[patch] Eliminate quadratic slow-down on number of solibs (part 2).

Paul Pluzhnikov ppluzhnikov@google.com
Thu May 21 16:17:00 GMT 2009


On Thu, May 21, 2009 at 8:15 AM, Joel Brobecker <brobecker@adacore.com> wrote:

> I lost track a little. Are there still some patches pending review
> in this series?

Yes, the "hacky patch":
  http://sourceware.org/ml/gdb-patches/2009-05/msg00097.html
saves additional 15% of CPU, because it avoids resetting the breakpoint
on every solib addition, when many solibs are added "at once" (this happens
at program startup/attach on UNIXen, but not on Windows).

Resetting a breakpoint involves non-trivial amount of computation:
bp->addr_string is re-parsed, symbol is looked up, sal's are evaluated, and
finally the breakpoint (on main) is reset to exactly the same location it
were before :-(

I am not happy with the "hackiness" of this patch, but don't see how to
fix this better.

Thanks,
-- 
Paul Pluzhnikov



More information about the Gdb-patches mailing list