Menu
▾
▴
perfsuite-bugs — Mailing list for PerfSuite bug reports
You can subscribe to this list here.
| 2004 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(1) |
Oct
|
Nov
|
Dec
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2005 |
Jan
(5) |
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
(2) |
Aug
|
Sep
|
Oct
|
Nov
(5) |
Dec
(2) |
| 2006 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(1) |
Jun
|
Jul
(2) |
Aug
(5) |
Sep
|
Oct
|
Nov
|
Dec
(4) |
| 2011 |
Jan
|
Feb
(1) |
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2012 |
Jan
(1) |
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2013 |
Jan
|
Feb
|
Mar
|
Apr
(1) |
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: Emmet C. <em...@ca...> - 2013-04-29 23:18:02
|
-- Emmet Caulfield (650) 733-6638 |
|
From: Mahin M. <mah...@ps...> - 2012-01-17 21:38:58
|
Hi,
I am building perfsuite on a UV system 'without' papi. Configuration
line is:
./configure --prefix=$prefix \
--with-tdom=$prefix/lib --without-libbfd \
--enable-java --disable-binutils --enable-rtc=mmtimer
But error is generated:
======================================
1: with PAPI, available PAPI preset event skipped
(papi-diff-loca.at:60)
2: with PAPI, unavailable PAPI preset event skipped
(papi-unav-pres.at:61)
3: with PAPI, native event skipped
(papi-native.at:61)
4: with PAPI, non-existent event skipped
(papi-non-exist.at:59)
5: without PAPI, PAPI preset event FAILED
(pres-diff-loca.at:57)
6: without PAPI, non-existent event FAILED
(non-exist.at:58)
## ------------- ##
## Test results. ##
## ------------- ##
ERROR: 2 tests were run,
2 failed unexpectedly.
4 tests were skipped.
## -------------------------- ##
## testsuite.log was created. ##
## -------------------------- ##
======================================
src/javalib/perfsuite/util/papi/tests/testsuite.log is attached FYI.
Should I be concerned about the error even not building with papi?
Thanks,
Mahin
|
|
From: Yuan T. <yua...@cs...> - 2011-02-25 04:05:58
|
-- - (\x.x x)(\x.x x) |
|
From: Rick K. <rk...@nc...> - 2006-12-02 00:11:20
|
Update on this issue: I have tracked down the problem and it looks as though the bug is in PerfSuite, not PAPI. It is a use of PAPI when profiling that did not trigger a runtime error in previous PAPI releases, but appears to do so now. A fix has also been developed that will appear in the next release of PerfSuite (hopefully not too far away). In the meantime, PerfSuite is still usable with PAPI 3.5.0, but in counting mode only, not profiling. Prior versions of PAPI should still work OK. Thanks very much for bringing this problem to my attention. Rick On Fri, 1 Dec 2006, Rick Kufrin wrote: > miwei, > > I have had a chance to build PAPI 3.5.0 on a Xeon workstation and test it > with PerfSuite. I'm able to confirm failures similar to what you > reported. > > The problem seems to arise when profiling, and the failure is due to an > error return from PAPI of PAPI_EINVAL, which means PAPI thinks it has > received an invalid argument to one of its library routines. This doesn't > happen in all uses of profiling, but does as you point out, and it happens > consistently. > > I'm not sure at this point whether the problem is in PerfSuite or in PAPI > 3.5.0 but will continue to investigate. > > I have a couple of suggestions in the meantime: > > 1. Use the latest PerfSuite 0.6.2 version, not 0.6.1 which is now rather > old. This won't fix the immediate problem but has many fixes and > enhancements that have been implemented. > > 2. Consider installing PAPI 3.2.1 in addition to 3.5.0 (place it in a > different directory) and try using PerfSuite with that. This > combination should be quite stable. > > Sorry for the problems, will post when I know more... > > Rick > >>> I run "make check" after perfsuite has been installed, but several fail >>> occurs. The tests to >>> papi and tdom works well after their installation. What 's wrong with it? >>> The testsuite.log >>> is attached. >>> Thanks for help >>> >>> -- > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share your > opinions on IT & business topics through brief surveys - and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > _______________________________________________ > PerfSuite-bugs mailing list > Per...@li... > https://lists.sourceforge.net/lists/listinfo/perfsuite-bugs > |
|
From: Rick K. <rk...@nc...> - 2006-12-01 19:30:36
|
miwei,
I have had a chance to build PAPI 3.5.0 on a Xeon workstation and test it
with PerfSuite. I'm able to confirm failures similar to what you
reported.
The problem seems to arise when profiling, and the failure is due to an
error return from PAPI of PAPI_EINVAL, which means PAPI thinks it has
received an invalid argument to one of its library routines. This doesn't
happen in all uses of profiling, but does as you point out, and it happens
consistently.
I'm not sure at this point whether the problem is in PerfSuite or in PAPI
3.5.0 but will continue to investigate.
I have a couple of suggestions in the meantime:
1. Use the latest PerfSuite 0.6.2 version, not 0.6.1 which is now rather
old. This won't fix the immediate problem but has many fixes and
enhancements that have been implemented.
2. Consider installing PAPI 3.2.1 in addition to 3.5.0 (place it in a
different directory) and try using PerfSuite with that. This
combination should be quite stable.
Sorry for the problems, will post when I know more...
