Help - Search - Members - Calendar
Full Version: First stab at some P4 optimized compiles of FLAC 1.1.4
Hydrogenaudio Forums > Lossless Audio Compression > FLAC
gharris999
Feb 25 2007, 10:38
I've posted a bunch of different compiles of FLAC 1.1.4 at http://www.hegardtfoundation.org/flacstuff...lacVersions.zip

The idea here is to provide various compiles of flac using various compile
and link time optimizations so that you can determine for yourself if one
version stands out on your machine as encoding or decoding significantly
faster than the stock version. The test regime provided in the post is crude,
but it should give you some idea of how these various versions stack up.

I have no idea how these compiles will perform on AMD machines and I
apologize for that.

With my own testing, no one version seems to be a clear winner across all
three P4 machines I'm testing on. It seems as though different Intel CPUs
prefer different optimizations. With some of these compiles, I'm seeing
40% to 50% encoding speed improvements over the stock compile. But
none of these compiles seem to significantly beat the stock 1.1.4 version in
decoding times.

Please post your results here if they seem interesting to you.

Thanks to Martin H for his helpful suggestions from previous threads.
guruboolez
Feb 25 2007, 12:45
Core2Duo E6300 on heavily fragmented SATA Disk
test.wav length = 54 minutes and 07 seconds (3247 seconds)
setting = default (-5)


CODE

FLAC tests on CORE2DUO
*
flac-1.1.4-IC9 encode testing
Global Time = 62.250 = 00:01:02.250 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 58.875 = 00:00:58.875 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 58.657 = 00:00:58.657 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 53.062 = 00:00:53.062 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 52.781 = 00:00:52.781 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 53.141 = 00:00:53.141 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 72.906 = 00:01:12.906 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 71.672 = 00:01:11.672 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 68.531 = 00:01:08.531 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 68.687 = 00:01:08.687 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 71.907 = 00:01:11.907 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 83.938 = 00:01:23.938 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 84.969 = 00:01:24.969 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 19.719 = 00:00:19.719 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 19.469 = 00:00:19.469 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 18.953 = 00:00:18.953 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 17.828 = 00:00:17.828 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 17.719 = 00:00:17.719 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 18.094 = 00:00:18.094 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 20.265 = 00:00:20.265 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 21.297 = 00:00:21.297 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 19.657 = 00:00:19.657 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 19.110 = 00:00:19.110 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 19.313 = 00:00:19.313 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 20.110 = 00:00:20.110 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 22.750 = 00:00:22.750 = 100%
*
*
25/02/2007 18:02 226ÿ566ÿ550 test.flac
25/02/2007 18:02 226ÿ570ÿ807 test.flac
25/02/2007 18:02 226ÿ570ÿ807 test.flac
25/02/2007 18:02 226ÿ570ÿ816 test.flac
25/02/2007 18:02 226ÿ570ÿ816 test.flac
25/02/2007 18:02 226ÿ570ÿ816 test.flac
25/02/2007 18:02 226ÿ566ÿ546 test.flac
25/02/2007 18:02 226ÿ566ÿ555 test.flac
25/02/2007 18:02 226ÿ570ÿ812 test.flac
25/02/2007 18:02 226ÿ570ÿ818 test.flac
25/02/2007 18:02 226ÿ566ÿ565 test.flac
25/02/2007 18:02 226ÿ570ÿ818 test.flac
25/02/2007 18:02 226ÿ570ÿ799 test.flac
*
*
Comparaison des fichiers test.wav et FLAC-1.1.4-IC9\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-IC9FAST\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-STOCK\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-VS2005\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-VS2005FAST\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e

Comparaison des fichiers test.wav et FLAC-1.1.4-VS2005OPT\TEST.WAV
FCÿ: aucune diff‚rence trouv‚e



in summary:
STOCK compile: 72.906 sec.
FASTEST compile: 52.781 sec. with IC9fastStaticQaxW
other fast compiles: IC9fastStaticQxN and IC9fastStaticOpt
slower compiles than STOCK: VS2005fastStaticOptSse (83.938 sec.) and VS2005opt (84.969)



With foobar2000 (using one CPU core) EDIT: and using a FLAC source:
1.1.3 CVS = 1:15.156, 43.19x realtime
1.1.3 IC9 = 1:03.688, 50.97x realtime
1.1.4 CVS = 0:59.375, 54.67x realtime
1.1.4 IC9 = 0:52.672, 61.63x realtime [IC9fastStaticQaxW]

1.1.4 CVS is already faster than 1.1.3 IC9 which was amazing on my computer. I immediately noticed it while using 1.1.4 from Josh on real usage. The good news is that further optimization are still possible with dedicated compiles of 1.1.4. Thank you!
drbeachboy
Feb 25 2007, 18:51
Here are my results for encoding only:
CODE
Encoding Only,  Global Time,  WAV=386,163 KB,  Processor=Intel P4 Williamette 1.3G, Instructions=MMX, SSE, SSE2,
Encoder Settings=FLAC -5 -V

112-Stock  2:26.240
113-Stock  2:26.501
113-IC9sseW  1:57.199
114-Stock  1:58.751
114-IC9  1:52.752
114-IC9fast  1:58.030
114-IC9fastStatic  1:51.350
114-IC9fastStaticOpt  1:55.015
114-IC9fastStaticQaxW  1:48.266
114-IC9fastStaticQxN  1:48.026
114-VS2005  2:09.046
114-VS2005fast  2:05.200
114-VS2005fastStatic  2:03.358
114-VS2005fastStaticNoSse  2:08.455
114-VS2005fastStaticOptSse  2:04.759
114-VS2005Opt  2:05.901
wraithdu
Feb 25 2007, 20:25
Processor - P4 2.4 GHz
RAM - 1 GB PC2700 DDR
Drive - 250 GB SATA

timeflac_results
CODE
FLAC tests
*
flac-1.1.4-IC9 encode testing
Global Time = 104.172 = 00:01:44.172 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 99.375 = 00:01:39.375 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 99.453 = 00:01:39.453 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 107.016 = 00:01:47.016 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 111.250 = 00:01:51.250 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 106.906 = 00:01:46.906 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 142.516 = 00:02:22.516 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 113.312 = 00:01:53.312 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 108.922 = 00:01:48.922 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 108.891 = 00:01:48.891 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 112.828 = 00:01:52.828 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 147.734 = 00:02:27.734 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 148.078 = 00:02:28.078 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 34.266 = 00:00:34.266 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 34.688 = 00:00:34.688 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 33.484 = 00:00:33.484 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 33.390 = 00:00:33.390 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 33.844 = 00:00:33.844 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 34.391 = 00:00:34.391 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 36.953 = 00:00:36.953 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 36.953 = 00:00:36.953 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 36.672 = 00:00:36.672 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 36.547 = 00:00:36.547 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 36.875 = 00:00:36.875 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 36.469 = 00:00:36.469 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 36.453 = 00:00:36.453 = 100%
*
*
02/25/2007 06:28 PM 465,995,863 test.flac
02/25/2007 06:28 PM 465,996,718 test.flac
02/25/2007 06:28 PM 465,996,718 test.flac
02/25/2007 06:28 PM 465,996,727 test.flac
02/25/2007 06:28 PM 465,996,727 test.flac
02/25/2007 06:28 PM 465,996,727 test.flac
02/25/2007 06:28 PM 465,995,859 test.flac
02/25/2007 06:28 PM 465,995,868 test.flac
02/25/2007 06:28 PM 465,996,723 test.flac
02/25/2007 06:28 PM 465,996,729 test.flac
02/25/2007 06:28 PM 465,995,878 test.flac
02/25/2007 06:28 PM 465,996,729 test.flac
02/25/2007 06:28 PM 465,996,710 test.flac
*
*
Comparing files test.wav and FLAC-1.1.4-IC9\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-STOCK\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: no differences encountered


testflac_results
CODE
FLAC tests
flac-1.1.4-IC9 Flac and MetaFlac encode test: 155.312
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 151.484
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 150.937
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 157.984
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 157.750
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 163.891
flac-1.1.4-stock Flac and MetaFlac encode test: 219.750
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 169.172
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 164.453
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 164.015
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 168.578
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 202.765
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 204.094
flac-1.1.4-IC9 decode test: 34.672
flac-1.1.4-IC9fast decode test: 34.375
flac-1.1.4-IC9fastStatic decode test: 34.141
flac-1.1.4-IC9fastStaticOpt decode test: 34.891
flac-1.1.4-IC9fastStaticQaxW decode test: 33.875
flac-1.1.4-IC9fastStaticQxN decode test: 33.703
flac-1.1.4-stock decode test: 37.063
flac-1.1.4-VS2005 decode test: 37.156
flac-1.1.4-VS2005fast decode test: 37.141
flac-1.1.4-VS2005fastStatic decode test: 36.641
flac-1.1.4-VS2005fastStaticNoSse decode test: 36.812
flac-1.1.4-VS2005fastStaticOptSse decode test: 36.172
flac-1.1.4-VS2005opt decode test: 36.672
02/25/2007 06:41 PM 465,995,863 test.flac
02/25/2007 06:43 PM 465,996,718 test.flac
02/25/2007 06:46 PM 465,996,718 test.flac
02/25/2007 06:48 PM 465,996,727 test.flac
02/25/2007 06:51 PM 465,996,727 test.flac
02/25/2007 06:54 PM 465,996,727 test.flac
02/25/2007 06:57 PM 465,995,859 test.flac
02/25/2007 07:00 PM 465,995,868 test.flac
02/25/2007 07:03 PM 465,996,723 test.flac
02/25/2007 07:06 PM 465,996,729 test.flac
02/25/2007 07:08 PM 465,995,878 test.flac
02/25/2007 07:12 PM 465,996,729 test.flac
02/25/2007 07:15 PM 465,996,710 test.flac
wraithdu
Feb 26 2007, 00:00
Processor - Core2Duo T2300 Notebook
RAM - 512 MB
Drive - 80 GB SATA

