Skip to content

[SYCL][Graph] Fix bugs with HIP/CUDA UR sync points. #14858

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.

Sign up for GitHub

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

Merged
steffenlarsen merged 1 commit into from
Aug 9, 2024
Merged

[SYCL][Graph] Fix bugs with HIP/CUDA UR sync points. #14858

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions sycl/cmake/modules/FetchUnifiedRuntime.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,13 +116,13 @@ if(SYCL_PI_UR_USE_FETCH_CONTENT)
endfunction()

set(UNIFIED_RUNTIME_REPO "https://github.com/oneapi-src/unified-runtime.git")
# commit 83f7ad9551ce0134931b44804b5a254ae2e93068
# Merge: ab9baf57 8fb6824d
# commit 3c12bbceec33cf5cf5fc4fa85e641b4f95e820b7
# Merge: 83f7ad95 ac7eb171
# Author: Omar Ahmed <[email protected]>
# Date: Thu Aug 8 11:11:13 2024 +0100
# Merge pull request #1860 from PietroGhg/pietro/fill
# [NATIVECPU] Fix pointer arithmetic in USMfill
set(UNIFIED_RUNTIME_TAG 83f7ad9551ce0134931b44804b5a254ae2e93068)
# Date: Fri Aug 9 10:51:05 2024 +0100
# Merge pull request #1910 from Bensuo/sync_point
# [CUDA][HIP] Improve command-buffer sync points
set(UNIFIED_RUNTIME_TAG 3c12bbceec33cf5cf5fc4fa85e641b4f95e820b7)

set(UMF_BUILD_EXAMPLES OFF CACHE INTERNAL "EXAMPLES")
# Due to the use of dependentloadflag and no installer for UMF and hwloc we need
Expand Down
16 changes: 16 additions & 0 deletions sycl/doc/design/CommandGraph.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -480,6 +480,14 @@ An executable CUDA Graph, which contains all commands and synchronization
information, is saved in the UR command-buffer to allow for efficient graph
resubmission.

#### Prefetch & Advise

The `urCommandBufferAppendUSMPrefetchExp` and
`urCommandBufferAppendUSMAdviseExp` UR entry-points used to implement
`handler::prefetch` and `handler::mem_advise` are implemented in the CUDA UR
adapter as empty nodes enforcing the node dependencies. As such the
optimization hints are a no-op.

### HIP

The HIP backend offers a graph management API very similar to CUDA Graph
Expand All @@ -502,6 +510,14 @@ An executable HIP Graph, which contains all commands and synchronization
information, is saved in the UR command-buffer to allow for efficient
graph resubmission.

#### Prefetch & Advise

The `urCommandBufferAppendUSMPrefetchExp` and
`urCommandBufferAppendUSMAdviseExp` UR entry-points used to implement
`handler::prefetch` and `handler::mem_advise` are implemented in the HIP UR
adapter as empty nodes enforcing the node dependencies. As such the
optimization hints are a no-op.

### OpenCL

SYCL-Graph is only enabled for an OpenCL backend when the
Expand Down