HydrogenAudio

Lossless Audio Compression => WavPack => Topic started by: wisodev on 2006-04-18 11:41:49

Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-18 11:41:49
NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.
Patched WavPack 4.31 sources (not the 4.32 sources for linux) with MMX optimizations posted by he-jo (http://www.hydrogenaudio.org/forums/index.php?showtopic=43617) with converted GCC specific MMX INTRINSICS to Microsoft Specific MMX INTRINSICS. This sources are for Visual C++ .NET 2003 (solution and project for WavPack encoder included).

[attachment=2177:attachment]

Windows binarys (Win32) of original and MMX optimized (Miscrosoft Visual C++ .NET 2003 build).

[attachment=2178:attachment]
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: rjamorim on 2006-04-18 12:00:26
Thank-you very much

Wouldn't it be nicer to just give up intrinsics and go for NASM?
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: Destroid on 2006-04-18 12:01:43
I'm sorry, what is the difference between Wavpack_mmx.EXE and Wavpack_org.EXE?
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-18 12:05:01
Quote
I\'m sorry, what is the difference between Wavpack_mmx.EXE and Wavpack_org.EXE?



Wavpack_mmx.EXE – optimized MMX build (he-jo optimizations converted by me to MS specific)
Wavpack_org.EXE – original 4.31 build
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: The Gaby on 2006-04-18 12:12:44
I'm sorry, what is the difference between Wavpack_mmx.EXE and Wavpack_org.EXE?


I haven´t seen them, but I guess:

Wavpack_org.EXE = Wavpack Original Version
Wavpack_mmx.EXE = Wavpack MMX Optimized Version
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-18 12:17:41
Quote
Thank-you very much

Wouldn\'t it be nicer to just give up intrinsics and go for NASM?


I have plain assembly version but my code (working well) was slower then intrinsics version.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: rjamorim on 2006-04-18 12:36:04
I have plain assembly version but my code (working well) was slower then intrinsics version.


Ah, that's curious.

That's a pity, NASM would allow for some compiler independency.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-18 12:46:09
Quote
Quote
I have plain assembly version but my code (working well) was slower then intrinsics version.


Ah, that\\\'s curious.

That\\\'s a pity, NASM would allow for some compiler independency.


My first goal was to create independent asm code, but I failed to write code faster then intrinsics version and finally I ended with this one. I will soon try again to write NASM version. I am not so experienced with asm programming, so do not blame me if something is wrong here.

It would be nice if someone can post speed comparison test results.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: he-jo on 2006-04-18 20:08:09
I'm also curious how it performs on different processors. Please test the extra modes, e.g. 'wavpack -h -x6'

Btw: You can additionally apply the following patch, which avoids a multiplication in the critical loops that are not MMX optimised yet. This gives a minor speedup in all modes.

Code: [Select]
--- wavpack.h
+++ wavpack.h
@@ -415,7 +415,7 @@

#if 1    // PERFCOND
#define update_weight(weight, delta, source, result) \
-    if (source && result) weight -= ((((source ^ result) >> 30) & 2) - 1) * delta;
+    if (source && result) { int32_t s = (int32_t) (source ^ result) >> 31; weight = (weight - s) + (s ^ delta); }
#else
#define update_weight(weight, delta, source, result) \
     if (source && result) (source ^ result) < 0 ? (weight -= delta) : (weight += delta);
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: Destroid on 2006-04-18 20:58:04
Quote

I\'m sorry, what is the difference between Wavpack_mmx.EXE and Wavpack_org.EXE?



Wavpack_mmx.EXE – optimized MMX build (he-jo optimizations converted by me to MS specific)
Wavpack_org.EXE – original 4.31 build


Thanks for clarifying for me.

A quick test using -h -x6:
original = 0.37x
mmx = 0.46x

Original and MMX files were identical using binary comparison. Good news!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-19 15:14:35
Quote
Quote

Quote

I\\\'m sorry, what is the difference between Wavpack_mmx.EXE and Wavpack_org.EXE?



Wavpack_mmx.EXE – optimized MMX build (he-jo optimizations converted by me to MS specific)
Wavpack_org.EXE – original 4.31 build


Thanks for clarifying for me.

A quick test using -h -x6:
original = 0.37x
mmx = 0.46x

Original and MMX files were identical using binary comparison. Good news!


Well 24 % speedup!
But more testing is required.
From my tests the files where always same as original for all -f -x1..6 modes and for -x1..6 and -h -x modes to.


Quote
I\'m also curious how it performs on different processors. Please test the extra modes, e.g. \'wavpack -h -x6\'

Btw: You can additionally apply the following patch, which avoids a multiplication in the critical loops that are not MMX optimised yet. This gives a minor speedup in all modes.

Code: [Select]
--- wavpack.h
+++ wavpack.h
@@ -415,7 +415,7 @@

#if 1    // PERFCOND
#define update_weight(weight, delta, source, result) \\
-    if (source && result) weight -= ((((source ^ result) >> 30) & 2) - 1) * delta;
+    if (source && result) { int32_t s = (int32_t) (source ^ result) >> 31; weight = (weight - s) + (s ^ delta); }
#else
#define update_weight(weight, delta, source, result) \\
     if (source && result) (source ^ result) < 0 ? (weight -= delta) : (weight += delta);



I will apply patch and post binarys here. I can test at PII 233Mhz , Celeron 466Mhz and Athlon XP 2000+.

I have made some changes (cleanups and reformatting, just for better look of code), this cleanups and patched code will be put here for download tomorrow.

I hope this will be useful for someone ;-)
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: PrakashP on 2006-04-19 15:27:51
I have plain assembly version but my code (working well) was slower then intrinsics version.


Ah, that's curious.

That's a pity, NASM would allow for some compiler independency.


I already pointed out in the original thread that using intrinsics is better and more portable (when done correctly) considering 32bit/64bit arch.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-19 15:36:30
Quote
Quote

Quote
I have plain assembly version but my code (working well) was slower then intrinsics version.


Ah, that\\\'s curious.

That\\\'s a pity, NASM would allow for some compiler independency.


I already pointed out in the original thread that using intrinsics is better and more portable (when done correctly) considering 32bit/64bit arch.


And how it can be done?

I am asking, because you have posted as below:

Quote
I did some \"magic\" in a header file to achieve this.


What precisely is the magic, I really would like to know!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-20 06:48:45
Quote
I already pointed out in the original thread that using intrinsics is better and more portable (when done correctly) considering 32bit/64bit arch.


Thanks, I checked the OpenAL MMX optimizations idea and implemented generally the same way at WavPack MMX optimized.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: pepoluan on 2006-04-20 11:32:29
Wow... even in the field of Lossless, we now witness an improvement... faster FLAC, faster WavPack, new yet-to-be-named Lossless encoder... 

Gee my Lossless Performance Test, almost complete, is now obsolescent, even before published... 

Oh well... I will post it anyway, with the version number clearly indicated, and also a MD5 hashes of the binaries to make sure that my encoders/decoders are truly what they are and not some tweaked-up version.

Maybe later on I will perform an amendment test with the new compressors with the exact same test corpus.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-20 12:16:15
NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.
Well as stated in previous posts, I have made some code cleanup and merged new he-jo patch (more info at wiso_changes.txt in src package).

Download updated binarys:
[attachment=2186:attachment]
Download updated sources:
[attachment=2187:attachment]

Do not use anymore  binarys and sources from post #1.

Below my test results (only tested one wav file) using new binarys!

NOTE: Tested on Athlon XP 2000+, 512MB RAM, 80GB SATA DISK (8 MB Cache), Windows XP SP2.

Code: [Select]
                                 Test file: 44kHz/16bit/Stereo, 34 435 676 bytes, 195 seconds

    WavPack 4.31 Options       Original 4.31 [s]     MMX Optimized 4.31 [s]      Difference [s]       Speedup [%]
             -f                      2,89                     2,92                    -0,03              -1,04
                                     3,41                     3,45                    -0,04              -1,17
             -h                      6,22                     6,13                    0,09               1,45
           -f -x                     32,27                   28,30                    3,97               12,30
           -f -x1                    6,34                     6,03                    0,31               4,89
           -f -x2                    9,44                     8,66                    0,78               8,26
           -f -x3                    9,75                     8,95                    0,80               8,21
           -f -x4                    20,55                   18,30                    2,25               10,95
           -f -x5                    26,77                   23,66                    3,11               11,62
           -f -x6                    32,30                   28,30                    4,00               12,38
             -x                      59,19                   50,42                    8,77               14,82
            -x1                      9,08                     8,45                    0,63               6,94
            -x2                      15,23                   13,63                    1,60               10,51
            -x3                      25,08                   22,03                    3,05               12,16
            -x4                      59,16                   50,33                    8,83               14,93
            -x5                      89,44                   75,69                    13,75              15,37
            -x6                     190,55                   159,42                   31,13              16,34
           -h -x                    144,02                   127,92                   16,10              11,18
           -h -x1                    18,95                   17,09                    1,86               9,82
           -h -x2                    33,95                   29,76                    4,19               12,34
           -h -x3                   144,03                   127,94                   16,09              11,17
           -h -x4                   229,61                   199,25                   30,36              13,22
           -h -x5                   331,49                   289,81                   41,68              12,57
           -h -x6                   730,89                   645,28                   85,61              11,71

NOTE: Original is Release Build from provided sources, and MMX Optimized is ReleaseMMX build.

