-
Notifications
You must be signed in to change notification settings - Fork 744
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[SYCL] SYCL 2020 sub_group algorithms #3786
[SYCL] SYCL 2020 sub_group algorithms #3786
Conversation
…arnings. Also, ONEAPI::reduction() should be converted to use SYCL2020 reduce. Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Some minor comments.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There are a few things that I think we should definitely remove, because they're not in SYCL 2020. There are a few others that I'm not sure about, because I'm unsure of whether they're mistakes in the SYCL 2020 specification or not.
@pvchupin: If this PR exposes an interface that is later removed because of a SYCL specification clarification, would we have to go through regular deprecation procedures?
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
…moved back to ONEAPI namespace. Rename 'sub-group-algorithm' to 'group-algorithm' where appropriate. Adjusted signatures as requested. Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
…hy git clang-format is not working correctly. Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
@vladimirlaz - can you help me understand this Jenkins/Precommit failure? I'm attempting to re-run the Parallel Stage. Not sure if that is the correct course of action. After re-running, it passes. Not sure how to get GitHub to update its status. |
The overall status is not updated if internal job is restarted. There was scheduled downtime for the service. The job can be treated as passed if the internal job passed after restart. |
…Not called. Originally removed by A. Batashev. Signed-off-by: Chris Perkins <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
joint_any_of
, joint_all_of
and joint_none_of
in SYCL 2020 don't have any requirements on the input types, because applying Predicate
must convert them to bool
: https://www.khronos.org/registry/SYCL/specs/sycl-2020/html/sycl-2020.html#_any_of_all_of_and_none_of
Sorry I missed these in my initial review. I didn't see anything else, so I'll approve after these three changes.
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
Signed-off-by: Chris Perkins <[email protected]>
* sycl: (320 commits) [SYCL] Silence a "local variable is initialized but not referenced" warning; NFC (intel#3870) [SYCL] Improve SYCL_DEVICE_ALLOWLIST (intel#3826) [SPIR-V] Change return value of mapType function (intel#3871) [SYCL] Fix post-commit failure in handler.hpp from unused-parameters. (intel#3874) [Driver][SYCL] Do not imply defaultlib msvcrt for Linux based driver on Windows (intel#3827) [SYCL] Unique stable name rebase (intel#3835) [SYCL] Align behavior of empty command groups with SYCL2020 (intel#3822) [SYCL][ESIMD] Make typenames and constants consistent with SYCL API style. (intel#3850) [SYCL] Allow __failed_assertion to support libstdc++-11 (intel#3774) [SYCL] Refactor stream class handing implementation (intel#3646) [SYCL] Fix syntax error introduced in intel#3401 (intel#3861) [SYCL] SYCL 2020 sub_group algorithms (intel#3786) [Buildbot][NFC] Add option to use LLD as linker (intel#3866) Revert "Emit correct location lists with basic block sections." [SPIRITTAnnotations] Fix debug info for ITT calls. (intel#3829) [SYCL][Doc] Fix build of Sphinx docs (intel#3863) [SYCL][FPGA][NFC] Tidy up intel_fpga_reg codegen test (intel#3810) [CODEOWNERS] Fix SPIRITTAnnnotations tests ownership (intel#3859) [SYCL][ESIMD] Host-compile simd.cpp test, fix errors & warnings. (intel#3846) [SYCL] Store pointers to memory allocations instead of iterators (intel#3860) ...
Treat a defined register as fully live "at" the instruction and update maximum pressure accordingly. Fixes #3786.
Getting the functionality of the ONEAPI subgroup algorithms into the SYCL namespace, along with API updates, new names, tests, deprecation warnings.
This is based on an earlier version of @alexbatashev PR #3765 which as of this writing is not yet merged. I'll need to update.
@alexbatashev - I hope we have not stepped on each other too much with these two PRs. Let me know if you spot anything.