Submitting your CBP submission
IMPORTANT: PLEASE READ THIS ENTIRE MESSAGE.
The web site for submissions can be accessed through the link on the CBP web
site (http://www.jilp.org/cbp/) or by
going directly to
http://systems.cs.colorado.edu/CBP/CRP/.
You need to register to the web site before submitting, so please do so as soon
as possible.
Your submission consists of two parts:
1) The abstract, due Friday Oct 15 9pm PST.
2) A tarball of the writeup, performance results, and code. This is due Friday
Oct 22 9pm PST.
The competition finalists' tarballs will be made publicly available through the
CBP website. In addition, we intend to automate the process of testing the
predictors. For these reasons, we request that the tarball be created in a
very specific way, as described below.
The tarball should be created through the following steps:
Preparing the directory
1) Make a copy (COPY) of your cbp-framework-version-3 directory; that is, the
one with your predictor in it. You will modify COPY in the remaining steps
(unless otherwise stated), and then submit a tarball of COPY. We ask you
to do this to preserve your original predictor.
2) Remove from COPY all files except those that are part of the framework and
the source files needed to build your predictor. For example, remove *.o's,
any dot files, predictor, .... Also remove the "traces" directory tree--we
don't need this. After this step, you should only have the original framework
files--minus the traces directory--and the source for your predictor.
3) We need to know whether you use make or scons to build your predictor.
If you build with make, keep Makefile and delete SConstruct. If you build
with scons, keep SConstruct and delete Makefile.
4) We need your predictor's performance results, and we need to know whether
your predictor uses the traces without data values and memory addresses or the
traces with them. Go to your original cbp-framework-version-3 directory.
If your predictor uses traces without data values and memory addresses, generate
the performance results with the following:
./gen_report.pl without-values > ./gen_report.output.without-values.txt
If your predictor uses traces with data values and memory addresses, use the
following:
./gen_report.pl with-values > ./gen_report.output.with-values.txt
Copy the resulting file (either ./gen_report.output.without-values.txt or ./gen_report.output.with-values.txt)
into the COPY directory.
5) Copy your 4-page PDF writeup into COPY. Call it "writeup.pdf".
Creating the tarball
6) Go to the directory that contains the COPY subdirectory; that is, the one
just above COPY.
7) Change the name of COPY to cbp-2004-paper-N, where N is your paper number.
For example:
mv my-cbp-framework-version-3
cbp-2004-paper-1
8) Create a tar file of cbp-2004-paper-N:
tar cvf cbp-2004-paper-1.tar
cbp-2004-paper-1
9) bzip2 (compress) the resulting tar file:
bzip2 cbp-2004-paper-1.tar
10) uuencode the resulting compressed tar file, turning it into an ASCII file:
uuencode cbp-2004-paper-1.tar.bz2
cbp-2004-paper-1.tar.bz2 > cbp-2004-paper-1.tar.bz2.uue
The resulting uuencoded bzipp'ed tar file (tarball), that is,
cbp-2004-paper-1.tar.bz2.uue, is what you need to upload to the server for your
submission. After uploading, you might want to download your submission,
unpackage it (uudecode, bunzip2, tar xvf), build it, and rerun it to verify that
it was uploaded correctly. Note: the server will strip the .tar.bz2.uue
extension and replace it with .txt, but the contents of the file will be
preserved.
-- Jared