Rick
>> I run "make check" after perfsuite has been installed, but several fail
>> occurs. The tests to
>> papi and tdom works well after their installation. What 's wrong with it?
>> The testsuite.log
>> is attached.
>> Thanks for help
>>
>> --
|
|
From: Rick K. <rk...@nc...> - 2006-12-01 17:45:23
|
Miwei, Thanks for sending the testsuite.log. Looking at it briefly, I wonder if the reason for the failures is that you are using PAPI version 3.5.0, which is the newest release of PAPI. PerfSuite has not yet been tested with PAPI 3.5.0 and so it's hard to say if any incompatibilities have arisen since the prior releases. I hesistate to recommend that you move back to PAPI 3.2.0 or PAPI 3.2.1 as I am sure there are many improvements in the newest release, but that is something to try if you choose. I'll take a look at PerfSuite's compatibility with the newest PAPI and update you when I know more. Rick On Fri, 1 Dec 2006, miwei wrote: > I run "make check" after perfsuite has been installed, but several fail > occurs. The tests to > papi and tdom works well after their installation. What 's wrong with it? > The testsuite.log > is attached. > Thanks for help > > -- > No virus found in this outgoing message. > Checked by AVG Free Edition. > Version: 7.1.409 / Virus Database: 268.15.2/560 - Release Date: 2006-11-30 > > > |
|
From: miwei <mi...@ic...> - 2006-12-01 08:14:20
|
I run "make check" after perfsuite has been installed, but several fail occurs. The tests to papi and tdom works well after their installation. What 's wrong with it? The testsuite.log is attached. Thanks for help -- No virus found in this outgoing message. Checked by AVG Free Edition. Version: 7.1.409 / Virus Database: 268.15.2/560 - Release Date: 2006-11-30 |
|
From: Rick K. <rk...@nc...> - 2006-08-31 13:02:57
|
Berkin, That error message: "a required software component is not installed" means that the PerfSuite build was not done correctly or that the configuration file is requesting a component (such as PAPI or Perfmon) that is not available on your system. If you run psrun without any configuration files, and if the XML document that results does not contain a listing of PAPI events, then you'll have to reinstall and you may want to review the INSTALL instruction file. Also, all of the supplied example configurations that being with "papi2_" are for PAPI release 2 (an older release which is probably not what you are using). Rick p.s. please remember to send email to the SourceForge mailing lists, which are the general mechanism for support. On Wed, 30 Aug 2006, Berkin Ozisikyilmaz wrote: > Hi, > > I have tried using a couple of the configuration files under > $PREFIX/share/perfsuite/xml/pshwpc, but i'm getting the following error. > How can I find what I'm missing? > > > berkin@CALC3:~/minebench/src/kmeans> psrun > -c /home/berkin/bin/share/perfsuite/xml/pshwpc/papi2_p6.xml /home/berkin/minebench/src/kmeans/example-icc9.1 -i / > home/berkin/minebench/src/kmeans/color -b -o -d -p 1 > libpsrun fatal error: a required software component not installed > > Regards, > Berkin |
|
From: Rick K. <rk...@nc...> - 2006-08-26 01:04:49
|
Berkin, On Fri, 25 Aug 2006, Berkin Ozisikyilmaz wrote: > Sorry for my late reply, I have been out of town for the last 2 weeks. Not to worry. It's summer and we are all in and out of town :) > I have used psprocess with the -e option, but still I can't get counter > values like branch instructions, conditional branch instructions > mispredictied, total translation lookaside buffer misses, level 1 > instruction cache misses etc. Is there a special flag that I need to use > while using psrun command? > It's still rather hard for me to see what the problem is or to diagnose, but I will tell you that based on the XML output you sent earlier in August, you had selected profiling using the "profil.xml" configuration file that's included in PerfSuite. That particular configuration does profiling based on the library routine profil(), which is a purely time-based profile, similar to what gprof uses. It will not profile based on hardware events such as you list but will only display time spent in the various portions of your program. To profile based on hardware events, you should select a different type of configuration file. Once you've installed PerfSuite, there will be a variety of example configurations installed in the directory: $PREFIX/share/perfsuite/xml/pshwpc You might want to take a look through these files (only concern yourself with the ones with a .xml extension) to see what they do. You can then use one of them, tailored as you like, by either: 1. setting the environment variable PS_HWPC_CONFIG to the pathname of the configuration file you've selected before running psrun on your executable, or 2. supplying the option "-c FILENAME" to psrun, where FILENAME is again the full path to the configuration file you want to use. Once you've run your program this way and obtained an XML output document, then use: psprocess -e YOURPROGRAM XML-OUTPUT-DOCUMENT to view the results. Please note that the current release of PerfSuite only allows profiling on one event per run. Hope this helps clarify operation, Rick |
|
From: Berkin O. <bo...@ec...> - 2006-08-25 16:59:43
|
