flpins = 0; MFLOPS = 0.0000

Open discussion of PAPI.

flpins = 0; MFLOPS = 0.0000

Postby zixzie » Wed Mar 31, 2010 2:09 pm

Hi,
I ran a very simple test flops.c (available here http://spiral.cs.drexel.edu/wiki/Papi370Ubuntu910).
This was what I got:
Real_time: 8.358779 Proc_time: 13.935387 Total flpins: 0 MFLOPS: 0.000000
Can anyone explain to me why the flpins and MFLOPS is 0?
My system is Ubuntu 9.10 (2.6.31) with an Intel i7.
Thanks.
Last edited by zixzie on Fri Apr 02, 2010 2:16 pm, edited 1 time in total.
zixzie
 
Posts: 2
Joined: Wed Mar 31, 2010 2:00 pm

Re: flpins = 0.0000; MFLOPS = 0.0000

Postby jagode00 » Wed Mar 31, 2010 3:05 pm

I ran the same flops.c test on our i7 machine:
$ uname -a
Linux gonzo 2.6.31-perfctr #1 SMP Thu Sep 24 10:17:37 EDT 2009 x86_64 GNU/Linux

testing both cases, perfctr and perf_events. In both cases I get reasonable values for flpins and MFLOPS.
perfctr v2.6:
Real_time: 7.691261 Proc_time: 7.681456 Total flpins: 2001175424 MFLOPS: 260.520325
perf_events:
Real_time: 7.628023 Proc_time: 7.625648 Total flpins: 2000166656 MFLOPS: 262.294647

Two questions:
1) What PAPI version are you using?
2) What do you get when you run one of PAPI's ctests, e.g. <papi-src-dir>/ctests/first?

Thanks,
heike.
jagode00
 
Posts: 29
Joined: Tue Aug 25, 2009 2:12 pm

Re: flpins = 0; MFLOPS = 0.0000

Postby zixzie » Fri Apr 02, 2010 2:24 pm

Hi,
1. My PAPI version is 3.7.2
2. The result of running first.c
Code: Select all
Test case 1: Non-overlapping start, stop, read.
-----------------------------------------------
Default domain is: 1 (PAPI_DOM_USER)
Default granularity is: 1 (PAPI_GRN_THR)
Using 20000000 iterations of c += a*b
-------------------------------------------------------------------------
Test type   :        1           2           3           4           5
PAPI_FP_INS:            0            0            0            0            0
PAPI_TOT_CYC:     63611959     60030384    120361818    180383830    180383830
-------------------------------------------------------------------------
Verification:
Row 1 Column 1 at least 20000000
% difference between PAPI_FP_INS: 1 & 2: nan
% difference between PAPI_TOT_CYC 1 & 2: 105.97
Column 1 approximately equals column 2
Column 3 approximately equals 2 * column 2
Column 4 approximately equals 3 * column 2
Column 4 exactly equals column 5
first.c                                  FAILED
Line # 150
Error: PAPI_FP_INS

It failed to run first.c, due to PAPI_FP_INS.
Thanks.
I was unable to upload an attachment.
zixzie
 
Posts: 2
Joined: Wed Mar 31, 2010 2:00 pm

Re: flpins = 0; MFLOPS = 0.0000

Postby jagode00 » Thu Apr 15, 2010 4:53 pm

Sorry that it took me so long to come back to you!

I have tested with the same PAPI version (3.7.2.) on our i7 machine. I still cannot reproduce what you are seeing. Here is the output I get for the first ctest:

Test case 1: Non-overlapping start, stop, read.
-----------------------------------------------
Default domain is: 1 (PAPI_DOM_USER)
Default granularity is: 1 (PAPI_GRN_THR)
Using 20000000 iterations of c += a*b
-------------------------------------------------------------------------
Test type : 1 2 3 4 5
PAPI_FP_INS: 40000541 40000573 80001099 120001671 120001671
PAPI_TOT_CYC: 62829633 62714677 125310882 188043650 188043650
-------------------------------------------------------------------------
Verification:
Row 1 Column 1 at least 20000000
% difference between PAPI_FP_INS: 1 & 2: 100.00
% difference between PAPI_TOT_CYC 1 & 2: 100.18
Column 1 approximately equals column 2
Column 3 approximately equals 2 * column 2
Column 4 approximately equals 3 * column 2
Column 4 exactly equals column 5


Please send me the config.log file. Hopefully this will shed light on the issues here.
Thanks,
heike.
jagode00
 
Posts: 29
Joined: Tue Aug 25, 2009 2:12 pm

Re: flpins = 0; MFLOPS = 0.0000

Postby victoria1986 » Wed Sep 05, 2012 1:17 pm

Hello everyone.
I decided to post my questionin the same thread.
When I run flops.c example, it's fine, i get the values.
But when I try to paste all papi directive and functions I get flps and mlops=0. May be someone had that issue already?
here is my code

//PAPI initialization
float real_time, proc_time, mflops;
long long flpins;
int retval, fip=0;
retval = PAPI_library_init( PAPI_VER_CURRENT );
if ( retval != PAPI_VER_CURRENT )
fprintf(stderr,"PAPI library version mismatch!\n");

if ( PAPI_query_event( PAPI_FP_INS ) == PAPI_OK )
fip = 1;
else if ( PAPI_query_event( PAPI_FP_OPS ) == PAPI_OK )
fip = 2;
else {

printf( "PAPI_FP_INS and PAPI_FP_OPS are not defined for this platform.\n" );
}

PAPI_shutdown( );
if ( fip > 0 )

{
if ( fip == 1 ) {
if ( ( retval = PAPI_flips( &real_time, &proc_time, &flpins, &mflops ) ) < PAPI_OK )
fprintf(stderr,"Problem reading flips\n");}
else {
if ( ( retval = PAPI_flops( &real_time, &proc_time, &flpins, &mflops ) ) < PAPI_OK )
fprintf(stderr,"Problem reading flops\n");

}
for ( i = 0; i < 2 * n; i = i + 2 )
{
z0 = 0.0;
z1 = 0.0;
x[i] = z0;
z[i] = z0;
x[i+1] = z1;
z[i+1] = z1;
}
if ( fip == 1 )
{
if ( ( retval = PAPI_flips( &real_time, &proc_time, &flpins, &mflops ) ) < PAPI_OK )
fprintf(stderr,"Problem reading flips2\n");
}
else {
if ( ( retval =PAPI_flops( &real_time, &proc_time, &flpins,&mflops ) ) < PAPI_OK )
fprintf(stderr,"Problem reading flops2\n");
}
if ( fip == 1 )
{
printf( "Real_time: %f Proc_time: %f Total flpins: ", real_time,proc_time );
}
else {
printf( "Real_time: %f Proc_time: %f Total flpops: ", real_time,proc_time );
}

printf("Flops: %lld ",flpins);
printf( " MFLOPS: %f\n", mflops );
}
I will appreciate any suggestion, thank you
victoria1986
 
Posts: 2
Joined: Wed Sep 05, 2012 1:11 pm


Return to General discussion

Who is online

Users browsing this forum: Bing [Bot] and 2 guests

cron