Repository navigation
(fix) rocthrust: cherry pick benchmark build fix into ROCm 10.0 - #10967
Merged
marbre merged 1 commit intoAug 18, 2026
Merged
Conversation
…ust benchmarks (#10956) An oversight from the inclusion of the rocRAND dependency into rocThrust is that they both share common CMake options. When building rocThrust benchmarks, `BUILD_BENCHMARK=ON` must be set. This same CMake option is present in rocRAND. Thus, anytime rocThrust must download/clone rocRAND as a dependency and builds benchmarks, it also causes the local copy of rocRAND to build its benchmarks as well. This, in combination with the recent 5.0.0 version bump of rocRAND in rocThrust's dependencies, triggered a primbench.hpp not found error. FetchContent runs in-process, so rocthrust's BUILD_BENCHMARK=ON leaks into rocrand and causes its benchmarks to build. I've suppressed that in rocThrust via an explicit `set(BUILD_BENCHMARK OFF)` immediatley before `FetchContent_Declare(rocrand)`. Then turn BUILD_BENCHMARK=ON immediately after so that rocThrust's benchmarks build. Ran the rocThrust benchmark build locally. it no longer build the rocRAND benchmark. The rocthrust benchmark build works. - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. JIRA ID: ROCM-29174
amd-jmahovsky
approved these changes
Aug 18, 2026
amd-jmahovsky
left a comment
amd-jmahovsky
left a comment
Contributor
There was a problem hiding this comment.
LGTM. Same as #10956
amd-hsong
approved these changes
Aug 18, 2026
marbre
approved these changes
Aug 18, 2026
Member
|
The CI signal on the original commit is enough to be confident enough to ignore CI signal here. |
marbre
enabled auto-merge (squash)
August 18, 2026 22:09
marbre
disabled auto-merge
August 18, 2026 22:21
1 task
stanleytsang-amd
added a commit
that referenced
this pull request
Aug 19, 2026
## Motivation Further investigation revealed some bugs in the fetch_dep call that is used to get rocPRIM and/or rocRAND as dependencies. Hardcoded references to rocPRIM were used, which obviously breaks when trying to fetch rocRAND. This was causing rocRAND installs not being found, specifically in [TheRock CI for ROCm 10](https://github.com/ROCm/TheRock/actions/runs/32192515118/job/95906621978?pr=7330), which in turn triggered rocThrust to download rocRAND, which in turn triggers #10967 and #10956, and also other behaviour that will soon be undesired, specifically the packaging of duplicate files in separate library packages. Unfortunately for this last behaviour, the fix is beyond the scope of ROCm 10. ## Technical Details Fix the `fetch_dep` function in cmake/Dependencies.cmake to properly handle both rocPRIM and rocRAND. ## Test Plan Run ROCm/TheRock#7475 in advance to see if TheRock CI passes completely. ## Test Result While it is still running at the time of writing, the Validate Artifact Structure job passed, which is what this PR aims to fix. I will check in the morning on the status. ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
1 task
marbre
added a commit
that referenced
this pull request
Aug 19, 2026
… 10.0 (#10991) This PR cherry picks #10978 to the 10.0 release branch. ## Motivation Further investigation revealed some bugs in the fetch_dep call that is used to get rocPRIM and/or rocRAND as dependencies. Hardcoded references to rocPRIM were used, which obviously breaks when trying to fetch rocRAND. This was causing rocRAND installs not being found, specifically in [TheRock CI for ROCm 10](https://github.com/ROCm/TheRock/actions/runs/32192515118/job/95906621978?pr=7330), which in turn triggered rocThrust to download rocRAND, which in turn triggers #10967 and #10956, and also other behaviour that will soon be undesired, specifically the packaging of duplicate files in separate library packages. Unfortunately for this last behaviour, the fix is beyond the scope of ROCm 10. ## Technical Details Fix the `fetch_dep` function in cmake/Dependencies.cmake to properly handle both rocPRIM and rocRAND. ## Test Plan Run ROCm/TheRock#7475 in advance to see if TheRock CI passes completely. ## Test Result While it is still running at the time of writing, the Validate Artifact Structure job passed, which is what this PR aims to fix. The windows jobs seem stuck but everything else passes. ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. JIRA ID: ROCM-29174 --------- Co-authored-by: Marius Brehler <marius.brehler@amd.com>
