[SYCL] spgemm #550
Draft
[SYCL] spgemm #550
Conversation
This file contains 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
Contributor
abagusetty
commented
Jan 4, 2022
abagusetty
commented
- Updates spgemm functionality for SYCL
- majority of the code remains the same as with CUDA for easier maintenance
- Depends on the infrastructure from [SYCL] Seq mv sycl #538
This is an initial commit that still needs some reworking and debugging.
This does a bunch of name changing of files, data structures, and variables from 'cuda' to 'device' in order to reflect which things are generic device functionality vs. tied to a specific language. In addition, this now compiles and runs a simple program that calls HYPRE_Init() and allocates/copies/frees memory on the device and host with unified memory.
Quick fix for compilation --with-cuda. Ran some tests on lassen and quartz as well to make sure I didn't break the cuda or cpu versions.
Modified csr matvec to choose the default execution policy instead of hard-coded device policy. This now passes tests and seems to run as expected using sycl unified memory and using host execution for everything.
Starting to put in boxloop sycl code. This compiles, but crashes.
I have fixed my compilation issues and can now run with my sycl boxloop1 implementation on frank's sever machine. The boxloop1 code seems to be giving correct results as well, though it seems somewhere along the line I screwed up the struct solvers tests, which yield a discrepancy in number of iterations for the first solvers.jobs job.
The non-reduction boxloops are all in and pass the struct tests. Performance is VERY slow, but this may just be due to the machine I am running on. Reduction boxloops are in progress.
The reduction boxloops are implemented and pass the struct solvers.sh tests. Cleanup of boxloop_sycl.h.
Uses shared memory pointer instead of buffers and accessors. Seems to work on iris, same error as before on arcticus.
…group collectives
…m() to align with hypre-space#549 2. CMakeLists support for SYCL build 3. remove dpct:: functions and include the implementations
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.