Messages in this thread | | | Date | Wed, 2 Apr 2008 15:51:56 -0700 (PDT) | From | Linus Torvalds <> | Subject | Re: [PATCH]: Fix SMP-reordering race in mark_buffer_dirty |
| |
On Thu, 3 Apr 2008, Mikulas Patocka wrote: > > When I added memory reads around mfence, the combined result was worse > than the sum of reads and mfence --- mfence alone 8 ticks, mfence+4reads > 18 ticks.
Sorry, my test-program was buggy.
> mfence+writes scales linearly, i.e. mfence + 4 writes does 12 ticks.
I get different results now that I fixed my memop test. I get a loop with reads and writes taking 3 cycles without mfence, and 21 cycles with (read after, write before).
Linus
| |