Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
C
fix(colibri.c): add Windows fallbacks for now_s and rss_gb
4ny3ld8abaae6809c
diff --git a/c/colibri.c b/c/colibri.cindex 656fdc64..5245410b 100644--- a/c/colibri.c+++ b/c/colibri.c@@ -349,13 +349,30 @@ static int parse_cuda_devices(const char *list, int *out){return n;}#endif-static double now_s(void){ struct timespec t; clock_gettime(CLOCK_MONOTONIC,&t); return t.tv_sec+t.tv_nsec*1e-9; }-static double rss_gb(void){ struct rusage r; getrusage(RUSAGE_SELF,&r);+static double now_s(void){+#ifdef _WIN32+ LARGE_INTEGER freq, count;+ QueryPerformanceFrequency(&freq);+ QueryPerformanceCounter(&count);+ return (double)count.QuadPart / (double)freq.QuadPart;+#else+ struct timespec t; clock_gettime(CLOCK_MONOTONIC,&t); return t.tv_sec+t.tv_nsec*1e-9;+#endif+}+static double rss_gb(void){+#ifdef _WIN32+ PROCESS_MEMORY_COUNTERS_EX pmc = {0}; pmc.cb = sizeof(pmc);+ if(GetProcessMemoryInfo(GetCurrentProcess(), (PROCESS_MEMORY_COUNTERS*)&pmc, sizeof(pmc)))+ return pmc.WorkingSetSize / (1024.0 * 1024.0 * 1024.0);+ return 0;+#else+ struct rusage r; getrusage(RUSAGE_SELF,&r);#ifdef __APPLE__return r.ru_maxrss/(1024.0*1024.0*1024.0); /* macOS: ru_maxrss in BYTE */#elsereturn r.ru_maxrss/(1024.0*1024.0); /* Linux: in KB */#endif+#endif}/* ---- PROF=1: opt-in performance profile ----------------------------------* Records per-forward decode latency and expert-file bytes fetched, then