building libjvm with -Os for space optimization - was : RE: RFR: 8234525: enable link-time section-gc for linux s390x to remove unused code

[Baesken, Matthias]

Hello,  a small update from my side .   I had  (on Linux)  for a couple of days our  jdk11u  tests + benchmarks  running with -Os instead of  the usual settings (-O3) .
With have  (independent of  -Os  /  -O3)   a bit of variance in our benchmarks  ( jbb15  , jvm2008, jvm98 ) .

However the benchmark results  of  -Os  vs.  -O3 were comparable .


>we discussed doing the opposite for Mac OS X recently, where builds are
>currently set to -Os by default. -O3 helped various networking
>(micro)benchmarks by up to 20%.


What microbenchmarks do you have in mind  that show  much better performance with -O3  compared to -Os  ?

Best regards, Matthias

> 
> Hi Martin,
> 
> On 2019-11-27 19:03, Doerr, Martin wrote:
> > Hi Claes,
> >
> > that kind of surprises me. I'd expect files which rather benefit from -O3 to
> be far less than those which benefit from -Os.
> > Most performance critical code lives inside the code cache and is not
> dependent on C++ compiler optimizations.
> > I'd expect GC code, C2's register allocation and a few runtime files to be the
> most performance critical C++ code.
> > So the list of files for -Os may become long.
> 
> that might very well be the end result, and once/if we've gone down this
> path long enough to see that -O3 becomes the exception, we can re-
> examine the default. Changing the default and then trying to recuperate
> would be hard/impossible to do incrementally.
> 
> >
> > Yeah, I think we should use native profiling information to find out what's
> really going on >
> > Your idea to change file by file and check for performance regression
> makes sense to me, though
> Hopefully we don't have to do one RFE per file.. :-)
> 
> /Claes