fix: Restore HPCG_MEMMGMT functionality by fixing memory allocation #2
Conversation
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
…eallocation This commit fixes the HPCG_MEMMGMT custom memory management system that was broken by commit abb79b3. The issue was that gpuMalloc (direct HIP/CUDA API) was incorrectly used instead of deviceMalloc (custom allocator), causing "invalid device pointer" errors when deviceRealloc/deviceDefrag tried to manage memory not allocated through the custom allocator. Changes: - Replace gpuMalloc with deviceMalloc in GenerateProblem.cu (7 allocations) - Replace gpuMalloc with deviceMalloc in GenerateCoarseProblem.cu (2 allocations) - Replace gpuMalloc with deviceMalloc in MultiColoring.cu (5 allocations) - Replace gpuFree with deviceFree for all memory allocated via deviceMalloc - Re-enable deviceRealloc calls in SparseMatrix.cu (previously commented out) The custom memory management system (gpuAllocator_t) provides performance optimizations through pre-allocation, memory defragmentation, and improved memory locality. This fix restores that functionality. Resolves issue #1 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…Vector This commit fixes additional "invalid device pointer" errors that occurred when deviceDefrag tried to manage memory allocated with gpuMalloc instead of deviceMalloc. These errors were discovered after fixing the initial HPCG_MEMMGMT issues. Changes: - Replace gpuMalloc with deviceMalloc in ExtractDiagonal (SparseMatrix.cu) for A.diag_idx and A.inv_diag arrays - Replace gpuMalloc with deviceMalloc in PermuteVector (Permute.cu) for vector buffer allocation - Replace gpuFree with deviceFree in PermuteVector for memory allocated through deviceMalloc (v.d_values) These arrays are used with deviceDefrag in OptimizeProblem.cu, so they must be allocated through the custom memory allocator. After this fix, the program runs successfully without crashes and produces correct results. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Summary
Fixes #1
This PR restores the
HPCG_MEMMGMTfunctionality that was broken by commit abb79b3 which replaced customdeviceMalloc/deviceFreecalls with directgpuMalloc/gpuFreeHIP/CUDA API calls.The custom memory management system (
gpuAllocator_t) requires that all GPU memory be allocated throughdeviceMallocto properly manage memory defragmentation viadeviceDefragand reallocation viadeviceRealloc.Changes
1. First commit: Restore HPCG_MEMMGMT core functionality
gpuMalloc→deviceMalloccalls inGenerateProblem.cugpuMalloc→deviceMalloccalls inGenerateCoarseProblem.cugpuMalloc→deviceMalloccalls inMultiColoring.cudeviceRealloccalls inSparseMatrix.cu(lines 284, 315)gpuFree→deviceFreecalls in all affected files2. Second commit: Fix remaining deviceDefrag errors
ExtractDiagonalinSparseMatrix.cu: ChangedgpuMalloc→deviceMallocfordiag_idxandinv_diagallocationsPermuteVectorinPermute.cu: ChangedgpuMalloc→deviceMallocfor buffer allocation andgpuFree→deviceFreefor deallocationTesting
Tested with
nx=104 ny=104 nz=104on HIP backend:Test plan
HPCG_MEMMGMTenabled🤖 Generated with Claude Code