https://github.com/kurtamohler/cusparsespmm-coo-bug

Example of how to correctly use CuSparse's cusparseSpMM() function
https://github.com/kurtamohler/cusparsespmm-coo-bug

Last synced: 11 days ago
JSON representation

Example of how to correctly use CuSparse's cusparseSpMM() function

• Host: GitHub
• URL: https://github.com/kurtamohler/cusparsespmm-coo-bug
• Owner: kurtamohler
• Created: 2020-02-11T17:55:42.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-02-13T16:51:42.000Z (over 4 years ago)
• Last Synced: 2024-10-13T05:42:55.016Z (26 days ago)
• Language: C++
• Homepage:
• Size: 6.84 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: Readme.md

Awesome Lists containing this project

README

        
# Explanation

This repo was created to reproduce an apparent bug in CuSparse's

sparse-dense multiply operation, `cusparseSpMM`.

However, someone from NVIDIA pointed out that I was using `uint64_t`

indices for my sparse matrices, which is not supported yet in

`cusparseSpMM`. When I switched to `uint32_t`, I started getting the

correct result.

# Build

```

$ make

```

# Run

```

$ ./a.out

```

# Dependencies

* CUDA 10.1