timeflac_results -
CODE
FLAC tests
*
flac-1.1.4-IC9 encode testing
Global Time = 44.609 = 00:00:44.609 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 44.687 = 00:00:44.687 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 42.937 = 00:00:42.937 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 44.610 = 00:00:44.610 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 45.093 = 00:00:45.093 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 43.985 = 00:00:43.985 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 47.875 = 00:00:47.875 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 48.344 = 00:00:48.344 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 47.593 = 00:00:47.593 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 47.109 = 00:00:47.109 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 47.797 = 00:00:47.797 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 59.906 = 00:00:59.906 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 60.219 = 00:01:00.219 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 27.156 = 00:00:27.156 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 27.593 = 00:00:27.593 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 32.718 = 00:00:32.718 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 26.344 = 00:00:26.344 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 27.312 = 00:00:27.312 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 28.266 = 00:00:28.266 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 34.937 = 00:00:34.937 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 31.922 = 00:00:31.922 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 27.093 = 00:00:27.093 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 26.422 = 00:00:26.422 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 26.953 = 00:00:26.953 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 26.906 = 00:00:26.906 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 26.469 = 00:00:26.469 = 100%
*
*
02/25/2007 10:11 PM 207,668,030 test.flac
02/25/2007 10:11 PM 207,668,920 test.flac
02/25/2007 10:11 PM 207,668,920 test.flac
02/25/2007 10:11 PM 207,668,929 test.flac
02/25/2007 10:11 PM 207,668,929 test.flac
02/25/2007 10:11 PM 207,668,929 test.flac
02/25/2007 10:11 PM 207,668,026 test.flac
02/25/2007 10:11 PM 207,668,035 test.flac
02/25/2007 10:11 PM 207,668,925 test.flac
02/25/2007 10:11 PM 207,668,931 test.flac
02/25/2007 10:11 PM 207,668,045 test.flac
02/25/2007 10:11 PM 207,668,931 test.flac
02/25/2007 10:11 PM 207,668,912 test.flac
*
*
Comparing files test.wav and FLAC-1.1.4-IC9\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-STOCK\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: no differences encountered


testflac_results
CODE
FLAC tests
flac-1.1.4-IC9 Flac and MetaFlac encode test: 119.172
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 98.531
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 96.156
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 102.109
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 107.344
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 111.609
flac-1.1.4-stock Flac and MetaFlac encode test: 97.219
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 99.719
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 96.218
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 98.703
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 124.047
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 124.453
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 107.282
flac-1.1.4-IC9 decode test: 26.890
flac-1.1.4-IC9fast decode test: 33.141
flac-1.1.4-IC9fastStatic decode test: 25.688
flac-1.1.4-IC9fastStaticOpt decode test: 30.844
flac-1.1.4-IC9fastStaticQaxW decode test: 25.078
flac-1.1.4-IC9fastStaticQxN decode test: 27.484
flac-1.1.4-stock decode test: 26.844
flac-1.1.4-VS2005 decode test: 31.329
flac-1.1.4-VS2005fast decode test: 42.500
flac-1.1.4-VS2005fastStatic decode test: 29.328
flac-1.1.4-VS2005fastStaticNoSse decode test: 29.016
flac-1.1.4-VS2005fastStaticOptSse decode test: 43.719
flac-1.1.4-VS2005opt decode test: 28.969
02/25/2007 10:48 PM 207,668,030 test.flac
02/25/2007 10:50 PM 207,668,920 test.flac
02/25/2007 10:51 PM 207,668,920 test.flac
02/25/2007 10:53 PM 207,668,929 test.flac
02/25/2007 10:55 PM 207,668,929 test.flac
02/25/2007 10:57 PM 207,668,929 test.flac
02/25/2007 10:58 PM 207,668,026 test.flac
02/25/2007 11:00 PM 207,668,035 test.flac
02/25/2007 11:01 PM 207,668,925 test.flac
02/25/2007 11:03 PM 207,668,931 test.flac
02/25/2007 11:05 PM 207,668,045 test.flac
02/25/2007 11:07 PM 207,668,931 test.flac
02/25/2007 11:09 PM 207,668,912 test.flac
Zoom
Feb 26 2007, 01:04
I did my own tests with the various encoders. I tested the same album image, with one setting for each version of FLAC. The only setting I used with each test was "-8". I tested in this manner because this is how I use FLAC in practice. I substituted the Rarewares 1.1.4 binary for the stock 1.1.4 binary because that is the one I currently use.

CODE
My Test System
Athlon 64 3500+ (Winchester Core) 2200MHz
1024MB RAM
WinXP SP2

Album Image:
The Shins - Wincing the Night Away
379,245,932 bytes

flac-1.1.2-stock
163.015 = 00:02:43.015

flac-1.1.3-stock
165.781 = 00:02:45.781

flac-1.1.4-stockRarewares-MSVC6
73.531 = 00:01:13.531

flac-1.1.3-IC9sseW
97.171 = 00:01:37.171

flac-1.1.4-IC9
87.734 = 00:01:27.734

flac-1.1.4-IC9fast
89.531 = 00:01:29.531

flac-1.1.4-IC9fastStatic
88.703 = 00:01:28.703

flac-1.1.4-IC9fastStaticOpt
78.640 = 00:01:18.640

flac-1.1.4-IC9fastStaticQaxW
93.718 = 00:01:33.718

flac-1.1.4-IC9fastStaticQxN***
N/A

flac-1.1.4-VS2005
103.281 = 00:01:43.281

flac-1.1.4-VS2005fast
105.234 = 00:01:45.234

flac-1.1.4-VS2005fastStatic
105.421 = 00:01:45.421

flac-1.1.4-VS2005fastStaticNoSse
103.000 = 00:01:43.000

flac-1.1.4-VS2005fastStaticOptSse
107.218 = 00:01:47.218

flac-1.1.4-VS2005opt
108.796 = 00:01:48.796

***flac-1.1.4-IC9fastStaticQxN failed to run on my processor.


EDIT: Just saw this post. It seems I'm having the same issues.
zombiewerewolf
Feb 26 2007, 05:44
Pentium M 740 (1.73GHz)
Hard Drive: Serial ATA, 5400 rpm
RAM: 512MB PC2-3200
test.wav Length = 01:01:11

CODE

FLAC tests on VAIO S45
flac-1.1.4-IC9 Flac and MetaFlac encode test: 230.047
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 235.734
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 236.438
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 231.297
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 236.625
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 229.016
flac-1.1.4-stock Flac and MetaFlac encode test: 235.719
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 230.453
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 243.250
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 274.219
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 283.109
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 312.328
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 342.344
flac-1.1.4-IC9 decode test: 63.360
flac-1.1.4-IC9fast decode test: 65.375
flac-1.1.4-IC9fastStatic decode test: 71.047
flac-1.1.4-IC9fastStaticOpt decode test: 68.954
flac-1.1.4-IC9fastStaticQaxW decode test: 65.860
flac-1.1.4-IC9fastStaticQxN decode test: 68.906
flac-1.1.4-stock decode test: 71.797
flac-1.1.4-VS2005 decode test: 94.234
flac-1.1.4-VS2005fast decode test: 73.000
flac-1.1.4-VS2005fastStatic decode test: 68.813
flac-1.1.4-VS2005fastStaticNoSse decode test: 70.406
flac-1.1.4-VS2005fastStaticOptSse decode test: 71.890
flac-1.1.4-VS2005opt decode test: 77.906
26/02/2007 13:05 426,278,629 test.flac
26/02/2007 13:09 426,282,899 test.flac
26/02/2007 13:13 426,282,899 test.flac
26/02/2007 13:17 426,282,908 test.flac
26/02/2007 13:21 426,282,908 test.flac
26/02/2007 13:25 426,282,908 test.flac
26/02/2007 13:29 426,278,625 test.flac
26/02/2007 13:32 426,278,634 test.flac
26/02/2007 13:36 426,282,904 test.flac
26/02/2007 13:41 426,282,910 test.flac
26/02/2007 13:46 426,278,644 test.flac
26/02/2007 13:51 426,282,910 test.flac
26/02/2007 13:57 426,282,891 test.flac


