Three problems pop out immediately:
a) the test.py file you modified was never intended to be run by
itself it is run my tests/run.py. To run it we would run:
./build/ALPHA_SE/m5.debug tests/run.py long/00.gzip/alpha/tru64/
simple-atomic
Note that test.py isn't actually in there, and the run.py script
figures out where test.py is based on the parameter to it.
b) The SPEC binary you have is a tru64 binary not a linux binary
(coff isn't used in linux), making the operating system parameter be
linux is incorrect, you'll get the wrong system calls.
c) cpu2000.py assumes a slightly different layout for the input
files. It assumes that all the directory structure looks like:
.../data/<benchmark (e.g. gzip)>/<input/output set (e.g.
ref,smred,etc)>/input/input.source
.../data/<benchmark (e.g. gzip)>/<input/output set (e.g.
ref,smred,etc)>/output/input.source.out
Hope that helps,
Ali
On Aug 7, 2007, at 10:30 AM, Dean Michael Ancajas wrote:
Hello everybody,
I'm still out of luck in running SPEC 2000 under M52.0b3(FS or
SE), I've already read [1] and I am a bit confused as to which
input to use. In my SPEC2000 CD i have ( from spec.org), the
structure in one of the directories(which i believe that has the
input set) goes like this
benchspec/
-CFP2000
-CINT2000
-164.gzip 181.mcf 252.eon 255.vortex CINT2000.bset
175.vpr 186.crafty 253.perlbmk 256.bzip2
176.gcc 197.parser 254.gap 300.twolf
-Makefile.default
-version
Let's say i want to test gzip.
Contents of 164.gzip directory:
data/ docs/ exe/ result/ run/ spec/ src/ version
Contents of data:
ref/ test/ train/
contents of ref:
input/ output/ reftime
Contents of input:
control input.graphic input.log input.program input.random
input.source
The important files that I think are:
1. input.program which is a "input.program: executable (RISC System/
6000 V3.1) or obj module not stripped" from the 'file' command.
2. input.random which is "input.random: data" from the 'file' comand
3. input.source which is a "input.source: POSIX tar archive (GNU)"
from the 'file' command
I assume all these files are inputs so I put them in a directory
using the structure stated in [1] but I get not-so-important-
messages when i invoke the following:
%./build/ALPHA_SE/m5.debug tests/long/00.gzip/test.py
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "build/ALPHA_SE/python/m5/main.py", line 326, in main
execfile(sys.argv[0], scope)
File "tests/long/00.gzip/test.py", line 1, in <module>
m5.AddToPath('../configs/common')
NameError: name 'm5' is not defined
I already made the necessary changes in spec_dist variable(in
cpu2000.py) and in the test.py file I changed "gzip_log('isa',
'opsys', 'smred')" to "gzip_log('alpha', 'linux', 'ref')" . which
means that I assume that the name of the INPUTSET is 'ref'. What am
i doing wrong?
Btw, I got the Alpha Binaries of SPEC 2000 from the SimpleScalar
website(and a little bit of searching around the bushes).
addtl info on ALPHA executable:
% file gzip00.peak.ev6
gzip00.peak.ev6: COFF format alpha executable paged not stripped -
version 3.11-10
I already searched the archive in the mailing list and found
nothing similar. Any inputs would be greatly appreciated...
[1] http://m5sim.org/wiki/index.php/SPEC2000_benchmarks
-Dean Michael Ancajas
--
Dean Michael B. Ancajas
_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
