On Sat, May 31, 2008 at 7:16 AM, David Mosberger-Tang
<[EMAIL PROTECTED]> wrote:
> Caching would improve performance greatly.  The ia64 version of
> libunwind has this implemented, but not the DWARF-based version (yet).

There is such a cache in dwarf/Gparser.c:get_rs_cache().

But I've tested it only on x86. Perhaps it's not getting enabled on
ARM for some reason?

On x86_64:

$ tests/Lperf-simple
unw_getcontext : cold avg=  259.876 nsec, warm avg=  209.808 nsec
unw_init_local : cold avg=  309.944 nsec, warm avg=  169.277 nsec
no cache        : unw_step : 1st=10104.406 min= 1750.674 avg= 1812.264 nsec
global cache    : unw_step : 1st= 7665.725 min=  122.615 avg=  133.057 nsec
per-thread cache: unw_step : 1st= 6103.516 min=  122.615 avg=  133.020 nsec

  -Arun


_______________________________________________
Libunwind-devel mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/libunwind-devel

Reply via email to