CODE

FLAC tests on VAIO S45
*
flac-1.1.4-IC9 encode testing
Global Time = 123.531 = 00:02:03.531 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 115.281 = 00:01:55.281 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 101.375 = 00:01:41.375 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 111.609 = 00:01:51.609 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 109.297 = 00:01:49.297 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 106.813 = 00:01:46.813 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 118.969 = 00:01:58.969 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 124.344 = 00:02:04.344 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 136.141 = 00:02:16.141 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 121.094 = 00:02:01.094 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 136.438 = 00:02:16.438 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 153.297 = 00:02:33.297 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 154.891 = 00:02:34.891 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 93.079 = 00:01:33.079 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 89.500 = 00:01:29.500 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 78.406 = 00:01:18.406 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 72.469 = 00:01:12.469 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 67.609 = 00:01:07.609 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 79.000 = 00:01:19.000 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 86.797 = 00:01:26.797 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 80.719 = 00:01:20.719 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 85.829 = 00:01:25.829 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 76.375 = 00:01:16.375 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 77.938 = 00:01:17.938 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 70.438 = 00:01:10.438 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 76.969 = 00:01:16.969 = 100%
*
*
26/02/2007 12:10 426,278,678 test.flac
26/02/2007 12:10 426,282,948 test.flac
26/02/2007 12:10 426,282,948 test.flac
26/02/2007 12:10 426,282,957 test.flac
26/02/2007 12:10 426,282,957 test.flac
26/02/2007 12:10 426,282,957 test.flac
26/02/2007 12:10 426,278,674 test.flac
26/02/2007 12:10 426,278,683 test.flac
26/02/2007 12:10 426,282,953 test.flac
26/02/2007 12:10 426,282,959 test.flac
26/02/2007 12:10 426,278,693 test.flac
26/02/2007 12:10 426,282,959 test.flac
26/02/2007 12:10 426,282,940 test.flac
*
*
Comparing files test.wav and FLAC-1.1.4-IC9\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-STOCK\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: no differences encountered

Moderation: CODE to CODEBOX
a_aa
Feb 26 2007, 15:16
Tested it on my old setup, which looks like this:

CPU: Intel Pentium 4 3.2 GHz (Northwood, FSB800, HTT, MMX, SSE, SSE2)
HDD: PATA100, Samsung Spinpoint 160 GB, almost unfragmented
RAM: 4 x 512 MB (timings: 2,3,2,6,2T)
MoBo: Gigabyte GA-8ISXT-FS (FS=Fujitsu Siemens)
NB: Intel i865P/PE/G/i848P rev. A2, SB: Intel 82801EB (ICH5) rev. 02

OS: MS XP Home SP2

test.wav duration 1:03:12.893

I've added the MSCV6-compile from Rarewares out of curiousity...


Testing with option -5 (56.63%):
timeflac_results
CODE
FLAC tests on P4 with -5 option
*
flac-1.1.4-IC9 encode testing
Global Time = 0.047 = 00:00:00.047 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 0.015 = 00:00:00.015 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 73.812 = 00:01:13.812 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 62.828 = 00:01:02.828 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 62.735 = 00:01:02.735 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 62.235 = 00:01:02.235 = 100%
*
flac-1.1.4-MSVC6 encode testing
Global Time = 102.157 = 00:01:42.157 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 102.313 = 00:01:42.313 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 0.032 = 00:00:00.032 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 0.032 = 00:00:00.032 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 76.281 = 00:01:16.281 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 80.266 = 00:01:20.266 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 86.250 = 00:01:26.250 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 87.688 = 00:01:27.688 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 0.031 = 00:00:00.031 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 0.016 = 00:00:00.016 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 32.141 = 00:00:32.141 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 72.204 = 00:01:12.204 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 60.500 = 00:01:00.500 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 48.297 = 00:00:48.297 = 100%
*
flac-1.1.4-MSVC6 decode testing
Global Time = 44.906 = 00:00:44.906 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 45.766 = 00:00:45.766 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 0.140 = 00:00:00.140 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 0.015 = 00:00:00.015 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 42.531 = 00:00:42.531 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 41.797 = 00:00:41.797 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 40.079 = 00:00:40.079 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 38.828 = 00:00:38.828 = 100%
*
*
25.02.2007 20:22 378ÿ888ÿ810 test.flac
25.02.2007 20:22 378ÿ888ÿ819 test.flac
25.02.2007 20:22 378ÿ888ÿ819 test.flac
25.02.2007 20:22 378ÿ888ÿ819 test.flac
25.02.2007 20:22 378ÿ888ÿ279 test.flac
25.02.2007 20:22 378ÿ888ÿ279 test.flac
25.02.2007 20:22 378ÿ888ÿ821 test.flac
25.02.2007 20:22 378ÿ888ÿ298 test.flac
25.02.2007 20:22 378ÿ888ÿ821 test.flac
25.02.2007 20:22 378ÿ888ÿ802 test.flac
*
*
Sammenligner filene test.wav og FLAC-1.1.4-IC9\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-IC9FAST\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-MSVC6\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-STOCK\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-VS2005FAST\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: Fant ingen forskjeller.
testflac_results
CODE
FLAC tests on P4 with -5 option
flac-1.1.4-IC9 Flac and MetaFlac encode test: 0.016
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 0.031
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 141.969
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 105.188
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 95.531
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 95.093
flac-1.1.4-MSVC6 Flac and MetaFlac encode test: 132.703
flac-1.1.4-stock Flac and MetaFlac encode test: 135.328
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 0.172
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 0.047
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 109.671
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 111.547
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 115.453
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 117.547
flac-1.1.4-IC9 decode test: 0.156
flac-1.1.4-IC9fast decode test: 0.032
flac-1.1.4-IC9fastStatic decode test: 62.296
flac-1.1.4-IC9fastStaticOpt decode test: 54.922
flac-1.1.4-IC9fastStaticQaxW decode test: 46.781
flac-1.1.4-IC9fastStaticQxN decode test: 40.531
flac-1.1.4-MSVC6 decode test: 42.094
flac-1.1.4-stock decode test: 41.094
flac-1.1.4-VS2005 decode test: 0.047
flac-1.1.4-VS2005fast decode test: 0.016
flac-1.1.4-VS2005fastStatic decode test: 44.515
flac-1.1.4-VS2005fastStaticNoSse decode test: 39.578
flac-1.1.4-VS2005fastStaticOptSse decode test: 44.094
flac-1.1.4-VS2005opt decode test: 36.657
26.02.2007 18:07 378ÿ888ÿ798 test.flac
26.02.2007 18:09 378ÿ888ÿ807 test.flac
26.02.2007 18:10 378ÿ888ÿ807 test.flac
26.02.2007 18:12 378ÿ888ÿ807 test.flac
26.02.2007 18:14 378ÿ888ÿ267 test.flac
26.02.2007 18:16 378ÿ888ÿ267 test.flac
26.02.2007 18:18 378ÿ888ÿ809 test.flac
26.02.2007 18:20 378ÿ888ÿ286 test.flac
26.02.2007 18:22 378ÿ888ÿ809 test.flac
26.02.2007 18:24 378ÿ888ÿ790 test.flac


Testing with option -8 (56.31%, my usual setting):
timeflac_results
CODE
FLAC tests on P4 with -8 option
*
flac-1.1.4-IC9 encode testing
Global Time = 0.015 = 00:00:00.015 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 0.015 = 00:00:00.015 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 197.203 = 00:03:17.203 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 186.204 = 00:03:06.204 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 186.266 = 00:03:06.266 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 185.219 = 00:03:05.219 = 100%
*
flac-1.1.4-MSVC6 encode testing
Global Time = 221.109 = 00:03:41.109 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 221.344 = 00:03:41.344 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 0.031 = 00:00:00.031 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 0.016 = 00:00:00.016 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 216.953 = 00:03:36.953 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 218.078 = 00:03:38.078 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 229.453 = 00:03:49.453 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 232.562 = 00:03:52.562 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 0.015 = 00:00:00.015 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 0.000 = 00:00:00.000 = 0%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 34.672 = 00:00:34.672 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 73.266 = 00:01:13.266 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 52.968 = 00:00:52.968 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 57.641 = 00:00:57.641 = 100%
*
flac-1.1.4-MSVC6 decode testing
Global Time = 47.047 = 00:00:47.047 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 45.563 = 00:00:45.563 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 0.047 = 00:00:00.047 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 0.031 = 00:00:00.031 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 42.469 = 00:00:42.469 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 48.578 = 00:00:48.578 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 43.984 = 00:00:43.984 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 49.062 = 00:00:49.062 = 100%
*
*
25.02.2007 20:22 376ÿ765ÿ675 test.flac
25.02.2007 20:22 376ÿ765ÿ684 test.flac
25.02.2007 20:22 376ÿ765ÿ684 test.flac
25.02.2007 20:22 376ÿ765ÿ684 test.flac
25.02.2007 20:22 376ÿ765ÿ667 test.flac
25.02.2007 20:22 376ÿ765ÿ667 test.flac
25.02.2007 20:22 376ÿ765ÿ686 test.flac
25.02.2007 20:22 376ÿ765ÿ686 test.flac
25.02.2007 20:22 376ÿ765ÿ686 test.flac
25.02.2007 20:22 376ÿ765ÿ667 test.flac
*
*
Sammenligner filene test.wav og FLAC-1.1.4-IC9\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-IC9FAST\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-MSVC6\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-STOCK\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-VS2005FAST\TEST.WAV
Sammenligner filene test.wav og FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: Fant ingen forskjeller.

