https://bugs.kde.org/show_bug.cgi?id=514297

Philippe Waroquiers <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |philippe.waroquiers@skynet.
                   |                            |be

--- Comment #17 from Philippe Waroquiers <[email protected]> ---
Comment on attachment 190171
  --> https://bugs.kde.org/attachment.cgi?id=190171
wip patch

Quickly (too quickly?) read the patch but IIUC:
When before the patch, is_valid_for was returning True,
it will now call belongs_to_guard_page when the address is in a segment that
has a guard page.

And belongs_to_guard_page is likely somewhat heavy (e.g. does several
syscalls).
I believe that some tools (memcheck leak checker for example) will do a high nr
of calls to VG_(am_is_valid_for_client). And the leak checker contains comments
telling that it is not using
the VA bit machinery because it is too slow, so clearly the perf there is
important.

This is also called in various other places (the debuginfo; the stack extension
detection; ...)

So, wondering of the performance impact on the leak check and/or on other
callers of VG_(am_is_valid_for_client)

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to