Revision: 1396
Author: [email protected]
Date: Fri Nov 19 03:57:16 2010
Log: Some tweaks to HACKING
http://code.google.com/p/perl-devel-nytprof/source/detail?r=1396
Modified:
/trunk/HACKING
=======================================
--- /trunk/HACKING Mon Oct 25 08:37:48 2010
+++ /trunk/HACKING Fri Nov 19 03:57:16 2010
@@ -233,17 +233,9 @@
Investigate style.css problem when using --outfile=some/other/dir
-Sub profiler should avoid sv_setpvf(subname_sv, "%s::%s", stash_name,
GvNAME(gv));
-because it's expensive (Perl_sv_setpvf_nocontext accounts for 29% of
pp_entersub_profiler).
-Use a two level hash: HvNAME(GvSTASH(gv)) then GvNAME(gv).
-Should then also be able to avoid newSV/free for subname_sv (which
accounts for 50% of its time).
-
Class::MOP should update %DB::sub (if $^P & 0x10 set) when it creates
methods.
Sub::Name should do same (extracting the file and line from the
ANON[...:...])
-Add refs so a string eval fid can be related to its 'siblings' (other
string
-eval fids from the same line in the 'parent' fid).
-
Profile should report _both_ the 'raw original' filename (possibly
relative)
used by the application being profiled, plus an absolute filename
determined
ASAP (to avoid problems with scripts that chdir).
@@ -314,6 +306,8 @@
Option to add sub call and return events into the data file as they happen.
Would enable a dprofpp -T like output. See
https://rt.cpan.org/Ticket/Display.html?id=50766
+Especially relevant as DProf is being removed from the code and this is
+something NYTProf still doesn't do.
String evals could/should be tied into the subroutine profiler.
That would give inclusive timings which we don't have at the moment.
@@ -341,11 +335,17 @@
Add "calls N subs" to treemap mouseover box
Upgrade treemap to JIT version 2 (which has transition animations).
+Sub profiler should avoid sv_setpvf(subname_sv, "%s::%s", stash_name,
GvNAME(gv));
+because it's expensive (Perl_sv_setpvf_nocontext accounts for 29% of
pp_entersub_profiler).
+Use a two level hash: HvNAME(GvSTASH(gv)) then GvNAME(gv).
+Should then also be able to avoid newSV/free for subname_sv (which
accounts for 50% of its time).
+
The subroutine profiler spends a lot of its time allocating the two
calle[rd]_subname_sv SVs in subr_entry_setup and freeing them in
subr_entry_destroy.
If, instead of freeing the SVs they were chained into a freelist that
subr_entry_setup could pull from, we'd get a significant boost in
performance
-for the subroutine profiler.
+for the subroutine profiler. It would also greatly reduce the risk of
+NYTProf overwriting a recently freed but still on the stack SV.
Would be good to be able to measure the time spent in require. The easy way
would be to treat it as a slowop. I recall enabling this in the early days
@@ -353,7 +353,7 @@
This would help with profiling issues like:
http://blog.moose.perl.org/2010/08/moose-110-and-classmop-105-now-compiling-10-faster.html
-We need a start=begin option to start at the _end_ of any INIT subs.
+We need a start=runtime option to start at the _end_ of any INIT subs.
(The current start=init option is the closest we have but it's not very
useful if lots of other work is done in INIT blocks.)
@@ -362,3 +362,5 @@
profile. Possibly straight-forward to implement. The trick is to add the
time
spent in the sub to the statement profiler overhead time, thus getting it
subtracted from the higher level sub time. Maybe.
+
+
--
You've received this message because you are subscribed to
the Devel::NYTProf Development User group.
Group hosted at: http://groups.google.com/group/develnytprof-dev
Project hosted at: http://perl-devel-nytprof.googlecode.com
CPAN distribution: http://search.cpan.org/dist/Devel-NYTProf
To post, email: [email protected]
To unsubscribe, email: [email protected]