Sammenligner filene test.wav og FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: Fant ingen forskjeller.
testflac_results
CODE
FLAC tests on P4 with -8 option
flac-1.1.4-IC9 Flac and MetaFlac encode test: 0.015
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 0.016
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 249.454
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 228.187
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 219.750
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 214.594
flac-1.1.4-MSVC6 Flac and MetaFlac encode test: 250.453
flac-1.1.4-stock Flac and MetaFlac encode test: 252.281
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 0.187
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 0.047
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 247.562
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 248.016
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 259.750
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 262.985
flac-1.1.4-IC9 decode test: 0.094
flac-1.1.4-IC9fast decode test: 0.047
flac-1.1.4-IC9fastStatic decode test: 66.344
flac-1.1.4-IC9fastStaticOpt decode test: 59.250
flac-1.1.4-IC9fastStaticQaxW decode test: 44.640
flac-1.1.4-IC9fastStaticQxN decode test: 39.594
flac-1.1.4-MSVC6 decode test: 47.843
flac-1.1.4-stock decode test: 47.110
flac-1.1.4-VS2005 decode test: 0.063
flac-1.1.4-VS2005fast decode test: 0.016
flac-1.1.4-VS2005fastStatic decode test: 42.093
flac-1.1.4-VS2005fastStaticNoSse decode test: 37.234
flac-1.1.4-VS2005fastStaticOptSse decode test: 36.625
flac-1.1.4-VS2005opt decode test: 36.031
26.02.2007 19:18 376ÿ765ÿ663 test.flac
26.02.2007 19:21 376ÿ765ÿ672 test.flac
26.02.2007 19:25 376ÿ765ÿ672 test.flac
26.02.2007 19:29 376ÿ765ÿ672 test.flac
26.02.2007 19:33 376ÿ765ÿ655 test.flac
26.02.2007 19:37 376ÿ765ÿ655 test.flac
26.02.2007 19:41 376ÿ765ÿ674 test.flac
26.02.2007 19:45 376ÿ765ÿ674 test.flac
26.02.2007 19:50 376ÿ765ÿ674 test.flac
26.02.2007 19:54 376ÿ765ÿ655 test.flac


Comments:
4 of the compiles would not run on my system - the times for IC9, IC9fast, VS2005 and VS2005fast in the resultfiles are of course wrong. (Btw, my system also refuses to run FlacGetV.exe - probably because of this?)

On my system, it seems that:
* the difference between the Rareware and stock compile is negligable
* the 2 ICL9fastStaticQ compiles are winners in terms of encoding speed
* the absolute gain in encoding speed (in seconds) is fairly similar for options -5 and -8
* the relative gain in encoding (in % related to stock) is much higher for option -5 (40%!)
* decoding speed seems to have a little less room for improvement, and it seems that different compression option turns the tables...

Anyways, since I use the -8 option, it seems that the IC9fastStaticQxN compile is something for me!

Thanks, gharris999! smile.gif
gharris999
Feb 26 2007, 17:21
QUOTE(a_aa @ Feb 26 2007, 14:16) *

4 of the compiles would not run on my system - the times for IC9, IC9fast, VS2005 and VS2005fast in the resultfiles are of course wrong. (Btw, my system also refuses to run FlacGetV.exe - probably because of this?)

The compiles you cite are the ones that aren't statically linked to msvcr80.dll and libmmd.dll. You can copy those dlls to someplace on your path (e.g. %SystemRoot%\system32 or, if you don't want to do that, you could try copying them to the parent folder of the "bin" folder for the compiles in question.

I *thought* that WinXP always checked the folder where an EXE image was located for any dlls missing at run-time. In your case, it seems like XP is just checking the cwd...the current working directory which is ..\bin as invoked by the test scripts.
JWolf
Feb 26 2007, 17:50
Can we have a compile that's optimized for the AMD Athlon XP please? Thanks!

Jon
a_aa
Feb 26 2007, 18:44
QUOTE(gharris999 @ Feb 27 2007, 00:21) *
The compiles you cite are the ones that aren't statically linked to msvcr80.dll and libmmd.dll. You can copy those dlls to someplace on your path (e.g. %SystemRoot%\system32 or, if you don't want to do that, you could try copying them to the parent folder of the "bin" folder for the compiles in question.

I *thought* that WinXP always checked the folder where an EXE image was located for any dlls missing at run-time. In your case, it seems like XP is just checking the cwd...the current working directory which is ..\bin as invoked by the test scripts.


Nope - still won't run... Placed dlls in System32 and parent folder(s) with no luck. Please don't tell me I'll have to fiddle around in the registry or something like that, I hate that! blink.gif

btw - the tags on the IC9fastStaticQaxW and IC9fastStaticQxN files says it was encoded by the IC9fastStaticOpt compile...


Best regards,

Asbjørn
gordo
Feb 26 2007, 19:32
3.2 GHz Pentium 4 (Prescott) with SSE3
200 GB SATA hard drive
Vista

I use flac -8 so I ran the tests for that:

CODE
FLAC tests on NOVA
*
flac-1.1.4-IC9 encode testing
Global Time  =   279.141 = 00:04:39.141 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time  =   271.360 = 00:04:31.360 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time  =   272.047 = 00:04:32.047 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time  =   259.953 = 00:04:19.953 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time  =   260.078 = 00:04:20.078 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time  =   260.172 = 00:04:20.172 = 100%
*
flac-1.1.4-stock encode testing
Global Time  =   257.156 = 00:04:17.156 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time  =   292.391 = 00:04:52.391 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time  =   292.891 = 00:04:52.891 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time  =   292.922 = 00:04:52.922 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time  =   295.406 = 00:04:55.406 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time  =   307.875 = 00:05:07.875 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time  =   309.875 = 00:05:09.875 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time  =    39.547 = 00:00:39.547 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time  =    41.547 = 00:00:41.547 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time  =    41.625 = 00:00:41.625 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time  =    43.094 = 00:00:43.094 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time  =    40.813 = 00:00:40.813 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time  =    38.062 = 00:00:38.062 = 100%
*
flac-1.1.4-stock decode testing
Global Time  =    41.843 = 00:00:41.843 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time  =    40.188 = 00:00:40.188 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time  =    38.938 = 00:00:38.938 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time  =    39.579 = 00:00:39.579 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time  =    53.375 = 00:00:53.375 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time  =    41.703 = 00:00:41.703 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time  =    40.687 = 00:00:40.687 = 100%
*
*


not much to gain here. I did see improvements with IC9faststatic... using -5 but they seem erased for -8
gharris999
Feb 27 2007, 00:18
QUOTE(JWolf @ Feb 26 2007, 16:50) *

Can we have a compile that's optimized for the AMD Athlon XP please?

I really and truly don't know how to do that! Why don't you run the test suite and see if any version seems to do better than the stock version on the AMD? (...and post your results, please!)

QUOTE(gordo @ Feb 26 2007, 18:32) *

3.2 GHz Pentium 4 (Prescott) with SSE3
200 GB SATA hard drive
Vista
I use flac -8 so I ran the tests for that:
not much to gain here. I did see improvements with IC9faststatic... using -5 but they seem erased for -8

Well that's disappointing. Anyone have a plausible explanation as to why these compiles perform so poorly on Vista?

QUOTE(a_aa @ Feb 26 2007, 17:44) *

btw - the tags on the IC9fastStaticQaxW and IC9fastStaticQxN files says it was encoded by the IC9fastStaticOpt compile...

I wasn't scrupulous about editing format.c in the libFLAC project to reflect all the various optimizations I was trying. So, you can ignore that.

---------------------------------------------------

Another disappointment is how irreproducible the results of the test scripts are. I can run the test scripts twice in a row and get markedly different numbers for the same compiles. I've posted a new version of the test CMD files here: http://www.hegardtfoundation.org/flacstuff/testtweaks.zip that will hopefully mitigate this by trying to run the binaries at a higher system priority. Hopefully, that will keep background junk from impinging on the test results.
JWolf
Feb 27 2007, 07:28
SSE3 is not compatible with the AMD Athlon XP. If you make a compiled version compatible with SSE, then you have a version compiled that will work for the Athlon XP.
Martin H
Feb 27 2007, 09:53
QUOTE(a_aa @ Feb 27 2007, 01:44) *

Nope - still won't run...
You need to generate a manifest file, when you use the VC8 runtimes as private assemblies. Copy msvcr80.dll into your apps folder and copy libmmd.dll either anywhere in %path%, or in the apps folder also. Copy and paste the following text into a file in the apps folder and name it: microsoft.vc80.crt.manifest
CODE
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<!-- Copyright © 1981-2001 Microsoft Corporation-->
<assembly xmlns="urn:schemas-microsoft-com:asm.v1"
                                 manifestVersion="1.0">
  <noInheritable/>
  <assemblyIdentity
    type="win32"
    name="Microsoft.VC80.CRT"
    version="8.0.50608.0"
    processorArchitecture="x86"
    publicKeyToken="1fc8b3b9a1e18e3b" />
  <file name="msvcr80.dll"/>
  <file name="msvcp80.dll"/>
  <file name="msvcm80.dll"/>
</assembly>
quellcore
Feb 27 2007, 10:15
Thanks for this batch of optimized Flac versions.
I'm on AMD here and wanted to try it anyway, here are the results.

