[geos-devel] Optimization

[Kurt Schwehr]

Amanda,

You can take a look at GDAL if you want to see an example of what a
different project has done, you can take a look at:

https://trac.osgeo.org/gdal/browser/trunk/gdal/alg/gdalwarpkernel.cpp

Search for USE_SSE and _mm.  Being that this is open source, this typically
boils down to:

1. Someone has the time / $ to implement it including tests
2. The project is willing to accept the contribution and the overhead it
may add to the project

There are a good number of issues hidden in #2.


On Tue, May 2, 2017 at 6:48 PM, Stephen Woodbridge <woodbri at swoodbridge.com>
wrote:

> I can't speak for the developers, but I think it is highly unlikely that
> they would do anything along these lines if it is not compatible across
> multiple architectures. There are lots of high performance CPUs/GPUs/etc
> out there each with their own special requirements. Who is going to fund
> development, ongoing support, debugging and bug fixing. And you have to
> have systems with these chips in them to support that. It is costly and
> time consuming to do this for even one of them. This also puts multiple
> different execution paths in the code that makes it harder to maintain.
>
> I think the community would rather the developers make really good,
> portable, well tested code that works on all platforms, and it release in
> regular stable way with new geospatial features rather than focus on
> support one platform in some specialized way instead of doing new feature
> development.
>
> My 2 cents anyway.
>   -Steve
>
> On 5/2/2017 5:51 PM, Amanda Osvaldo wrote:
>
>> Hi Steve, more than the compiler it is necessary for the data to fit into
>> the processor correctly.
>>
>> And more, I don't have found GPU code yet.
>> I don't know if it's will take some advantage, however for a large scale
>> processing it should be necessary.
>>
>> *Tell me, I not confounding the things?*
>>
>> I see Geos has a geometry engine, however, it doesn't mean that it's a
>> high-performance engine or designed for massive computation.
>>
>> -- Amanda Osvaldo
>>
>> On Tue, 2017-05-02 at 14:02 -0400, Stephen Woodbridge wrote:
>>
>>> On 5/2/2017 1:53 PM, Amanda Osvaldo wrote:
>>>
>>>> Hi everybody. I'm reading the Geos source code and have no found any
>>>> low-level optimization, like the use AVX2 instruction set for Intel
>>>> processors. It's an important question because without these and others
>>>> optimization I have no significant difference between a Intel Core and
>>>> Intel Xeon processors. Can someone tell me if the lack of optimization is
>>>> intentional or just a lack of programmers? -- Amanda Osvaldo
>>>>
>>>
>>>
>>> Amanda,
>>>
>>> I would think that this is the responsibility of the compiler, not the
>>> application. I believe you can get an Intel compiler that is designed to
>>> optimize for these various HW differences.
>>>
>>> https://software.intel.com/en-us/blogs/2013/01/03/free-intel
>>> -c-compilers-for-students-and-related-parallel-programming-tools
>>>
>>> -Steve W
>>>
>>> ---
>>> This email has been checked for viruses by Avast antivirus software.
>>> https://www.avast.com/antivirus
>>>
>>> _______________________________________________
>>> geos-devel mailing list
>>> geos-devel at lists.osgeo.org <mailto:geos-devel at lists.osgeo.org>
>>> https://lists.osgeo.org/mailman/listinfo/geos-devel
>>>
>>
>>
>> _______________________________________________
>> geos-devel mailing list
>> geos-devel at lists.osgeo.org
>> https://lists.osgeo.org/mailman/listinfo/geos-devel
>>
>>
>
> ---
> This email has been checked for viruses by Avast antivirus software.
> https://www.avast.com/antivirus
>
> _______________________________________________
> geos-devel mailing list
> geos-devel at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/geos-devel
>



-- 
--
http://schwehr.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/geos-devel/attachments/20170504/c2e517c8/attachment.html>