Skip to content

Commit

Permalink
Initial commit: migrating from UW server
Browse files Browse the repository at this point in the history 
... since I'll be leaving UW pretty soon
  • Loading branch information
@hcho3
hcho3 committed Feb 9, 2018
0 parents commit dce8a1f
Show file tree
Hide file tree
Showing 27 changed files with 654 additions and 0 deletions.
79 changes: 79 additions & 0 deletions cholesky.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,79 @@
<!DOCTYPE html>
<html>
<head>
<meta name="viewport" content="width=device-width" />
<link rel="shortcut icon" href="favicon.ico" />
<script type="text/javascript" async src="fonts-min.js"></script>
<link rel="stylesheet" href="style.css" />
<meta charset="utf-8">
<meta name="description" content="We use computing power of general-purpose
GPUs to accelerate a dense linear algebra routine known as Cholesky
decomposition." />
<title>
An efficient out-of-core implementation of block Cholesky decomposition on
a multi-GPU system
</title>
</head>
<body>
<div id="container" itemscope
itemtype="http://www.schema.org/ScholarlyArticle">
<p>[<a href=".">&larr; Go back to profile</a>]</p>
<h1 itemprop="name">An efficient out-of-core implementation of block
Cholesky decomposition on a multi-GPU system</h1>
<p>
with
<a href="http://internet2.trincoll.edu/facProfiles/Default.aspx?fid=1335119">
Lin Cheng</a>,
<a href="http://turing.cs.trincoll.edu/~pyoon/">Peter Yoon</a> and
<a href="http://www.linkedin.com/pub/jiajia-zhao/34/a32/8ab">Jiajia Zhao</a>
</p>
<p>
Paper presented at
<a href="http://www.iasted.org/conferences/pastinfo-789.html">IASTED
PDCS</a>
(2012)<br>
Poster presented at <a href="https://www.ieee.org/conferences_events/conferences/conferencedetails/index.html?Conf_ID=14633">IEEE EMBS</a>
(2012)
</p>
<h2>Download</h2>
<ul>
<li>IASTED PDCS talk:
[<a href="talks/cholesky_talk.pptx">PPTX</a>]
[<a href="talks/cholesky_talk.pdf">PDF</a>]</li>
<li>IASTED PDCS paper:
[<a href="http://www.actapress.com/PaperInfo.aspx?PaperID=454758">Abstract</a>]
[<a href="preprints/cholesky_iasted.pdf">Paper</a>]
</li>
</ul>
<h2>Synopsis</h2>
<p itemprop="description">We use computing power of general-purpose
GPUs to accelerate a dense linear algebra routine known as <strong>
Cholesky decomposition</strong>. Our implementation eliminates the
limitation in memory space by storing the system matrix in hard disk
and loading only parts of it into main memory.
</p>
<h2>Publication Details</h2>
<ul>
<li>
Conference Paper:<br>
Lin Cheng, Hyunsu Cho, Peter Yoon, and Jiajia
Zhao. &ldquo;An Efficient Out-of-Core Implementation of Block
Cholesky Decomposition on a Multi-GPU System,&rdquo;
<em>Proceedings of the 24th IASTED International Conference:
Parallel and Distributed Computing and Systems</em>, Las Vegas,
NV, November 13, 2012.
</li>
<li>
Poster:<br>
Lin Cheng, Hyunsu Cho, Peter Yoon, and Jiajia Zhao. &ldquo;A
Multi-GPU Implementation of Incomplete Cholesky Decomposition for
Image Segmentation,&rdquo;
<em>The 34th Annual International Conference
of the IEEE Engineering in Medicine and Biology Society</em>,
San Diego, CA, August 30, 2012.
</li>
</ul>
<p>[<a href=".">&larr; Go back to profile</a>]</p>
</div>
</body>
</html>
Binary file added cv.pdf
Binary file not shown.
