[aittalam]

Description

Bump the llama.cpp submodule from b10052 (b2dd28a) to b10083 (846e991) and refresh the patch set.

Claude's report:

38 of 39 patches applied cleanly (line-number fuzz only). The Vulkan patch (ggml_src_ggml-vulkan_ggml-vulkan.cpp.patch) needed a one-hunk reconciliation: upstream added vk_queue_handle / *_synchronized / unsynchronized structs before struct vk_queue, shifting the five ggml_backend_vk_buffer_type forward declarations the first hunk annotates with GGML_CALL. Intent is unchanged; only the surrounding context and line numbers moved.

No BUILD.mk source-list changes (no new/deleted/renamed buildable sources) and no llamafile integration changes (llama.h/common.h/mtmd public API unchanged). New GPU-only sources (ggml-cuda/dsv4-hc.cu/.cuh, vulkan-shaders/dequant_q2_0.comp) are glob-collected by the runtime GPU build scripts.

PR Type

• 🦙 llama.cpp sync

Checklist

• I understand the code I am submitting.
• I have run this code locally and verified the change.
• New and existing tests pass locally, or I have explained why tests were not run.
• Documentation was updated where necessary.
• If I changed code in llama.cpp/, whisper.cpp/, or stable-diffusion.cpp/, I also updated the matching *.patches/ files.
• I have read and followed the contribution guidelines.
• AI Usage:
  • No AI was used.
  • AI was used in an assistive capacity.
  • This PR includes substantial AI-generated content.

AI Usage Information

• AI Model used: Claude Opus 4.8
• AI Developer Tool used: Claude code

[aittalam]

Test results:

• integration tests on linux succeeded (see logs)
• successful smoke test on Mac (Metal), linux (CUDA, Vulkan), Windows (CUDA, Vulkan)

[aittalam]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

🤖 Generated with Claude Code