cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

ajk
Journeyman III

Incorrect results when multiple processes use the same GPU

Hello. I've noticed strange behaviour of our scientific application when launching several instances on the same GPU - sometimes it gives incorrect results.

I've reproduced this behaviour on the small program (see attachment).

1. It creates two slightly different matrix multiplication kernels.

2. Creates three buffers a, b, c.

3. First kernel multiplies a * b and writes result to c. Second kernel multiplies b * a and writes result to the same c.

4. It launches kernel 1, kernel 2 in the loop several times (at the end of the loop, kernel 1 always), then gets the result back and compares it with gold a * b.

When executing single.sh (see attachment): correct result on all platforms (AMD, Intel, NVIDIA).

When executing multi.sh (which launches 9 instances of the program in parallel): some of the instances gave wrong result on AMD platform (tested on FirePro W9100 with fglrx-13.352.1014 and fglrx-14.20), on Intel and NVIDIA - always correct results.

0 Likes
4 Replies
bobwhitecotton
Adept III

HI Alexey,

I've asked one of our moderators to take a look at your sample code and get back to you.

Best regards,

Bob

0 Likes
dipak
Big Boss

Hi,

I was unable to reproduce the error using driver fglrx 14.20 on HD 8670D (as don't have the FirePro W9100). I performed the following steps:

1. Built the executable from test.c

[ to avoid compilation error from gcc, I changed the following line of code:

for (int ii = 0; ii <= ((i_test & 1) ? i_test + 1 : i_test); ii++) {

to

int ii;

for (ii = 0; ii <= ((i_test & 1) ? i_test + 1 : i_test); ii++) {

]

2. Ran the multi.sh several times but no "err.log" file was generated.

My setup details:

AMD HD 8670D (Devastator)

driver: 14.20-140529a-172427E-ATI

APP SDK 2.9

Ubuntu 14.04 LTS

I've attached output of clinfo and the log files generated after running the multi.sh.

I'll try to arrange a FirePro card and test the same. Meanwhile, if possible, please can you test the same on other cards and let us know your findings.

Regards,

0 Likes
ajk
Journeyman III

Hi,

I've tested Radeon 7970 and A10 7850 - everything was ok, but on FirePro W9100 there were errors.

0 Likes

Hi ajk,

I tested on a W9100 and I can reproduce the problem.

I will investigate the problem and let you the outcome of my investigation

Regards

Ben

0 Likes