2000 frames, 35.67 fps, 1983.92 kb/s
average: 35.68 fps

x264.1360.gcc.4.4.2.make.fprofiled
encoded 2000 frames, 35.78 fps, 1983.92 kb/s
encoded 2000 frames, 35.87 fps, 1983.92 kb/s
encoded 2000 frames, 35.87 fps, 1983.92 kb/s
encoded 2000 frames, 35.72 fps, 1983.92 kb/s
average: 35.81 fps

So yes, GCC 4.4.2 is slightly faster than GCC 3.4.6.