llama.cpp b11266 loads F32 A matrix 2 at a time for Vulkan on Intel

The llama.cpp project released build b11266, which includes a Vulkan optimization that loads F32 A matrices two elements at a time when they are 2-aligned. This change addresses performance issues on Intel hardware where loading floats one by one is inefficient, and it also corrects a missing validation for `a_offset` alignment in the original patch.

The modification leverages existing 2-aligned load logic in `mul_mat_vec` to improve throughput on Intel BMG architectures.
Benchmark results on an Intel B60 show significant speedups for specific matrix multiplication shapes, with performance increasing from 153.08 GFLOPS to 221.66 GFLOPS for one test case and up to 1.63 TFLOPS for another.
The release provides binaries for macOS (Apple Silicon and Intel), Linux (CPU, Vulkan, CUDA, ROCm, OpenVINO, SYCL, Snapdragon), Android, Windows, and openEuler.

This update improves inference performance on Intel GPUs using the Vulkan backend by optimizing memory access patterns for floating-point matrix operations.

llama.cpp adds GLM-5.3-Flash support with k-pool and MTP fixes

llama.cpp b11265 fixes Vulkan MOE matmul tile selection

llama.cpp b11238 adds ggml_pad_ext for left padding and skips DFlash2 taps

llama.cpp b11232 enables tiled flash attention on x86

llama.cpp b11228 adds left and circular padding support in Metal GGML_OP_PAD