This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Pass 4 (non-)optimization speedup
- From: Mark Wielaard <mjw at redhat dot com>
- To: systemtap at sources dot redhat dot com
- Date: Sat, 11 Jul 2009 20:32:02 +0200
- Subject: Pass 4 (non-)optimization speedup
Hi,
To make pass 4 a bit more flexible I added -O[0123s] as arguments to
stap (commit 5a5732). This allows you to pass through this option to
gcc. So users can determine whether they want fast compilation or
optimized code for speed or size.
The default is -O0 which makes pass 4 a lot faster, so I think this is a
good default. But maybe the default could be tuned, for example when
using the compile server, which can be assumed to generate systemtap
script modules that will be often reused.
The -O0 default speeds up some things a lot.
$ stap -v -k -p4 -e 'probe syscall.* { log(argstr); }'
Before: Pass 4: compiled C [...] in 7520usr/900sys/8415real ms.
After: Pass 4: compiled C [...] in 4980usr/730sys/5782real ms.
make installcheck RUNTESTFLAGS=buildok.exp is 15 seconds faster,
was 4 minutes 10 seconds, now 3 minutes 55 seconds
(both with clean caches).
Cheers,
Mark