gmtan

ACML-GPU & HD4870X2

Discussion created by gmtan on Dec 11, 2009

I want to run two mpi process on a node with a hd4870x2. Each mpi process call dgemm. I tried to set environment ACML-GPU to 0x1 and 0x2 in mpi process to decide which GPU is used. However, this method can not work: 

noderank = 1

gpumask = 1

noderank = 2
gpumask = 2
ERROR: unable to allocate minimum GPU memory

              Total   Available   Last Request
Local:       1024 MB     1001 MB     16777216 (  16 MB) FAILED
Remote (NC): 1233 MB     1211 MB            0 (   0 MB) FAILED
Remote (C):   489 MB      489 MB            0 (   0 MB) FAILED

ERROR: unable to allocate minimum GPU memory

              Total   Available   Last Request
Local:       1024 MB     1001 MB     16777216 (  16 MB) FAILED
Remote (NC): 1233 MB     1212 MB            0 (   0 MB) FAILED
Remote (C):   489 MB      489 MB            0 (   0 MB) FAILED

ERROR: Failed to initialize GPUs
ERROR: Failed to initialize GPU(s) for DGEMM

Warning: libCALBLAS is not reentrant!
Multi-threaded use of this library is not supported.
When GPU-accellerated ACML routines are called on multiple threads
concurrently, the requests will be executed serially, even if multiple
GPUs are present.

 

Outcomes