Binary file added cv/cv.docx
Binary file not shown.
Binary file added cv/cv.pdf
Binary file not shown.
87 changes: 87 additions & 0 deletions dstedc.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,87 @@
<!DOCTYPE html>
<html>
<head>
<meta name="viewport" content="width=device-width" />
<link rel="shortcut icon" href="favicon.ico" />
<script type="text/javascript" async src="fonts-min.js"></script>
<link rel="stylesheet" href="style.css" />
<meta charset="utf-8">
<meta name="description" content="We present a memory-efficient
implementation of divide-and-conquer eigenvalue algorithm. It features
automatic profiling that dynamicallly partitions work according to machine
configuration." />
<title>
A memory-efficient algorithm for large-scale symmetric tridiagonal
eigenvalue problem on multi-GPU systems
</title>
</head>
<body>
<div id="container" itemscope
itemtype="http://www.schema.org/ScholarlyArticle">
<p>[<a href=".">&larr; Go back to profile</a>]</p>
<h1 itemprop="name">
A memory-efficient algorithm for large-scale symmetric tridiagonal
eigenvalue problem on multi-GPU systems
</h1>
<p>
with
<a href="http://turing.cs.trincoll.edu/~pyoon/">Peter Yoon</a>
</p>
<p>
Paper presented at
<a href="http://worldacademyofscience.org/worldcomp14/ws/conferences/pdpta14">
WorldComp 2014: PDPTA</a> (2014)<br>
Poster presented at the
<a href="http://www.gputechconf.com/page/home.html">
GPU Technology Conference</a> (2014)
</p>
<h2>Download</h2>
<ul>
<li>Code: <a href="https://github.com/hcho3/dstedc_mgpu">
https://github.com/hcho3/dstedc_mgpu</a></li>
<li>GTC poster: [<a href="posters/dstedc_gtc.pdf">PDF</a>]</li>
<li>WorldComp PDPTA talk:
[<a href="talks/dstedc_talk.pptx">PPTX</a>]
[<a href="talks/dstedc_talk.pdf">PDF</a>]
</li>
<li>WorldComp PDPTA paper:
[<a href="preprints/dstedc_worldcomp.pdf">Paper</a>]</li>
</ul>
<h2>Synopsis</h2>
<p itemprop="description">
<strong>Divide-and-conquer algorithm</strong> is a numerically stable
and efficient algorithm that computes the eigenvalues and eigenvectors
of a symmetric tridiagonal matrix. We often face the situation where
the input matrix fits into the main memory but not into the on-chip
memory of a GPU device. We present an <strong>out-of-core
implementation</strong> where only part of the input matrix is
resident in GPU memory at any point in time. It works independently of
the physical size of GPU memory, handling any size of input as long as
it fits into the main memory. Work is dynamically allocated to
multiple GPUs and CPU cores, taking account of available workspaces
and progress of the algorithm. In addition, it delivers a performance
comparable to that of conventional multi-GPU implementations for cases
where workspaces fit into the GPU memory.
</p>
<h2>Publication Details</h2>
<ul>
<li>
Conference Paper:<br>
Hyunsu Cho and Peter Yoon. &ldquo;A Memory-Efficient Algorithm for
Large-Scale Symmetric Tridiagonal Eigenvalue Problem on Multi-GPU
Systems,&rdquo; <em>Proceedings of the 2014 International
Conference on Parallel and Distributed Processing Techniques
and Applications</em>, pp. 568-573, Las Vegas, NV, July 24, 2014.
</li>
<li>
Poster:<br>
Hyunsu Cho and Peter Yoon. &ldquo;Symmetric Tridiagonal Eigenvalue
Problem on Multi-GPU Systems,&rdquo;
