Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
C
fix(metal): METAL=1 could not build on a stock macOS
Nicolas6e9af36fbcdd
diff --git a/c/colibri.c b/c/colibri.cindex aa8bc12c..40fcf93d 100644--- a/c/colibri.c+++ b/c/colibri.c@@ -66,8 +66,15 @@#ifdef _OPENMP#include <omp.h> /* scratch per-thread nell'attention */#else+/* Shims for a build without an OpenMP runtime (Apple clang without Homebrew+ * libomp is the common case; the Makefile warns and carries on). Every omp_*+ * the engine calls needs one, or the fallback only appears to work: the CPU+ * build compiled because it happens not to reach omp_in_parallel(), while+ * METAL=1 does and failed. */static inline int omp_get_max_threads(void){ return 1; }static inline int omp_get_thread_num(void){ return 0; }+static inline int omp_in_parallel(void){ return 0; } /* no runtime: never inside a team */+static inline void omp_set_num_threads(int n){ (void)n; }#endif#include "omp_tune.h"#ifdef COLI_CUDA@@ -78,7 +85,12 @@ static inline int omp_get_thread_num(void){ return 0; }#endif#ifdef COLI_METAL#include "backend_metal.h"-#include <omp.h>+/* No <omp.h> here: the guarded include above already provides it under _OPENMP+ * and shims omp_get_max_threads/omp_get_thread_num without it. Including it+ * unconditionally defeated that fallback and made METAL=1 unbuildable on a+ * stock macOS -- exactly the platform this backend targets -- even though the+ * Makefile advertises the single-threaded path ("libomp not found: building+ * single-threaded"). */static int g_metal_enabled;static int g_metal_gemm_min=16; /* COLI_METAL_GEMM_MIN: min rows to send a matmul_qt GEMM to GPU *//* output dello shared expert gia' calcolato su GPU (solo Metal layer-CB) */