Refactor/kernel interfaces #804
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
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
from
736bff3 to
6b179af
Compare
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
from
6b179af to
d426eb2
Compare
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
4 times, most recently
from
f489b60 to
921624e
Compare
12 tasks
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
from
2d98ded to
b2ed65d
Compare
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
3 times, most recently
from
f0e4ed0 to
211d5d8
Compare
Contributor
Author
|
@mingjie-intel @chudur-budur All existing tests except caching work with the new API. I have updated the description to capture pending TODOs. Any early feedback will be very helpful. |
| extra_compile_flags=extra_compile_flags, | ||
| ) | ||
|
|
||
| self._target_context = cres.target_context |
Contributor
There was a problem hiding this comment.
@diptorupd Why we are not doing the same as it has been done in _compile()?
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
3 times, most recently
from
f99df29 to
b43201c
Compare
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
4 times, most recently
from
27490c8 to
17676d2
Compare
chudur-budur
reviewed
Dec 7, 2022
| np.copyto(obj._orig_val, obj._packed_val) | ||
|
|
||
| def __init__( | ||
| self, kernel_name, arg_list, argty_list, access_specifiers_list, queue |
Contributor
There was a problem hiding this comment.
I think we need to use consistent variable naming, like pyfunc_name instead of kernel_name
| compile_flags=None, | ||
| array_access_specifiers=None, | ||
| ): | ||
| self.typingctx = dpex_target.typing_context |
Contributor
There was a problem hiding this comment.
dpex_target is a global object coming from an external module, why do we always keep it in some local variable? Is there any specific reason for this?
| ) | ||
| func = cres.library.get_function(cres.fndesc.llvm_func_name) | ||
| cres.target_context.mark_ocl_device(func) | ||
| devfn = DpexFunction(cres) |
Contributor
There was a problem hiding this comment.
Do we need to cache compiled func?
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
2 times, most recently
from
26c4334 to
4656193
Compare
Contributor
|
Documentation preview: show. |
- The compute follows data checking is now based on queue
equality.
- USMNdArray no longer requires usm_type and device
during construction. It allows us to specialize an usm_ndarray
only on ndims, layout and dtype.
- No check for compute follows data for eager compilation.
- Change caching to not require backend and device-type.
- Fixes to test cases.
- The DEFAULT_LOCAL_SIZE is deprecated and users warned to
provided a valid local range for nd_range kernels.
- Removed the global_range and local_range kw args from
JitKernel.__call__().
- Undeprecate the JitKernel.__getitem__ call.
- Fix and improve how arguments to JitKernel.__call__() are
parsed to extract the global_range and local_range.
diptorupd
force-pushed
the
refactor/kernel_interfaces
branch
from
637a04d to
c74ea24
Compare
Contributor
|
Documentation preview: show. |
Contributor
Author
|
Merging as TeamCity CI is all green. |
Contributor
|
Documentation preview removed. |
github-actions bot
added a commit
that referenced
this pull request
Jan 18, 2023
Refactor/kernel interfaces 187782d
5 tasks
This was referenced Jan 29, 2023
fcharras
added a commit
to fcharras/numba-dpex
that referenced
this pull request
Feb 3, 2023
I don't have a minimal reproducer yet but I can say that this fixes the issues I've had after IntelPython#804
This was referenced Feb 3, 2023
5 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Have you provided a meaningful PR description?
numba_dpex\compiler.pymixes both things, making hard the separation of compute-follows-data based kernel launch from legacydpctl.device_contextbased behavior.dpctl.device_contextfor kernels.__getitem__to provide global and local ranges for a kernel launch. (to be reevaluated Deprecate __getitem__ support in numba_dpex.kernel #790)numba_dpex.core.kernel_interface.DpexFuncto new APInumba_dpex.compiler.pyHave you added a test, reproducer or referred to an issue with a reproducer?
Have you tested your changes locally for CPU and GPU devices?
Have you made sure that new changes do not introduce compiler warnings?
If this PR is a work in progress, are you filing the PR as a draft?
Fixes #814, #816, #780, #810