-
Notifications
You must be signed in to change notification settings - Fork 86
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
[Modes] DPC++ backend #494
Conversation
…that we can discuss don't know how it reflects to our implementation. I also solved all the problems for compiling this file in many other files. The file memory.cpp can now be compiled without error with dpcpp compiler.
…nel doesn't accept virtual functions.
This reverts commit dffee51.
…ueue arguments" This reverts commit ecfc203.
…SYCL kernel doesn't accept virtual functions." This reverts commit 26e7398.
…a way to enqueue arguments"" This reverts commit 8e2aacd.
…y to enqueue arguments"" This reverts commit 8e2aacd.
This reverts commit 0d4e35b.
…ample is still not compiling
…/occa-dev into dpcpp
DPC++ backend clases.
Codecov Report
@@ Coverage Diff @@
## main #494 +/- ##
==========================================
+ Coverage 75.68% 76.24% +0.56%
==========================================
Files 253 261 +8
Lines 19109 19399 +290
==========================================
+ Hits 14462 14791 +329
+ Misses 4647 4608 -39
|
The MacOS gcc-9 test fails curiously with
I am not sure what is out of sync here. |
Co-authored-by: David Medina <dmed256@gmail.com>
Description
This PR implements a new OCCA backend using Data Parallel C++ (DPC++).
Highlights
@outer
and@inner
loop indices are mapped to work-group IDs and work-items, respectively — similar to OpenCL.Building
Currently only building with CMake is supported.
Coverage
During development the Intel oneAPI DPC++ compiler was used. Intel GPU hardware was used for testing, including the JLSE Aurora testbeds at ALCF. In the near future testing will be expanded to included other DPC++ implementations and other vendor hardware.
Acknowledgements
This work was completed jointly by ALCF and Intel, with contributions from