[Starpu-devel] StarPU changeable sizes and StarPU purge

[Rabab Al-omairy <rabab.omairy@kaust.edu.sa>]

Dear StarPU developer,

I am Rabab Alomairy, PhD student working with Hatem Ltaief at KAUST. I would like to have answers to the following two problems:

First issue:
I am working on hierarchical matrices and I am running onto some problem related to changing  sizes after matrix compression. More precisely, hierarchical matrices are divided up onto tiles. I am using starpu_matrix_interface for defining each tile of the matrix. Initially, all tiles contain the same number of rows and columns. However, after matrix compression each tile ends up having different sizes. In order to achieve this on shared memory, I am using the same recommendation that you proposed later on: https://lists.gforge.inria.fr/pipermail/starpu-devel/2019-June/002646.html. Therefore, I realloc interface pointer and change number of column. Shared memory results having very promising outcomes. However, on distributed memory on which tiles are 2D cyclically distributed among mpi nodes  I have  the following error:

[1][starpu_mpi][_starpu_mpi_test_detached_requests][assert failure] MPI_Test returning MPI_ERR_TRUNCATE: message truncated
time_zgetrf_tile: starpu_mpi.c:1153: _starpu_mpi_test_detached_requests: Assertion `req->ret == 0' failed.

How could I resolve this issue?

Second issue:
I am running LU algorithm on low rank matrix on distributed memory architecture (CPU only). For some matrix size which I can say matrix with medium size, I have the following message which cause application to stuck and not continue execution.

[starpu][starpu_memchunk_tidy] Low memory left on node RAM 0 (6405MiB over 128808MiB). Your application data set seems too huge to fit on the device, StarPU will cope by trying to purge 6474 MiB out. This message will not be printed again for further purges. The thresholds can be tuned using the STARPU_MINIMUM_AVAILABLE_MEM and STARPU_TARGET_AVAILABLE_MEM environment variables.
[starpu][_starpu_memory_reclaim_generic] Not enough memory left on node RAM 0. Your application data set seems too huge to fit on the device, StarPU will cope by trying to purge 32202 MiB out. This message will not be printed again for further purges


What could cause this issue?, What can I do to resolve this issue? and How could I avoid this issue?


Please let me know if you need any more details about those questions.
Thank you,
Rabab

--

Best Regards,

Rabab Alomairy

PhD Student

King Abdullah University of Science and Technology

Kingdom of Saudi Arabia
Extreme Computing Research Center
Al-Khawarizmi (Bldg.1)
Floor 0, Office 0203-WS06 

Mobile KSA: +966 (0) 548884792

E-mail: rabab.omairy@kaust.edu.sa

---

This message and its contents, including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email.