Mailing-List: contact user-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hadoop.apache.org
Received-SPF: pass (athena.apache.org: domain of bepcyc@gmail.com designates
 74.125.82.41 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAEAZmyhjg0E6g5nDX25KLc_JWWuwKZdjPU+j-xvXqGUmaY3PgA@mail.gmail.com>
References: 
 <CAEAZmyhjg0E6g5nDX25KLc_JWWuwKZdjPU+j-xvXqGUmaY3PgA@mail.gmail.com>
Date: Wed, 15 Jan 2014 14:09:31 +0100
Message-ID: 
 <CAF-KAGm_0ArxqWnp47b0E+rsqEpv0no7HGXMG7eeCj14AXTtgw@mail.gmail.com>
Subject: Re: HADOOP & CUDA
From: Slava Rodionov <bepcyc@gmail.com>
To: user@hadoop.apache.org
Content-Type: multipart/alternative; boundary=047d7ba975c808736704f00205d1

--047d7ba975c808736704f00205d1
Content-Type: text/plain; charset=ISO-8859-1

Hi Massimo,

I don't see why should anyone actually need any interface between Hadoop
and CUDA in case that you're using C++?
Why don't you just make sure that all of your data nodes are able to run
CUDA locally, then write some function that uses CUDA and then just call
this function inside your mapper or reducer?

Best regards,
Viacheslav


2014/1/15 Massimo Simoniello <massimo.simoniello@gmail.com>

> Hi, I'm working on hadoop 2.2.0 and I'm writting a C++ application with
> pipes. I would like to use also CUDA in my mapper/reducer but I'm not able
> to find enough info.
> I'have found this interesting slide:
> http://www.slideshare.net/airbots/cuda-29330283 and this page in the wiki
> http://wiki.apache.org/hadoop/CUDA%20On%20Hadoop.
>
> I don't understand where are these functions:
>
>    - void processHadoopData(string& input);
>    - void cudaCompute(std::map<string,string>& output);
>
>
> I don't understand where I have to write the code and if it is really
> possible.
>
> Is there any tutorial to learning how to configure hadoop+cuda, write the
> code, compile it, execute it?
>
> Is there any example?
>

--047d7ba975c808736704f00205d1
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hi Massimo,<div><br></div><div>I don&#39;t see why should =
anyone actually need any interface between Hadoop and CUDA in case that you=
&#39;re using C++?</div><div>Why don&#39;t you just make sure that all of y=
our data nodes are able to run CUDA locally, then write some function that =
uses CUDA and then just call this function inside your mapper or reducer?</=
div>
<div><br></div><div>Best regards,</div><div>Viacheslav</div></div><div clas=
s=3D"gmail_extra"><br><br><div class=3D"gmail_quote">2014/1/15 Massimo Simo=
niello <span dir=3D"ltr">&lt;<a href=3D"mailto:massimo.simoniello@gmail.com=
" target=3D"_blank">massimo.simoniello@gmail.com</a>&gt;</span><br>
<blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1p=
x #ccc solid;padding-left:1ex"><div dir=3D"ltr">Hi, I&#39;m working on hado=
op 2.2.0 and I&#39;m writting a C++ application with pipes. I would like to=
 use also CUDA in my mapper/reducer but I&#39;m not able to find enough inf=
o.<div>
I&#39;have found this interesting slide:=A0<a href=3D"http://www.slideshare=
.net/airbots/cuda-29330283" target=3D"_blank">http://www.slideshare.net/air=
bots/cuda-29330283</a>=A0and this page in the wiki=A0<a href=3D"http://wiki=
.apache.org/hadoop/CUDA%20On%20Hadoop" target=3D"_blank">http://wiki.apache=
.org/hadoop/CUDA%20On%20Hadoop</a>.</div>


<div><br></div><div>I don&#39;t understand where are these functions:=A0</d=
iv><div><ul><li><font face=3D"courier new, monospace">void processHadoopDat=
a(string&amp; input);<br></font></li><li><font face=3D"courier new, monospa=
ce">void cudaCompute(std::map&lt;string,string&gt;&amp; output);</font><br>


</li></ul></div><div><br></div><div>I don&#39;t understand where I have to =
write the code and if it is really possible.</div><div><br></div><div>Is th=
ere any tutorial to learning how to configure hadoop+cuda, write the code, =
compile it, execute it?<br>


<br>Is there any example?</div></div>
</blockquote></div><br></div>

--047d7ba975c808736704f00205d1--