Dave Marquardt wrote: > "Dave" == Dave Marquardt <Dave.Marquardt at Sun.COM> writes: > >>> More recently, more work was done for LPOOB, but I'm not as familiar > >>> with that code. > > Roland> Ok.. but even for B61 64k pages are not used by default (or $ ksh93 -c > Roland> 'pmap -s -x $$ ; true' # shows wrong values (note the "true", > otherwise > Roland> ksh93 will |exec()| the last command and pmap shows the page map for > Roland> itself)). > > Dave> So it appears there hasn't been tuning done to help this particular > Dave> case. > > And it appears I'm wrong about this. I've been told by one of the > LPOOB phase 2 developers that UltraSPARC-II support has been in since > snv_52. I'd be curious to see the output you're talking about here. > If the stack or heap aren't large enough, they won't be mapped with > larger pages. So there may be additional tuning that needs to be > done, or it may just not be as aggressive as you'd like.
Or it may not be able to do the right thing "in time". As said in http://mail.opensolaris.org/pipermail/ksh93-integration-discuss/2007-June/002716.html the "hit&&run" usage of a shell or small tools like gzip2/bzip2 may not trigger the LPOOP (=Large Page Out Of the Box) stuff fast enougth. And on some platforms like Solaris 11/B61+B65 on UltraSPARC-3/4 machines like a Blade1000 or a SF68k/SF69k no 64k pages are used by LPOOP even if it's usefull (for example to map the stack with 64k pages or the limited amount of memory used for in/out buffers of gzip/bzip2/etc.). > >>> Finally, you've locked yourself into 64K pages, and it's possible the > >>> kernel will continue to be improved in the selection of large pages. > >>> These improvements may not work if you've already selected 64K pages, > >>> or perhaps the selection of 64K pages will interfere with good > >>> performance. > > Dave> I've opened an RFE, CR 6569725: > > Dave> 6569725 Add better large page out of box support for UltraSPARC I and II > > And it turns out the code is already there, so this RFE has been > closed. Erm... are you sure it's working in B61 or B65 ? I don't see the use 64k pages on an Ultra5 for the stack even if I force the stack usage of ksh93 up to several MB with something like $ for ((j=0 ; j < 5000 ; j++ )) ; do integer i=0 ; function incnum { (( i++ , i < 900 )) && incnum ; } ; incnum ; done' # ... ;-( ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 7950090 (;O/ \/ \O;)
