llama.cpp b11203：CUDA FWHT 新增 F16 输入支持
原标题：b11203
AI 摘要
llama.cpp 发布 b11203 版本，主要变更为 CUDA 后端的 FWHT（快速沃尔什-哈达玛变换）内核新增 F16 输入支持。该改动将源数据类型改为模板参数，使内核可直接读取 F16 源而无需先转换为 F32 副本，原有 F32 路径保持不变。同时引入共享谓词 ggml_cuda_op_mul_mat_use_fwht，统一 supports_op 与调度逻辑，避免类型检查与形状断言之间的不一致。在 A10 上运行 test-backend-ops，MUL_MAT 全部 1297 项通过，含 24 个 Hadamard 用例。
正文节选
<details open> cuda: add F16 input to the FWHT (#29096) * cuda: add F16 input to the FWHT The CUDA FWHT accepts F32 input only. This makes the source type a template parameter, so the kernel reads an F16 source directly instead of requiring a converted copy. The F32 path is unchanged. supports_op accepts an F16 src1 against an F32 src0 for the Hadamard hint. Every other F16 src1 against a non-F16 src0 is still refused. ggml_cuda_op_mul_mat_use_fwht is the single predicate both supports_op a