[Scilab-Dev] Raw performance or good backtraces for Scilab 6.0.0 ?

[Clément David]

Hello Scilab developers,

On Linux, I discovered (or re-discovered) that it is really hard to get a precise stacktrace on
Scilab 6 using the default build configuration '-02 -g' due to the nice C++ compiler tricks. I tried
to get a nicer one using -fno-omit-frame-pointer and got a complete one with the following benchmark
differences (not precise enough but still some variation). What do you think about enabling it by
default [1] ?



[1]: https://codereview.scilab.org/#/c/18815/

Thanks,

PS: some screenshot of the Linux `perf report` utility for for_increment.tst

--
Clément
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/dev/attachments/20161223/94bc27e6/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: perf-report_before_after.png
Type: image/png
Size: 69193 bytes
Desc: not available
URL: <https://lists.scilab.org/pipermail/dev/attachments/20161223/94bc27e6/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: unknown-D3WFSY
Type: image/png
Size: 38894 bytes
Desc: not available
URL: <https://lists.scilab.org/pipermail/dev/attachments/20161223/94bc27e6/attachment-0001.png>