On Wed, Aug 27, 2008 at 8:54 PM, C. Scott Ananian <[EMAIL PROTECTED]> wrote: > On Wed, Aug 27, 2008 at 1:34 PM, Tomeu Vizoso <[EMAIL PROTECTED]> wrote: >> 3.0 MiB + 277.5 KiB = 3.3 >> MiB /usr/bin/python2.5/usr/sbin/olpc-update-query--auto-s10 >> Normal build: >> >> ... >> 3.3 MiB + 333.0 KiB = 3.6 MiB python/usr/bin/sugar-shell-service >> --------------------------------- >> 103.9 MiB > > You added an olpc-update-query in your first set of numbers (-OO), but > there wasn't one running in your 'normal' run. Also, your addition > seems to be off, probably because you are eliding other rows which > aren't actually matching up between runs. When I add the matching > columns that you included, I get 73.7M for -OO and 79.7M for 'normal', > for a 6M (~8%) savings.
Yes, I just wanted to post in a hurry what I got yesterday. That's what "preliminar" wanted to convey. >> Had to backport two patches for -OO to work: >> >> https://bugzilla.redhat.com/show_bug.cgi?id=460334 > > See also http://scipy.org/scipy/numpy/ticket/893, which contains a > patch we'd need to apply to numpy if we were to actually use -OO. True, I'll add a fedora bug today if we are still going to pursue this path. >> As Scott wrote initially, we still need to dig a bit more in order to >> see all the tradeoffs in play here. > > So far it seems we have "modest" memory savings (perhaps more as we > write more documentation!) -- I'd like to see numbers quantifying any > speed improvement (if any) which -OO provides, as well as the NAND > cost: how big are all those .pyo files? > > Do folks think it's worth trying out -OO in joyride? I'm going to spend some hours today in order to quantify all this better. More later. Regards and big thanks to Riccardo and Scott for suggesting this possibility, Tomeu _______________________________________________ Sugar mailing list [email protected] http://lists.laptop.org/listinfo/sugar