CPU/System specs:
CODE
Name                    AMD Opteron 175 (UP)
Codename                Toledo
Specification           Dual Core AMD Opteron(tm) Processor 175
Package                 Socket 939
Core Speed              2750 MHz (11.0 x 250.0 MHz)
HT Link speed           1000 MHz
Stock frequency         2200 MHz
Instructions sets       MMX (+), 3DNow! (+), SSE, SSE2, SSE3, x86-64
L1 Data cache           2 x 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache    2 x 64 KBytes, 2-way set associative, 64-byte line size
L2 cache                2 x 1024 KBytes, 16-way set associative, 64-byte line size
RAM                     1 GB of DDR-Ram at 200 MHz
Harddisk                320 GB Sata-II


Test.wav
CODE
Filesize:      642.100.748 Bytes
Duration:      1:00:40.027
Flac Settings: -8


FLAC tests on Dual Core AMD Opteron™ Processor 175 @ 2750 MHz
CODE
flac-1.1.4-IC9 Flac and MetaFlac encode test: 153.906
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 157.782
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 154.046
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 144.828
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 163.281
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 0.109
flac-1.1.4-stock Flac and MetaFlac encode test: 147.078
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 176.578
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 178.890
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 177.547
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 173.766
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 193.672
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 198.062
flac-1.1.4-IC9 decode test: 21.375
flac-1.1.4-IC9fast decode test: 26.328
flac-1.1.4-IC9fastStatic decode test: 21.297
flac-1.1.4-IC9fastStaticOpt decode test: 23.172
flac-1.1.4-IC9fastStaticQaxW decode test: 21.922
flac-1.1.4-IC9fastStaticQxN decode test: 0.203
flac-1.1.4-stock decode test: 21.953
flac-1.1.4-VS2005 decode test: 24.015
flac-1.1.4-VS2005fast decode test: 21.859
flac-1.1.4-VS2005fastStatic decode test: 26.453
flac-1.1.4-VS2005fastStaticNoSse decode test: 21.734
flac-1.1.4-VS2005fastStaticOptSse decode test: 21.578
flac-1.1.4-VS2005opt decode test: 22.266
27.02.2007 00:38 337.943.266 test.flac
27.02.2007 00:41 337.986.380 test.flac
27.02.2007 00:43 337.986.380 test.flac
27.02.2007 00:46 337.986.389 test.flac
27.02.2007 00:49 337.986.389 test.flac
27.02.2007 00:51 337.943.262 test.flac
27.02.2007 00:54 337.943.271 test.flac
27.02.2007 00:57 337.986.385 test.flac
27.02.2007 01:00 337.986.391 test.flac
27.02.2007 01:03 337.943.281 test.flac
27.02.2007 01:06 337.986.391 test.flac
27.02.2007 01:09 337.986.372 test.flac


Conclusion:
With Flac setting "-8" only flac-1.1.4-IC9fastStaticOpt is faster than flac-1.1.4-stock and the speed gain is only about 1.5% ( = 2.3 seconds ).
Better than nothing, especially considering that it was not intended to be optimized for Athlon64 series.

flac-1.1.4-IC9fastStaticQxN didn't run at all.

Regards,
Quellcore
Synthetic Soul
Feb 27 2007, 10:22
QUOTE(JWolf @ Feb 27 2007, 13:28) *
SSE3 is not compatible with the AMD Athlon XP. If you make a compiled version compatible with SSE, then you have a version compiled that will work for the Athlon XP.
I think gharris999 is aware of that. smile.gif

Most of the compiles are using the FLAC__SSE_OS switch (most, not all). If you're not prepared to test all builds then maybe try flac-1.1.4-IC9fastStaticQaxW. Check this post for an explanation for my vague thinking (hey, at least I'm trying).

As a FLAC user it's well worth your time experimenting properly though. If enough users did so the data could be invaluable to Josh.
a_aa
Feb 27 2007, 11:03
QUOTE(Martin H @ Feb 27 2007, 16:53) *

QUOTE(a_aa @ Feb 27 2007, 01:44) *

Nope - still won't run...
You need to generate a manifest file...

Naaah, too late - I saw your post from yesterday here: http://www.hydrogenaudio.org/forums/index....st&p=475132

Downloaded the Microsoft Visual C++ 2005 Redistributable Package (x86), installed it and now all compiles (including FlacGetV.exe) are working just fine! biggrin.gif


Takker og bukker!

Asbjørn
Martin H
Feb 28 2007, 08:19
Yes, as Synthetic Soul said previously, then AMD users should not even bother trying to run compiles made with "/Qx(x)" switches. AMD users can use compiles made with "/Qax(x)" switches if their CPU's also supports the specific SIMD target of the switch used, like e.g. "/QaxW", which targets SSE+SSE2 on both Intel and AMD systems. For Intel P4 users, then "/QxN" compiles should theoretically be the fastest.

@gharris999

Thank you for your kind words in your first post my friend smile.gif I have in the mean time changed to using WavPack instead of FLAC, but i'm sure that all FLAC users are most gratefull to you for all your continued efforts on this great project of yours i.e. to provide optimized binaries of the FLAC command-line tools smile.gif

@Asbjørn

HAHA laugh.gif For a minutte there, then i thought that you where from Denmark wink.gif Anyway, i'm glad to hear that it works for you now smile.gif

CU, Martin.
Synthetic Soul
Feb 28 2007, 08:34
QUOTE(Martin H @ Feb 28 2007, 14:19) *
Yes, as Synthetic Soul said previously, then AMD users should not even bother trying to run compiles made with "/Qx(x)" switches. AMD users can use compiles made with "/Qax(x)" switches if their CPU's also supports the specific SIMD target of the switch used, like e.g. "/QaxW", which targets SSE+SSE2 on both Intel and AMD systems.
Thanks for the confirmation Martin.

NB: I will try to test them all on my Athlon XP in the next day or so.

As I said in the previous thread, I would love to see a decent amount of testing done here, to produce a conclusive result. If we don't get enough results from a variety of machines few conclusions can be reached, and it will all just fall by the wayside.

It amazes/frustrates me that, given the number of FLAC users on this board, we cannot produce results from hundreds of systems, and from that provide clear and confident results which Josh, and hopefully other developers, can then use.

By all means, prove me wrong.
JWolf
Feb 28 2007, 09:40
QUOTE(Synthetic Soul @ Feb 27 2007, 11:22) *

QUOTE(JWolf @ Feb 27 2007, 13:28) *
SSE3 is not compatible with the AMD Athlon XP. If you make a compiled version compatible with SSE, then you have a version compiled that will work for the Athlon XP.
I think gharris999 is aware of that. smile.gif

Most of the compiles are using the FLAC__SSE_OS switch (most, not all). If you're not prepared to test all builds then maybe try flac-1.1.4-IC9fastStaticQaxW. Check this post for an explanation for my vague thinking (hey, at least I'm trying).

As a FLAC user it's well worth your time experimenting properly though. If enough users did so the data could be invaluable to Josh.

I'll give that version a go sometime today and see how it works out. Thanks!
wisodev
Feb 28 2007, 10:03
Are this P4 build using PGO (/Qprof_genx and /Qprof_use switches) and IPO (/Qipo switch) optimizations? The PGO and IPO optimizations (used in Intel C++ compiler) are providing bigest speedups.

For AMD CPUs with SSE I'd recommend using /QxK switch, for AMD CPUs with SSE2 I'd recommend using /QxW switch.

For ICL builds I'd recommend using this switches (of course in pair with the /QxK or /QxW switches):
CODE
/GA /Ox /Og /Ob1 /Ot

and specific to version of compiler, with 9.1 you should use:
CODE
/Qipo4 /fp:fast /Qprec-div- /Qcomplex-limited-range

I use this settings for my Aften and Flake builds and it seems to be the best configuration.

This is only my personal view, maybe it will help in finding the best solution.

Thanks,
wisodev
Martin H
Feb 28 2007, 10:28
It's official now - I'm completely retarded it seems crying.gif Please forgive me for posting wrong information and thanks to wisodev for his recommendations, which i now can see by rereading the documentation, that he is perfectly right in. When i previously said that all "/Qx(x)" switches wouldn't work on AMD's, then that was not true, since two of them does work i.e. "/QxK" and "/QxW" as wisodev also said.

Again, please accept my sincere apology for this error of mine smile.gif

CU, Martin.
jcoalson
Feb 28 2007, 17:43
ok, here's mine. I used testtweaks.zip for the run. also I've cut out the flavors that didn't run.

