Cythonize cuda.core more
#1070
Cythonize cuda.core more
#1070
Conversation
leofang
added
enhancement
This comment was marked as resolved.
This comment was marked as resolved.
leofang
force-pushed
the
cythonize_more
branch
from
7641f18 to
fa88b28
Compare
leofang
added
P0
|
/ok to test b713c5c |
This comment has been minimized.
This comment has been minimized.
| cdef int _get_device_and_context(self) except?-1: | ||
| cdef cydriver.CUcontext curr_ctx | ||
| if self._device_id == cydriver.CU_DEVICE_INVALID: | ||
| # TODO: It is likely faster/safer to call cuCtxGetCurrent? |
There was a problem hiding this comment.
I don't think this is universally true, where a context can be non-current without being destroyed. Retrieving the device and context from that stream object would be valid.
There was a problem hiding this comment.
I did not bother doing a profiling here, just thought that it might be good to reduce a few CUDA calls... 😛
There was a problem hiding this comment.
Can we remove this TODO before merge? It's not logically correct to call cuCtxGetCurrent here as the Stream object could be from a different context.
There was a problem hiding this comment.
I see why this TODO is confusing. We actually need the current context in order to do a safe cuCtxGetDevice dance (set the stream ctx to current -> get device -> revert to the current ctx). Let me clarify in the code.
There was a problem hiding this comment.
I think we can just use cuCtxGetDevice_v2 which allows passing a ctx explicitly and doesn't require messing with the current ctx?
There was a problem hiding this comment.
I think we can just use
cuCtxGetDevice_v2which allows passing a ctx explicitly and doesn't require messing with the current ctx?
Unfortunately it seems to be CUDA 13 only... 😢
|
/ok to test a5e6bcf |
|
/ok to test 3809a33 |
|
/ok to test 22b0e2e |
|
Description
Close #1065
Close #1063
Checklist