Script 'mail_helper' called by obssrc Hello community, here is the log from the commit of package llamacpp for openSUSE:Factory checked in at 2026-04-25 21:38:12 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/llamacpp (Old) and /work/SRC/openSUSE:Factory/.llamacpp.new.11940 (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "llamacpp" Sat Apr 25 21:38:12 2026 rev:27 rq:1349242 version:8814 Changes: -------- --- /work/SRC/openSUSE:Factory/llamacpp/llamacpp.changes 2026-03-05 17:18:11.546212897 +0100 +++ /work/SRC/openSUSE:Factory/.llamacpp.new.11940/llamacpp.changes 2026-04-25 21:43:16.072237159 +0200 @@ -1,0 +2,27 @@ +Fri Apr 17 05:58:59 UTC 2026 - Adrian Schröter <[email protected]> + +- Update to version 8814: + - Quantization: Introduced 1-bit (Q1_0) quantization support + across CPU, CUDA, Vulkan, Metal, and SYCL backends. + - Model Support: Added support for Gemma 4 (text, vision, and + audio), Qwen 3 (Omni, ASR, and Next), DeepSeek v3.2, + HunyuanOCR, Phi-4, and Granite 4.0. + - Hardware Acceleration: Implemented 128-bit RISC-V Vector + (RVV) and SIMD kernels for enhanced CPU performance. + - GPU Backends: Added AMD CDNA4 (gfx950) support in HIP and + optimized Flash Attention for CUDA, Metal, and WebGPU. + - WebGPU: Significant updates including compute pass batching, + JIT operator porting, and f32 accumulation for register tiling. + - Multimodal: Expanded audio/vision support, including Gemma 4 + Conformer encoders and DeepSeekOCR integration. + - Server/API: Added OpenAI-compatible audio transcriptions API + support and improved router mode for multi-model management. + - WebUI: Introduced MCP (Model Context Protocol) diagnostics, + conversation branching, and dedicated reasoning content + rendering. + - RPC Backend: Implemented native RDMA (RoCEv2) transport for + high-performance remote procedure calls. + - DevOps: Updated core dependencies including BoringSSL to + 0.20260413.0 and added KleidiAI-enabled ARM release artifacts. + +------------------------------------------------------------------- Old: ---- llamacpp-8189.tar.gz New: ---- llamacpp-8814.tar.gz ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ llamacpp.spec ++++++ --- /var/tmp/diff_new_pack.WRKE7A/_old 2026-04-25 21:43:16.792266535 +0200 +++ /var/tmp/diff_new_pack.WRKE7A/_new 2026-04-25 21:43:16.792266535 +0200 @@ -25,11 +25,11 @@ %global mtmd_sover 0.0.%{version} %global mtmd_sover_suffix 0 -%global ggml_sover 0.9.7 +%global ggml_sover 0.9.11 %global ggml_sover_suffix 0 Name: llamacpp -Version: 8189 +Version: 8814 Release: 0 Summary: Inference of Meta's LLaMA model (and others) in pure C/C++ License: MIT @@ -41,6 +41,7 @@ BuildRequires: ninja BuildRequires: pkgconfig BuildRequires: shaderc +BuildRequires: spirv-headers BuildRequires: pkgconfig(OpenCL) BuildRequires: pkgconfig(libcurl) BuildRequires: pkgconfig(vulkan) ++++++ llamacpp-8189.tar.gz -> llamacpp-8814.tar.gz ++++++ /work/SRC/openSUSE:Factory/llamacpp/llamacpp-8189.tar.gz /work/SRC/openSUSE:Factory/.llamacpp.new.11940/llamacpp-8814.tar.gz differ: char 12, line 1