(P4 1.8GHz, 768MB RAM)

CODE
encode:
flac-1.1.3-stock                  Global Time  =   298.469 = 00:04:58.469 = 100%
flac-1.1.4-IC9fastStatic          Global Time  =   182.672 = 00:03:02.672 = 100%
flac-1.1.4-IC9fastStaticOpt       Global Time  =   214.359 = 00:03:34.359 = 100%
flac-1.1.4-IC9fastStaticQaxW      Global Time  =   205.047 = 00:03:25.047 = 100%
flac-1.1.4-IC9fastStaticQxN       Global Time  =   204.203 = 00:03:24.203 = 100%
flac-1.1.4-stock                  Global Time  =   252.922 = 00:04:12.922 = 100%
flac-1.1.4-VS2005                 Global Time  =   210.313 = 00:03:30.313 = 100%
flac-1.1.4-VS2005fast             Global Time  =   217.141 = 00:03:37.141 = 100%
flac-1.1.4-VS2005fastStatic       Global Time  =   204.609 = 00:03:24.609 = 100%
flac-1.1.4-VS2005fastStaticNoSse  Global Time  =   208.578 = 00:03:28.578 = 100%
flac-1.1.4-VS2005fastStaticOptSse Global Time  =   251.312 = 00:04:11.312 = 100%
flac-1.1.4-VS2005opt              Global Time  =   253.406 = 00:04:13.406 = 100%

decode:
flac-1.1.3-stock                  Global Time  =    85.515 = 00:01:25.515 = 100%
flac-1.1.4-IC9fastStatic          Global Time  =    58.297 = 00:00:58.297 = 100%
flac-1.1.4-IC9fastStaticOpt       Global Time  =    58.438 = 00:00:58.438 = 100%
flac-1.1.4-IC9fastStaticQaxW      Global Time  =    60.906 = 00:01:00.906 = 100%
flac-1.1.4-IC9fastStaticQxN       Global Time  =    60.516 = 00:01:00.516 = 100%
flac-1.1.4-stock                  Global Time  =    69.766 = 00:01:09.766 = 100%
flac-1.1.4-VS2005                 Global Time  =    65.188 = 00:01:05.188 = 100%
flac-1.1.4-VS2005fast             Global Time  =    67.360 = 00:01:07.360 = 100%
flac-1.1.4-VS2005fastStatic       Global Time  =    70.391 = 00:01:10.391 = 100%
flac-1.1.4-VS2005fastStaticNoSse  Global Time  =    67.188 = 00:01:07.188 = 100%
flac-1.1.4-VS2005fastStaticOptSse Global Time  =    71.125 = 00:01:11.125 = 100%
flac-1.1.4-VS2005opt              Global Time  =    73.985 = 00:01:13.985 = 100%


what non-default switches did you use for the vs2005 compiles? I want to try and duplicate your gains but my builds are all at least 10-20% slower that with VC6.

Josh
Synthetic Soul
Mar 1 2007, 03:53
Following wisodev's (much valued) input, can we maybe get a /QxK build for the Athlon XP's please?

I think my testing is going to have to be done over the weekend, when I know no-one will touch the machine. I will test on my Athlon XP and Turion 64 Mobile ML-36 if possible.
gharris999
Mar 1 2007, 13:20
QUOTE(jcoalson @ Feb 28 2007, 16:43) *

what non-default switches did you use for the vs2005 compiles? I want to try and duplicate your gains but my builds are all at least 10-20% slower that with VC6.

Josh

Josh: I've posted what I hope are the appropriate files at http://www.hegardtfoundation.org/flacstuff/VSOpts.zip

This was the command I used to gather up the option files. Please let me know if you need more.

CODE

for /D %I in (flac-1.1.4-VS*) do robocopy .\%I c:\temp\VSOpts\%I /s *.sln *.suo *.dsp *.vcproj

gharris999
Mar 1 2007, 15:59
QUOTE(wisodev @ Feb 28 2007, 09:03) *

This is only my personal view, maybe it will help in finding the best solution.

I'm going to try to include wisodev's suggestions into my next compile, but it's time for me to write a real makefile and stop building within the VS2005 IDE. Anyone know how to easily translate a SLN and VCPROJ and ICPROJ files into a makefile?


Funkdude
Mar 1 2007, 16:29
I have a Prescott P4 3.00ghz (HTT, MMX, SSE, SSE2, SSE3) and oddly the fastest compiles were the non-optimized stock and rarewares compiles. huh.gif Ran the tests twice BTW, got similar results on both, here are the results of the first run. Used -8 BTW, because that's what I use on my music.

testflac
CODE

FLAC tests on EULYSSES-1A39C6
flac-1.1.4-IC9 Flac and MetaFlac encode test: 864.906
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 880.562
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 928.281
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 900.812
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 862.235
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 1513.984
flac-1.1.4-rw Flac and MetaFlac encode test: 925.156
flac-1.1.4-stock Flac and MetaFlac encode test: 960.860
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 944.640
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 929.859
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 919.547
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 915.172
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 956.953
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 956.063
flac-1.1.4-IC9 decode test: 55.219
flac-1.1.4-IC9fast decode test: 57.610
flac-1.1.4-IC9fastStatic decode test: 58.172
flac-1.1.4-IC9fastStaticOpt decode test: 59.468
flac-1.1.4-IC9fastStaticQaxW decode test: 59.391
flac-1.1.4-IC9fastStaticQxN decode test: 59.547
flac-1.1.4-rw decode test: 65.172
flac-1.1.4-stock decode test: 66.266
flac-1.1.4-VS2005 decode test: 70.188
flac-1.1.4-VS2005fast decode test: 62.890
flac-1.1.4-VS2005fastStatic decode test: 62.563
flac-1.1.4-VS2005fastStaticNoSse decode test: 64.297
flac-1.1.4-VS2005fastStaticOptSse decode test: 62.765
flac-1.1.4-VS2005opt decode test: 65.047
28/02/2007 11:14 AM 510,590,159 test.flac
28/02/2007 11:29 AM 510,590,163 test.flac
28/02/2007 11:45 AM 510,590,163 test.flac
28/02/2007 12:00 PM 510,590,172 test.flac
28/02/2007 12:14 PM 510,590,172 test.flac
28/02/2007 12:39 PM 510,590,172 test.flac
28/02/2007 12:55 PM 510,590,155 test.flac
28/02/2007 01:11 PM 510,590,155 test.flac
28/02/2007 01:26 PM 510,590,164 test.flac
28/02/2007 01:42 PM 510,590,168 test.flac
28/02/2007 01:57 PM 510,590,174 test.flac
28/02/2007 02:12 PM 510,590,174 test.flac
28/02/2007 02:28 PM 510,590,174 test.flac
28/02/2007 02:44 PM 510,590,155 test.flac


timeflac
CODE

FLAC tests on EULYSSES-1A39C6
*
flac-1.1.4-IC9 encode testing
Global Time = 766.125 = 00:12:46.125 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 771.281 = 00:12:51.281 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 919.656 = 00:15:19.656 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 724.203 = 00:12:04.203 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 706.016 = 00:11:46.016 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 703.703 = 00:11:43.703 = 100%
*
flac-1.1.4-rw encode testing
Global Time = 683.297 = 00:11:23.297 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 683.656 = 00:11:23.656 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 822.203 = 00:13:42.203 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 823.297 = 00:13:43.297 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 817.047 = 00:13:37.047 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 816.000 = 00:13:36.000 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 898.781 = 00:14:58.781 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 931.672 = 00:15:31.672 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 68.000 = 00:01:08.000 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 67.703 = 00:01:07.703 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 66.890 = 00:01:06.890 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 68.500 = 00:01:08.500 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 66.234 = 00:01:06.234 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 67.093 = 00:01:07.093 = 100%
*
flac-1.1.4-rw decode testing
Global Time = 74.640 = 00:01:14.640 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 72.032 = 00:01:12.032 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 82.781 = 00:01:22.781 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 65.391 = 00:01:05.391 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 63.344 = 00:01:03.344 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 65.031 = 00:01:05.031 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 63.328 = 00:01:03.328 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 64.688 = 00:01:04.688 = 100%
*
*
28/02/2007 10:47 AM 510,590,171 test.flac
28/02/2007 10:47 AM 510,590,175 test.flac
28/02/2007 10:47 AM 510,590,175 test.flac
28/02/2007 10:47 AM 510,590,184 test.flac
28/02/2007 10:47 AM 510,590,184 test.flac
28/02/2007 10:47 AM 510,590,184 test.flac
28/02/2007 10:47 AM 510,590,167 test.flac
28/02/2007 10:47 AM 510,590,167 test.flac
28/02/2007 10:47 AM 510,590,176 test.flac
28/02/2007 10:47 AM 510,590,180 test.flac
28/02/2007 10:47 AM 510,590,186 test.flac
28/02/2007 10:47 AM 510,590,186 test.flac
28/02/2007 10:47 AM 510,590,186 test.flac
28/02/2007 10:47 AM 510,590,167 test.flac
*
*
Comparing files test.wav and FLAC-1.1.4-IC9\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-RW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-STOCK\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: no differences encountered


Edit: just noticed that stock was the slowest on timeflac but fastest on testflac huh.gif
gharris999
Mar 1 2007, 20:34
QUOTE(Funkdude @ Mar 1 2007, 15:29) *

just noticed that stock was the slowest on timeflac but fastest on testflac huh.gif

Yeah. I can't explain that either. I saw strange anomalies like that in my own testing. Double huh. Did you use the testtweaks cmd files mentioned above?
Funkdude
Mar 1 2007, 20:39
QUOTE(gharris999 @ Mar 1 2007, 21:34) *

QUOTE(Funkdude @ Mar 1 2007, 15:29) *

just noticed that stock was the slowest on timeflac but fastest on testflac huh.gif

Yeah. I can't explain that either. I saw strange anomalies like that in my own testing. Double huh. Did you use the testtweaks cmd files mentioned above?


