llama.cpp
683f0c72
- Only index by compile times + always multiply/add (#25445)
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
13 days ago
Only index by compile times + always multiply/add (#25445) The first one avoids relying on compile to optimize local memory away, and the second is cheaper than issuing control flow statements
References
#25445 - CUDA: Small optimizations to MMVQ
Author
ORippler
Parents
259f2e2a
Loading