since the complain happens inside
mcx_core.cu, so, mcxlab
was likely complaining about GPU memory, not the host
or CPU memory.
if you run
gpuinfo=mcxlab('gpuinfo');
it will return the GPU memory specs.
Generally, even a low-end GPU can run MCX without memory issues
unless you simulate a huge number of time gates (because the global
memory needed by MCX is largely sizeof(float) x size of volume x
number of time gates).
Skimming through your script, I noticed that your # of time gate is
just
1. However, your tend/tstep are .5 seconds! Typically, for small
domains,
time gates of width on the scale of ns are more than sufficient.
There is
no need to use such a long time period. Also after several ns, the
photon
weights will be close to numerical zero. So, there is really no
contribution
from running it for long time.
Perhaps you can fix this and see if you are getting the same error.
By the way, which mcxlab release are you using? I do not see
Line#1090
in the current
mcx_core.cu, so I suspect you were using an old
release.
Qianqian
--
You received this message because you are subscribed to the Google
Groups "mcx-users" group.
To unsubscribe from this group and stop receiving emails from it,
send an email to mcx-users+...@googlegroups.com.
To post to this group, send email to mcx-...@googlegroups.com.
Visit this group at https://groups.google.com/group/mcx-users.
For more options, visit https://groups.google.com/d/optout.