[Debian-med-packaging] Bug#1041451: [Sender Not Verified] Re: Bug#1041451: gmap: FTBFS on all !amd64 archs

[Thomas Wu]

Hi Andreas,

I have restored the non-SIMD binaries in the latest version 2023-10-01 of
GMAP, which includes various other improvements and bug fixes.  This
version can be downloaded from our Web server at
http://research-pub.gene.com/gmap.  Please let me know if you find any
problems with this release.

Thanks,

Tom


On Tue, Sep 19, 2023 at 11:50 AM Thomas Wu <twu at gene.com> wrote:

>
> Hi Andreas,
>
> I have decided to restore the non-SIMD binaries.  I have just written code
> to do that, and will release a new version once some large-scale test runs
> complete successfully.
>
> Thanks,
>
> Thomas Wu
>
>
>
> On Tue, Sep 19, 2023 at 8:48 AM Thomas Wu <twu at gene.com> wrote:
>
>>
>> Hi Andreas,
>>
>> The GMAP package should compile on both Intel and Apple architectures.  I
>> think we are now requiring SIMD or its equivalent Neon, though, which is
>> why we are no longer building the *.nosimd binaries.  I don't think that's
>> a problem since every modern computer supports SIMD or Neon.  Does Debian
>> require a non-SIMD target machine?  I will try to look at your bug report
>> to understand what is going on.
>>
>> Thanks,
>>
>> Thomas Wu
>>
>>
>> On Tue, Sep 19, 2023 at 7:44 AM Andreas Tille <andreas at an3as.eu> wrote:
>>
>>> **Warning** The sender address (Andreas Tille ) can not be verified,
>>> sender email address could be spoofed. Please take care to proceed.
>>> Control: tags -1 upstream
>>> Control: forwarded -1 Thomas Wu <twu at gene.com>, Colin K. Watanabe <
>>> ckw at gene.com>
>>>
>>>
>>> Hi Thomas and Colin,
>>>
>>> the Debian packaged gmap fails to build since version 2023-06-01 for all
>>> release architectures besides amd64.  You can read about this bug on our
>>> bug tracker[1].  I think the analysis from Étienne below is a sensible
>>> explanation for the issue.
>>>
>>> It would be really helpful if you could clarify why you disabled SIMD?
>>> Does this mean you suggest we should provide gmap for amd64 only?
>>>
>>> Kind regards
>>>     Andreas.
>>>
>>>
>>> [1] https://bugs.debian.org/1041451
>>>
>>> Am Tue, Aug 15, 2023 at 12:42:09PM +0200 schrieb Étienne Mollier:
>>> > Hi,
>>> >
>>> > The relevant part of the error message shows that the generic
>>> > fully scalar gmap.nosimd executable is never built for any cpu
>>> > architecture:
>>> >
>>> >       Note: /<<PKGBUILDDIR>>/build/src/gmap.avx2 does not exist.  For
>>> faster speed, may want to compile package on an AVX2 machine
>>> >       Note: /<<PKGBUILDDIR>>/build/src/gmap.sse42 does not exist.  For
>>> faster speed, may want to compile package on an SSE4.2 machine
>>> >       Note: /<<PKGBUILDDIR>>/build/src/gmap.sse41 does not exist.  For
>>> faster speed, may want to compile package on an SSE4.1 machine
>>> >       Note: /<<PKGBUILDDIR>>/build/src/gmap.ssse3 does not exist.  For
>>> faster speed, may want to compile package on an SSSE3 machine
>>> >       Note: /<<PKGBUILDDIR>>/build/src/gmap.sse2 does not exist.  For
>>> faster speed, may want to compile package on an SSE2 machine
>>> >       Note: /<<PKGBUILDDIR>>/build/src/gmap.nosimd does not exist.
>>> For faster speed, may want to compile package on an non-SIMD machine
>>> >       Error: appropriate GMAP version not found
>>> >
>>> > Looking into src/Makefile.am, indeed they seem disabled upstream
>>> > for the current gmap versions:
>>> >
>>> >       # intersect-uint2.c requires SIMD
>>> >       #bin_PROGRAMS += gmap.nosimd
>>> >       #bin_PROGRAMS += gmapl.nosimd
>>> >       #bin_PROGRAMS += gsnap.nosimd
>>> >       #bin_PROGRAMS += gsnapl.nosimd
>>> >
>>> > My quick attempts to bring the necessary support in the
>>> > aforementioned intersect-uint2.c file were not very fruitful so
>>> > far.  Something in there looks to prevent use of simde.
>>> >
>>> > Anyways, in hope this helps further investigations,
>>> > --
>>> >   .''`.  Étienne Mollier <emollier at debian.org>
>>> >  : :' :  gpg: 8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
>>> >  `. `'   sent from /dev/pts/4, please excuse my verbosity
>>> >    `-
>>>
>>>
>>>
>>> > _______________________________________________
>>> > Debian-med-packaging mailing list
>>> > Debian-med-packaging at alioth-lists.debian.net
>>> >
>>> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-med-packaging
>>>
>>>
>>> --
>>> http://fam-tille.de
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20231004/8018e340/attachment.htm>