Changed precondition for edge case in serial_merge to prevent assertion error #622
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
NguyenNhuDi
requested review from
stanleytsang-amd,
umfranzw,
RobsonRLemos and
lawruble13
as code owners
NguyenNhuDi
requested review from
Naraenda,
umfranzw,
RobsonRLemos and
stanleytsang-amd
and removed request for
umfranzw,
lawruble13,
RobsonRLemos and
stanleytsang-amd
Naraenda
reviewed
Oct 11, 2024
There was a problem hiding this comment.
I'm not sure it makes sense to use begin = min(begin, end) here. What it synthetically describes and the actual goal it tries to achieve are different.
Assuming AllowUnsafe = false, it'd be much more logical to check in void serial_merge(...) if both begin1 and begin2 are valid. If both are invalid we just return without executing anything. If only one is invalid, we still want to continue with merging as a singular invalid begin pointer is already caught by the out-of-bounds checks (here and here).
This way we also prevent this from happening when merge path gets reused.
NguyenNhuDi
changed the title
stanleytsang-amd
approved these changes
Oct 11, 2024
stanleytsang-amd
merged commit b9a2da9
into
ROCm:release-staging/rocm-rel-6.3
16 checks passed
NguyenNhuDi
added a commit
that referenced
this pull request
Nov 1, 2024
* Remove website URL from comments (#600) Referencing or using code from some websites is prohibited in this repository. This change removes an informational reference in the comments. * Fix rare memory access faults when using internal serial merge (#597) * test: add tests for internal serial merge function * refactor(detail/merge_path.hpp): removed code duplication * fix(detail/merge_path.hpp): stricter boundary checking in serial merge * fix(detail/block_sort_merge.hpp): fix missing block-wide sync During a previous refactor, serial_merge does no longer do a block sync. This has now been re-added. * feat: add unsafe variant of serial merge * fix: use bounded version for serial merge to fix rare page faults * test(test_internal_merge_path): clean up internal merge path tests * style: standardize range_t<> construction * fix(detail/merge_path.hpp): fix 'range_t<>::count1()' and 'range_t<>::count2()' return types to be same as encapsulated type * perf(detail/merge_path.hpp): use const ref in function parameters * refactor(detail/merge_path.hpp): replace redundant use of 'OffsetT' with 'unsigned int' * chore: update changelog * fix: restore missing thread sync This got removed during a rebase. * Add gfx1151 target (#601) (#603) Co-authored-by: Stanley Tsang <stanley.tsang@amd.com> * Merge back 6.2 hotfixes (#607) (#620) * Update dependency names for static builds (#557) This also removes the line setting `BUILD_SHARED_LIBS` to `ON`, which was previously required to get the correctly named packages when not specifically compiling for a static build. Updates to the ROCmCMakeBuildTools (rocm-cmake) should mean this is no longer necessary. * Fix BUILD_SHARED_LIBS for packaging (#558) * Fix the dependencies of the static packages (#563) * cmake: don't set CMAKE_C_COMPILER, as rocPRIM is a CXX project (#567) * add developer guidelines (#555) (#574) * Update Read the Docs config to Python 3.10 and latest rocm-docs-core (#564) (#579) * Cherry-pick: Optimize block_reduce_warp_reduce when block size is the same as warp size (#599) * Optimize block_reduce_warp_reduce when block size == warp size * Make conditional constexpr * Fix conflict in concepts.rst --------- Co-authored-by: Lauren Wrubleski <Lauren.Wrubleski@amd.com> Co-authored-by: Steve Leung <Steve.Leung@amd.com> Co-authored-by: randyh62 <42045079+randyh62@users.noreply.github.com> Co-authored-by: Nol Moonen <nol@streamhpc.com> Co-authored-by: Sam Wu <22262939+samjwu@users.noreply.github.com> * Changed precondition for edge case in serial_merge to prevent assertion error (#622) * added std::min to ensure no out of bound acess * fixed typo keys->keys1 * updated changelog * reverted std::min * implemented suggested logic * edited to conform to standards (#618) * Memory leak fix for multiple rocPRIM unit tests (#614) * fixed mem leak in test_config_dispatch.cpp * added missing hip free for method==4 in test_block_scan.kernels * added graphHelpeer class that does not cause memory leak due to using hipGraphCreate * replaced old hipGraph helpers with new class in device_bin_search * changed HIP_CHECK_NON_VOID to HIP_CHECK * fixed mem leak in device_bin_search * added additional functions * changed out old calls to hipGraphCrete to new GraphHelper class * added missing stream sync for hipgrag_algs * n * added missing hipFree and HIP_CHECK for lookback_reproducibility * added missing hipFree in test_discard_iterator * fixed test failures * removed extra hipFree * removed unused variables * updated change log * removed redundant function --------- Co-authored-by: Your Name <you@example.com> Co-authored-by: root <root@x1001c6s0b1n0.hostmgmt2001.cm.lockhart.amd.com> * changed to using gHelper * merged 6.3 version of merge_sort_merge_path (failing large indices test) * updated mergepath to have fix for large indices * updated the changelog for 6.3 (#632) * updated changelog for test_utils_hipgraph change * updated changelog --------- Co-authored-by: Wayne Franz <wayfranz@amd.com> Co-authored-by: Nara <nara@streamhpc.com> Co-authored-by: amd-garydeng <garydeng@amd.com> Co-authored-by: Stanley Tsang <stanley.tsang@amd.com> Co-authored-by: Lauren Wrubleski <Lauren.Wrubleski@amd.com> Co-authored-by: Steve Leung <Steve.Leung@amd.com> Co-authored-by: randyh62 <42045079+randyh62@users.noreply.github.com> Co-authored-by: Nol Moonen <nol@streamhpc.com> Co-authored-by: Sam Wu <22262939+samjwu@users.noreply.github.com> Co-authored-by: spolifroni-amd <Sandra.Polifroni@amd.com> Co-authored-by: Your Name <you@example.com> Co-authored-by: root <root@x1001c6s0b1n0.hostmgmt2001.cm.lockhart.amd.com>
