runtime: fix misaligned 64-bit atomic
Fixes #4869.
Fixes #5007.
Update #5005.

R=golang-dev, 0xe2.0x9a.0x9b, bradfitz
CC=golang-dev
https://golang.org/cl/7534044
diff --git a/src/pkg/runtime/runtime.h b/src/pkg/runtime/runtime.h
index 8ae6e6a..d9afd5b 100644
--- a/src/pkg/runtime/runtime.h
+++ b/src/pkg/runtime/runtime.h
@@ -483,6 +483,7 @@
 	bool wait;			// if true, wait while all threads finish processing,
 					// otherwise parfor may return while other threads are still working
 	ParForThread *thr;		// array of thread descriptors
+	uint32 pad;			// to align ParForThread.pos for 64-bit atomic operations
 	// stats
 	uint64 nsteal;
 	uint64 nstealcnt;