Re: [RFA 2/4] dwarf2_physname

[Daniel Jacobowitz <dan@codesourcery.com>]

On Wed, Jan 20, 2010 at 12:37:23PM -0800, Keith Seitz wrote:
> 1. time ./gdb -nx -q --batch-silent --pid $PID 2> /dev/null
> HEAD:
> real    0m20.242s  0m2.407s   0m2.418s   0m2.390s   0m2.416s
> user    0m3.565s   0m2.224s   0m2.253s   0m2.225s   0m2.240s
> sys     0m0.680s   0m0.184s   0m0.163s   0m0.165s   0m0.177s
> 
> dwarf2_physname:
> real    0m3.747s   0m2.452s   0m2.447s   0m2.447s   0m2.457s
> user    0m2.293s   0m2.283s   0m2.268s   0m2.267s   0m2.272s
> sys     0m0.185s   0m0.168s   0m0.178s   0m0.178s   0m0.184s

Consistently more, but not a major impact.

> 2. time ./gdb -nx -q --batch-silent --pid $PID -ex "thread apply all
> bt full" 2> /dev/null
> HEAD:
> real    0m3.001s   0m2.422s   0m2.440s   0m2.462s   0m2.447s
> user    0m2.277s   0m2.258s   0m2.250s   0m2.261s   0m2.251s
> sys     0m0.184s   0m0.163s   0m0.189s   0m0.199s   0m0.190s
> 
> dwarf2_physname:
> real    0m2.593s   0m2.502s   0m2.510s   0m2.513s   0m2.502s
> user    0m2.335s   0m2.322s   0m2.349s   0m2.321s   0m2.334s
> sys     0m0.166s   0m0.179s   0m0.160s   0m0.186s   0m0.157s

Ditto.  A little bigger than previously, but still not much.

> 3. time ./gdb -nx -q --batch-silent --readnow --pid $PID 2> /dev/null
> HEAD:
> real    0m28.275s   0m10.996s   0m11.060s   0m11.035s   0m11.038s
> user    0m12.346s   0m10.378s   0m10.473s   0m10.450s   0m10.436s
> sys     0m1.082s    0m0.613s    0m0.567s    0m0.570s    0m0.583s
> 
> dwarf2_physname:
> real    0m36.998s   0m17.212s   0m17.295s   0m17.304s   0m17.288s
> user    0m18.684s   0m16.502s   0m16.537s   0m16.572s   0m16.585s
> sys     0m1.188s    0m0.702s    0m0.740s    0m0.719s    0m0.689s

Not nearly as bad as I expected.

I think that this means we should move on past the speed concern for
now, and focus on any other issues in the patch.  Where were we on
that front?

-- 
Daniel Jacobowitz
CodeSourcery