Nope. Just downloaded them, starting a a new test run with these for the night. Results to come tomorrow.
Synthetic Soul
Mar 4 2007, 02:04
Results for my Althlon XP (CPUZ report)

CODE
Name            AMD Athlon XP
Codename        Thoroughbred
Specification   AMD Athlon(tm) XP 2400+
Package         Socket A (462)

Build                                   Enc       Dec       Both
================================================================
flac-1.1.4-IC9fastStatic             96.038    45.215    141.253
flac-1.1.4-IC9fastStaticQaxW         97.82     44.884    142.704
flac-1.1.4-IC9fast                   97.471    45.596    143.067
flac-1.1.4-IC9                       99.082    45.335    144.417
flac-1.1.4-VS2005                    96.399    48.36     144.759
flac-1.1.4-VS2005fast               102.447    44.694    147.141
flac-1.1.4-stock                    107.624    45.806    153.43
flac-1.1.4-VS2005fastStaticNoSse    108.336    45.296    153.632
flac-1.1.4-VS2005fastStatic         110.94     45.185    156.125
flac-1.1.3-IC9sseW                  110.88     48.71     159.59
flac-1.1.3-stock                    120.343    51.044    171.387

Results are sorted by the combined encode/decode time. Test was conducted with The Stranglers Greatest Hits 1977 - 1990 (548MB WAVE). Only the builds that run on the XP are listed.
Synthetic Soul
Mar 4 2007, 14:26
Results for my AMD Turion 64 laptop (CPUZ report)

CODE
Name             AMD Turion 64 Mobile ML-36
Codename         Taylor
Specification    AMD Turion(tm) 64 Mobile Technology MK-36
Package          Socket S1 (638)

Build                                    Enc       Dec       Both
=================================================================
flac-1.1.4-IC9fastStaticOpt           80.812    59.046    139.858
flac-1.1.4-IC9fastStaticQaxW          82.687    57.656    140.343
flac-1.1.4-VS2005fastStaticNoSse      83.906    57.093    140.999
flac-1.1.4-IC9fast                    78.609    62.594    141.203
flac-1.1.4-VS2005fastStatic           83.219    66.922    150.141
flac-1.1.4-VS2005fast                 83.578    69.625    153.203
flac-1.1.4-stock                      93.594    61.657    155.251
flac-1.1.3-stock                     100.812    68.000    168.812
flac-1.1.4-VS2005                    109.937    61.922    171.859
flac-1.1.4-IC9fastStatic             110.594    64.906    175.500
flac-1.1.4-IC9                       104.563    72.078    176.641
flac-1.1.4-VS2005opt                 129.265    58.125    187.390
flac-1.1.4-VS2005fastStaticOptSse    131.079    62.281    193.360

Results are sorted by the combined encode/decode time. Test was conducted with The Stranglers Greatest Hits 1977 - 1990 (548MB WAVE).
gharris999
Mar 7 2007, 15:19
QUOTE(wisodev @ Feb 28 2007, 09:03) *

Are this P4 build using PGO (/Qprof_genx and /Qprof_use switches) and IPO (/Qipo switch) optimizations? The PGO and IPO optimizations (used in Intel C++ compiler) are providing bigest speedups.

For AMD CPUs with SSE I'd recommend using /QxK switch, for AMD CPUs with SSE2 I'd recommend using /QxW switch.

For ICL builds I'd recommend using this switches (of course in pair with the /QxK or /QxW switches):
CODE
/GA /Ox /Og /Ob1 /Ot

and specific to version of compiler, with 9.1 you should use:
CODE
/Qipo4 /fp:fast /Qprec-div- /Qcomplex-limited-range

I use this settings for my Aften and Flake builds and it seems to be the best configuration.

This is only my personal view, maybe it will help in finding the best solution.

Thanks,
wisodev

OK, I've tried to incorporate most of these suggestions in a couple of new versions.

http://www.hegardtfoundation.org/flacstuff...-IC9wisodev.zip
http://www.hegardtfoundation.org/flacstuff...IC9wisodevK.zip

In the flac-1.1.4-IC9wisodev.zip version, you'll find the .mak make files I used to compile and link that version. Suggestions welcome.

I've also posted http://www.hegardtfoundation.org/flacstuff...aticSSEQaxW.zip for comparison sake. These versions include some changes found in the FLAC CVS over the past week or so.



JWolf
Mar 8 2007, 20:45
Why is it some of the builds make FLAC files that are larger then FLAC 1.1.4?
gharris999
Mar 9 2007, 01:47
QUOTE(JWolf @ Mar 8 2007, 19:45) *

Why is it some of the builds make FLAC files that are larger then FLAC 1.1.4?

I don't know Josh's code well enough to say why exactly. But superficially, any compile made with the "FLAC__SSE_OS" define turned on produces slightly larger flac files (~1.5k for a whole CD in my experience.) I've viewed that as a minor size penalty for the speed increase that seems to go with SSE support.

Josh has introduced some new code recently into the CVS that seems to do SSE OS support checking making "FLAC__SSE_OS" unnecessary. The next batch 'o compiles I'll make will not use that define. We'll see what effect that has on size 'n speed.
jcoalson
Mar 9 2007, 02:11
yes, you've found it. after coming up empty searching through other projects, I finally figured out a tricky way to do thread-safe SSE OS detection at runtime, so in the future FLAC__SSE_OS will not be needed anymore. but it is not tested too well, let me know if it's crashy.

SSE floating point is 64-bit (not 80-bit like the FPU) and not IEEE compliant, so there are small differences in the constructed filter kernel, hence the tiny size difference.

Josh
Funkdude
Mar 9 2007, 13:56
New results with the test tweaks and the wisodev compile. The wisodev compile is a clear winner on my machine, but stock and rw are still the follow-ups, with all the other compiles slower. Results this time around are consistent across both timeflac and testflac.

Prescott P4 3.00ghz (HTT, MMX, SSE, SSE2, SSE3), 1024mb RAM, used -8

CODE

FLAC tests on EULYSSES-1A39C6
flac-1.1.4-IC9 Flac and MetaFlac encode test: 949.046
flac-1.1.4-IC9fast Flac and MetaFlac encode test: 898.750
flac-1.1.4-IC9fastStatic Flac and MetaFlac encode test: 895.344
flac-1.1.4-IC9fastStaticOpt Flac and MetaFlac encode test: 828.562
flac-1.1.4-IC9fastStaticQaxW Flac and MetaFlac encode test: 826.829
flac-1.1.4-IC9fastStaticQxN Flac and MetaFlac encode test: 824.578
flac-1.1.4-IC9wisodevK Flac and MetaFlac encode test: 698.171
flac-1.1.4-rw Flac and MetaFlac encode test: 815.938
flac-1.1.4-stock Flac and MetaFlac encode test: 820.750
flac-1.1.4-VS2005 Flac and MetaFlac encode test: 967.485
flac-1.1.4-VS2005fast Flac and MetaFlac encode test: 970.109
flac-1.1.4-VS2005fastStatic Flac and MetaFlac encode test: 961.250
flac-1.1.4-VS2005fastStaticNoSse Flac and MetaFlac encode test: 961.235
flac-1.1.4-VS2005fastStaticOptSse Flac and MetaFlac encode test: 1002.469
flac-1.1.4-VS2005opt Flac and MetaFlac encode test: 1013.969
flac-1.1.4-IC9 decode test: 75.579
flac-1.1.4-IC9fast decode test: 58.906
flac-1.1.4-IC9fastStatic decode test: 61.891
flac-1.1.4-IC9fastStaticOpt decode test: 64.969
flac-1.1.4-IC9fastStaticQaxW decode test: 62.062
flac-1.1.4-IC9fastStaticQxN decode test: 63.359
flac-1.1.4-IC9wisodevK decode test: 67.704
flac-1.1.4-rw decode test: 70.641
flac-1.1.4-stock decode test: 71.125
flac-1.1.4-VS2005 decode test: 67.813
flac-1.1.4-VS2005fast decode test: 67.985
flac-1.1.4-VS2005fastStatic decode test: 71.390
flac-1.1.4-VS2005fastStaticNoSse decode test: 67.625
flac-1.1.4-VS2005fastStaticOptSse decode test: 68.110
flac-1.1.4-VS2005opt decode test: 68.188
08/03/2007 11:25 PM 510,590,159 test.flac
08/03/2007 11:40 PM 510,590,163 test.flac
08/03/2007 11:55 PM 510,590,163 test.flac
09/03/2007 12:09 AM 510,590,172 test.flac
09/03/2007 12:23 AM 510,590,172 test.flac
09/03/2007 12:36 AM 510,590,172 test.flac
09/03/2007 12:48 AM 510,590,166 test.flac
09/03/2007 01:02 AM 510,590,155 test.flac
09/03/2007 01:15 AM 510,590,155 test.flac
09/03/2007 01:32 AM 510,590,164 test.flac
09/03/2007 01:48 AM 510,590,168 test.flac
09/03/2007 02:04 AM 510,590,174 test.flac
09/03/2007 02:20 AM 510,590,174 test.flac
09/03/2007 02:36 AM 510,590,174 test.flac
09/03/2007 02:53 AM 510,590,155 test.flac


