Bug report for cudaTensorCoreGemm
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 9.6k
- Forks
- 2.4k
- Avg merge
- 53m
- Merged PRs (30d)
- 1
Description
I've identified an error in line 227 of the cudaTensorCoreGemm.cu file. The code should replace BLOCK_COL_TILES with BLOCK_ROW_TILES. Under the original parameter settings, these two values happen to be the same, so no error occurs. However, if the parameters are modified such that they are no longer equal, errors are likely to arise.
https://github.com/NVIDIA/cuda-samples/blob/master/Samples/3_CUDA_Features/cudaTensorCoreGemm/cudaTensorCoreGemm.cu
227 const unsigned int block_tile_j = (block_pos * BLOCK_COL_TILES) % N_TILES;
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Open Samples/3_CUDA_Features/cudaTensorCoreGemm/cudaTensorCoreGemm.cu and inspect line 227 in the block tile index calculation. Compare the use of BLOCK_COL_TILES with the surrounding row and column tile parameters. Done means the calculation uses the intended parameter and the sample still works when the block row and column tile values differ.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- hpc
- Issue type
- Bug
- Difficulty
- 1/5
- Estimated time
- Under an hour
- Activity status
- Stale
- Clarity
- Clearly specified
- Newbie friendliness
- 58/100