How to limit computation to one CPU core? #6464
Unanswered
louis-shao
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi, I use jax as the frontend to dump a xla hlo module, following the example here, then compile and execute it with this openxla library. Is there a way I can limit the computation to one CPU core? I tried setting the number of cpu devices and max number of inflight computations both to 1, but it didn't work. What I observe is, if I add the
taskset N
prefix to bind the execution to one core, the computation time is about 2x longer than the case withouttaskset N
, which seem indicating that it is not running on one core.Update:
HloModuleConfig::set_intra_op_parallelism_threads
and settingasynchronous
to False when creating the client. Neither works.XLA_FLAGS="--xla_cpu_multi_thread_eigen=false intra_op_parallelism_threads=0"
according to this makes the execution slower, but still quite faster than addingtaskset N
.Beta Was this translation helpful? Give feedback.
All reactions