llama.cpp/ggml
Miro Bucko c3e73990ec cmake: fix ARM feature detection hang on platforms without SVE/SME
Add a compile-time pre-check before the runtime feature test in
check_arm_feature. Before running a test binary with -mcpu=native+<tag>,
first check if -mcpu=native (without the forced +tag) defines the
compiler's __ARM_FEATURE_* macro.

If the compiler doesn't think the native CPU has the feature, skip the
runtime test entirely. This prevents hangs on platforms where the
compiler can compile the feature code (via forced +tag) but the binary
traps at an OS level without delivering SIGILL to userspace — as happens
on macOS/Apple Silicon with SVE instructions.

When the compiler does report native support, the runtime test still
runs as before to verify actual hardware capability.
2026-03-03 17:34:14 +07:00
..
cmake ggml: Skip backend library linking code when GGML_BACKEND_DL=ON (#15094) 2025-08-07 13:45:41 +02:00
include ggml/gguf : prevent integer overflows (#19856) 2026-02-24 20:17:11 +02:00
src cmake: fix ARM feature detection hang on platforms without SVE/SME 2026-03-03 17:34:14 +07:00
.gitignore vulkan : cmake integration (#8119) 2024-07-13 18:12:39 +02:00
CMakeLists.txt ggml : bump version to 0.9.7 (ggml/1425) 2026-02-15 22:24:29 +02:00