Sorry for my late reply, I have been out of town for the last 2 weeks.=20 I have used psprocess with the -e option, but still I can't get counter values like branch instructions, conditional branch instructions mispredictied, total translation lookaside buffer misses, level 1 instruction cache misses etc. Is there a special flag that I need to use while using psrun command? Thank you Berkin Ozisikyilmaz -----Original Message----- From: Rick Kufrin [mailto:rk...@nc...]=20 Sent: Saturday, August 12, 2006 9:15 AM To: Berkin Ozisikyilmaz Cc: per...@li... Subject: RE: [PerfSuite-bugs] [PerfSuite 0.6.2a5] testsuite: 1 2 3 4 5 8 = 11 12 15 16 19 20 22 23 26 27 28 29 failed Berkin, The output XML document looks fine, but I am guessing the reason why you = did not get profiling information is that the "-e" option was omitted in = the psprocess command line. You might notice a warning prompt to that=20 effect in your second command below. The typical use of psprocess with profiling information (from, say, a run of "a.out") is: psprocess -e a.out a.out.12356.xml Rick On Mon, 31 Jul 2006, Berkin Ozisikyilmaz wrote: > Hi, > > It seems the problem was with perfctr's installation. Now the tests = seem to > be ok, but when I ran a program I don't get any statistics. The = runline and > the output xml file is attached. > > Thanks > Berkin > > berkin@CALC3:~/minebench/src/kmeans> psrun > /home/berkin/minebench/src/kmeans/example-icc9.1 -I color -o -b -d -p = 1 > > berkin@CALC3:~/minebench/src/kmeans> psprocess example-icc9.1.9447.CALC3.xml > no executable supplied (use the "-e PROGRAM" option) > > berkin@CALC3:~/perfsuite-0.6.2a5> make -s check Making check in src = Making > check in libperfsuite Making check in . Making check in tests ## |
|
From: Rick K. <rk...@nc...> - 2006-08-12 14:15:08
|
Berkin, The output XML document looks fine, but I am guessing the reason why you did not get profiling information is that the "-e" option was omitted in the psprocess command line. You might notice a warning prompt to that effect in your second command below. The typical use of psprocess with profiling information (from, say, a run of "a.out") is: psprocess -e a.out a.out.12356.xml Rick On Mon, 31 Jul 2006, Berkin Ozisikyilmaz wrote: > Hi, > > It seems the problem was with perfctr's installation. Now the tests seem to > be ok, but when I ran a program I don't get any statistics. The runline and > the output xml file is attached. > > Thanks > Berkin > > berkin@CALC3:~/minebench/src/kmeans> psrun > /home/berkin/minebench/src/kmeans/example-icc9.1 -I color -o -b -d -p 1 > > berkin@CALC3:~/minebench/src/kmeans> psprocess example-icc9.1.9447.CALC3.xml > no executable supplied (use the "-e PROGRAM" option) > > berkin@CALC3:~/perfsuite-0.6.2a5> make -s check Making check in src Making > check in libperfsuite Making check in . Making check in tests ## |
|
From: Rick K. <rk...@nc...> - 2006-08-12 14:05:11
|
Resending the reply below as it seems to have been lost in the SourceForge mailing list system. On Fri, 28 Jul 2006, Rick Kufrin wrote: > Berkin, > > This is most likely not a bug, but instead a problem due to either a > non-functional installation of PAPI on your system or misconfiguration of > PerfSuite. The test results show a pattern in which only PAPI-related > tests are failing. > > You might want to check that PAPI is working on your system by running its > test suite (check the PAPI installation notes for instructions). If all > PAPI tests run well, then try reviewing the PerfSuite INSTALL file. It's > important that PAPI is actually installed (not just built) and that the > --with-papi=DIR option to configure points to the top-level of the PAPI > installation directory (there should be an "include" and a "lib" directory > there). > > Rick > > On Fri, 28 Jul 2006, bo...@ec... wrote: > >> Hi, >> >> I could not find a testsuite.log. >> Here is what i get on the command line. >> >> Thanks >> Berkin >> > |
|
From: Rick K. <rk...@nc...> - 2006-07-29 00:21:11
|
Berkin, This is most likely not a bug, but instead a problem due to either a non-functional installation of PAPI on your system or misconfiguration of PerfSuite. The test results show a pattern in which only PAPI-related tests are failing. You might want to check that PAPI is working on your system by running its test suite (check the PAPI installation notes for instructions). If all PAPI tests run well, then try reviewing the PerfSuite INSTALL file. It's important that PAPI is actually installed (not just built) and that the --with-papi=DIR option to configure points to the top-level of the PAPI installation directory (there should be an "include" and a "lib" directory there). Rick On Fri, 28 Jul 2006, bo...@ec... wrote: > Hi, > > I could not find a testsuite.log. > Here is what i get on the command line. > > Thanks > Berkin > > berkin@CALC3:~/perfsuite-0.6.2a5> make -s check > Making check in src > Making check in libperfsuite > Making check in . > Making check in tests > ## ------------------------------------------- ## > ## PerfSuite 0.6.2a5 test suite: Core Library. ## > ## ------------------------------------------- ## > 1: Hello, world! ok > 2: CPU identification ok > 3: CPU speed information ok > 4: CPU speed information (Fortran) ok > 5: CPU usage ok > 6: tokenizer ok > 7: error reporting ok > 8: error reporting (Fortran) ok > 9: argument processing ok > 10: proc filesystem ok > 11: memory usage reporting ok > 12: memory usage reporting (Fortran) ok > 13: dynamic memory usage reporting ok > 14: hash tables ok > 15: load map address inquiry ok > > ## ------------- ## > ## Test results. ## > ## ------------- ## > > All 15 tests were successful. > Making check in libpshwpc > Making check in . > Making check in arch-defaults > Making check in xml > Making check in css > Making check in dtds > Making check in xsl > Making check in tests > ## ----------------------------------------------------------- ## > ## PerfSuite 0.6.2a5 test suite: Hardware Performance Library. ## > ## ----------------------------------------------------------- ## > 1: base PAPI functionality FAILED (PAPI.at:47) > 2: PAPI compatibility FAILED (papicompat.at:47) > 3: PAPI to PerfSuite conversion (C/Fortran) FAILED (PAPI-PS.at:47) > 4: PAPI with POSIX threads (non-multiplexed) FAILED (PAPI-nompx.at:47) > 5: PAPI with POSIX threads (multiplexed) FAILED (PAPI-mpx.at:47) > 6: XML input document parsing (counting) ok > 7: XML input document parsing (profiling) ok > 8: inquiry (numevents) FAILED (inquiry.at:61) > 9: inquiry (numevents) error handling ok > 10: inquiry (numevents = 0) ok > 11: serial program (counting) FAILED (serial.at:73) > 12: serial program (profiling) FAILED (serial.at:85) > 13: serial program (profil) ok > 14: serial program (itimer) ok > 15: serial program (counting auto) FAILED (serial.at:121) > 16: serial program (profiling auto) FAILED (serial.at:133) > 17: serial program (profil auto) ok > 18: serial program (itimer auto) ok > 19: serial program (counting Fortran) FAILED (serial.at:167) > 20: serial program (profiling Fortran) FAILED (serial.at:179) > 21: serial program (profil Fortran) ok > 22: serial program (counting Fortran auto) FAILED (serial.at:202) > 23: serial program (profiling Fortran auto) FAILED (serial.at:214) > 24: serial program (profil Fortran auto) ok > 25: serial program (itimer Fortran auto) ok > 26: fork (counting) FAILED (fork.at:53) > 27: fork (profiling) FAILED (fork.at:63) > 28: POSIX threads program (counting) FAILED (threads.at:56) > 29: POSIX threads program (profiling) FAILED (threads.at:69) > > ## ------------- ## > ## Test results. ## > ## ------------- ## > > ERROR: All 29 tests were run, > 18 failed unexpectedly. > ## -------------------------- ## > ## testsuite.log was created. ## > ## -------------------------- ## > > Please send `testsuite.log' and all information you think might help: > > To: <per...@li...> > Subject: [PerfSuite 0.6.2a5] testsuite: 1 2 3 4 5 8 11 12 15 16 19 20 22 23 > 26 27 28 29 failed > > > You may investigate any problem if you feel able to do so, in which > case the test suite provides a good starting point. > > make[4]: *** [check-local] Error 1 > make[3]: *** [check-am] Error 2 > make[2]: *** [check-recursive] Error 1 > make[1]: *** [check-recursive] Error 1 > make: *** [check-recursive] Error 1 > > _________________________________________________________ > This message was sent through the NU ECE webmail gateway. > |
|
From: <bo...@ec...> - 2006-07-28 16:15:41
|
Hi, I could not find a testsuite.log. Here is what i get on the command line. Thanks Berkin berkin@CALC3:~/perfsuite-0.6.2a5> make -s check Making check in src Making check in libperfsuite Making check in . Making check in tests ## ------------------------------------------- ## ## PerfSuite 0.6.2a5 test suite: Core Library. ## ## ------------------------------------------- ## 1: Hello, world! ok 2: CPU identification ok 3: CPU speed information ok 4: CPU speed information (Fortran) ok 5: CPU usage ok 6: tokenizer ok 7: error reporting ok 8: error reporting (Fortran) ok 9: argument processing ok 10: proc filesystem ok 11: memory usage reporting ok 12: memory usage reporting (Fortran) ok 13: dynamic memory usage reporting ok 14: hash tables ok 15: load map address inquiry ok ## ------------- ## ## Test results. ## ## ------------- ## All 15 tests were successful. Making check in libpshwpc Making check in . Making check in arch-defaults Making check in xml Making check in css Making check in dtds Making check in xsl Making check in tests ## ----------------------------------------------------------- ## ## PerfSuite 0.6.2a5 test suite: Hardware Performance Library. ## ## ----------------------------------------------------------- ## 1: base PAPI functionality FAILED (PAPI.at:47) 2: PAPI compatibility FAILED (papicompat.at:47) 3: PAPI to PerfSuite conversion (C/Fortran) FAILED (PAPI-PS.at:47) 4: PAPI with POSIX threads (non-multiplexed) FAILED (PAPI-nompx.at:47) 5: PAPI with POSIX threads (multiplexed) FAILED (PAPI-mpx.at:47) 6: XML input document parsing (counting) ok 7: XML input document parsing (profiling) ok 8: inquiry (numevents) FAILED (inquiry.at:61) 9: inquiry (numevents) error handling ok 10: inquiry (numevents = 0) ok 11: serial program (counting) FAILED (serial.at:73) 12: serial program (profiling) FAILED (serial.at:85) 13: serial program (profil) ok 14: serial program (itimer) ok 15: serial program (counting auto) FAILED (serial.at:121) 16: serial program (profiling auto) FAILED (serial.at:133) 17: serial program (profil auto) ok 18: serial program (itimer auto) ok 19: serial program (counting Fortran) FAILED (serial.at:167) 20: serial program (profiling Fortran) FAILED (serial.at:179) 21: serial program (profil Fortran) ok 22: serial program (counting Fortran auto) FAILED (serial.at:202) 23: serial program (profiling Fortran auto) FAILED (serial.at:214) 24: serial program (profil Fortran auto) ok 25: serial program (itimer Fortran auto) ok 26: fork (counting) FAILED (fork.at:53) 27: fork (profiling) FAILED (fork.at:63) 28: POSIX threads program (counting) FAILED (threads.at:56) 29: POSIX threads program (profiling) FAILED (threads.at:69) ## ------------- ## ## Test results. ## ## ------------- ## ERROR: All 29 tests were run, 18 failed unexpectedly. ## -------------------------- ## ## testsuite.log was created. ## ## -------------------------- ## Please send `testsuite.log' and all information you think might help: To: <per...