ggml/llama.cpp releases: b10908

TL;DR

The ggml/llama.cpp project released a fix for idle threads in specific multiplication kernels used for neural networks with fewer than 1024 elements. This update generalizes a previous row split to additional related kernels, improving thread utilization and potentially enhancing performance.

Detailed Summary

The ggml/llama.cpp project released a new version addressing issues with idle threads in specific multiplication kernels for input sizes less than 1024. This update generalizes row splitting across multiple related kernels to improve performance. The broader impact includes enhanced computational efficiency, particularly benefiting systems processing smaller input sizes.

Key Points

• • metal : fix idle threads in the remaining iq mul_mv kernels for ne00 < 1024
• • Generalize row split from #28086 to six other kernels