Here is my batch file [test.cmd], used for testing (works on WinXP SP2):
Code: [Select]
@echo off

rem WavPack MMX OPTIMIZED SPEED TESTS by WISO

rem COMMANDLINE OPTIONS FROM FILE (each line is one test)
set OptionsFile=options.txt

rem INPUT (*.wav) TEST FILE
set InFile=test.wav

rem OUTPUT (*.wv) TEST FILES
set OutFileORG=T_ORG.wv
set OutFileMMX=T_MMX.wv

rem PATH TO ORIGINAL EXE
set ExeFileORG=wavpack_ORG.exe

rem PATH TO MMX OPTIMIZED EXE
set ExeFileMMX=wavpack_MMX.exe

rem RUN SPEED TESTS
FOR /F \"tokens=*\" %%i in (%OptionsFile%) do (
  @echo TESTING: %%i
  %ExeFileORG% %%i %InFile% %OutFileORG%
  %ExeFileMMX% %%i %InFile% %OutFileMMX%
  fc /B %OutFileORG% %OutFileMMX%
  del %OutFileORG%
  del %OutFileMMX%
  @echo ################################################################################
)

echo ALL TESTS DONE
pause


and [options.txt] file, note that empty line is not empty, there is one ASCII SPACE char:
Code: [Select]
-f

-h
-f -x
-f -x1
-f -x2
-f -x3
-f -x4
-f -x5
-f -x6
-x
-x1
-x2
-x3
-x4
-x5
-x6
-h -x
-h -x1
-h -x2
-h -x3
-h -x4
-h -x5
-h -x6


Thats all,
WISO (wisodev @ HA.org).
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: waileongyeo on 2006-04-20 15:43:39
Erm... Thanks for the mmx optimization mod. I've some QUICK test results.

Input file: "8_Channel_Sound.wav"  (~ 8 secs)
-------------------------------------------
bitrate = 9216
samplerate = 48000
channels = 8
codec = PCM
encoding = lossless
bitspersample = 24
386383 samples @ 48000Hz
File size: 9,273,236 Bytes (8.84 MB)
-------------------------------------------

Command switch used: -h -x6
Code: [Select]
[Build version]   [Time (s)] [Speed Increased %]
org (wisodev)     24.94       0 (All others build are normalized to this)
mmx (wisodev)     19.41       22.17
mmx_v2 (wisodev)  19.00       23.82
mmx (wly)         23.00       7.78
mmx_v2 (wly)      24.58       1.44

org (wisodev) - Original build (from 'wavpack_4.31_bin_mmx.zip')
mmx (wisodev) - MMX optimized build (from 'wavpack_4.31_bin_mmx.zip')
mmx_v2 (wisodev) - MMX optimized build version 2 (from 'wavpack_4.31_bin_mmx_wiso.zip')
mmx (wly) - My build, MMX optimized build base on 'wavpack_4.31_src_mmx.zip' using VC2005
mmx_v2 (wly) - My build, MMX optimized build base on 'wavpack_4.31_src_mmx_wiso.zip' using VC2005

wisodev's builds are working fine but...
What's wrong with my builds?!?!?!?!?

My computer spec:
Pentium 4 - 2.4GHz
RAM 756MB
WinXP SP2
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-04-22 09:32:03
wisodev's builds are working fine but...
What's wrong with my builds?!?!?!?!?

My computer spec:
Pentium 4 - 2.4GHz
RAM 756MB
WinXP SP2


Maybe you have selected Release from Build configurations, you need to select ReleaseMMX to build MMX optimized version.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: waileongyeo on 2006-04-24 12:01:26
Maybe you have selected Release from Build configurations, you need to select ReleaseMMX to build MMX optimized version.

Oopps... My bad. That's my mistake.
Thanks for the tips.

However, the binary is only 9~10% faster (sound reasonable now) under the build of VS8. You build (VC7.1) is the king!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-05 20:03:24
NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.
OK, included latest bsr ASM optimizations by he-jo (wavpack_MMX-BSR.exe) and done some quick test, all is included in file below (alsow the sources):

[attachment=2217:attachment]
For more information see this thread (http://www.hydrogenaudio.org/forums/index.php?act=ST&f=19&t=43226).
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-23 18:58:44
[!--sizeo:2--][span style=\"font-size:10pt;line-height:100%\"][!--/sizeo--]NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.[/size]
Latest optimazation (JFL2B) from he-jo (posted here (http://www.hydrogenaudio.org/forums/index.php?showtopic=43226&st=25&p=394165&#entry394165)), including sources:

[attachment=2238:attachment]

Binarys included :
Code: [Select]
Release\\wavpack.exe          - original 4.31
ReleaseBSR\\wavpack.exe       - original 4.31 + he-jo BSR ASM optimizations
ReleaseJFL2B\\wavpack.exe     - original 4.31 + he-jo JFL2B ASM optimizations
ReleaseMMX\\wavpack.exe       - original 4.31 + he-jo MMX optimizations
ReleaseMMX-BSR\\wavpack.exe   - original 4.31 + he-jo MMX optimizations + he-jo BSR ASM optimizations
ReleaseMMX-JFL2B\\wavpack.exe - original 4.31 + he-jo MMX optimizations + he-jo JFL2B ASM optimizations
My quick speed test:
Code: [Select]
Test file: 44kHz/16bit/Stereo, 34 435 676 bytes, 195 seconds
Tested on Athlon XP 2000+, 512MB RAM, 80GB SATA DISK (8 MB Cache), Windows XP SP2
Results are in seconds, numbers in (...) are in 100 nanosecod units.

OPTIONS: -f -x6

ORG:
Kernel Time  = 0.109 (1093750)
User Time = 32.109 (321093750)
Process Time = 32.218 (322187500)
Global Time  = 32.484 (324843750)

BSR: 3,30%
Kernel Time  = 0.156 (1562500)
User Time = 31.000 (310000000)
Process Time = 31.156 (311562500)
Global Time  = 31.359 (313593750)

JFL2B: 0,19%
Kernel Time  = 0.328 (3281250)
User Time = 31.828 (318281250)
Process Time = 32.156 (321562500)
Global Time  = 32.312 (323125000)

MMX: 12,07%
Kernel Time  = 0.296 (2968750)
User Time = 28.031 (280312500)
Process Time = 28.328 (283281250)
Global Time  = 28.515 (285156250)

MMX-BSR: 15,86%
Kernel Time  = 0.218 (2187500)
User Time = 26.890 (268906250)
Process Time = 27.109 (271093750)
Global Time  = 27.328 (273281250)

MMX-JFL2B: 12,75%
Kernel Time  = 0.203 (2031250)
User Time = 27.906 (279062500)
Process Time = 28.109 (281093750)
Global Time  = 28.312 (283125000)

On my Ahtlon XP JFL2B optimizations are slower then BSR optimazations. Maybe I have done something wrong, but it seems that all is OK with my builds. Output files are same for all builds (binary comparison).

EDIT: Codebox for test results.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: he-jo on 2006-05-23 22:18:12
Hm, this is really strange! But it seems, that you always used the same file for the tests you posted here. Since the speed difference of my asm routines to the original code depends on the data in the input file, it would be nice, if you could do some more testing. Please use tracks with different kinds of music, maybe some louder tracks?!

Thanks for your work anyway!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-24 05:41:56
Quote
Hm, this is really strange! But it seems, that you always used the same file for the tests you posted here. Since the speed difference of my asm routines to the original code depends on the data in the input file, it would be nice, if you could do some more testing. Please use tracks with different kinds of music, maybe some louder tracks?!

Thanks for your work anyway!