@li...> Subject: [PerfSuite 0.6.2a5] testsuite: 1 2 3 4 5 8 11 12 15 16 19 20 22 23 26 27 28 29 failed You may investigate any problem if you feel able to do so, in which case the test suite provides a good starting point. make[4]: *** [check-local] Error 1 make[3]: *** [check-am] Error 2 make[2]: *** [check-recursive] Error 1 make[1]: *** [check-recursive] Error 1 make: *** [check-recursive] Error 1 _________________________________________________________ This message was sent through the NU ECE webmail gateway. |
|
From: Rick K. <rk...@nc...> - 2006-05-04 02:14:23
|
Henry,
Thanks very much for testing this out and reporting the problem. You
located the problem spot perfectly and yes, unfortunately it is a bug in
psprocess. I'm going to copy the list on this in case others run across
the same issue and will package an a4 release to fix this problem as soon
as possible. The combine operation is very common for people to use so
this shouldn't stay unaddressed for long. Fortunately, there is a quick
fix that can be applied to combine.tcl either before or after
installation. At the beginning of the source code in combine.tcl,
lines 56-60 are:
proc readAndValidateMultiple {fileList ignoreErrors} {
global timers
set res {}
set doctype ""
The following line should be added after line 60 ("set doctype"):
set timers(total,domparse) 0
Within the distribution, combine.tcl can be found in the directory:
src/tools/psprocess/tcl
After installation, it will be located in:
$PREFIX/share/perfsuite/tclbin/psprocess
This is a bug that was introduced with a new capability of this release
(measuring the time necessary to do XML document processing).
Sorry I missed this before the release! But thanks again for the report
and also the very good news that things appear to work otherwise on
Columbia.
Rick
On Wed, 3 May 2006, Haoqiang H. Jin wrote:
> Hi Rick,
>
> Glad to see you had the new version released. I installed
> this version on our Columbia system. Except for a minor
> problem, it worked pretty much out of the box. The minor
> glitch occurred when I tried psprocess in the combined mode.
>
> % psprocess -c a.out.0.m.xml a.out.1.m.xml
> can't read "timers(total,domparse)": no such variable
>
> The error seems to be generated from 'combine.tcl'
> in line 78:
> incr timers(total,domparse) [expr {$stoptime - $starttime}]
>
> If I comment out this line or replace 'incr' with 'set',
> psprocess ran successfully. Is this a bug or something
> else that I should specify in the command line?
>
> Thanks for your great work to make the package available!
>
> -Henry
|
|
From: Rick K. <rk...@nc...> - 2005-12-07 03:14:21
|
On Tue, 6 Dec 2005 gg...@un... wrote: > I am using perfsuite-0.6.1 with GNU compilers gcc and g77 and with mpich.1.2.7. > It seems that all works properly, but when I run > > % /usr/local/mpich-1.2.7/bin/mpirun -np 2 psrun cpi-mpi > > the answer is > > Unrecognized argument psrun ignored. > Process 0 on gemelli > Process 1 on gemelli > pi is approximately 3.1415926535899814, Error is 0.0000000000001883 > wall clock time = 1.040470 > > What happens? Thanks in advance. Giuseppe, This is a situation that depends on the version of MPI you are using and is a common problem. The output "Unrecognized argument" implies that the MPI launch command mpirun is interpreting psrun as an option (which it doesn't recognize and therefore ignores). You might try supplying the full pathname to psrun, for example: /usr/local/mpich-1.2.7/bin/mpirun -np 2 /usr/local/bin/psrun cpi-mpi Hope this helps, Rick p.s. I am redirecting your question from perfsuite-bugs to perfsuite-users because I think this is of general interest. |
|
From: <gg...@un...> - 2005-12-06 09:28:45
|
I am using perfsuite-0.6.1 with GNU compilers gcc and g77 and with mpich.= 1.2.7. It seems that all works properly, but when I run % /usr/local/mpich-1.2.7/bin/mpirun -np 2 psrun cpi-mpi the answer is Unrecognized argument psrun ignored. Process 0 on gemelli Process 1 on gemelli pi is approximately 3.1415926535899814, Error is 0.0000000000001883 wall clock time =3D 1.040470 What happens? Thanks in advance. Dr. Giuseppe Grieco DIFA Universit=E0 della Basilicata |
|
From: Rick K. <rk...@nc...> - 2005-11-30 15:05:14
|
