On Tue, 11 Nov 2014, Bruce Lilly wrote:
==> time.32.1 <==
real 0.43
user 0.38
sys 0.03
Such short run-times are not particularly accurate since times(2) only
counts time in terms of kernel clock ticks (e.g. 100hz) and other
factors may also dominate the run times. Use OpenIndiana/Solaris
'ptime -m' if you need more precise time accounting than traditional
'time' can provide.
Here are some simple performance tests of a heavily floating point and
heavily parallel algorithm with GraphicsMagick (using its built-in
benchmark driver) compiled as 64-bit and 32-bit using GCC:
64-bit:
gm benchmark -duration 10 convert -size 4000x3000 xc:gray -gaussian 0x3 null:
Results: 32 threads 20 iter 309.75s user 10.45s total 1.914 iter/s 0.065
iter/cpu
32-bit:
gm benchmark -duration 10 convert -size 4000x3000 xc:gray -gaussian 0x3 null:
Results: 32 threads 18 iter 302.67s user 10.29s total 1.749 iter/s 0.059
iter/cpu
The key factors above are the 'iter/s' and 'iter/cpu' values.
In the above cases, the 64-bit version ran a bit faster.
If the amount of data required exceeds what the 32-bit version can
handle, then the 64-bit version runs a *lot* faster.
Bob
--
Bob Friesenhahn
bfrie...@simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
_______________________________________________
openindiana-discuss mailing list
openindiana-discuss@openindiana.org
http://openindiana.org/mailman/listinfo/openindiana-discuss
