BBQ OpenCL benchmark: contrib-ocl-samples

[Edoardo Paone]

Hi,

I'm working on the bops-next branch (259cbb4) with TPD on NSJ.

I see some rescheduling actions from BBQ to the RTLib, but the onConfigure method of my ExC implementation is not called:

0000.616557 30019:MandelbrotOCL I rpc     : SyncP_4 (Post-Change) EXC [0]

0002.485960 30019:MandelbrotOCL I rpc     : SyncP_1 (Pre-Change) EXC [0], Action [1], Assigned AWM [3]

0002.489924 30019:MandelbrotOCL I rpc     : SyncP_4 (Post-Change) EXC [0]

0004.209745 30019:MandelbrotOCL I rpc     : SyncP_1 (Pre-Change) EXC [0], Action [1], Assigned AWM [7]

0004.213244 30019:MandelbrotOCL I rpc     : SyncP_4 (Post-Change) EXC [0]

0005.875575 30019:MandelbrotOCL I rpc     : SyncP_1 (Pre-Change) EXC [0], Action [1], Assigned AWM [3]

0005.878334 30019:MandelbrotOCL I rpc     : SyncP_4 (Post-Change) EXC [0]

This problem happens on NSJ when the synchronization point is not forced, while the onConfigure method is correctly called if the synchronization is forced.

It is not a problem for me to force the synchronization point, I just wanted to report this issue.

/Edoardo

[Patrick Bellasi]

Thanks Edo for the report... I'll look into that problem and report to you as soon as possible!

Ciao Patrick

 

/Edoardo

--
You received this message because you are subscribed to the Google Groups "BOSP Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bosp-devel+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

--
#include <best/regards.h>

Patrick Bellasi
Post-Doc at Politecnico di Milano
http://home.dei.polimi.it/bellasi

[Edoardo Paone]

Hi Patrick,

I'm working on the bosp-next branch (259cbb4) with Test Platform Data enabled.

By forcing the synchronization point, the RTLib actually calls the onConfigure method whenever the AWM changes. In the attached file, the onConfigure is called three times with AWM 4, 5 and 7 but the final report at the end of the execution context shows only one line, for AWM 4 (the first assigned). While this is not a problem for my work, I wonder whether it is related to the bug that I posted in my previous message.

-Edoardo

[Patrick Bellasi]

On Mon, Jul 1, 2013 at 12:00 PM, Edoardo Paone <pa...@elet.polimi.it> wrote:

Hi Patrick,

Hi Edoardo!

 

I'm working on the bosp-next branch (259cbb4) with Test Platform Data enabled.

By forcing the synchronization point, the RTLib actually calls the onConfigure method whenever the AWM changes. In the attached file, the onConfigure is called three times with AWM 4, 5 and 7 but the final report at the end of the execution context shows only one line, for AWM 4 (the first assigned). While this is not a problem for my work, I wonder whether it is related to the bug that I posted in my previous message.

You are right, it seems to be correlated to the previous bug... which however it is still in the pipeline for fixing. Thanks again for the report, I'll let you know when a fix will be release, hopefully pretty soon.

 

-Edoardo

Patrick

[Edoardo Paone]

I was running lots of tests on NSJ with BBQ and I realized too late that were many run-time errors... all look like in the attached log file.

I was working on the bosp-next branch (259cbb4) with Test Platform Data enabled and Synchronization Point Forcing.

Which is the most stable branch of BBQ you recommend me to use? The master branch?

-Edoardo

[Edoardo Paone]

A different error, but with much lower rate, is the one in the attached file.
This error is probably not related to the synchronization issue.

-Edoardo