Make improvements

- Introduce portable sched_getcpu() api
- Support GCC's __target_clones__ feature
- Make fma() go faster on x86 in default mode
- Remove some asan checks from core libraries
- WinMain() now ensures $HOME and $USER are defined

third_party/nsync/futex.c

@@ -143,7 +143,7 @@ static int nsync_futex_polyfill_ (atomic_int *w, int expect, struct timespec *ab
 		if (abstime && timespec_cmp (timespec_real (), *abstime) >= 0) {
 			return -ETIMEDOUT;
 		}
-		pthread_yield ();
+		pthread_yield_np ();
 	}
 }
 
@@ -373,7 +373,7 @@ int nsync_futex_wake_ (atomic_int *w, int count, char pshare) {
 		}
 	} else {
 	Polyfill:
-		pthread_yield ();
+		pthread_yield_np ();
 		rc = 0;
 	}
 

third_party/nsync/yield.c

@@ -22,6 +22,6 @@
 #include "third_party/nsync/common.internal.h"
 
 void nsync_yield_ (void) {
-	pthread_yield ();
+	pthread_yield_np ();
 	STRACE ("nsync_yield_()");
 }