PULP Community
Running HERO polybench tests with larger datasets - Printable Version

+- PULP Community (https://pulp-platform.org/community)
+-- Forum: PULP's Community forum (https://pulp-platform.org/community/forumdisplay.php?fid=1)
+--- Forum: PULP General questions (https://pulp-platform.org/community/forumdisplay.php?fid=2)
+--- Thread: Running HERO polybench tests with larger datasets (/showthread.php?tid=387)



Running HERO polybench tests with larger datasets - ajgaspar - 08-05-2026

When running the polybench 2mm test with the default HERO setup, the EXTRAMINI and MINI datasets both run without issue after defining them in the 2mm.h file. However, any size larger than those results in some kind of silent erroring out, where there is no output after "[0,0] ---> omp_set_default_device" and the run seemingly ends without executing the calculation or producing any error message. Is there a memory parameter or something I need to change in order to get the larger datasets to run?


RE: Running HERO polybench tests with larger datasets - marktuan - 09-19-2026

(08-05-2026, 09:39 PM)ajgaspar Wrote: When running the polybench 2mm test with the default HERO setup, the EXTRAMINI and MINI datasets both run without issue after defining them in the 2mm.h file. However, any size larger than those results in some kind of silent erroring out, where there is no output after "[0,0] ---> omp_set_default_device" and the run seemingly ends without executing the calculation or producing any error message. Is there a memory parameter or something I need to change in order to get the larger datasets to run? hot games

I’ve run into something similar with larger datasets. Since EXTRAMINI and MINI work but the larger ones stop immediately after
Code:
omp_set_default_device
, I’d suspect a memory/allocation limit rather than the PolyBench calculation itself.
I’d check the HERO configuration for the available L2/cluster memory and external-memory settings, and also look at the linker map or runtime output to see where the larger arrays are being placed. If the dataset exceeds the memory region used by the default setup, increasing the dataset size alone probably won’t be enough—you may need to adjust the memory layout or explicitly place the larger arrays in external memory.