Giuseppe, On Wed, 30 Nov 2005 gg...@un... wrote: > I arrived at the same conclusion when looking that it works with g77. Now I > would like to know if I can configure with both ifort and icc and if I have just > to add -F77=ifort -GCC=icc after all in the configure line. > In addition, may I have perfsuite configured for both GNU and Intel compilers? > Thanks, Giuseppe. This is a reasonable question and in fact something that we have to do here, in a multi-compiler environment. The thing to remember is that the important point is the Fortran-C calling convention, which is different for g77 and ifort. For C programs, it shouldn't matter. PerfSuite should always be built with gcc (not icc). What works for us is to build two separate versions of the libraries, one to be used with g77-compiled Fortran programs, the other for ifort-compiled Fortran programs. We request a separate subdirectory in the final install that is specific to the Intel version. The general "template" is something like this: 1% ./configure <options> 2% make 3% make install 4% ./configure <options> --libdir=/usr/local/lib/intel F77=ifort 5% make 6% make install Steps 1-3 build a g77-compatible version of PerfSuite and installs the libraries in /usr/local/lib. Steps 4-6 build an ifort-compatible version of PerfSuite and installs the libraries in /usr/local/lib/intel. When linking an application, you select the appropriate one and also set the LD_LIBRARY_PATH variable according to the Fortran compiler you are using. Commands like psrun should work regardless, what I am describing here is only needed for applications that use the PerfSuite API (like the matvec example). I hope this clears up a bit more of the confusion. Thanks for hanging in there! Rick |
|
From: Rick K. <rk...@nc...> - 2005-11-30 14:21:44
|
On Wed, 30 Nov 2005 gg...@un... wrote: > Hi Rick, > I made a syntax error but the music does not change, even if I link with -lpapi > -lpshwpc or with -lpapi -lpshwpc -libperfsuite. I think I have these libraries > in /usr/local/lib. I attach a file listing the content of the directory. > Thanks, Giuseppe. Giuseppe, Well, I think I missed one other thing in your config.log. It is the selection of the Fortran compiler to be used when building the libraries. It looks as if you are using the default, which is GNU g77, not Intel ifort. This will cause link time problems similar to what you are seeing. The solution for this is to reconfigure and rebuild PerfSuite exactly as you've already done, but add the flag F77=ifort at the very end of your "configure" line. Then do "make clean", "make", and "make install" as before. There is more information about this in the INSTALL file. Hope this helps! Rick |
|
From: Rick K. <rk...@nc...> - 2005-11-30 13:54:39
|
Giuseppe, This sounds like good progress. Thanks for trying the link again and letting me know that your compiler is new. Your config.log looks fine, I see that it successfully found tDOM, Tcl, PAPI, and MPI, which is good news. Regarding the link error you mention, I see on your compile/link line that you request "-lpwhwpc", which is incorrect (it should be "-lpshwpc"). Also, you might want to make sure that the PerfSuite libraries have been installed into /usr/local (-L/usr/local/lib) through "make install". The errors you are getting are typical of the case when libraries and routines contained within them are not found. One other thing - it is possible that you may need to explicitly link in libperfsuite as well. You will know that you need to do this if you find similar errors that cannot find the routine "psf_strerror_" and/or "ps_strerror". You can request libperfsuite by adding the flag -lperfsuite to your link line immediately after -lpshwpc. I hope this helps resolve the problem - good luck! Rick On Wed, 30 Nov 2005 gg...@un... wrote: > Dear Rick, > first of all thanks for your support. > When I made > > % make -s check > > I did not find any message of error. > I can build and run the matvec example without perfsuite call. > When I print > > % nm -g matvec | grep -i main > > the answer is > > U __libc_start_main@@GLIBC_2.0 > 0804a280 T main > 0804a2d4 T MAIN__ > > Another thing that happens and that I cannot explain is that now I am not more > able to reproduce an executable for ps that gives me again the same message of > error I wrote you. I can compile by > > % ifort -c matvec-ps.f -o matvec-ps.o -g -ip -xK -pad -sox -I/usr/local/include > > but when I try to link with > > % ifort matvec-ps.o -L/usr/local/lib -L/usr/local/lib -lpwhwpc -lpapi > > the answer is > > IPO link: can not find -lpwhwpc > IPO Error: unresolved : psf_hwpc_init_ > Referenced in matvec-ps.o > IPO Error: unresolved : psf_hwpc_start_ > Referenced in matvec-ps.o > IPO Error: unresolved : psf_hwpc_stop_ > Referenced in matvec-ps.o > IPO Error: unresolved : psf_hwpc_shutdown_ > Referenced in matvec-ps.o > ld: cannot find -lpwhwpc > > Yesterday I installed an Intel 9.0 C++ compiler. Did it do any thing? I attach > you the file you asked me. Thanks again, Giuseppe. > > |
|
From: Rick K. <rk...@nc...> - 2005-11-29 23:16:18
|
Giuseppe,
I've tried to reproduce your error today using the Intel 9.0 compilers on=
=20
both Itanium and Xeon systems but unfortunately cannot get the same=20
behavior. That is, I get no link or runtime errors when building the
matvec example.
The message that you are referring to in the INSTALL file is a little bit=
=20
confusing, I think. It is primarily directed towards problems that were=20
encountered in building the PerfSuite test suite, not example programs or=
=20
actual user applications. The issue comes down to: how do you link=20
programs properly that mix Fortran and C? For the matvec example, this=20
should not be an issue as it is completely a Fortran program.
Can you build and run the matvec example that contains no calls to the=20
PerfSuite library? It should be as simple as doing:
# ifort -o matvec matvec.f
If you can, then we can assume your Fortran compiler is working properly.