marbre
added a commit
to ROCm/TheRock
that referenced
this pull request
Aug 19, 2026
Bumps rocm-libraries to include cherry-picks for JIRA ID AIRDEL-36 * ROCm/rocm-libraries#10637 * ROCm/rocm-libraries#10879 * ROCm/rocm-libraries#10967 * ROCm/rocm-libraries#10991
marbre
added a commit
to ROCm/TheRock
that referenced
this pull request
Aug 20, 2026
Bumps rocm-libraries to include cherry-picks for JIRA ID AIRDEL-36 * ROCm/rocm-libraries#10637 * ROCm/rocm-libraries#10879 * ROCm/rocm-libraries#10967 * ROCm/rocm-libraries#10991
stanleytsang-amd
added a commit
that referenced
this pull request
Sep 18, 2026
## Motivation Further investigation revealed some bugs in the fetch_dep call that is used to get rocPRIM and/or rocRAND as dependencies. Hardcoded references to rocPRIM were used, which obviously breaks when trying to fetch rocRAND. This was causing rocRAND installs not being found, specifically in [TheRock CI for ROCm 10](https://github.com/ROCm/TheRock/actions/runs/32192515118/job/95906621978?pr=7330), which in turn triggered rocThrust to download rocRAND, which in turn triggers #10967 and #10956, and also other behaviour that will soon be undesired, specifically the packaging of duplicate files in separate library packages. Unfortunately for this last behaviour, the fix is beyond the scope of ROCm 10. ## Technical Details Fix the `fetch_dep` function in cmake/Dependencies.cmake to properly handle both rocPRIM and rocRAND. ## Test Plan Run ROCm/TheRock#7475 in advance to see if TheRock CI passes completely. ## Test Result While it is still running at the time of writing, the Validate Artifact Structure job passed, which is what this PR aims to fix. I will check in the morning on the status. ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
illsilin
pushed a commit
to igorluppi/rocm-libraries
that referenced
this pull request
Oct 1, 2026
## Motivation Further investigation revealed some bugs in the fetch_dep call that is used to get rocPRIM and/or rocRAND as dependencies. Hardcoded references to rocPRIM were used, which obviously breaks when trying to fetch rocRAND. This was causing rocRAND installs not being found, specifically in [TheRock CI for ROCm 10](https://github.com/ROCm/TheRock/actions/runs/32192515118/job/95906621978?pr=7330), which in turn triggered rocThrust to download rocRAND, which in turn triggers ROCm#10967 and ROCm#10956, and also other behaviour that will soon be undesired, specifically the packaging of duplicate files in separate library packages. Unfortunately for this last behaviour, the fix is beyond the scope of ROCm 10. ## Technical Details Fix the `fetch_dep` function in cmake/Dependencies.cmake to properly handle both rocPRIM and rocRAND. ## Test Plan Run ROCm/TheRock#7475 in advance to see if TheRock CI passes completely. ## Test Result While it is still running at the time of writing, the Validate Artifact Structure job passed, which is what this PR aims to fix. I will check in the morning on the status. ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
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
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.
See #10956.
An oversight from the inclusion of the rocRAND dependency into rocThrust is that they both share common CMake options. When building rocThrust benchmarks,
BUILD_BENCHMARK=ONmust be set. This same CMake option is present in rocRAND. Thus, anytime rocThrust must download/clone rocRAND as a dependency and builds benchmarks, it also causes the local copy of rocRAND to build its benchmarks as well. This, in combination with the recent 5.0.0 version bump of rocRAND in rocThrust's dependencies, triggered a primbench.hpp not found error.JIRA ID: ROCM-29174
Technical Details
FetchContent runs in-process, so rocthrust's BUILD_BENCHMARK=ON leaks into rocrand and causes its benchmarks to build. I've suppressed that in rocThrust via an explicit
set(BUILD_BENCHMARK OFF)immediatley beforeFetchContent_Declare(rocrand). Then turn BUILD_BENCHMARK=ON immediately after so that rocThrust's benchmarks build.Test Plan
Ran the rocThrust benchmark build locally. CI ran on #10956
Test Result
CI passed, benchmark build worked.
Submission Checklist