valgrind-users

[Valgrind-users] my cache factoring saga continues

[Jerry S. <js...@fi...>]

Hi again,

This is just such an awful thing to be doing but such are the cards I get handed.

The program I am trying to measure the cache performance on is actually a win32 progran that I am running under wine on redhat linux. I thought for a while the problem might have been that I had a stripped version of the program and that was causing a problem. After much aggravation, wine is now built locally with -g -O so I can debug if really necessary.

When I run it under cachegrind, it doesn't seem to generate the statistics.

here is the wine program:
$ file `which wine`
/usr/local/bin/wine: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.18, not stripped

here is the run:
$ valgrind --tool=cachegrind --LL=16777216,16,64 wine ../../../../Program\ Files/LTC/LTspiceIV/scad3.exe -wine -b xxx.cir 
==13347== Cachegrind, a cache and branch-prediction profiler
==13347== Copyright (C) 2002-2012, and GNU GPL'd, by Nicholas Nethercote et al.
==13347== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
==13347== Command: wine ../../../../Program\ Files/LTC/LTspiceIV/scad3.exe -wine -b xxx.cir
==13347== 
--13347-- warning: Unknown Intel cache config value (0x63), ignoring
--13347-- warning: L3 cache found, using its data for the LL simulation.
==13347== Auto-detected LL cache configuration not supported: Cache set count is not a power of two.
==13347==   LL: 26,214,400 B, 20-way, 64 B lines

that's the end of the output. On other programs, there is the stats and miss percentage.
 valgrind --tool=cachegrind --LL=16777216,16,64 pwd
==13662== Cachegrind, a cache and branch-prediction profiler
==13662== Copyright (C) 2002-2012, and GNU GPL'd, by Nicholas Nethercote et al.
==13662== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
==13662== Command: pwd
==13662== 
--13662-- warning: Unknown Intel cache config value (0x63), ignoring
--13662-- warning: L3 cache found, using its data for the LL simulation.
==13662== Auto-detected LL cache configuration not supported: Cache set count is not a power of two.
==13662==   LL: 26,214,400 B, 20-way, 64 B lines
/yyy/.wine/drive_c/users/root/phitest/release
==13662== 
==13662== I   refs:      196,379
==13662== I1  misses:      1,003
==13662== LLi misses:        990
==13662== I1  miss rate:    0.51%
==13662== LLi miss rate:    0.50%
==13662== 
==13662== D   refs:       66,222  (50,912 rd   + 15,310 wr)
==13662== D1  misses:      2,464  ( 2,099 rd   +    365 wr)
==13662== LLd misses:      1,936  ( 1,603 rd   +    333 wr)
==13662== D1  miss rate:     3.7% (   4.1%     +    2.3%  )
==13662== LLd miss rate:     2.9% (   3.1%     +    2.1%  )
==13662== 
==13662== LL refs:         3,467  ( 3,102 rd   +    365 wr)
==13662== LL misses:       2,926  ( 2,593 rd   +    333 wr)
==13662== LL miss rate:      1.1% (   1.0%     +    2.1%  )

This works with both 32 and 64 bit executables I have tested, but not wine. The program has successfully run and generated the right output.


what can I do?

thanks,
jerry

Re: [Valgrind-users] my cache factoring saga continues

[Josef W. <Jos...@gm...>]

Am 07.07.2014 01:28, schrieb Jerry Scharf:
> Hi again,
> 
> This is just such an awful thing to be doing but such are the cards I get handed.
> 
> The program I am trying to measure the cache performance on is actually a win32 progran that I am running under wine on redhat linux. I thought for a while the problem might have been that I had a stripped version of the program and that was causing a problem. After much aggravation, wine is now built locally with -g -O so I can debug if really necessary.
> 
> When I run it under cachegrind, it doesn't seem to generate the statistics.

I definitely do not want to just run the "wine" binary within the
simulator, as this is most
probable just a startup wrapper. Try with "--trace-children=yes".
This should give you an output file for every process started during the
program run.

Josef


> 
> here is the wine program:
> $ file `which wine`
> /usr/local/bin/wine: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.18, not stripped
> 
> here is the run:
> $ valgrind --tool=cachegrind --LL=16777216,16,64 wine ../../../../Program\ Files/LTC/LTspiceIV/scad3.exe -wine -b xxx.cir 
> ==13347== Cachegrind, a cache and branch-prediction profiler
> ==13347== Copyright (C) 2002-2012, and GNU GPL'd, by Nicholas Nethercote et al.
> ==13347== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
> ==13347== Command: wine ../../../../Program\ Files/LTC/LTspiceIV/scad3.exe -wine -b xxx.cir
> ==13347== 
> --13347-- warning: Unknown Intel cache config value (0x63), ignoring
> --13347-- warning: L3 cache found, using its data for the LL simulation.
> ==13347== Auto-detected LL cache configuration not supported: Cache set count is not a power of two.
> ==13347==   LL: 26,214,400 B, 20-way, 64 B lines
> 
> that's the end of the output. On other programs, there is the stats and miss percentage.
>  valgrind --tool=cachegrind --LL=16777216,16,64 pwd
> ==13662== Cachegrind, a cache and branch-prediction profiler
> ==13662== Copyright (C) 2002-2012, and GNU GPL'd, by Nicholas Nethercote et al.
> ==13662== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
> ==13662== Command: pwd
> ==13662== 
> --13662-- warning: Unknown Intel cache config value (0x63), ignoring
> --13662-- warning: L3 cache found, using its data for the LL simulation.
> ==13662== Auto-detected LL cache configuration not supported: Cache set count is not a power of two.
> ==13662==   LL: 26,214,400 B, 20-way, 64 B lines
> /yyy/.wine/drive_c/users/root/phitest/release
> ==13662== 
> ==13662== I   refs:      196,379
> ==13662== I1  misses:      1,003
> ==13662== LLi misses:        990
> ==13662== I1  miss rate:    0.51%
> ==13662== LLi miss rate:    0.50%
> ==13662== 
> ==13662== D   refs:       66,222  (50,912 rd   + 15,310 wr)
> ==13662== D1  misses:      2,464  ( 2,099 rd   +    365 wr)
> ==13662== LLd misses:      1,936  ( 1,603 rd   +    333 wr)
> ==13662== D1  miss rate:     3.7% (   4.1%     +    2.3%  )
> ==13662== LLd miss rate:     2.9% (   3.1%     +    2.1%  )
> ==13662== 
> ==13662== LL refs:         3,467  ( 3,102 rd   +    365 wr)
> ==13662== LL misses:       2,926  ( 2,593 rd   +    333 wr)
> ==13662== LL miss rate:      1.1% (   1.0%     +    2.1%  )
> 
> This works with both 32 and 64 bit executables I have tested, but not wine. The program has successfully run and generated the right output.
> 
> 
> what can I do?
> 
> thanks,
> jerry
> 
> ------------------------------------------------------------------------------
> Open source business process management suite built on Java and Eclipse
> Turn processes into business applications with Bonita BPM Community Edition
> Quickly connect people, data, and systems into organized workflows
> Winner of BOSSIE, CODIE, OW2 and Gartner awards
> http://p.sf.net/sfu/Bonitasoft
> _______________________________________________
> Valgrind-users mailing list
> Val...@li...
> https://lists.sourceforge.net/lists/listinfo/valgrind-users
>