The "MAIN_" symbol should actually be contained in your executable, in=20
this case "matvec". You can see if it is there by running the command
"nm -g matvec | grep -i main". My output for that command is:
U __libc_start_main@@GLIBC_2.0
08049b70 T main
08049bbc T MAIN__
I'll wait to hear back from you about these experiments before we try=20
further. Thanks for your patience and reports.
Rick
p.s. just to reiterate, you can ignore the sentences in the INSTALL file=20
about modifying "config.h". This was applicable only to the build of=20
PerfSuite itself. If your test suite runs to completion after you have=20
built PerfSuite (make -s check), your build should be OK. If not, then I=
=20
would like to see your "config.log" file if you still have it available.
> Hi,
> I installed perfsuite 0.6.1 with Intel Fortran 9.0 for Linux. I try to =
run the
> examples matvec-ps which links to lpshwpc and lpapi. The message I rece=
ive is=20
>=20
> /opt/intel/fc/9.0/lib/for_main.o(.text+0x41): In function `main':
> : undefined reference to `MAIN__'
>=20
> This kind of error is reported in the INSTALL file of perfsuite, and it=
suggests
> to hand-edit the config.h file with
>=20
> #define F77_MAIN MAIN__
>=20
> and making it again.=20
> The fact is that my config.h file has this line, but the error still re=
mains.
> What to do? Thanks in advance.
>=20
> Dr. Giuseppe Grieco
> DIFA Universit=E0 della Basilicata
>=20
>=20
|
|
From: <gg...@un...> - 2005-11-28 14:05:49
|
Hi, I installed perfsuite 0.6.1 with Intel Fortran 9.0 for Linux. I try to ru= n the examples matvec-ps which links to lpshwpc and lpapi. The message I receiv= e is=20 /opt/intel/fc/9.0/lib/for_main.o(.text+0x41): In function `main': : undefined reference to `MAIN__' This kind of error is reported in the INSTALL file of perfsuite, and it s= uggests to hand-edit the config.h file with #define F77_MAIN MAIN__ and making it again.=20 The fact is that my config.h file has this line, but the error still rema= ins. What to do? Thanks in advance. Dr. Giuseppe Grieco DIFA Universit=E0 della Basilicata |
|
From: Rick K. <rk...@nc...> - 2005-07-09 05:01:30
|
Edwin, Thank you for reporting this and also for the nice comment about PerfSuite, I'm very glad you find it useful and took the time to give it a try. Any of this type of report or suggestions for improvement are always welcome. I agree with you that it is probably not a bug per se, but a configuration file issue that should be corrected for future releases. I'm glad you were able to work it through without much problem. In general, the PAPI development team often updates their event definitions to improve them and make sure they're counting things as best as possible, and I bet this is what happened for PAPI_L2_DCR on the P4, and PerfSuite got a bit behind on what's the latest! Thanks again, Rick > Thank you for a very interesting and powerful tool! > > I'm not sure if this is a bug or not, but on my system I > had to disable the following line in papi3_p4.xml, v 1.4.2.3: > > <!-- =================================================== > This does not work on my P4!! > <ps_hwpc_event type="preset" name="PAPI_L2_DCR" /> > --> > > When I run a program using psrun and the unaltered (original) XML file, I > get this error message: > > => psrun -c papi3_p4.xml testCompare > psrun fatal error: hardware event not available > > When I remove the PAPI_L2_DCR event, it runs fine. > > I am using the following configuration: > > - Red Hat Enterprise Linux, kernel version 2.4.21-20.EL, patched with > perfctr-2.6.15 > - PAPI 3.0.7 > - PerfSuite 0.6.1 > - gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-42) > > The PerfSuite test suite did not have any errors. However, the PAPI > tests did show initially fail a test with overflow3_pthreads.c, > but I just ran them again and it now passes all of them. Maybe a recompile > fixed it, > who knows? > > Here are my system details (from psinv): > > System Information - > Processors: 1 > Total Memory (MB): 1002.93 > System Page Size (KB): 4.00 > > Processor Information - > Vendor: Intel > Processor family: Pentium 4 > Brand: Intel(R) Pentium(R) 4 CPU 2.66GHz > Model (Type): Pentium 4 (OEM) > Revision: 7 > Clock Speed: 2657.88 MHz > > Cache and TLB Information - > Cache levels: 2 > Caches/TLBs: 5 > > Cache Details - > Level 1: > Type: Data > Size: 8 KB > Line size: 64 bytes > Associativity: 4-way set associative > > Type: Instruction Trace > Size: 12K uOps > Associativity: 8-way set associative > > Level 2: > Type: Unified > Size: 512 KB > Line size: 64 bytes > Associativity: 8-way set associative > > TLB Details - > Level 1: > Type: Instruction > Entries: 128 > Pagesize (KB): 4 2048 4096 > Associativity: Fully associative > > Type: Data > Entries: 64 > Pagesize (KB): 4 4096 > Associativity: Fully associative > > > Regards, > Edwin Fine > Tel: (813) 978-4718 (W) > e-mail: edw...@ve... > ------------------------------ |
|
From: <edw...@ve...> - 2005-07-07 14:14:04
|
Thank you for a very interesting and powerful tool!
I'm not sure if this is a bug or not, but on my system I
had to disable the following line in papi3_p4.xml, v 1.4.2.3:
<!-- ===================================================
This does not work on my P4!!
