PATCH: new GC implementation
Mark Mitchell
mark@codesourcery.com
Thu Sep 30 23:58:00 GMT 1999
>>>>> "Andi" == Andi Kleen <ak@muc.de> writes:
Andi> Yes there is actually. Sorry I was wrong. MAP_ANONYMOUS does
Andi> not manipulate the the page tables directly, but only
Andi> creates a vma (a highlevel memory object) for the mapped
Andi> range. The actual pte is only updated on access. /dev/zero
Andi> directly sets up all the ptes, which can be a quite slow
Andi> operation on x86 especially on SMP (the CPU always flushes
Andi> page table changes to the memory bus).
Andi> This is actually a bug, it should really use the same method
Andi> as MAP_ANONYMOUS. Thank you for finding it.
Thanks for the explanation. I think that means we should leave the
use of MAP_ANONYMOUS in the new GC (in contrast to what RTH suggested
in his mail to Alex); lots of us are going to be running our current
kernels for quite some time.
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com
More information about the Gcc-patches
mailing list