Yes, the same file was used in all tests. I was short on time and only used one file. This file was actually loud metal music. But if it really depends on input data, then I will do more testing on different music genres. Beside I know that testing one file is not enough ;-)
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: he-jo on 2006-05-24 08:24:33
Just to be sure - for the Celeron tests I used this file set from http://www.rarewares.org/test_samples/ (http://www.rarewares.org/test_samples/) :

41_30sec.wav Bartok_strings2.wav BigYellow.wav DaFunk.wav EnolaGay.wav Leahy.wav Mama.wav NewYorkCity.wav OrdinaryWorld.wav Quizas.wav SinceAlways.wav TheSource.wav Twelve.wav Waiting.wav bodyheat.wav rosemary.wav thear1.wav trust.wav

and three 24 bit files from the last chapter of this page: ttp://www.mytekdigital.com/compare/comparison1.htm

dBTech_122-96_24bit_web.wav mytek_8X96_24bit_web.wav prism_AD124_24bit_web.wav

Could you also do a test on this file set? I run the test with this command line:

timer wavpack -q -f -x6 "*.wav"
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-24 09:12:03
OK, this is very good suggestion to use same test files!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-26 07:40:56
I have tested the files (using suggested command-line with timer) that you have suggested (he-jo) on my two machines!

TESTED FILES:
Code: [Select]
41_30sec.wav
Bartok_strings2.wav
BigYellow.wav
bodyheat.wav
DaFunk.wav
dBTech_122-96_24bit_web.wav
EnolaGay.wav
Leahy.wav
Mama.wav
mytek_8X96_24bit_web.wav
NewYorkCity.wav
OrdinaryWorld.wav
prism_AD124_24bit_web.wav
Quizas.wav
rosemary.wav
SinceAlways.wav
thear1.wav
TheSource.wav
trust.wav
Twelve.wav
Waiting.wav
CELERON MACHINE SPECS:
Code: [Select]
OS:       Microsoft Windows XP Professional Service Pack 2
CPU:      Intel Celeron-A, 466 MHz (7 x 67)
MEMORY:   256 MB (SDRAM)
MOBO:     Intel 82440BX/ZX
VIDEO:    NVIDIA RIVA TNT (16 MB)
MONITOR:  Compaq V55 [15 inch CRT]
SOUND:    ESS Technology ES1938/ES1941/ES1946 Solo-1(E) Sound Card
HDD:      ST310211A (10 GB, 5400 RPM, Ultra-ATA/100)
OPTICAL:  HL-DT-ST CD-RW GCE-8480B (48x/16x/48x CD-RW)
ATHLON MACHINE SPECS:
Code: [Select]
OS:       Microsoft Windows XP Professional Service Pack 2
CPU:      AMD Athlon XP, 1666 MHz (12.5 x 133) 2000+
MEMORY:   512 MB (DDR SDRAM)
MOBO:     ECS KT600-A (4 PCI, 1 AGP, 1 CNR, 3 DDR DIMM, Audio, LAN)
VIDEO:    WinFast A340
MONITOR:  Samsung SyncMaster 750(M)s(T) [17 inch CRT]
SOUND:    VIA AC97 Enhanced Audio Controller
HDD:      Maxtor 6 Y080M0 SCSI Disk Device (80 GB, 7200 RPM, SATA)
OPTICAL:  HL-DT-ST CD-RW GCE-8520B (52x/24x/52x CD-RW)
RESULTS FOR CELERON:
Code: [Select]
OVERALL RESULTS [-f -x6]:
----------------------------------------------
TEST_ORIGINAL
Kernel Time  =    3.304 = 00:00:03.304 =  0%
User Time    =  471.017 = 00:07:51.017 =  98%
Process Time =  474.322 = 00:07:54.322 =  98%
Global Time  =  479.960 = 00:07:59.960 = 100%
----------------------------------------------
TEST_BSR [+7,84%]
Kernel Time  =    3.424 = 00:00:03.424 =  0%
User Time    =  433.733 = 00:07:13.733 =  98%
Process Time =  437.158 = 00:07:17.158 =  98%
Global Time  =  442.477 = 00:07:22.477 = 100%
----------------------------------------------
TEST_JFL2B [+2,08%]
Kernel Time  =    3.324 = 00:00:03.324 =  0%
User Time    =  461.123 = 00:07:41.123 =  98%
Process Time =  464.447 = 00:07:44.447 =  98%
Global Time  =  469.715 = 00:07:49.715 = 100%
----------------------------------------------
TEST_MMX [+14,30%]
Kernel Time  =    2.954 = 00:00:02.954 =  0%
User Time    =  403.560 = 00:06:43.560 =  98%
Process Time =  406.514 = 00:06:46.514 =  98%
Global Time  =  411.352 = 00:06:51.352 = 100%
----------------------------------------------
TEST_MMX-BSR [+22,64%]
Kernel Time  =    3.304 = 00:00:03.304 =  0%
User Time    =  363.642 = 00:06:03.642 =  97%
Process Time =  366.947 = 00:06:06.947 =  98%
Global Time  =  372.085 = 00:06:12.085 = 100%
----------------------------------------------
TEST_MMX-JFL2B [+16,58%]
Kernel Time  =    2.984 = 00:00:02.984 =  0%
User Time    =  392.684 = 00:06:32.684 =  97%
Process Time =  395.668 = 00:06:35.668 =  98%
Global Time  =  400.857 = 00:06:40.857 = 100%
RSEULTS FOR ATHLON:
Code: [Select]
OVERALL RESULTS [-f -x6]:
----------------------------------------------
TEST_ORIGINAL
Kernel Time  =    0.765 = 00:00:00.765 =  0%
User Time    =  100.734 = 00:01:40.734 =  98%
Process Time =  101.500 = 00:01:41.500 =  99%
Global Time  =  102.281 = 00:01:42.281 = 100%
----------------------------------------------
TEST_BSR [+3,65%]
Kernel Time  =    0.953 = 00:00:00.953 =  0%
User Time    =    96.843 = 00:01:36.843 =  98%
Process Time =    97.796 = 00:01:37.796 =  99%
Global Time  =    98.484 = 00:01:38.484 = 100%
----------------------------------------------
TEST_JFL2B [+0,55%]
Kernel Time  =    0.921 = 00:00:00.921 =  0%
User Time    =  100.015 = 00:01:40.015 =  98%
Process Time =  100.937 = 00:01:40.937 =  99%
Global Time  =  101.688 = 00:01:41.688 = 100%
----------------------------------------------
TEST_MMX [+12,78%]
Kernel Time  =    0.812 = 00:00:00.812 =  0%
User Time    =    87.718 = 00:01:27.718 =  98%
Process Time =    88.531 = 00:01:28.531 =  99%
Global Time  =    89.078 = 00:01:29.078 = 100%
----------------------------------------------
TEST_MMX-BSR [+16,49%]
Kernel Time  =    0.796 = 00:00:00.796 =  0%
User Time    =    83.968 = 00:01:23.968 =  98%
Process Time =    84.765 = 00:01:24.765 =  99%
Global Time  =    85.375 = 00:01:25.375 = 100%
----------------------------------------------
TEST_MMX-JFL2B [+13,50%]
Kernel Time  =    0.968 = 00:00:00.968 =  1%
User Time    =    86.828 = 00:01:26.828 =  98%
Process Time =    87.796 = 00:01:27.796 =  99%
Global Time  =    88.484 = 00:01:28.484 = 100%
Well the conclusion is after done this tests that MMX with BSR is fastest on slower machine ~22% with -f -x6 options, the JFL2B optimization is just tiny faster, but is this caused by my build, almost certainly YES, because he-jo have stated that this optimization is ~8% faster. I do not know where is the problem with my implementation, it will be nice if some one familiar with VC++ and NASM could check my sources.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: he-jo on 2006-05-26 13:14:38
Thank you very much for your tests! I think I've found the problem. It's caused by a memory stall, which is very expensive on Athlons. I've already written a workaround, and now I'm getting a speedup of 10% on the Celeron machine. I'll probably post the modified patch tomorrow.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: askoff on 2006-05-27 04:16:04
Hi. I made the test with my P4. I used bunch of random samples and few random songs in my test. The resulted test wav was 236MB and 23min 28sec long. I took two test rounds. First with -f -x6 settings and second with -h -x settings.