<ps_hwpc_event type="preset" name="PAPI_L2_DCR" />
-->
When I run a program using psrun and the unaltered (original) XML file, I
get this error message:
=> psrun -c papi3_p4.xml testCompare
psrun fatal error: hardware event not available
When I remove the PAPI_L2_DCR event, it runs fine.
I am using the following configuration:
- Red Hat Enterprise Linux, kernel version 2.4.21-20.EL, patched with
perfctr-2.6.15
- PAPI 3.0.7
- PerfSuite 0.6.1
- gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-42)
The PerfSuite test suite did not have any errors. However, the PAPI
tests did show initially fail a test with overflow3_pthreads.c,
but I just ran them again and it now passes all of them. Maybe a recompile
fixed it,
who knows?
Here are my system details (from psinv):
System Information -
Processors: 1
Total Memory (MB): 1002.93
System Page Size (KB): 4.00
Processor Information -
Vendor: Intel
Processor family: Pentium 4
Brand: Intel(R) Pentium(R) 4 CPU 2.66GHz
Model (Type): Pentium 4 (OEM)
Revision: 7
Clock Speed: 2657.88 MHz
Cache and TLB Information -
Cache levels: 2
Caches/TLBs: 5
Cache Details -
Level 1:
Type: Data
Size: 8 KB
Line size: 64 bytes
Associativity: 4-way set associative
Type: Instruction Trace
Size: 12K uOps
Associativity: 8-way set associative
Level 2:
Type: Unified
Size: 512 KB
Line size: 64 bytes
Associativity: 8-way set associative
TLB Details -
Level 1:
Type: Instruction
Entries: 128
Pagesize (KB): 4 2048 4096
Associativity: Fully associative
Type: Data
Entries: 64
Pagesize (KB): 4 4096
Associativity: Fully associative
Regards,
Edwin Fine
Tel: (813) 978-4718 (W)
e-mail: edw...@ve...
------------------------------ |
|
From: Rick K. <rk...@nc...> - 2005-01-08 18:17:39
|
Hello again, I'll try to combine information about your two reports with testsuite output (thank you for including) into one reply. There are several things going on here. Regarding building PerfSuite with Intel compilers: you're correct that the inline assembly in the timing routines (and, on x86/x86-64, the processor detection routines) will cause it to choke. The solution to this is to build PerfSuite with gcc but to cause the Intel Fortran/C calling convention to be used through the variable F77. In other words, PerfSuite itself should always be built with gcc, but that doesn't restrict its use with other compilers. There's very little computation done by PerfSuite, so any speed differences on IA-64 with icc/gcc don't really matter (that's not the case, of course with the application code!). So all you should have to do is remove "CC=icc" and "CXX=icpc" from your configure line, but retain "F77=ifort", in order to pick up the ifort/C calling convention. At NCSA, we typically build dual builds of PerfSuite to handle both g77 and Intel compilers, with the Intel libraries installed in $PREFIX/lib/intel. Regarding the other test case failures: these, I believe, are due to the version of PAPI you're using, which appears to be 2.3.4.3. There have been a number of changes in PAPI, and the most recent version (3.0.7) has many fixes to problems that existed before. In fact, just yesterday a problem was identified and fixed that caused incorrect results (failures) on Altix/Itanium 2. My experience has been that the two problem spots in PAPI 2 have been a) multiplexing and b) pthreads. Again, the PAPI development team has made numerous improvements in PAPI to address these spots and so the best thing to do is to upgrade to PAPI 3.0.7 from the UTK/ICL web site. You might even want to wait for 3.0.8 since changes like the one yesterday are probably still only in the PAPI CVS tree. I hope these suggestions help - please let me know if problems still persist, and thanks for the very complete problem report! Rick On Fri, 7 Jan 2005, Constantinos Evangelinos wrote: > On an IA-64 system running RedHat EL AS 3, with a patched 2.4.21 kernel.org > kernel (with the perfmon and SCORE patches) and PAPI 2.4.3.4 build with the > Intel compiler (v. 8.1), configuring Perfsuite with the Intel C/C++ compiler > (icc -no-gcc -cxxlib-gcc) fails when building timer.c because of and tricks > to use gcc in libtool only for timers.c fall flat on their face later on... > > Perfsuite was configured as: > > MPICPPFLAGS="-I/usr/local/pkg-ia64/mpich-vmi/mpich-vmi-1.2.5/intel/include" > CC=icc CFLAGS=-O2 CXX=icpc CXXFLAGS=-O2 F77=ifort FFLAGS=-O2 > PTHREAD_LIBS=-lpthread ./configure > --with-tdom=/usr/local/pkg-ia64/tdom/tdom-0.8.1 > --with-papi=/usr/local/pkg-ia64/papi/papi-2.3.4.3 --enable-tests --enable-mpi > --prefix=/usr/local/pkg-ia64/perfsuite/perfsuite-0.6.1rc1/papi2 > > The error is as follows: > icc -DHAVE_CONFIG_H -I. -I. -I../.. -O2 -MT timers.lo -MD -MP -MF > .deps/timers.Tpo -c timers.c -fPIC -DPIC -o .libs/timers.o > timers.c(315): error: asm statements not supported in this environment > rdtscll(tmp); > ^ > > timers.c(315): error: expected a ")" > rdtscll(tmp); > ^ > > and reconfiguring with icc -no-gcc -cxxlib-gcc etc. does not solve the > problem. > > |