CODE

FLAC tests on EULYSSES-1A39C6
*
flac-1.1.4-IC9 encode testing
Global Time = 816.860 = 00:13:36.860 = 100%
*
flac-1.1.4-IC9fast encode testing
Global Time = 817.859 = 00:13:37.859 = 100%
*
flac-1.1.4-IC9fastStatic encode testing
Global Time = 814.516 = 00:13:34.516 = 100%
*
flac-1.1.4-IC9fastStaticOpt encode testing
Global Time = 743.937 = 00:12:23.937 = 100%
*
flac-1.1.4-IC9fastStaticQaxW encode testing
Global Time = 741.297 = 00:12:21.297 = 100%
*
flac-1.1.4-IC9fastStaticQxN encode testing
Global Time = 742.718 = 00:12:22.718 = 100%
*
flac-1.1.4-IC9wisodevK encode testing
Global Time = 607.469 = 00:10:07.469 = 100%
*
flac-1.1.4-rw encode testing
Global Time = 732.203 = 00:12:12.203 = 100%
*
flac-1.1.4-stock encode testing
Global Time = 726.672 = 00:12:06.672 = 100%
*
flac-1.1.4-VS2005 encode testing
Global Time = 876.469 = 00:14:36.469 = 100%
*
flac-1.1.4-VS2005fast encode testing
Global Time = 877.625 = 00:14:37.625 = 100%
*
flac-1.1.4-VS2005fastStatic encode testing
Global Time = 875.703 = 00:14:35.703 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse encode testing
Global Time = 872.953 = 00:14:32.953 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse encode testing
Global Time = 917.781 = 00:15:17.781 = 100%
*
flac-1.1.4-VS2005opt encode testing
Global Time = 920.750 = 00:15:20.750 = 100%
*
*
*
flac-1.1.4-IC9 decode testing
Global Time = 64.375 = 00:01:04.375 = 100%
*
flac-1.1.4-IC9fast decode testing
Global Time = 77.891 = 00:01:17.891 = 100%
*
flac-1.1.4-IC9fastStatic decode testing
Global Time = 61.750 = 00:01:01.750 = 100%
*
flac-1.1.4-IC9fastStaticOpt decode testing
Global Time = 67.094 = 00:01:07.094 = 100%
*
flac-1.1.4-IC9fastStaticQaxW decode testing
Global Time = 63.328 = 00:01:03.328 = 100%
*
flac-1.1.4-IC9fastStaticQxN decode testing
Global Time = 75.141 = 00:01:15.141 = 100%
*
flac-1.1.4-IC9wisodevK decode testing
Global Time = 65.047 = 00:01:05.047 = 100%
*
flac-1.1.4-rw decode testing
Global Time = 70.500 = 00:01:10.500 = 100%
*
flac-1.1.4-stock decode testing
Global Time = 80.906 = 00:01:20.906 = 100%
*
flac-1.1.4-VS2005 decode testing
Global Time = 66.594 = 00:01:06.594 = 100%
*
flac-1.1.4-VS2005fast decode testing
Global Time = 68.235 = 00:01:08.235 = 100%
*
flac-1.1.4-VS2005fastStatic decode testing
Global Time = 75.734 = 00:01:15.734 = 100%
*
flac-1.1.4-VS2005fastStaticNoSse decode testing
Global Time = 66.984 = 00:01:06.984 = 100%
*
flac-1.1.4-VS2005fastStaticOptSse decode testing
Global Time = 67.812 = 00:01:07.812 = 100%
*
flac-1.1.4-VS2005opt decode testing
Global Time = 68.047 = 00:01:08.047 = 100%
*
*
28/02/2007 10:47 AM 510,590,171 test.flac
28/02/2007 10:47 AM 510,590,175 test.flac
28/02/2007 10:47 AM 510,590,175 test.flac
28/02/2007 10:47 AM 510,590,184 test.flac
28/02/2007 10:47 AM 510,590,184 test.flac
28/02/2007 10:47 AM 510,590,184 test.flac
28/02/2007 10:47 AM 510,590,178 test.flac
28/02/2007 10:47 AM 510,590,167 test.flac
28/02/2007 10:47 AM 510,590,167 test.flac
28/02/2007 10:47 AM 510,590,176 test.flac
28/02/2007 10:47 AM 510,590,180 test.flac
28/02/2007 10:47 AM 510,590,186 test.flac
28/02/2007 10:47 AM 510,590,186 test.flac
28/02/2007 10:47 AM 510,590,186 test.flac
28/02/2007 10:47 AM 510,590,167 test.flac
*
*
Comparing files test.wav and FLAC-1.1.4-IC9\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICOPT\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQAXW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9FASTSTATICQXN\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-IC9WISODEVK\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-RW\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-STOCK\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FAST\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATIC\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICNOSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005FASTSTATICOPTSSE\TEST.WAV
FC: no differences encountered

Comparing files test.wav and FLAC-1.1.4-VS2005OPT\TEST.WAV
FC: no differences encountered

a_aa
Mar 12 2007, 15:52
Tested with new script and included the new builds, on the same old setup:

CPU: Intel Pentium 4 3.2 GHz (Northwood, FSB800, HTT, MMX, SSE, SSE2)
HDD: PATA100, (IDE) Samsung Spinpoint 160 GB, almost unfragmented
RAM: 4 x 512 MB DDR400 (timings: 2,3,2,6,2T)
MoBo: Gigabyte GA-8ISXT-FS (FS=Fujitsu Siemens)
NB: Intel i865P/PE/G/i848P rev. A2, SB: Intel 82801EB (ICH5) rev. 02

OS: MS XP Home SP2

I've tested builds with both -5 and -8, and rated them according to total time (testflac5+timeflac5+testflac8+timeflac8, encoding+metadata+decoding).

CODE
test.wav: 3792.893 s (1:03:12.893), 638 MB (669 066 428 byte)


testflac + timeflac results || Enc & Meta[X] | Decode #1[X] || Encode[X] | Decode #2[X] |
----------------------------------------------------------------------------------------------------
FLAC 1.1.4-build | Tot [s]|| -5 | -8 | -5 | -8 || -5 | -8 | -5 | -8 |
----------------------------------------------------------------------------------------------------
IC9FastStaticSSEQaxW | 719,454|| 41,9X| 18,7X| 77,4X| 87,6X|| 62,3X| 21,7X| 76,4X| 78,0X|
IC9wisodev | 724,139|| 40,6X| 18,4X| 77,8X| 84,7X|| 61,5X| 21,7X| 79,2X| 80,5X|
IC9fastStaticQxN | 731,125|| 42,7X| 17,9X| 84,3X| 94,3X|| 61,3X| 20,5X| 77,8X| 76,7X|
IC9wisodevK | 734,515|| 40,6X| 18,3X| 74,8X| 73,1X|| 61,6X| 21,7X| 80,6X| 79,9X|
IC9fastStaticQaxW | 739,639|| 41,8X| 17,7X| 93,8X| 89,9X|| 61,1X| 20,3X| 75,2X| 71,9X|
IC9fastStaticOpt | 761,609|| 40,3X| 17,7X| 77,1X| 77,4X|| 60,7X| 20,4X| 71,4X| 71,0X|
IC9fastStatic | 819,361|| 36,2X| 16,5X| 75,5X| 72,1X|| 55,0X| 19,3X| 65,9X| 64,7X|
VS2005fast | 820,141|| 35,8X| 15,3X| 80,6X| 100,4X|| 49,5X| 17,4X| 96,5X| 79,3X|
VS2005fastStatic | 824,642|| 35,2X| 15,3X| 87,7X| 90,8X|| 49,6X| 17,5X| 79,9X| 86,3X|
VS2005 | 831,374|| 34,9X| 15,3X| 83,5X| 91,3X|| 47,4X| 17,4X| 79,8X| 88,9X|
VS2005fastStaticNoSse | 833,670|| 34,8X| 15,2X| 91,4X| 80,4X|| 47,2X| 17,4X| 79,9X| 93,4X|
VS2005fastStaticOptSse | 858,546|| 33,0X| 14,7X| 84,5X| 83,2X|| 43,9X| 16,5X| 98,1X| 92,5X|
VS2005opt | 860,594|| 31,9X| 14,5X| 100,8X| 90,3X|| 43,3X| 16,3X| 96,6X| 91,2X|
IC9 | 868,500|| 27,5X| 14,4X| 68,6X| 65,2X|| 48,8X| 19,2X| 89,0X| 105,0X|
stock | 885,796|| 29,0X| 15,1X| 93,3X| 79,9X|| 37,1X| 17,1X| 79,8X| 84,9X|
IC9fast | 946,125|| 30,4X| 15,1X| 54,1X| 54,5X|| 54,6X| 19,2X| 46,9X| 46,1X|
All the 3 new builds are among the top five which distinguishes themselves from the rest of the pack. There are still some peculiarities (e.g. IC9-decoding), but I guess adding it all up may even things out.

Great job, thanks!


Asbjørn

PS: I may do a last upgrade to P4 Gallatin on my Skt478 board - if I do, I'll run these test again.
a_aa
Mar 26 2007, 14:29
QUOTE(a_aa @ Mar 12 2007, 22:52) *
PS: I may do a last upgrade to P4 Gallatin on my Skt478 board - if I do, I'll run these test again.

Done. Same setup as above except CPU which is a P4 3.4 GHz Extreme Editition (Gallatin, FSB800, HTT, MMX, SSE, SSE2).

CODE
test.wav: 3792,893 s (1:03:12,893), 638 MB (669 066 428 byte)


testflac + timeflac results || Enc & Meta[X] | Decode #1[X] || Encode[X] | Decode #2[X] |
-------------------------------------------------------------------------------------------------