Here are the machine specs.
Code: [Select]
OS:       Microsoft Windows XP Professional Service Pack 2
CPU:      Intel Pentium 4, 2533 MHz (533 MHz FSB) and its overclocked to 2866 MHz (19 x 151)
MEMORY:   1024 MB (DDR-SDRAM) (2x512 in dual channel)
MOBO:     Intel E7205

Here are the test results with -f -x6
Code: [Select]
Release\wavpack.exe -f -x6 Image.wav 

Kernel Time  = 1.171 = 00:00:01.171 =  0%
User Time =  163.359 = 00:02:43.359 =  95%
Process Time =  164.531 = 00:02:44.531 =  96%
Global Time  =  170.781 = 00:02:50.781 = 100%


ReleaseBSR\wavpack.exe -f -x6 Image.wav

Kernel Time  = 0.906 = 00:00:00.906 =  0%
User Time =  135.781 = 00:02:15.781 =  94%
Process Time =  136.687 = 00:02:16.687 =  95%
Global Time  =  142.984 = 00:02:22.984 = 100%

ReleaseJFL2B\wavpack.exe -f -x6 Image.wav

Kernel Time  = 1.187 = 00:00:01.187 =  0%
User Time =  233.312 = 00:03:53.312 =  95%
Process Time =  234.500 = 00:03:54.500 =  96%
Global Time  =  243.750 = 00:04:03.750 = 100%

ReleaseMMX\wavpack.exe -f -x6 Image.wav

Kernel Time  = 0.875 = 00:00:00.875 =  0%
User Time =  132.078 = 00:02:12.078 =  95%
Process Time =  132.953 = 00:02:12.953 =  95%
Global Time  =  138.906 = 00:02:18.906 = 100%

ReleaseMMX-BSR\wavpack.exe -f -x6 Image.wav

Kernel Time  = 1.203 = 00:00:01.203 =  1%
User Time =  103.953 = 00:01:43.953 =  94%
Process Time =  105.156 = 00:01:45.156 =  95%
Global Time  =  110.422 = 00:01:50.422 = 100%

ReleaseMMX-JFL2B\wavpack.exe -f -x6 Image.wav

Kernel Time  = 1.125 = 00:00:01.125 =  0%
User Time =  201.703 = 00:03:21.703 =  95%
Process Time =  202.828 = 00:03:22.828 =  96%
Global Time  =  211.109 = 00:03:31.109 = 100%

And here are the test results with -h -x
Code: [Select]
Release\wavpack.exe -h -x Image.wav 

Kernel Time  = 1.125 = 00:00:01.125 =  0%
User Time =  782.984 = 00:13:02.984 =  96%
Process Time =  784.109 = 00:13:04.109 =  96%
Global Time  =  808.532 = 00:13:28.532 = 100%

ReleaseBSR\wavpack.exe -h -x Image.wav

Kernel Time  = 1.265 = 00:00:01.265 =  0%
User Time =  657.890 = 00:10:57.890 =  97%
Process Time =  659.156 = 00:10:59.156 =  97%
Global Time  =  677.000 = 00:11:17.000 = 100%

ReleaseJFL2B\wavpack.exe -h -x Image.wav

Kernel Time  = 1.328 = 00:00:01.328 =  0%
User Time =  1098.031 = 00:18:18.031 =  97%
Process Time =  1099.359 = 00:18:19.359 =  97%
Global Time  =  1131.922 = 00:18:51.922 = 100%

ReleaseMMX\wavpack.exe -h -x Image.wav

Kernel Time  = 1.328 = 00:00:01.328 =  0%
User Time =  599.156 = 00:09:59.156 =  96%
Process Time =  600.484 = 00:10:00.484 =  97%
Global Time  =  617.922 = 00:10:17.922 = 100%

ReleaseMMX-BSR\wavpack.exe -h -x Image.wav

Kernel Time  = 1.171 = 00:00:01.171 =  0%
User Time =  471.296 = 00:07:51.296 =  96%
Process Time =  472.468 = 00:07:52.468 =  97%
Global Time  =  486.954 = 00:08:06.954 = 100%

ReleaseMMX-JFL2B\wavpack.exe -h -x Image.wav

Kernel Time  = 1.421 = 00:00:01.421 =  0%
User Time =  909.218 = 00:15:09.218 =  96%
Process Time =  910.640 = 00:15:10.640 =  96%
Global Time  =  940.078 = 00:15:40.078 = 100%

It clearly shows that JFL2B build's are no good wit P4
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-29 14:26:32
Quote
Thank you very much for your tests! I think I\'ve found the problem. It\'s caused by a memory stall, which is very expensive on Athlons. I\'ve already written a workaround, and now I\'m getting a speedup of 10% on the Celeron machine. I\'ll probably post the modified patch tomorrow.


he, i have found the update on your website (wp-4.32-jfl2b.diff.gz, 28-May-2006 12:03) the date is suggesting this ;-)

so next builds and test results will be released soon
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: he-jo on 2006-05-29 22:07:43
Thanks to askoff for his P4 tests! I'm very interested in this. Would be nice, if you could run them again after wisodev has released his new binaries.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-05-30 06:55:48
[!--sizeo:2--][span style=\"font-size:10pt;line-height:100%\"][!--/sizeo--]NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.[/size]
Well, the fresh builds are here and look very promising!

Using as always latest optimizations posted in this thread (http://www.hydrogenaudio.org/forums/index.php?showtopic=43226) by he-jo, including sources (using JFL2B, called by me JFL2B version 2).

New binarys are included in ReleaseJFL2B and ReleaseMMX-JFL2B dirs, other binarys are the same as in post #21, but the sources are updated.

Download the newest binarys and sources:
[attachment=2250:attachment]
Below are RAW test results I have done yesterday. Completed three runs to be sure of the results. I have only tested on my Athlon machine, the specs are the same as in the previous test (well I have yesterday lot of fun playing with Xgl and compiz in my new and shiny SUSE Linux 10.1 and done only Athlon test). Using in all tests -f -x6 options.
Code: [Select]
Run 1 results:

TEST_ORIGINAL
Kernel Time  =    0.937 = 00:00:00.937 =  0%
User Time    =  100.953 = 00:01:40.953 =  96%
Process Time =  101.890 = 00:01:41.890 =  97%
Global Time  =  104.656 = 00:01:44.656 = 100%

TEST_BSR [+3,80%]
Kernel Time  =    0.906 = 00:00:00.906 =  0%
User Time    =    97.109 = 00:01:37.109 =  98%
Process Time =    98.015 = 00:01:38.015 =  99%
Global Time  =    98.843 = 00:01:38.843 = 100%

TEST_JFL2B [+10,73%]
Kernel Time  =    0.703 = 00:00:00.703 =  0%
User Time    =    90.250 = 00:01:30.250 =  98%
Process Time =    90.953 = 00:01:30.953 =  99%
Global Time  =    91.547 = 00:01:31.547 = 100%

TEST_MMX [+13,08%]
Kernel Time  =    0.843 = 00:00:00.843 =  0%
User Time    =    87.718 = 00:01:27.718 =  98%
Process Time =    88.562 = 00:01:28.562 =  99%
Global Time  =    89.328 = 00:01:29.328 = 100%

TEST_MMX-BSR [+16,93%]
Kernel Time  =    0.781 = 00:00:00.781 =  0%
User Time    =    83.859 = 00:01:23.859 =  98%
Process Time =    84.640 = 00:01:24.640 =  99%
Global Time  =    85.375 = 00:01:25.375 = 100%

TEST_MMX-JFL2B [+24,20%]
Kernel Time  =    0.671 = 00:00:00.671 =  0%
User Time    =    76.562 = 00:01:16.562 =  98%
Process Time =    77.234 = 00:01:17.234 =  99%
Global Time  =    77.828 = 00:01:17.828 = 100%

---

Run 2 results:

TEST_ORIGINAL
Kernel Time  =    0.859 = 00:00:00.859 =  0%
User Time    =  100.640 = 00:01:40.640 =  98%
Process Time =  101.500 = 00:01:41.500 =  99%
Global Time  =  102.219 = 00:01:42.219 = 100%

TEST_BSR [+3,63%]
Kernel Time  =    0.921 = 00:00:00.921 =  0%
User Time    =    96.890 = 00:01:36.890 =  98%
Process Time =    97.812 = 00:01:37.812 =  99%
Global Time  =    98.547 = 00:01:38.547 = 100%

TEST_JFL2B [+10,78%]
Kernel Time  =    0.875 = 00:00:00.875 =  0%
User Time    =    89.687 = 00:01:29.687 =  98%
Process Time =    90.562 = 00:01:30.562 =  99%
Global Time  =    91.187 = 00:01:31.187 = 100%

TEST_MMX [+12,93%]
Kernel Time  =    1.125 = 00:00:01.125 =  1%
User Time    =    87.250 = 00:01:27.250 =  97%
Process Time =    88.375 = 00:01:28.375 =  99%
Global Time  =    89.062 = 00:01:29.062 = 100%

TEST_MMX-BSR [+16,63%]
Kernel Time  =    0.734 = 00:00:00.734 =  0%
User Time    =    83.890 = 00:01:23.890 =  98%
Process Time =    84.625 = 00:01:24.625 =  99%
Global Time  =    85.281 = 00:01:25.281 = 100%

