Shoot, that's what I get for trying to fix things too fast. Thanks for the fix.
One thing Will, can you rebuild PAPI with --with-debug to configure and then set PAPI_DEBUG=SUBSTRATE? That will tell us a lot more information. But before you do that please run papi_native_avail and papi_avail from the test suite. They should each show a few events available....But I know there are problems with the umasks, will try and track that down. I wish I had access to some of this hardware. Phil On Tue, 2006-10-31 at 11:19 -0500, William Cohen wrote: > Philip J. Mucci wrote: > > Ok, I found the memory information error and fixed it in CVS, but the > > next error is troubling and it is due to the way we handle umasked > > events in PAPI. It fails on AMD because the first event has a umask. > > I have fixed the core dump though. > > > > I'm attempting a fix now. Can you give her another spin? let me know how > > it goes. > > > > Thanks > > Did a cvs update. > > strstr() returns a pointer if it is successful and NULL if there is NO match. > The attached patch should do the correct thing. > > Still get the following error when trying papi_command_line: > > $ ./utils/papi_command_line PAPI_TOT_CYC > Successfully added: PAPI_TOT_CYC > > PAPI Error: pfm_write_pmcs(3,0x2ba7cbd3c110,1): not supported. > PAPI Error: Error Code -3,PAPI_ESYS,A System/C library call failed: Invalid > argument. > command_line.c FAILED > Line # 64 > System error in PAPI_start: Invalid argument > > -Will > _______________________________________________ > perfmon mailing list > [email protected] > http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/ _______________________________________________ perfmon mailing list [email protected] http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
