ggerganov/llama.cpp Released b11804 with OpenCL ARGMAX Support
Llama.cpp release b11804 introduces an OpenCL F32-to-I32 ARGMAX implementation alongside routine cross-platform binary distributions. This update incorporates contribution #30897 to optimize tensor operations on supported hardware accelerators. This routine point release improves hardware acceleration capabilities for devices utilizing OpenCL. Such incremental updates ensure broader compatibility and better execution efficiency across diverse computing environments. The release features PR #30897 which adds F32-to-I32 ARGMAX support for OpenCL, with variables initialized to zero for CPU consistency. It also provides precompiled binary packages for macOS, Linux, Android, and Windows targeting various CPU and GPU architectures.
## BACKGROUND
llama.cpp is an open-source software library that performs inference on various large language models like Meta's Llama. OpenCL is an open, royalty-free standard for cross-platform, parallel programming of diverse accelerators found in modern processors.