<em>The GPU Technology Conference 2014</em>, San Jose, CA,
March 24, 2014.
</li>
</ul>
<p>[<a href=".">&larr; Go back to profile</a>]</p>
</div>
</body>
</html>
87 changes: 87 additions & 0 deletions eigenmap_gpu.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,87 @@
<!DOCTYPE html>
<html>
<head>
<meta name="viewport" content="width=device-width" />
<link rel="shortcut icon" href="favicon.ico" />
<script type="text/javascript" async src="fonts-min.js"></script>
<link rel="stylesheet" href="style.css" />
<meta charset="utf-8">
<meta name="description" content="Laplacian eigenmap is an image
segmentation algorithm that began to gain traction in recent years. We make
use of general-purpose GPUs to accelerate the algorithm." />
<title>
GPU accelerated vessel segmentation using Laplacian eigenmaps
</title>
</head>
<body>
<div id="container" itemscope
itemtype="http://www.schema.org/ScholarlyArticle">
<p>[<a href=".">&larr; Go back to profile</a>]</p>
<h1 itemprop="name">GPU accelerated vessel segmentation using
Laplacian eigenmaps</h1>
<p>
with
<a href="http://internet2.trincoll.edu/facProfiles/Default.aspx?fid=1335119">
Lin Cheng</a>,
<a href="http://turing.cs.trincoll.edu/~pyoon/">Peter Yoon</a> and
<a href="http://www.linkedin.com/pub/jiajia-zhao/34/a32/8ab">Jiajia Zhao</a>
</p>
<p>
Paper presented at
<a href="http://www.iasted.org/conferences/pastinfo-811.html">
IASTED PDCN</a> (2014)<br>
Poster presented at the <a href="http://www.gputechconf.com/page/home.html">
GPU Technology Conference</a> (2013)
</p>
<h2>Download</h2>
<ul>
<li>
Code:
<a href="https://github.com/hcho3/eigenmap_gpu">
https://github.com/hcho3/eigenmap_gpu
</a>
</li>
<li>
GTC poster: [<a href="posters/eigenmap_gpu_gtc.pdf">PDF</a>]
</li>
<li>
IASTED PDCN talk:
[<a href="talks/eigenmap_gpu_talk.pptx">PPTX</a>]
[<a href="talks/eigenmap_gpu_talk.pdf">PDF</a>]
</li>
<li>IASTED PDCN paper:
[<a href="http://www.actapress.com/PaperInfo.aspx?paperId=455890">Abstract</a>]
[<a href="preprints/eigenmap_gpu_iasted.pdf">Paper</a>]
</li>
</ul>
<h2>Synopsis</h2>
<p itemprop="description">
<strong>Laplacian eigenmap</strong> is an image segmentation algorithm
that began to gain traction in recent years. It involves a generalized
eigenvalue problem which extracts high-level features from local
neighborhood information. Unfortunately, it is computationally costly
to compute eigenvalues of a large linear systems. We make use of
general-purpose GPUs to accelerate the segmentation process.
</p>
<h2>Publication Details</h2>
<ul>
<li>
Conference Paper:<br>
Lin Cheng, Hyunsu Cho, and Peter Yoon. &ldquo;GPU Accelerated
Vessel Segmentation Using Laplacian Eigenmaps,&rdquo;
<em>Proceedings of the IASTED International Conference on Parallel
and Distributed Computing and Networks</em>, pp. 177-184,
Innsbruck, Austria, February 17, 2014.
</li>
<li>
Poster:<br>
Lin Cheng, Hyunsu Cho, Peter Yoon, and Jiajia Zhao. &ldquo;GPU
Accelerated Vessel Segmentation Using Laplacian Eigenmaps,&rdquo;
<em>The GPU Technology Conference 2013</em>, San Jose, CA,
March 18, 2013.
</li>
</ul>
<p>[<a href=".">&larr; Go back to profile</a>]</p>
</div>
</body>
</html>
Binary file added favicon.ico
Binary file not shown.
1 change: 1 addition & 0 deletions fonts-min.js

