<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent posts to Discussion</title><link href="https://sourceforge.net/p/viennacl/discussion/" rel="alternate"/><link href="https://sourceforge.net/p/viennacl/discussion/feed.atom" rel="self"/><id>https://sourceforge.net/p/viennacl/discussion/</id><updated>2024-06-17T07:10:50.153000Z</updated><subtitle>Recent posts to Discussion</subtitle><entry><title>Is ViennaCL alive?</title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/9a966b9881/?limit=25#3e0d" rel="alternate"/><published>2024-06-17T07:10:50.153000Z</published><updated>2024-06-17T07:10:50.153000Z</updated><author><name>Marcin Błażejowski</name><uri>https://sourceforge.net/u/marcingretl/</uri></author><id>https://sourceforge.nete8413655e4375dcf37b5b6c40e5283c1bb8d00f1</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi,&lt;br/&gt;
I'd like to ask about the current projects' status: is it still developed or already abandoned?&lt;/p&gt;
&lt;p&gt;Cheers,&lt;br/&gt;
Marcin&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Tesla V100 slower than GTX1080</title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/3c35cc783c/?limit=25#2da3" rel="alternate"/><published>2020-10-12T19:22:53.012000Z</published><updated>2020-10-12T19:22:53.012000Z</updated><author><name>Yucheng</name><uri>https://sourceforge.net/u/yuchengs/</uri></author><id>https://sourceforge.neta19a6a85b5005f6bebc6bc98f0c95eb753de0a5a</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I have the same program running on both Tesla V100 and GTX1080. However, it seems Tesla V100 is significantly slower than GTX1080 on solving sparse linear systems (bicgstab with viennacl::linalg::chow_patel_ilu_precond)&lt;/p&gt;
&lt;p&gt;I think the problem is the library is not tuned for Tesla V100 according to its number of SM. Can anyone confirm this?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>how to send exchange data between two cuda devices</title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/617fcaaaa6/?limit=25#5e63" rel="alternate"/><published>2020-08-22T23:41:05.724000Z</published><updated>2020-08-22T23:41:05.724000Z</updated><author><name>p ding</name><uri>https://sourceforge.net/u/ztdep/</uri></author><id>https://sourceforge.net9eee49719bf2205f4abeee202d05559c3b7de99a</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Thanks. but which command should i use to select corresponding GPU devecies? &lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>how to send exchange data between two cuda devices</title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/617fcaaaa6/?limit=25#c3d5" rel="alternate"/><published>2020-08-22T14:28:32.710000Z</published><updated>2020-08-22T14:28:32.710000Z</updated><author><name>Karl Rupp</name><uri>https://sourceforge.net/u/koarl0815/</uri></author><id>https://sourceforge.net3f23a8b9dc71a8062be247bc4e3c30de2b7eeaeb</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;You'll have to do the data copying manually through the host via viennacl::copy (for fast_copy) and switch between the decives accordingly.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>how to send exchange data between two cuda devices</title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/617fcaaaa6/?limit=25#8871" rel="alternate"/><published>2020-08-22T01:19:51.155000Z</published><updated>2020-08-22T01:19:51.155000Z</updated><author><name>p ding</name><uri>https://sourceforge.net/u/ztdep/</uri></author><id>https://sourceforge.net50d73bc323a7da2b6bdeaff314fec66571414e9b</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;If i have 2 GPU cards. how to exchange data with ViennaCL. &lt;br/&gt;
Does ViennalCL support 2 GPU cards default.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>ViennaCL::copy function from ublas to native </title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/130f02a39c/?limit=25#2afe" rel="alternate"/><published>2020-03-28T10:02:01.070000Z</published><updated>2020-03-28T10:02:01.070000Z</updated><author><name>Dmitriy</name><uri>https://sourceforge.net/u/dmitriyp/</uri></author><id>https://sourceforge.netc0921bd6e55ce95f051f3837952861e34b8035e7</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi carl. It is a great news. Let me sat thanks for your work. It really great.&lt;/p&gt;
&lt;p&gt;Actually I replaced unsigned int occurrences in compressed_matrix.hpp, sparse_matrix_hpp and in many places for couple of iterative solvers and now it seems to work fine eve for system of:&lt;br/&gt;