stanleytsang-amd
added a commit
that referenced
this pull request
Nov 21, 2024
* Remove website URL from comments (#600) Referencing or using code from some websites is prohibited in this repository. This change removes an informational reference in the comments. * Fix rare memory access faults when using internal serial merge (#597) * test: add tests for internal serial merge function * refactor(detail/merge_path.hpp): removed code duplication * fix(detail/merge_path.hpp): stricter boundary checking in serial merge * fix(detail/block_sort_merge.hpp): fix missing block-wide sync During a previous refactor, serial_merge does no longer do a block sync. This has now been re-added. * feat: add unsafe variant of serial merge * fix: use bounded version for serial merge to fix rare page faults * test(test_internal_merge_path): clean up internal merge path tests * style: standardize range_t<> construction * fix(detail/merge_path.hpp): fix 'range_t<>::count1()' and 'range_t<>::count2()' return types to be same as encapsulated type * perf(detail/merge_path.hpp): use const ref in function parameters * refactor(detail/merge_path.hpp): replace redundant use of 'OffsetT' with 'unsigned int' * chore: update changelog * fix: restore missing thread sync This got removed during a rebase. * Add gfx1151 target (#601) (#603) Co-authored-by: Stanley Tsang <stanley.tsang@amd.com> * Merge back 6.2 hotfixes (#607) (#620) * Update dependency names for static builds (#557) This also removes the line setting `BUILD_SHARED_LIBS` to `ON`, which was previously required to get the correctly named packages when not specifically compiling for a static build. Updates to the ROCmCMakeBuildTools (rocm-cmake) should mean this is no longer necessary. * Fix BUILD_SHARED_LIBS for packaging (#558) * Fix the dependencies of the static packages (#563) * cmake: don't set CMAKE_C_COMPILER, as rocPRIM is a CXX project (#567) * add developer guidelines (#555) (#574) * Update Read the Docs config to Python 3.10 and latest rocm-docs-core (#564) (#579) * Cherry-pick: Optimize block_reduce_warp_reduce when block size is the same as warp size (#599) * Optimize block_reduce_warp_reduce when block size == warp size * Make conditional constexpr * Fix conflict in concepts.rst --------- Co-authored-by: Lauren Wrubleski <Lauren.Wrubleski@amd.com> Co-authored-by: Steve Leung <Steve.Leung@amd.com> Co-authored-by: randyh62 <42045079+randyh62@users.noreply.github.com> Co-authored-by: Nol Moonen <nol@streamhpc.com> Co-authored-by: Sam Wu <22262939+samjwu@users.noreply.github.com> * Changed precondition for edge case in serial_merge to prevent assertion error (#622) * added std::min to ensure no out of bound acess * fixed typo keys->keys1 * updated changelog * reverted std::min * implemented suggested logic * edited to conform to standards (#618) * Memory leak fix for multiple rocPRIM unit tests (#614) * fixed mem leak in test_config_dispatch.cpp * added missing hip free for method==4 in test_block_scan.kernels * added graphHelpeer class that does not cause memory leak due to using hipGraphCreate * replaced old hipGraph helpers with new class in device_bin_search * changed HIP_CHECK_NON_VOID to HIP_CHECK * fixed mem leak in device_bin_search * added additional functions * changed out old calls to hipGraphCrete to new GraphHelper class * added missing stream sync for hipgrag_algs * n * added missing hipFree and HIP_CHECK for lookback_reproducibility * added missing hipFree in test_discard_iterator * fixed test failures * removed extra hipFree * removed unused variables * updated change log * removed redundant function --------- Co-authored-by: Your Name <you@example.com> Co-authored-by: root <root@x1001c6s0b1n0.hostmgmt2001.cm.lockhart.amd.com> * updated the changelog for 6.3 (#632) * updated default gpu to include gfx12 and gfx1151 * updated changelog * fixed minor grammar mistake in changelog * Update CHANGELOG.md Co-authored-by: spolifroni-amd <Sandra.Polifroni@amd.com> * Remove gfx940,gfx941 targets (#639) * Update rocPRIM version --------- Co-authored-by: Wayne Franz <wayfranz@amd.com> Co-authored-by: Nara <nara@streamhpc.com> Co-authored-by: amd-garydeng <garydeng@amd.com> Co-authored-by: Lauren Wrubleski <Lauren.Wrubleski@amd.com> Co-authored-by: Steve Leung <Steve.Leung@amd.com> Co-authored-by: randyh62 <42045079+randyh62@users.noreply.github.com> Co-authored-by: Nol Moonen <nol@streamhpc.com> Co-authored-by: Sam Wu <22262939+samjwu@users.noreply.github.com> Co-authored-by: Di Nguyen <dinguyennhu@gmail.com> Co-authored-by: spolifroni-amd <Sandra.Polifroni@amd.com> Co-authored-by: Your Name <you@example.com> Co-authored-by: root <root@x1001c6s0b1n0.hostmgmt2001.cm.lockhart.amd.com> Co-authored-by: Val Movsik <160653499+vamovsik@users.noreply.github.com>
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.
When creating ranges for serial_merge in block_sort_merge, device_merge_sort_mergepath, device_merge, and warp_sort_stable there were no checks for out of range cases (when the begin key is bigger than the end key)