Feature/async kernel submition #1219
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
3 times, most recently
from
c0cfee5 to
5c1a60d
Compare
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
5 times, most recently
from
914dcd5 to
c7b22d1
Compare
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
2 times, most recently
from
036e301 to
1ebbe2e
Compare
ZzEeKkAa
changed the title
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
numba_dpex/experimental/launcher.py
Outdated
| index_space, | ||
| kernel_args, | ||
| ) | ||
| device_event.wait() # pylint: disable=E1101 |
Contributor
There was a problem hiding this comment.
For this case, how will the device_event object get cleaned up? Previosuly, we had sycl.dpctl_event_delete(self._builder, eref).
Contributor
Author
There was a problem hiding this comment.
wrap_event_reference allocates meminfo and sycl event object becomes smart pointer. Destructor of meminfo will call DPCTLEvent_Delete which is the same function called by dpctl_event_delete https://github.com/IntelPython/numba-dpex/blob/main/numba_dpex/core/runtime/_eventstruct.c#L19
Contributor
Author
There was a problem hiding this comment.
Here is the log of async call:
DPEXRT-DEBUG: In DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 805
DPEXRT-DEBUG: usm type = 1 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 282.
DPEXRT-DEBUG: NRT_MemInfo_init mi=0x561537ebfc00 external_allocator=0x561537e91fb0 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 476
DPEXRT-DEBUG: Done with unboxing call to DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 885
DPEXRT-DEBUG: In DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 805
DPEXRT-DEBUG: usm type = 1 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 282.
DPEXRT-DEBUG: NRT_MemInfo_init mi=0x561537b7c1c0 external_allocator=0x561537df11b0 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 476
DPEXRT-DEBUG: Done with unboxing call to DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 885
DPEXRT-DEBUG: In DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 805
DPEXRT-DEBUG: usm type = 1 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 282.
DPEXRT-DEBUG: NRT_MemInfo_init mi=0x561537b49c50 external_allocator=0x561537fb8120 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 476
DPEXRT-DEBUG: Done with unboxing call to DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 885
DPEXRT-DEBUG: creating new dpctl event meminfo.
DPEXRT-DEBUG: scheduling nrt meminfo release.
DPEXRT-DEBUG: acquired meminfo.
DPEXRT-DEBUG: creating new dpctl event meminfo.
DPEXRT-DEBUG: creating new event object.
DPEXRT-DEBUG: released meminfo from host_task.
DPEXRT-DEBUG: deleting dpctl event reference.
DPEXRT-DEBUG: creating new event object.
DPEXRT-DEBUG: deleting dpctl event reference.
DPEXRT-DEBUG: released meminfo from host_task.
DPEXRT-DEBUG: released meminfo from host_task.
both dpctl event reference getting deleted.
And for call_kernel:
DPEXRT-DEBUG: In DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 805
DPEXRT-DEBUG: usm type = 1 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 282.
DPEXRT-DEBUG: NRT_MemInfo_init mi=0x55f94f0452d0 external_allocator=0x55f94f4a6260 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 476
DPEXRT-DEBUG: Done with unboxing call to DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 885
DPEXRT-DEBUG: In DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 805
DPEXRT-DEBUG: usm type = 1 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 282.
DPEXRT-DEBUG: NRT_MemInfo_init mi=0x55f94f3f3470 external_allocator=0x55f94f403860 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 476
DPEXRT-DEBUG: Done with unboxing call to DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 885
DPEXRT-DEBUG: In DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 805
DPEXRT-DEBUG: usm type = 1 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 282.
DPEXRT-DEBUG: NRT_MemInfo_init mi=0x55f94f044c40 external_allocator=0x55f94f0e5b30 at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 476
DPEXRT-DEBUG: Done with unboxing call to DPEXRT_sycl_usm_ndarray_from_python at /home/yevhenii/Projects/numba-dpex/numba_dpex/core/runtime/_dpexrt_python.c, line 885
DPEXRT-DEBUG: creating new dpctl event meminfo.
DPEXRT-DEBUG: deleting dpctl event reference.
As you can see - it is getting deleted.
Contributor
There was a problem hiding this comment.
Ok. But, we need not add the overhead of NRT_MemInfo allocation when we know it is not going to be used. We should just keep the direct call to sycl.dpctl_event_delete for the call_kernel scenario.
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
3 times, most recently
from
31b1122 to
425dabb
Compare
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
from
425dabb to
cb53e93
Compare
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
2 times, most recently
from
0ca5372 to
d704dd1
Compare
diptorupd
reviewed
Nov 29, 2023
diptorupd
reviewed
Nov 29, 2023
diptorupd
approved these changes
Nov 29, 2023
Contributor
diptorupd
left a comment
diptorupd
left a comment
There was a problem hiding this comment.
Please merge after adding the last two docstrings.
ZzEeKkAa
force-pushed
the
feature/async_kernel_submition
branch
from
d704dd1 to
707412b
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add new
call_kenel_asynfunction toexperimentalpackage to submit kernel without waiting till it's done execution.