[
https://issues.apache.org/jira/browse/LUCENE-7745?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15952375#comment-15952375
]
vikash commented on LUCENE-7745:
--------------------------------
Hello all,
I have been reading a lot about GPU working and GPU parallelization in
particularly about General Purpose computing on Graphics Processing Units and
have also looked into in detail the source code of the BooleanScorer.java file
, its a nice thing and I am having no difficulty understanding its working
since Java is my speciality so the job was quite fun . There are a few things
that seem unclear to me but I am reading and experimenting so I will resolve
them soon. It is a nice idea to use gpu to perform the search and indexing
operations on a document using the GPU and that would be faster using the GPU.
And regarding the licencing issue, since we are generating code and as it was
said above the code that we generate may not go to Lucene necessarily so
assuming this happens then will licencing still be an issue if we use the
libraries in our code? And as Uwe Schindler said we may host the code on
github and certainly it would not be a good idea to develop code for special
hardware, but still we can give it a try and try to make it compatible with
most of the hardwares. I dont mind if this code does not go to Lucene, but we
can try to change lucene and make it better and I am preparing myself for it
and things would stay on track with your kind mentorship .
So should I submit my proposal now or do I need to complete all the four steps
that Ishaan told to do in the last comment and then submit my proposal? And
which one of the ideas would you prefer to mentor me on that is which one do
you think would be a better one to continue with?
>Copy over and index lots of points and corresponding docids to the GPU as an
>offline, one time operation. Then, given a query point, return top-n nearest
>indexed points.
>Copy over and index lots of points and corresponding docids to the GPU as an
>offline, one time operation. Then, given a polygon (complex shape), return all
>points that lie inside the polygon.
>Benchmarking an aggregation over a DocValues field and comparing the
>corresponding performance when executed on the GPU.
>Benchmarking the speed of calculations on GPU vs. speed observed when doing
>the same through the BooleanScorer. Preferably, on a large result set with the
>time for copying results and scores in and out of the device memory from/to
>the main memory included?
-Vikash
> Explore GPU acceleration
> ------------------------
>
> Key: LUCENE-7745
> URL: https://issues.apache.org/jira/browse/LUCENE-7745
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Ishan Chattopadhyaya
> Labels: gsoc2017, mentor
>
> There are parts of Lucene that can potentially be speeded up if computations
> were to be offloaded from CPU to the GPU(s). With commodity GPUs having as
> high as 12GB of high bandwidth RAM, we might be able to leverage GPUs to
> speed parts of Lucene (indexing, search).
> First that comes to mind is spatial filtering, which is traditionally known
> to be a good candidate for GPU based speedup (esp. when complex polygons are
> involved). In the past, Mike McCandless has mentioned that "both initial
> indexing and merging are CPU/IO intensive, but they are very amenable to
> soaking up the hardware's concurrency."
> I'm opening this issue as an exploratory task, suitable for a GSoC project. I
> volunteer to mentor any GSoC student willing to work on this this summer.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