TEST_MMX-JFL2B [+24,15%]
Kernel Time  =    1.093 = 00:00:01.093 =  1%
User Time    =    75.890 = 00:01:15.890 =  97%
Process Time =    76.984 = 00:01:16.984 =  99%
Global Time  =    77.656 = 00:01:17.656 = 100%

---

Run 3 results:

TEST_ORIGINAL
Kernel Time  =    0.734 = 00:00:00.734 =  0%
User Time    =  100.812 = 00:01:40.812 =  98%
Process Time =  101.546 = 00:01:41.546 =  99%
Global Time  =  102.188 = 00:01:42.188 = 100%

TEST_BSR [+3,82%]
Kernel Time  =    1.109 = 00:00:01.109 =  1%
User Time    =    96.562 = 00:01:36.562 =  98%
Process Time =    97.671 = 00:01:37.671 =  99%
Global Time  =    98.484 = 00:01:38.484 = 100%

TEST_JFL2B [+10,91%]
Kernel Time  =    0.890 = 00:00:00.890 =  0%
User Time    =    89.578 = 00:01:29.578 =  98%
Process Time =    90.468 = 00:01:30.468 =  99%
Global Time  =    91.156 = 00:01:31.156 = 100%

TEST_MMX [+13,08%]
Kernel Time  =    0.765 = 00:00:00.765 =  0%
User Time    =    87.500 = 00:01:27.500 =  98%
Process Time =    88.265 = 00:01:28.265 =  99%
Global Time  =    89.047 = 00:01:29.047 = 100%

TEST_MMX-BSR [+16,59%]
Kernel Time  =    0.890 = 00:00:00.890 =  1%
User Time    =    83.812 = 00:01:23.812 =  98%
Process Time =    84.703 = 00:01:24.703 =  99%
Global Time  =    85.234 = 00:01:25.234 = 100%

TEST_MMX-JFL2B [+24,10%]
Kernel Time  =    1.046 = 00:00:01.046 =  1%
User Time    =    76.031 = 00:01:16.031 =  97%
Process Time =    77.078 = 00:01:17.078 =  99%
Global Time  =    77.656 = 00:01:17.656 = 100%
NOTE: TEST_JFL2B and TEST_MMX-JFL2B are version 2 of he-jo JFL2B optimizations (the updated one).

Tests summary for -f -x6 options, all values are positive percentage speedups:
Code: [Select]
            CPU: AHTLON XP 2000+
   TEST NAME    run 1 [%]  run 2 [%]  run 3 [%]  avg [%]
TEST_ORIGINAL      0          0          0           0
TEST_BSR          3,80       3,63       3,82       3,75
TEST_JFL2B       10,73      10,78      10,91      10,81
TEST_MMX         13,08      12,93      13,08      13,03
TEST_MMX-BSR     16,93      16,63      16,59      16,72
TEST_MMX-JFL2B   24,20      24,15      24,10      24,15
So clearly the version 2 of JFL2B optimizations is very fast and combined with MMX reach 24% speedup with -f -x6, maybe on my Celeron this version could be even faster (appropriate tests will be done).

Thanks for your great work he-jo!!!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: askoff on 2006-05-31 08:33:34
Here's my results with P4.
First with -f -x6
Code: [Select]
Release\wavpack.exe -f -x6 Image.wav 
-------------------------------------
Kernel Time  =    0.953 = 00:00:00.953 =  0%
User Time    =  162.578 = 00:02:42.578 =  96%
Process Time =  163.531 = 00:02:43.531 =  96%
Global Time  =  169.109 = 00:02:49.109 = 100%

ReleaseBSR\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.125 = 00:00:01.125 =  0%
User Time    =  134.656 = 00:02:14.656 =  95%
Process Time =  135.781 = 00:02:15.781 =  96%
Global Time  =  140.579 = 00:02:20.579 = 100%

ReleaseJFL2B\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.000 = 00:00:01.000 =  0%
User Time    =  162.343 = 00:02:42.343 =  96%
Process Time =  163.343 = 00:02:43.343 =  96%
Global Time  =  169.015 = 00:02:49.015 = 100%

ReleaseMMX\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.765 = 00:00:00.765 =  0%
User Time    =  131.343 = 00:02:11.343 =  96%
Process Time =  132.109 = 00:02:12.109 =  96%
Global Time  =  136.797 = 00:02:16.797 = 100%

ReleaseMMX-BSR\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.875 = 00:00:00.875 =  0%
User Time    =  103.859 = 00:01:43.859 =  95%
Process Time =  104.734 = 00:01:44.734 =  96%
Global Time  =  108.515 = 00:01:48.515 = 100%

ReleaseMMX-JFL2B\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.031 = 00:00:01.031 =  0%
User Time    =  128.546 = 00:02:08.546 =  95%
Process Time =  129.578 = 00:02:09.578 =  96%
Global Time  =  134.015 = 00:02:14.015 = 100%

And the second test with -h -x
Code: [Select]
Release\wavpack.exe -h -x Image.wav 
-------------------------------------
Kernel Time  =    1.281 = 00:00:01.281 =  0%
User Time    =  776.218 = 00:12:56.218 =  96%
Process Time =  777.500 = 00:12:57.500 =  97%
Global Time  =  800.703 = 00:13:20.703 = 100%

ReleaseBSR\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.312 = 00:00:01.312 =  0%
User Time    =  659.390 = 00:10:59.390 =  96%
Process Time =  660.703 = 00:11:00.703 =  96%
Global Time  =  683.172 = 00:11:23.172 = 100%

ReleaseJFL2B\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.140 = 00:00:01.140 =  0%
User Time    =  762.328 = 00:12:42.328 =  97%
Process Time =  763.468 = 00:12:43.468 =  97%
Global Time  =  784.672 = 00:13:04.672 = 100%

ReleaseMMX\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.187 = 00:00:01.187 =  0%
User Time    =  593.734 = 00:09:53.734 =  97%
Process Time =  594.921 = 00:09:54.921 =  97%
Global Time  =  611.016 = 00:10:11.016 = 100%

ReleaseMMX-BSR\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    0.984 = 00:00:00.984 =  0%
User Time    =  468.406 = 00:07:48.406 =  97%
Process Time =  469.390 = 00:07:49.390 =  97%
Global Time  =  481.750 = 00:08:01.750 = 100%

ReleaseMMX-JFL2B\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.125 = 00:00:01.125 =  0%
User Time    =  578.656 = 00:09:38.656 =  97%
Process Time =  579.781 = 00:09:39.781 =  97%
Global Time  =  595.406 = 00:09:55.406 = 100%
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-06-02 05:36:49
NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.
I have added latest full MMX optimization from wp-4.32-jh_mmx.diff.gz but on my tests in was slower (tiny margin) then previous version. Maybe is this due to removal of the unions? The test where done on my athlon machine.

Download sources and new MMX only optimized binary, you can use binarys from post #31 for comparison:
[attachment=2260:attachment]
Test results for -f -x6:
Code: [Select]
TEST_ORIGINAL (Release\wavpack.exe)

Kernel Time  =     0.765 = 00:00:00.765 =   0%
User Time    =   100.781 = 00:01:40.781 =  96%
Process Time =   101.546 = 00:01:41.546 =  97%
Global Time  =   104.187 = 00:01:44.187 = 100%

TEST_MMX_OLD (ReleaseMMX\wavpack.exe)

Kernel Time  =     0.828 = 00:00:00.828 =   0%
User Time    =    87.578 = 00:01:27.578 =  98%
Process Time =    88.406 = 00:01:28.406 =  99%
Global Time  =    89.109 = 00:01:29.109 = 100%

TEST_MMX_FULL (ReleaseMMX\wavpack_MMX_FULL.exe)

Kernel Time  =     0.625 = 00:00:00.625 =   0%
User Time    =    88.421 = 00:01:28.421 =  98%
Process Time =    89.046 = 00:01:29.046 =  99%
Global Time  =    89.703 = 00:01:29.703 = 100%
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-06-02 19:07:05
[!--sizeo:2--][span style=\"font-size:10pt;line-height:100%\"][!--/sizeo--]NOTE: This are old binarys and sources! Use updated binarys and sources from POST #40.[/size]
Here are fixed binarys! The binarys from previous post have a small bug and you should not use and test them.

