[Debian-science-sagemath] Linbox optimisations

[Tobias Hansen]

Hi,

the Linbox 1.4.1 build now fails at the fflas-ffpack version check
because I have version 2.2.2 installed. Did the flags maybe just come
from the fflas-ffpack and/or givaro pkg-config files? I think they are
cleaned up now so Linbox doesn't get the flags there anymore. I'll
upload Linbox 1.4.2 to the repo and then you can see if it works for you.

BTW, there's a --enable-optimization flag in Linbox' debian/rules, I
guess that should just enable the normal -O2 flag or similar?

Thanks, I didn't see the _wip_docbuild branch before. README.Debian in
the master branch says the docbuild works after creating a symlink.

Best,
Tobias


On 10/04/2016 03:20 PM, Ximin Luo wrote:
> (moving back to the list)
> 
> Tobias Hansen:
>> Hi Ximin,
>>
>> it's strange, I'm now working with Linbox 1.4.2 and without changing
>> anything these commands give no output:
>>
>> objdump -S /usr/lib/x86_64-linux-gnu/liblinbox.so | grep
>> '\b\(ymm\|zmm\|vfm\)'
>> objdump -S /usr/lib/x86_64-linux-gnu/liblinboxsage.so | grep
>> '\b\(ymm\|zmm\|vfm\)'
>>
>> It's the same computer that produced the optimized builds before.
>> I didn't find anything regarding a fix for this in the Linbox changelog
>> or in the commit messages between 1.4.1 and 1.4.2.
>>
> 
> This is strange... I did try to disable optimisations before myself in linbox, but I couldn't find any sign of optimisations being applied - -mcpu or -march or ./configure flags, etc. Which is why I didn't do this myself already :p
> 
> If you rebuild 1.4.1 again now (with your current computer) does it still apply the optimisations? I hope it does, so we have something reproducible to work with..
> 
> If even 1.4.1 now builds without optimisations for you, perhaps do you think your previous build had some extra environment variables that enabled optimisations?
> 
> If we really can't figure it out, I suggest you add the above snippet to debian/rules as a post-build check, e.g. in override_dh_auto_test or something, just as a "sanity check" for future builds.
> 
>> But my sage build fails in the docbuild because sage segfaults. Also
>> can't run the test because sage-starts fails.
>>
> 
> I have a _wip_docbuild branch in git that should fix the segfault. However, it is now running out of memory for my machine as well, that has previously run the docbuild fine. (Sometimes it even locks up my machine and I have to restart. :/)
> 
> I don't know why sage-starts fails for you though. It works for me (master branch). What's the error?
> 
> X
>