thx for answers
and
why is 1 thread faster 16 thread
Cpu= AMD 8 Core Linux=Fedora 21
gmic with OpenMP
TIME WITH 1 THREAD : [orti@localhost ~]$ time gmic 1.png -unsharp_octave 4,5,3,0 -o 22.png [gmic]-0./ Start G'MIC interpreter. [gmic]-0./ Input file '1.png' at position 0 (1 image 1280x720x1x3). [gmic]-1./ Apply octave sharpening on image [0], with 4 scales, radius 5, amount 3 and threshold 0. [gmic]-1./ Output image [0] as file '22.png' (1 image 1280x720x1x3). [gmic]-1./ End G'MIC interpreter.
real 0m1.294s user 0m1.776s sys 0m0.115s
thx
TIME WITH 16 THREAD :
-apply_parallel_overlap [orti@localhost ~]$ time gmic 1.png -apply_parallel_overlap "-unsharp_octave 4,5,3,0",0,nb_threads=16 -o 22.png [gmic]-0./ Start G'MIC interpreter. [gmic]-0./ Input file '1.png' at position 0 (1 image 1280x720x1x3). [gmic]-1./ Apply parallelized command '-unsharp_octave 4,5,3,0,0,nb_threads=16' on image [0], with overlap 0 and 8 threads. [gmic]-1./ Output image [0] as file '22.png' (1 image 1280x720x1x3). [gmic]-1./ End G'MIC interpreter.
real 0m1.272s user 0m1.734s sys 0m0.250s
[orti@localhost ~]$ time gmic 1.png -apply_parallel "-unsharp_octave 4,5,3,0",16 -o 22.png [gmic]-0./ Start G'MIC interpreter. [gmic]-0./ Input file '1.png' at position 0 (1 image 1280x720x1x3). [gmic]-1./ Execute command '-unsharp_octave 4,5,3,0,16' on all image [0] in parallel, using 8 threads. [gmic]-1./ Output image [0] as file '22.png' (1 image 1280x720x1x3). [gmic]-1./ End G'MIC interpreter.
real 0m1.348s user 0m1.812s sys 0m0.154s
-apply_parallel [orti@localhost ~]$ time gmic 1.png -apply_parallel "-unsharp_octave 4,5,3,0",0,nb_threads=16 -o 22.png [gmic]-0./ Start G'MIC interpreter. [gmic]-0./ Input file '1.png' at position 0 (1 image 1280x720x1x3). [gmic]-1./ Execute command '-unsharp_octave 4,5,3,0,0,nb_threads=16' on all image [0] in parallel, 8 threads. [gmic]-1./ Output image [0] as file '22.png' (1 image 1280x720x1x3). [gmic]-1./ End G'MIC interpreter.
real 0m1.347s user 0m1.820s sys 0m0.128s
|