Re: HOWTO: benchmarking monotone (was Re: [Monotone-devel] "memory exhausted" error for 'mtn list status' command)

[Nathaniel Smith]

On Sat, Jul 29, 2006 at 03:52:19PM -0700, Eric Anderson wrote:
> The difference is that valgrind is size(allocated_bytes -
> freed_bytes), whereas memtime is looking at (allocated_bytes -
> freed_bytes + external_fragmentation).  It is the external
> fragmentation caused by the allocation of a little memory after
> allocating a large amount that means that even if the large amount is
> freed, since the little memory is allocated after it, the resident
> size doesn't drop.

Huh, that's a useful example.

Of course, things are slightly more subtle even than that... I believe
massif is looking at
  allocated bytes + per currently allocated block overhead + stack size
  - freed bytes
while memtime is looking at
  allocated bytes - large freed blocks - re-used freed bytes
  + heap overhead + stack + mmap()'d executable/library size
or so?

I wonder if there's any convenient way to mask off the read-only mmaps
from memtime, it would make the numbers easier to interpret (and in
particular, make it much easier to compare massif and memtime in order
to measure fragmentation; none of these tools measure fragmentation
directly).

-- Nathaniel

-- 
"But in Middle-earth, the distinct accusative case disappeared from
the speech of the Noldor (such things happen when you are busy
fighting Orcs, Balrogs, and Dragons)."