[Users] PAPI on stampede/mic

Erik Schnetter eschnetter at perimeterinstitute.ca
Sun Jan 12 21:08:19 CST 2014


Updating to PAPI 5.3 should be fine, assuming it compiles on the other systems as well.

I think I added "unset CPP" before "unset LIBS" for PAPI 5.2; maybe this is also necessary for PAPI 5.3.

-erik

On Jan 12, 2014, at 21:52 , Roland Haas <rhaas at tapir.caltech.edu> wrote:

> Signed PGP part
> Hello Erik, all,
> 
> it seems that PAPI 5.3 has options to compile for the mics. One only
> has to pass --with-mic to configure. Are you aware of
> incompatibilities of PAPI 5.3 or should it be fine to update the
> tarball in ExternalLibraries? (CC'd to users at einsteintoolkit.org)
> 
> I can compile PAPI 5.3 on stampede and run the papi.par test in
> ExternalLibraries/PAPI and get Cache misses (but no timing, but lots
> of event-does-not exist warnings):
> 
> INFO (PAPI): PAPI_register_clock
> INFO (PAPI): Running local single-threaded DGEMM benchmark with N=1000...
> INFO (PAPI): Floating point operations: 2 Gflop over 1 threads
> INFO (PAPI): Pure performance:          0.651531 Gflop/sec
> INFO (PAPI): Overall performance:       0.643122 Gflop/sec
> INFO (PAPI): Running local multi-threaded DGEMM benchmark with N=1000...
> INFO (PAPI): Floating point operations: 488 Gflop over 244 threads
> INFO (PAPI): Pure performance:          36.7471 Gflop/sec
> INFO (PAPI): Overall performance:       27.8211 Gflop/sec
> INFO (PAPI): NOTE: These performance numbers do not have to be large,
> but they need to agree with what PAPI reports
> INFO (PAPI): PAPI Statistics:
>    CPU time                5042.929817665 sec
>    CPU cycles              5520.440487605 Gcyc
>    PAPI_TOT_CYC               0.324841654 cyc^(-1)
>    PAPI_TOT_INS               0.027041240 cyc^(-1)
>    PAPI_L1_ICA                0.033998239 cyc^(-1)
>    PAPI_L1_ICM                0.000011858 cyc^(-1)
>    PAPI_L1_DCA                0.011631986 cyc^(-1)
>    PAPI_L1_DCM                0.000893145 cyc^(-1)
>    PAPI_LD_INS              562.116089268 cyc^(-1)
>    PAPI_SR_INS                0.001194933 cyc^(-1)
> INFO (PAPI): PAPI Statistics:
>    CPU time                   4.165410795 sec
>    CPU cycles                 4.559431145 Gcyc
>    PAPI_TOT_CYC             393.906331451 cyc^(-1)
>    PAPI_TOT_INS              32.842575465 cyc^(-1)
>    PAPI_L1_ICA               41.205762993 cyc^(-1)
>    PAPI_L1_ICM                0.014362943 cyc^(-1)
>    PAPI_L1_DCA               14.119073876 cyc^(-1)
>    PAPI_L1_DCM                1.093455247 cyc^(-1)
>    PAPI_LD_INS           680595.542337835 cyc^(-1)
>    PAPI_SR_INS                1.448845164 cyc^(-1)
> 
> 
> Yours,
> Roland
> 
> --
> My email is as private as my paper mail. I therefore support encrypting
> and signing email messages. Get my PGP key from http://keys.gnupg.net.
> 
> _______________________________________________
> Users mailing list
> Users at einsteintoolkit.org
> http://lists.einsteintoolkit.org/mailman/listinfo/users

-- 
Erik Schnetter <eschnetter at perimeterinstitute.ca>
http://www.perimeterinstitute.ca/personal/eschnetter/

My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu/.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 203 bytes
Desc: Message signed with OpenPGP using GPGMail
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20140112/9b9a9402/attachment.bin 


More information about the Users mailing list