rows = 210 mlns, columns = 210mls and with about 50 nonzeros in each row. I tested it in my fem geomechanics and hydrodynaics code, seems to work fine.&lt;/p&gt;
&lt;p&gt;But of course, I have not capabilities to do it for each preconditioner and each iterative solver.&lt;br/&gt;
It will be a great if you will release a branch with 64-bit integers.&lt;br/&gt;
You a doing a great job.&lt;br/&gt;
Best regards,&lt;br/&gt;
Dmitry&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>ViennaCL::copy function from ublas to native </title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/130f02a39c/?limit=25#02d5" rel="alternate"/><published>2020-03-28T03:58:39.642000Z</published><updated>2020-03-28T03:58:39.642000Z</updated><author><name>Karl Rupp</name><uri>https://sourceforge.net/u/koarl0815/</uri></author><id>https://sourceforge.net3974f4a936488369adba3a07ac5fa94c289a399b</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi Dmitriy,&lt;br/&gt;
I'll give the conversion to 64-bit integers a shot in a separate branch in the developer repository. It will take a day or two until I get to it, though.&lt;br/&gt;
Best regards,&lt;br/&gt;
Karli&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>ViennaCL::copy function from ublas to native </title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/130f02a39c/?limit=25#4ed6" rel="alternate"/><published>2020-03-25T05:35:50.753000Z</published><updated>2020-03-25T05:35:50.753000Z</updated><author><name>Karl Rupp</name><uri>https://sourceforge.net/u/koarl0815/</uri></author><id>https://sourceforge.net1f3692fa6810aaa5f40ebe11138719591ace020f</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Ah, so the suspected 32-bit overflow. :-/&lt;/p&gt;
&lt;p&gt;If you are on Windows, you should attempt a 64-bit build in order to get 64-bit long. On Linux and MacOS 'long' is usually 64 bits anyway. It won't help with the unsigned int overflow, though.&lt;/p&gt;
&lt;p&gt;There is no easy way to get around the &lt;code&gt;unsigned int&lt;/code&gt; issue without making a significant impact on the overall code. It's certainly doable, but will take some time. ViennaCL had primarily GPUs in mind, for which memory is limited to a few GB, thus the GPU runs out of RAM before &lt;code&gt;unsigned int&lt;/code&gt; overflows. With OpenMP on a workstation the situation is apparently different. Let me check the code and I'll get back to you.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>ViennaCL::copy function from ublas to native </title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/130f02a39c/?limit=25#cc9b" rel="alternate"/><published>2020-03-25T03:59:25.252000Z</published><updated>2020-03-25T03:59:25.252000Z</updated><author><name>Dmitriy</name><uri>https://sourceforge.net/u/dmitriyp/</uri></author><id>https://sourceforge.netfecd20a575adcb3af54e061230c43eac11bae6c6</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;So, as I see this problems really connected with very large indexes.&lt;/p&gt;
&lt;p&gt;In file: cpu_ram.hpp (line 94)&lt;/p&gt;
&lt;p&gt;My size_in_bytes was about 2.3e9. &lt;br/&gt;
After classical cast:     long(size_in_bytes)  it becomes negative and data don't copy.&lt;/p&gt;
&lt;p&gt;So I replaced (long) to (long long) and everything is ok.&lt;/p&gt;
&lt;p&gt;But are there any good way to work with indexes larger that (unsigned int)?&lt;br/&gt;
I mean if I have number of nonzeros &amp;gt; 4.6e9?&lt;/p&gt;
&lt;p&gt;Of course, I can rewrite function copy and a lot of othwer functions in compressed_matrix.hpp  to replace (unsigned int) to (vcl_size_t) but it is not good decision))&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>ViennaCL::copy function from ublas to native </title><link href="https://sourceforge.net/p/viennacl/discussion/1143678/thread/130f02a39c/?limit=25#9684" rel="alternate"/><published>2020-03-24T23:15:25.341000Z</published><updated>2020-03-24T23:15:25.341000Z</updated><author><name>Dmitriy</name><uri>https://sourceforge.net/u/dmitriyp/</uri></author><id>https://sourceforge.net9208f26217016e127ea3ffd58cd828285714ddf6</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I had long time of debugging, and saw unsigned_int in many functions of viennacl::compressed_matrix. For example in function copy type of row_jumper and col_buffer must be unsigned int as I see.  Does it mean that I hacve a limitation on number of nonzeros connected with unsigned int maximum?&lt;/p&gt;&lt;/div&gt;</summary></entry></feed>