This HTML automatically generated with rman for NEMO
Table of Contents

Name

ccdbench - benchmark taking moments along an axis of an image

Synopsis

ccdbench [parameter=value]

Description

ccdbench currently reads a cube (ideally the same dimensions in X, Y and Z) in 6 different looping patterns, not surprisingly showing huge variations in CPU as the stride misses the cache.

Parameters

The following parameters are recognized in order; they may be given in any order if the keyword is also given. Use --help to confirm this man page is up to date.
in=
Input image(5NEMO) file. No default.
order=
Order to loop over cube, where the last dimension is the fastest running in memory. xyz should be the fastest. [Default: xyz]
iter=
How often to loop [1]

Bench

The associated Benchfile runs all 6 access patterns on a 512 cube image. The size of the cube can be varied using SIZE=512,512,512:

cd $NEMO/src/image/misc
make -f Benchfile SIZE=512,512,512 clean bench1

with the following CPU results. The sys column in the overhead read time, quite possible disk cashing limited the damage here.

SIZE ccdgen sys xyz yxz xzy yzx zxy zyx CPU ---- ------ ---- ---- ---- ---- ------ ----- ------ ------------- 512 2.16 0.40 0.07 0.10 0.45 1.31 1.72 2.16 i7-1260P
1024 20.78 0.70 0.60 0.83 4.47 11.86 15.06 20.64 i7-1260P 512 27.15 0.20 0.12 0.16 0.96 24.39 0.78 23.67 AMD EPYC 7302
1024 247.59 1.46 1.01 1.17 7.64 194.57 6.08 192.77 AMD EPYC 7302

See Also

snapbench(1NEMO) , image(5NEMO)

Files

src/image/misc

Author

Peter

Teuben

History


18-mar-2024    Created in a rush    PJT

Table of Contents