This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Some timings after LCM, resp. alignment of stack doubles on i686


L.S.,

Here are some timings for our NWP code after addition of LCM and after
the alignment code for stack doubles went into egcs.

                  Pre-LCM            with LCM          with alignment

0SUPOBS TOOK :    0.30999946594      0.32000064850        0.31000041962
0DATACH TOOK :  147.2699890137     150.6799926758       148.7799987793
0ANAEVA TOOK :  214.6200103760     209.6599884033       200.5200042725
0GRPEVA TOOK : 1259.9199218750    1973.4698486328      1227.4000244141
0HUMSUP TOOK :    0.05993652344      0.05004882812        0.05004882812
0DATACH TOOK :   46.0799560547      46.5000000000        46.8900146484
0HUMEVA TOOK :   23.9699707031      23.1499023438        23.2700195312
0GRPEVA TOOK :   31.8000488281      32.6699218750        31.3900146484

See the fourth line !  This code contains _some_ DOUBLE PRECISION
variables and automatic arrays.  Apparently something in the LCM code
generation threw it off-alignment, which was repaired by the alignment
updates.

Unfortunately, the forecast code crashes (it generates NaNs and then
dies) since the LCM change; I'll have to dig into that.

HTH,

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
g77 Support: fortran@gnu.org; egcs: egcs-bugs@cygnus.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]