From: Jan Smets <jan.smets@al...> - 2013-04-26 09:09:58
First of all, thank you for this awesome tool. I've been sucessfully
using it for a large project.
I would love to see performance improvements in lcov and genhtml. Would
it be possible to implement some parallelism ?
It currently takes 20min to run a single lcov analysis. (and that is
with branch_coverage disabled.)
From: Peter Oberparleiter <oberpar@li...> - 2013-04-29 12:37:57
On 26.04.2013 11:09, Jan Smets wrote:
> First of all, thank you for this awesome tool. I've been sucessfully
> using it for a large project.
Glad to hear it is of use to you!
> I would love to see performance improvements in lcov and genhtml. Would
> it be possible to implement some parallelism ?
> It currently takes 20min to run a single lcov analysis. (and that is
> with branch_coverage disabled.)
I agree that data collection and processing of lcov is too slow,
especially for measuring the code coverage of large projects.
Introducing parallelism, for example when running gcov during data
collection, is likely to improve the run-time significantly, but an
actual implementation faces several problems, such as using a portable
way for implementing parallelism (especially IPC), but also ensuring
that multiple instances of gcov running at the same time don't interfere
with each other.
That said, this topic is high on my list of TODOs for future
improvements of lcov.
Linux on System z Development - IBM Germany
Get latest updates about Open Source Projects, Conferences and News.