GEMM Strides/Extents to CBLAS mapping
Open
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 167
- Forks
- 31
- PR merge metrics
- No merged PRs in 30d
Description
Does this seem right? Posting for each matrix (extent(0),extent(1)), (stride(0),stride(1))
| C | A | B | defacto layouts | CBLAS call |
|---|---|---|---|---|
| (M,N), (1,M) | (M,K), (1,M) | (K,N), (1,K) | (left,left,left) | gemm('N','N', M, N, K, 1., A.data(), K, B.data(), N, 1., C.data(), N) |
| (M,N), (1,M) | (M,K), (K,1) | (K,N), (1,K) | (left,right,left) | gemm('T','N', N, M, K, 1., A.data(), M, B.data(), N, 1., C.data(), N) |
| (M,N), (1,M) | (M,K), (1,M) | (K,N), (N,1) | (left,left,right) | gemm('N','T', M, N, K, 1., A.data(), K, B.data(), K, 1., C.data(), N) |
| (M,N), (1,M) | (M,K), (K,1) | (K,N), (N,1) | (left,right,right) | gemm('T','T', M, N, K, 1., A.data(), M, B.data(), K, 1., C.data(), N) |
| (M,N), (N,1) | (M,K), (1, M) | (K,N), (1,K) | (right,left,left) | gemm('T','T', N, M, K, 1., B.data(), N, A.data(), K, 1., C.data(), M) |
| (M,N), (N,1) | (M,K), (K, 1) | (K,N), (1,K) | (right,right,left) | gemm('T','N', N, M, K, 1., B.data(), N, A.data(), M, 1., C.data(), M) |
| (M,N), (N,1) | (M,K), (1, M) | (K,N), (N,1) | (right,left,right) | gemm('N','T', N, M, K, 1., B.data(), K, A.data(), K, 1., C.data(), M) |
| (M,N), (N,1) | (M,K), (K,1) | (K,N), (N,1) | (right,right,right) | gemm('N','N', N, M, K, 1., B.data(), K, A.data(), M, 1., C.data(), M) |
Contributor guide
No contributing guide indexed for this repository
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
Review the eight layout cases and CBLAS calls in the issue against the CBLAS GEMM specification, checking transpose flags, dimensions, and leading dimensions. Done means reaching an agreed, correct mapping for every stride/extent combination and recording any required changes.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- backend
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100