Download (sources, binarys - all of them!, test results):
[attachment=2262:attachment]
test_results_small_files.txt:
Code: [Select]
TEST FILES: set of small files suggested by he-jo 
[-f -x6]

TEST_ORIGINAL
Kernel Time  =    0.937 = 00:00:00.937 =  0%
User Time    =  100.953 = 00:01:40.953 =  96%
Process Time =  101.890 = 00:01:41.890 =  97%
Global Time  =  104.656 = 00:01:44.656 = 100%

TEST_MMX [+13,08%]
Kernel Time  =    0.843 = 00:00:00.843 =  0%
User Time    =    87.718 = 00:01:27.718 =  98%
Process Time =    88.562 = 00:01:28.562 =  99%
Global Time  =    89.328 = 00:01:29.328 = 100%

TEST_MMX_FULL [+12,01%]
Kernel Time  =    0.937 = 00:00:00.937 =  1%
User Time    =    88.718 = 00:01:28.718 =  98%
Process Time =    89.656 = 00:01:29.656 =  99%
Global Time  =    90.266 = 00:01:30.266 = 100%

TEST_MMX-BSR [+16,93%]
Kernel Time  =    0.781 = 00:00:00.781 =  0%
User Time    =    83.859 = 00:01:23.859 =  98%
Process Time =    84.640 = 00:01:24.640 =  99%
Global Time  =    85.375 = 00:01:25.375 = 100%

TEST_MMX_FULL-BSR [+15,86%]
Kernel Time  =    0.734 = 00:00:00.734 =  0%
User Time    =    85.000 = 00:01:25.000 =  98%
Process Time =    85.734 = 00:01:25.734 =  99%
Global Time  =    86.468 = 00:01:26.468 = 100%

TEST_MMX-JFL2B [+24,20%]
Kernel Time  =    0.671 = 00:00:00.671 =  0%
User Time    =    76.562 = 00:01:16.562 =  98%
Process Time =    77.234 = 00:01:17.234 =  99%
Global Time  =    77.828 = 00:01:17.828 = 100%

TEST_MMX_FULL-JFL2B [+23,40%]
Kernel Time  =    0.937 = 00:00:00.937 =  1%
User Time    =    77.109 = 00:01:17.109 =  97%
Process Time =    78.046 = 00:01:18.046 =  99%
Global Time  =    78.828 = 00:01:18.828 = 100%

[binary comparison: all result files VALID - no problems here]
test_results_big_file.txt:
Code: [Select]
TEST FILE: 34435676 bytes, 44100 Hz, 16 bits, Stereo
[-f -x6]

TEST_ORG
Kernel Time  =    0.359 = 00:00:00.359 =  1%
User Time    =    31.781 = 00:00:31.781 =  98%
Process Time =    32.140 = 00:00:32.140 =  99%
Global Time  =    32.265 = 00:00:32.265 = 100%

TEST_BSR [+3,26%]
Kernel Time  =    0.250 = 00:00:00.250 =  0%
User Time    =    30.843 = 00:00:30.843 =  98%
Process Time =    31.093 = 00:00:31.093 =  99%
Global Time  =    31.282 = 00:00:31.282 = 100%

TEST_JFL2B [-0,20%]
Kernel Time  =    0.359 = 00:00:00.359 =  1%
User Time    =    31.843 = 00:00:31.843 =  98%
Process Time =    32.203 = 00:00:32.203 =  99%
Global Time  =    32.437 = 00:00:32.437 = 100%

TEST_MMX [+12,64%]
Kernel Time  =    0.312 = 00:00:00.312 =  1%
User Time    =    27.765 = 00:00:27.765 =  98%
Process Time =    28.078 = 00:00:28.078 =  99%
Global Time  =    28.297 = 00:00:28.297 = 100%

TEST_MMX_FULL [+10,55%]
Kernel Time  =    0.328 = 00:00:00.328 =  1%
User Time    =    28.421 = 00:00:28.421 =  98%
Process Time =    28.750 = 00:00:28.750 =  99%
Global Time  =    28.984 = 00:00:28.984 = 100%

TEST_MMX-BSR [+15,80%]
Kernel Time  =    0.359 = 00:00:00.359 =  1%
User Time    =    26.703 = 00:00:26.703 =  98%
Process Time =    27.062 = 00:00:27.062 =  99%
Global Time  =    27.234 = 00:00:27.234 = 100%

TEST_MMX_FULL-BSR [+15,31%]
Kernel Time  =    0.296 = 00:00:00.296 =  1%
User Time    =    26.921 = 00:00:26.921 =  97%
Process Time =    27.218 = 00:00:27.218 =  99%
Global Time  =    27.485 = 00:00:27.485 = 100%

TEST_MMX-JFL2B [+12,40%]
Kernel Time  =    0.406 = 00:00:00.406 =  1%
User Time    =    27.750 = 00:00:27.750 =  97%
Process Time =    28.156 = 00:00:28.156 =  99%
Global Time  =    28.344 = 00:00:28.344 = 100%

TEST_MMX_FULL-JFL2B [+20,47%]
Kernel Time  =    0.218 = 00:00:00.218 =  0%
User Time    =    25.343 = 00:00:25.343 =  98%
Process Time =    25.562 = 00:00:25.562 =  99%
Global Time  =    25.687 = 00:00:25.687 = 100%

[binary comparison: all result files VALID - no problems here]
test_results_big_files.txt:
Code: [Select]
TEST FILES: 13 Tracks, 431891276 bytes, 44100 Hz, 16 bits, Stereo
[-f -x6]

TEST_ORG
Kernel Time  =    2.562 = 00:00:02.562 =  0%
User Time    =  395.937 = 00:06:35.937 =  98%
Process Time =  398.500 = 00:06:38.500 =  98%
Global Time  =  402.797 = 00:06:42.797 = 100%

TEST_MMX_FULL-JFL2B [+22,16%]
Kernel Time  =    2.578 = 00:00:02.578 =  0%
User Time    =  307.609 = 00:05:07.609 =  97%
Process Time =  310.187 = 00:05:10.187 =  98%
Global Time  =  314.406 = 00:05:14.406 = 100%

[binary comparison: all result files VALID - no problems here]
This are very intresting results I think!!! But he-jo has posted already new patch, witch is not included here! The more is to come on monday from me.

NOTE: The MMX_FULL means MMX optimizations + MMX optimizations for negative values in extra2.c, from latest patch by he-jo!
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: vhl on 2006-06-12 18:08:50
sorry for offtop, but where can i download "timer.exe"?
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: Synthetic Soul on 2006-06-12 22:51:43
sorry for offtop, but where can i download "timer.exe"?
http://www.7-zip.org/igor.html (http://www.7-zip.org/igor.html)
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: askoff on 2006-06-13 11:55:20
And my P4 tests. It took some time but here they are. -h -x results aren't done yet, but as soon as I have time to do them I will.
-f -x6:
Code: [Select]
Release\wavpack.exe -f -x6 Image.wav 
-------------------------------------
Kernel Time  =    0.796 = 00:00:00.796 =  0%
User Time    =  162.468 = 00:02:42.468 =  95%
Process Time =  163.265 = 00:02:43.265 =  96%
Global Time  =  169.422 = 00:02:49.422 = 100%

ReleaseBSR\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.734 = 00:00:00.734 =  0%
User Time    =  135.187 = 00:02:15.187 =  96%
Process Time =  135.921 = 00:02:15.921 =  96%
Global Time  =  140.687 = 00:02:20.687 = 100%

ReleaseJFL2B\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.859 = 00:00:00.859 =  0%
User Time    =  232.375 = 00:03:52.375 =  96%
Process Time =  233.234 = 00:03:53.234 =  96%
Global Time  =  240.782 = 00:04:00.782 = 100%

ReleaseMMX\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.906 = 00:00:00.906 =  0%
User Time    =  131.484 = 00:02:11.484 =  95%
Process Time =  132.390 = 00:02:12.390 =  96%
Global Time  =  137.078 = 00:02:17.078 = 100%

ReleaseMMX-FULL\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.843 = 00:00:00.843 =  0%
User Time    =  152.578 = 00:02:32.578 =  96%
Process Time =  153.421 = 00:02:33.421 =  96%
Global Time  =  158.657 = 00:02:38.657 = 100%

ReleaseMMX-BSR\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.937 = 00:00:00.937 =  0%
User Time    =  103.750 = 00:01:43.750 =  95%
Process Time =  104.687 = 00:01:44.687 =  96%
Global Time  =  108.312 = 00:01:48.312 = 100%

ReleaseMMX-BSR-FULL\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.140 = 00:00:01.140 =  0%
User Time    =  124.578 = 00:02:04.578 =  95%
Process Time =  125.718 = 00:02:05.718 =  96%
Global Time  =  130.078 = 00:02:10.078 = 100%

