5
90%
master: 90%

Ran 23 Oct 2017 04:11AM UTC

Files 37

Run time 1s

Badge

Embed ▾

Committed 23 Oct 2017 04:00AM UTC coverage: 89.101% (+0.05%) from 89.05%

Job # COMPILER=g++-6 CXXFLAGS="-coverage" LDFLAGS="-coverage"

Build Type

push

travis-ci

Committed by

rtobar

Commit Message

Adding OpenMP support to BruteForceConvolver

Just like we perform convolution for individual output pixels in
parallel in OpenCL, we now also support OpenMP-accellerated 2d
brute-force convolution. The previous incarnation of the brute-force
convolution needed some small adjustments to be OpenMP-safe (i.e,
declaring variables inside the for loop for proper thread isolation),
but the structure of the code overall was already correct, looping first
over the output pixels (which now we do with "omp parallel for") and
then over the source images.

A small new test double-checks that for any data we get the same results
when using one thread or two threads. I also separately double-checked
that the new code structure produces the same results than the old one
for the case of one thread. There is obviously now no such test; that
would have required having the two copies of the code living alongside,
and therefore a new Convolver class just to cover the one thread case.
The preformance when compiled properly was the same, and therefore there
was no real reason to keep the old code around.

Signed-off-by: Rodrigo Tobar <rtobar@icrar.org>

Run Details

981 of 1101 relevant lines covered (89.1%)

3645406.28 hits per line

ICRAR / libprofit / 121 / 5
90%
master: 90%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job 121.5 (COMPILER=g++-6 CXXFLAGS="-coverage" LDFLAGS="-coverage")

ICRAR / libprofit / 121 / 5 90% master: 90%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job 121.5 (COMPILER=g++-6 CXXFLAGS="-coverage" LDFLAGS="-coverage")

ICRAR / libprofit / 121 / 5
90%
master: 90%

README BADGES
x