Fwd: [GRASS-dev] multi-threaded r.mapcalc

[Hamish]

Glynn wrote:
>>> I have added experimental multi-threading support to r.mapcalc in 7.0

Markus:
> > r.shaded.relief seems to have the most complex r.mapcalc expression of
> > any of the supplied scripts.
> 
> I have built r.mapcalc with 'make USE_PTHREAD=1' in 7,
> this are the results:
....
> GRASS 7.0.svn (nc_spm_07):~ > time r.shaded.relief elevation
....
> 9.22 user;  1.54 system;  0:09.63 elapsed;  111%CPU
....
> GRASS 6.5.svn (nc_spm_07):~ > time r.shaded.relief elevation
....
> 10.68 user;  0.40 system;  0:11.13 elapsed;  99%CPU


not too much of a speedup. I wonder how it looks if you try with a more
"difficult" res=1 ?  I guess it is still more of an I/O disk bound
bottleneck than a computational one. what's a nice expensive r.mapcalc
expression to try? pow()? atan2()? graph()? mode()?


Hamish