head 1.3;
access;
symbols;
locks; strict;
comment @# @;
1.3
date 2000.05.18.22.17.22; author dbc; state Exp;
branches;
next 1.2;
1.2
date 2000.05.18.19.33.37; author dbc; state Exp;
branches;
next 1.1;
1.1
date 2000.01.02.21.48.04; author dbc; state Exp;
branches;
next ;
desc
@@
1.3
log
@Modified Files:
Makefile addFrame beginnings.html.in beijing.html.in
index.html.in mpiJava.html.in
@
text
@Lectures at Peking University, January 2000
Lectures on Data Parallel, SPMD programming
Peking University, Beijing, China, January 2000.
-
Lecture 1. Development of data-parallel programming.
This lecture reviews the historical development of programming languages
for data-parallel programming, and introduces High Performance
Fortran.
Slides:
Notes:
-
Lecture 2. Issues in translation of HPF.
Some issues in the translation of an HPF program to a SPMD program
are reviewed. The need for a run-time array descriptor is discussed.
An abstract model of a suitable Distributed Array Descriptor is developed.
Illustrative translations of HPF fragments to Fortran and C++ are
presented.
Slides:
Notes:
-
Lecture 3. Communication in data-parallel languages.
We start by discussing the patterns of communication needed to implement
various constructs in High Performance Fortran. Some communication
libraries that have been developed to support these kinds of
communications are reviewed.
Slides:
Notes:
-
Lecture 4. An HPspmd programming model.
We motivate a parallel programming model that combines data parallel
features from HPF with an explicitly SPMD programming style, then review
in detail a specific HPspmd language called HPJava.
Slides:
Notes:
-
Lecture 5. Java for High-performance computing.
Survey of related work, mpiJava, MPJ.
Slides:
@
1.2
log
@Modified Files:
addFrame beijing.html.in index.html.in mpiJava.html.in
@
text
@d19 1
a19 1
Powerpoint |
d21 1
a21 1
HTML |
d28 1
a28 1
Postscript |
d30 1
a30 1
PDF |
d32 1
a32 1
HTML |
d49 1
a49 1
Powerpoint |
d51 1
a51 1
HTML |
d58 1
a58 1
Postscript |
d60 1
a60 1
PDF |
d62 1
a62 1
HTML |
d78 1
a78 1
Powerpoint |
d80 1
a80 1
HTML |
d87 1
a87 1
Postscript |
d89 1
a89 1
PDF |
d91 1
a91 1
HTML |
d107 1
a107 1
Powerpoint |
d109 1
a109 1
HTML |
d116 1
a116 1
Postscript |
d118 1
a118 1
PDF |
d120 1
a120 1
HTML |
d133 1
a133 1
Powerpoint |
d135 1
a135 1
HTML |
@
1.1
log
@Modified Files:
Makefile
Added Files:
beijing.html.in
@
text
@d16 9
d36 10
d49 3
a51 1
Powerpoint |
a54 10
Lecture 2. Issues in translation of HPF.
This lecture reviews some issues in the translation of an HPF
program to a SPMD program. The need for a run-time array descriptor
is discussed. An abstract model of a suitable Distributed Array
Descriptor is developed. Illustrative translations of HPF fragments
to Fortran and C++ are presented.
a65 7
Slides:
d71 1
a71 1
various constructs in High Performance Fortran. Then several communication
d73 10
a82 3
communications will be reviewed. Finally, as a case study a, scheme
for implementing the array remapping operation used in general array
assignments is described in detail.
a95 10
d104 9
d124 6
d133 3
a135 1
Powerpoint |
a138 11
Lecture 5. Java for High-performance computing.
Java Grande activities, mpiJava.
@