ggml/llama.cpp releases: b10901

TL;DR

The ggml/llama.cpp project released version b10901, which includes a Vulkan update allowing CPU writes in asynchronous tensor copies when the context is idle. This update aims to optimize performance by leveraging CPU resources more efficiently during idle GPU contexts.

Detailed Summary

The ggml/llama.cpp project released a new version, incorporating a Vulkan update that allows for CPU writes in `ggml_backend_vk_cpy_tensor_async` when the context is idle. This update was addressed in issue #28618 and impacts users of the macOS Apple Silicon (arm64) platform through the ggml/llama.cpp GitHub repository. The broader impact includes enhanced performance and efficiency for users running specific workloads on compatible hardware.

Key Points

• • ggml/llama.cpp release includes Vulkan updates
• • Uses CPU writes in ggml_backend_vk_cpy_tensor_async when context is idle
• • Attestations available for verification