ReleaseMMX-JFL2B\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.828 = 00:00:00.828 =  0%
User Time    =  200.796 = 00:03:20.796 =  96%
Process Time =  201.625 = 00:03:21.625 =  96%
Global Time  =  208.328 = 00:03:28.328 = 100%

ReleaseMMX-JFL2B-FULL\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.015 = 00:00:01.015 =  0%
User Time    =  149.640 = 00:02:29.640 =  96%
Process Time =  150.656 = 00:02:30.656 =  96%
Global Time  =  155.828 = 00:02:35.828 = 100%

Here's the -h -x result.
Code: [Select]
Release\wavpack.exe -h -x Image.wav 
-------------------------------------
Kernel Time  =    1.218 = 00:00:01.218 =  0%
User Time    =  776.515 = 00:12:56.515 =  96%
Process Time =  777.734 = 00:12:57.734 =  97%
Global Time  =  801.062 = 00:13:21.062 = 100%

ReleaseBSR\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.078 = 00:00:01.078 =  0%
User Time    =  651.000 = 00:10:51.000 =  96%
Process Time =  652.078 = 00:10:52.078 =  97%
Global Time  =  671.594 = 00:11:11.594 = 100%

ReleaseJFL2B\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.281 = 00:00:01.281 =  0%
User Time    =  1088.562 = 00:18:08.562 =  96%
Process Time =  1089.843 = 00:18:09.843 =  96%
Global Time  =  1127.016 = 00:18:47.016 = 100%

ReleaseMMX\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.078 = 00:00:01.078 =  0%
User Time    =  593.468 = 00:09:53.468 =  96%
Process Time =  594.546 = 00:09:54.546 =  97%
Global Time  =  611.922 = 00:10:11.922 = 100%

ReleaseMMX-FULL\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.234 = 00:00:01.234 =  0%
User Time    =  698.218 = 00:11:38.218 =  96%
Process Time =  699.453 = 00:11:39.453 =  97%
Global Time  =  720.547 = 00:12:00.547 = 100%

ReleaseMMX-BSR\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.015 = 00:00:01.015 =  0%
User Time    =  470.109 = 00:07:50.109 =  97%
Process Time =  471.125 = 00:07:51.125 =  97%
Global Time  =  484.391 = 00:08:04.391 = 100%

ReleaseMMX-BSR-FULL\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.031 = 00:00:01.031 =  0%
User Time    =  573.562 = 00:09:33.562 =  96%
Process Time =  574.593 = 00:09:34.593 =  96%
Global Time  =  592.516 = 00:09:52.516 = 100%

ReleaseMMX-JFL2B\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.312 = 00:00:01.312 =  0%
User Time    =  902.515 = 00:15:02.515 =  96%
Process Time =  903.828 = 00:15:03.828 =  96%
Global Time  =  931.812 = 00:15:31.812 = 100%

ReleaseMMX-JFL2B-FULL\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.609 = 00:00:01.609 =  0%
User Time    =  682.937 = 00:11:22.937 =  96%
Process Time =  684.546 = 00:11:24.546 =  97%
Global Time  =  705.125 = 00:11:45.125 = 100%
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: vhl on 2006-06-13 18:48:46
Ok, there are my results on Athlon 64 X2 4200+ (on WinXP Sp2 32bit)
Sample file ~140Megs. Keys: -f -x6
Code: [Select]
TEST_ORIGINAL
Global Time  =    78.969 = 00:01:18.969 = 100%
TEST_BSR
Global Time  =    73.859 = 00:01:13.859 = 100%
TEST_JFL2B
Global Time  =    65.703 = 00:01:05.703 = 100%
TEST_MMX
Global Time  =    67.125 = 00:01:07.125 = 100%
TEST_MMX-BSR
Global Time  =    61.891 = 00:01:01.891 = 100%
TEST_MMX-JFL2B
Global Time  =    53.782 = 00:00:53.782 = 100%
How about next optimizations for multicore/multithreading  ? Because all process of compressing take only 50% CPU time. And one question - can you place all he-jo samples in one archive file for more easy downloading? (it's so laszy to downloading many little files)
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: PrakashP on 2006-06-13 22:30:49
Just a quick test with the bins in Linux using wine shows the MMX-BSR to be fastest for me on my Athlon-XP using about 23% less time than orig. (MMX-JFL2B takes 18% less time.)
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-06-14 18:02:49
Sorry for so late reply, but I had lot of work in my paid job.

I have added the newest patch to my builds. Now the MMX code works very nice in MSVC++ without modification. Thanks for that improvement he-jo.

Download latest binarys and sources:
[attachment=2288:attachment]
Test results (samples):
Code: [Select]
TEST FILES: set of small files suggested by he-jo 
[-f -x6]

TEST_ORG

Kernel Time  =    0.796 = 00:00:00.796 =  0%
User Time    =  100.968 = 00:01:40.968 =  98%
Process Time =  101.765 = 00:01:41.765 =  99%
Global Time  =  102.453 = 00:01:42.453 = 100%

TEST_BSR [+3,87%]

Kernel Time  =    0.859 = 00:00:00.859 =  0%
User Time    =    96.968 = 00:01:36.968 =  98%
Process Time =    97.828 = 00:01:37.828 =  99%
Global Time  =    98.703 = 00:01:38.703 = 100%

TEST_JFL2B [+0,68%]

Kernel Time  =    0.750 = 00:00:00.750 =  0%
User Time    =  100.328 = 00:01:40.328 =  98%
Process Time =  101.078 = 00:01:41.078 =  99%
Global Time  =  101.735 = 00:01:41.735 = 100%

TEST_MMX [+12,96%]

Kernel Time  =    1.031 = 00:00:01.031 =  1%
User Time    =    87.546 = 00:01:27.546 =  98%
Process Time =    88.578 = 00:01:28.578 =  99%
Global Time  =    89.203 = 00:01:29.203 = 100%

TEST_MMX_FULL [+12,51%]

Kernel Time  =    0.921 = 00:00:00.921 =  1%
User Time    =    88.109 = 00:01:28.109 =  98%
Process Time =    89.031 = 00:01:29.031 =  99%
Global Time  =    89.812 = 00:01:29.812 = 100%

TEST_MMX-BSR [+16,74%]

Kernel Time  =    0.937 = 00:00:00.937 =  1%
User Time    =    83.796 = 00:01:23.796 =  98%
Process Time =    84.734 = 00:01:24.734 =  99%
Global Time  =    85.328 = 00:01:25.328 = 100%

TEST_MMX_FULL-BSR [+16,04%]

Kernel Time  =    1.140 = 00:00:01.140 =  1%
User Time    =    84.296 = 00:01:24.296 =  97%
Process Time =    85.437 = 00:01:25.437 =  99%
Global Time  =    86.062 = 00:01:26.062 = 100%

TEST_MMX-JFL2B [+13,73%]

Kernel Time  =    0.937 = 00:00:00.937 =  1%
User Time    =    86.859 = 00:01:26.859 =  98%
Process Time =    87.796 = 00:01:27.796 =  99%
Global Time  =    88.515 = 00:01:28.515 = 100%

TEST_MMX_FULL-JFL2B [+22,09%]

Kernel Time  =    1.093 = 00:00:01.093 =  1%
User Time    =    78.187 = 00:01:18.187 =  97%
Process Time =    79.281 = 00:01:19.281 =  99%
Global Time  =    79.922 = 00:01:19.922 = 100%
Test results (one big file):
Code: [Select]
TEST FILE: 34435676 bytes, 44100 Hz, 16 bits, Stereo
[-f -x6]

TEST_ORG

Kernel Time  =    0.281 = 00:00:00.281 =  0%
User Time    =    31.734 = 00:00:31.734 =  98%
Process Time =    32.015 = 00:00:32.015 =  99%
Global Time  =    32.203 = 00:00:32.203 = 100%

TEST_BSR [+3,37%]

Kernel Time  =    0.218 = 00:00:00.218 =  0%
User Time    =    30.718 = 00:00:30.718 =  98%
Process Time =    30.937 = 00:00:30.937 =  99%
Global Time  =    31.125 = 00:00:31.125 = 100%

TEST_JFL2B [+0,10%]

Kernel Time  =    0.265 = 00:00:00.265 =  0%
User Time    =    31.718 = 00:00:31.718 =  98%
Process Time =    31.984 = 00:00:31.984 =  99%
Global Time  =    32.141 = 00:00:32.141 = 100%

TEST_MMX [+11,86%]

Kernel Time  =    0.296 = 00:00:00.296 =  1%
User Time    =    27.921 = 00:00:27.921 =  98%
Process Time =    28.218 = 00:00:28.218 =  99%
Global Time  =    28.312 = 00:00:28.312 = 100%

