a big reason for why mapbench is slow on xv6: mfence!

the kernel spends about 35% of its time in mfence. commented-out #define to get rid of __sync_synchronize for testing. maybe having many atomics is going to be expensive.

a big reason for why mapbench is slow on xv6: mfence!
40c96d6a · Nickolai Zeldovich · be61fd64 · 40c96d6a
--- a/include/atomic.hh
+++ b/include/atomic.hh
 #pragma once

+#if 0
+#define __sync_synchronize() do { __asm__ __volatile__("" ::: "memory"); } while (0)
+#endif
+
 #define _GLIBCXX_VISIBILITY(x)
 #define _GLIBCXX_BEGIN_NAMESPACE_VERSION
 #define _GLIBCXX_END_NAMESPACE_VERSION