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

41 changes: 41 additions & 0 deletions fonts.js
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
/* Detect IE */
var tmp = document.documentMode, e, isIE;
// Try to force this property to be a string.
try{
document.documentMode = "";
} catch(e){ };
// If document.documentMode is a number, then it is a read-only property, and
// so we have IE 8+.
// Otherwise, if conditional compilation works, then we have IE < 11.
// Otherwise, we have a non-IE browser.
isIE = typeof document.documentMode == "number" ? !0 : eval("/*@cc_on!@*/!1");
// Switch back the value to be unobtrusive for non-IE browsers.
try{
document.documentMode = tmp;
} catch(e){ };

function addCSSRule(sheet, selector, rules, index) {
if(sheet.insertRule)
sheet.insertRule(selector + "{" + rules + "}", index);
else
sheet.addRule(selector, rules, index);
}


/* For IE, use Segoe UI; for others, use Open Sans. */
if (!isIE) {
WebFontConfig = {
google: {
families: [ 'Open+Sans:400,400italic,700:latin' ]
}
};
(function() {
var wf = document.createElement('script');
wf.src = ('https:' == document.location.protocol ? 'https' : 'http') +
'://ajax.googleapis.com/ajax/libs/webfont/1/webfont.js';
wf.type = 'text/javascript';
wf.async = 'true';
var s = document.getElementsByTagName('script')[0];
s.parentNode.insertBefore(wf, s);
})();
}
74 changes: 74 additions & 0 deletions hypergraph.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,74 @@
<!DOCTYPE html>
<html>
<head>
<meta name="viewport" content="width=device-width" />
<link rel="shortcut icon" href="favicon.ico" />
<script type="text/javascript" async src="fonts-min.js"></script>
<link rel="stylesheet" href="style.css" />
<meta charset="utf-8">
<meta name="description" content="One of the obstacles in accelerating sparse
graph applications using GPUs is load imbalance, which in certain cases causes
threads to stall. We investigate a specific application known as hypergraph
coarsening and explore a technique for addressing load imbalance." />
<title>
An accelerated procedure for hypergraph coarsening on the GPU
</title>
</head>
<body>
<div id="container" itemscope
itemtype="http://www.schema.org/ScholarlyArticle">
<p>[<a href=".">&larr; Go back to profile</a>]</p>
<h1 itemprop="name">
An accelerated procedure for hypergraph coarsening on the GPU
</h1>
<p>
with
<a href="http://internet2.trincoll.edu/facProfiles/Default.aspx?fid=1335119">
Lin Cheng</a> and
<a href="http://turing.cs.trincoll.edu/~pyoon/">Peter Yoon</a>
</p>
<p>
Paper presented at
<a href="http://www.ieee-hpec.org/2015/">IEEE HPEC</a> (2015)
</p>
<h2>Download</h2>
<ul>
<li>IEEE HPEC talk:
[<a href="talks/hypergraph_talk.pptx">PPTX</a>]
[<a href="talks/hypergraph_talk.pdf">PDF</a>]
</li>
<li>IEEE HPEC paper:
[<a href="preprints/hypergraph_hpec.pdf">Paper</a>]</li>
</ul>
<h2>Synopsis</h2>
<p itemprop="description">
One of the obstacles in accelerating sparse graph
applications using GPUs is load imbalance, which in certain
cases causes threads to stall. We investigate a specific application
known as <strong>hypergraph coarsening</strong> and explore a technique for
addressing load imbalance. The hypergraph is a generalization
of the graph where one edge may connect more than two nodes.
Many problems of interest may be expressed in terms of optimal
partitioning of hypergraphs where the edge cut is minimized.
The most costly step in hypergraph partitioning is hypergraph
coarsening, the process of grouping nodes with similar connectivity
patterns into one node to yield a new hypergraph with
fewer nodes. Hypergraph coarsening proves to be computationally
challenging on GPUs because many hypergraphs exhibit an
irregular distribution of connections. To address the resulting
load imbalance, we explore a novel task allocation scheme to
distribute work more evenly among GPU threads.
</p>
<h2>Publication Details</h2>
<ul>
<li>
Conference Paper:<br>
Lin Cheng, Hyunsu Cho, and Peter Yoon. &ldquo;An Accelerated Procedure
for Hypergraph Coarsening on the GPU,&rdquo; <em>IEEE High Performance
Extreme Computing Conference</em>, Waltham, MA, September 16, 2015.
</li>
</ul>
<p>[<a href=".">&larr; Go back to profile</a>]</p>
</div>
</body>
</html>
Loading

0 comments on commit dce8a1f

Please sign in to comment.