TEST_MMX_FULL [+11,57%]

Kernel Time  =    0.250 = 00:00:00.250 =  0%
User Time    =    28.062 = 00:00:28.062 =  98%
Process Time =    28.312 = 00:00:28.312 =  99%
Global Time  =    28.547 = 00:00:28.547 = 100%

TEST_MMX-BSR [+15,62%]

Kernel Time  =    0.328 = 00:00:00.328 =  1%
User Time    =    26.687 = 00:00:26.687 =  98%
Process Time =    27.015 = 00:00:27.015 =  99%
Global Time  =    27.203 = 00:00:27.203 = 100%

TEST_MMX_FULL-BSR [+14,69%]

Kernel Time  =    0.265 = 00:00:00.265 =  0%
User Time    =    27.046 = 00:00:27.046 =  98%
Process Time =    27.312 = 00:00:27.312 =  99%
Global Time  =    27.484 = 00:00:27.484 = 100%

TEST_MMX-JFL2B [+12,25%]

Kernel Time  =    0.328 = 00:00:00.328 =  1%
User Time    =    27.765 = 00:00:27.765 =  98%
Process Time =    28.093 = 00:00:28.093 =  99%
Global Time  =    28.266 = 00:00:28.266 = 100%

TEST_MMX_FULL-JFL2B [+21,03%]

Kernel Time  =    0.234 = 00:00:00.234 =  0%
User Time    =    25.046 = 00:00:25.046 =  98%
Process Time =    25.281 = 00:00:25.281 =  99%
Global Time  =    25.437 = 00:00:25.437 = 100%
Test results (album):
Code: [Select]
TEST FILES: 13 Tracks, 431891276 bytes, 44100 Hz, 16 bits, Stereo
[-f -x6]

TEST_ORG

Kernel Time  =    2.828 = 00:00:02.828 =  0%
User Time    =  394.984 = 00:06:34.984 =  98%
Process Time =  397.812 = 00:06:37.812 =  98%
Global Time  =  402.187 = 00:06:42.187 = 100%

TEST_MMX_FULL-JFL2B [+22,68%]

Kernel Time  =    2.906 = 00:00:02.906 =  0%
User Time    =  304.687 = 00:05:04.687 =  51%
Process Time =  307.593 = 00:05:07.593 =  51%
Global Time  =  592.672 = 00:09:52.672 = 100%
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: ssjkakaroto on 2006-06-14 19:49:08
wisodev do you know why the global time for the Test results (album) was so high?
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: askoff on 2006-06-15 15:58:11
P4 tests -f -x6:
Code: [Select]
Release\wavpack.exe -f -x6 Image.wav 
-------------------------------------
Kernel Time  =    0.906 = 00:00:00.906 =  0%
User Time    =  162.546 = 00:02:42.546 =  95%
Process Time =  163.453 = 00:02:43.453 =  96%
Global Time  =  170.188 = 00:02:50.188 = 100%

ReleaseBSR\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.109 = 00:00:01.109 =  0%
User Time    =  134.859 = 00:02:14.859 =  95%
Process Time =  135.968 = 00:02:15.968 =  96%
Global Time  =  141.250 = 00:02:21.250 = 100%

ReleaseJFL2B\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.875 = 00:00:00.875 =  0%
User Time    =  232.390 = 00:03:52.390 =  96%
Process Time =  233.265 = 00:03:53.265 =  96%
Global Time  =  241.109 = 00:04:01.109 = 100%

ReleaseMMX\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.890 = 00:00:00.890 =  0%
User Time    =  131.359 = 00:02:11.359 =  95%
Process Time =  132.250 = 00:02:12.250 =  96%
Global Time  =  137.015 = 00:02:17.015 = 100%

ReleaseMMX-FULL\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.968 = 00:00:00.968 =  0%
User Time    =  147.125 = 00:02:27.125 =  95%
Process Time =  148.093 = 00:02:28.093 =  96%
Global Time  =  153.407 = 00:02:33.407 = 100%

ReleaseMMX-BSR\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.796 = 00:00:00.796 =  0%
User Time    =  103.953 = 00:01:43.953 =  95%
Process Time =  104.750 = 00:01:44.750 =  96%
Global Time  =  108.547 = 00:01:48.547 = 100%

ReleaseMMX-BSR-FULL\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    1.140 = 00:00:01.140 =  0%
User Time    =  119.609 = 00:01:59.609 =  95%
Process Time =  120.750 = 00:02:00.750 =  96%
Global Time  =  125.141 = 00:02:05.141 = 100%

ReleaseMMX-JFL2B\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.781 = 00:00:00.781 =  0%
User Time    =  200.750 = 00:03:20.750 =  96%
Process Time =  201.531 = 00:03:21.531 =  96%
Global Time  =  208.437 = 00:03:28.437 = 100%

ReleaseMMX-JFL2B-FULL\wavpack.exe -f -x6 Image.wav
-------------------------------------
Kernel Time  =    0.781 = 00:00:00.781 =  0%
User Time    =  144.984 = 00:02:24.984 =  96%
Process Time =  145.765 = 00:02:25.765 =  96%
Global Time  =  150.734 = 00:02:30.734 = 100%
-h -x.
Code: [Select]
Release\wavpack.exe -h -x Image.wav 
-------------------------------------
Kernel Time  =    1.484 = 00:00:01.484 =  0%
User Time    =  776.109 = 00:12:56.109 =  96%
Process Time =  777.593 = 00:12:57.593 =  97%
Global Time  =  801.563 = 00:13:21.563 = 100%

ReleaseBSR\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.156 = 00:00:01.156 =  0%
User Time    =  651.531 = 00:10:51.531 =  96%
Process Time =  652.687 = 00:10:52.687 =  97%
Global Time  =  672.766 = 00:11:12.766 = 100%

ReleaseJFL2B\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.421 = 00:00:01.421 =  0%
User Time    =  1087.031 = 00:18:07.031 =  96%
Process Time =  1088.453 = 00:18:08.453 =  97%
Global Time  =  1122.000 = 00:18:42.000 = 100%

ReleaseMMX\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.140 = 00:00:01.140 =  0%
User Time    =  593.046 = 00:09:53.046 =  96%
Process Time =  594.187 = 00:09:54.187 =  97%
Global Time  =  612.360 = 00:10:12.360 = 100%

ReleaseMMX-FULL\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.312 = 00:00:01.312 =  0%
User Time    =  657.312 = 00:10:57.312 =  96%
Process Time =  658.625 = 00:10:58.625 =  97%
Global Time  =  677.937 = 00:11:17.937 = 100%

ReleaseMMX-BSR\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.406 = 00:00:01.406 =  0%
User Time    =  467.984 = 00:07:47.984 =  96%
Process Time =  469.390 = 00:07:49.390 =  97%
Global Time  =  483.093 = 00:08:03.093 = 100%

ReleaseMMX-BSR-FULL\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.265 = 00:00:01.265 =  0%
User Time    =  532.312 = 00:08:52.312 =  96%
Process Time =  533.578 = 00:08:53.578 =  97%
Global Time  =  549.781 = 00:09:09.781 = 100%

ReleaseMMX-JFL2B\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    1.171 = 00:00:01.171 =  0%
User Time    =  902.671 = 00:15:02.671 =  96%
Process Time =  903.843 = 00:15:03.843 =  97%
Global Time  =  931.781 = 00:15:31.781 = 100%

ReleaseMMX-JFL2B-FULL\wavpack.exe -h -x Image.wav
-------------------------------------
Kernel Time  =    0.937 = 00:00:00.937 =  0%
User Time    =  645.062 = 00:10:45.062 =  96%
Process Time =  646.000 = 00:10:46.000 =  97%
Global Time  =  665.437 = 00:11:05.437 = 100%
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: wisodev on 2006-06-16 09:39:51
wisodev do you know why the global time for the Test results (album) was so high?


I have bin copying output files to other folder to compare them later with original output files, this is done becose I want to know if results are correct. This results in later tests with I/O slowdown. But now I think I should do checksum of files for comparisson.
Title: MMX Optimized WavPack Encoder (for Windows)
Post by: vhl on 2006-06-17 08:45:24
Hmm, interesting things, all results with new binaries are little slower:
Code: [Select]
TEST_MMX_FULL
Process Time =    66.640 = 00:01:06.640 =  99%
TEST_MMX-JFL2B
Process Time =    65.625 = 00:01:05.625 =  99%
TEST_MMX_FULL-JFL2B
Process Time =    58.546 = 00:00:58.546 =  99%
I retest with old binaries and get the same results
Code: [Select]
TEST_MMX_FULL-JFL2B
Process Time =    53.796
! What's happend?