llama.cpp b10092 Released with PowerPC on AIX Support and SGEMM Fix
The llama.cpp b10092 release enables PowerPC CPU backend variants on the AIX operating system and fixes a missing semicolon in the sgemm implementation. This allows the reuse of existing PowerPC backend implementations on AIX by updating the CMake configuration. This release expands the compatibility of llama.cpp to legacy or enterprise environments running AIX on PowerPC hardware. It also resolves a syntax issue that could prevent successful compilation or execution of matrix multiplication routines on these platforms. The update extends the platform check in CMake to support AIX without altering the behavior of the existing PowerPC backend. Additionally, the syntax fix addresses a missing semicolon in the PowerPC Q0 matmul implementation within sgemm.cpp.
## BACKGROUND
llama.cpp is an open-source LLM inference framework designed for high performance across various hardware using the ggml tensor library. AIX is IBM's proprietary Unix operating system, which runs on PowerPC processors. SGEMM (Single-precision General Matrix Multiply) is a critical linear algebra operation used to perform matrix multiplications in machine learning models.