Sanchayan Maity
2fcc51c2c1
While at it also add the libpthread static library amd m5op_x86 for matrix multiplication test code as well. Note that the splash2 benchmark code does not comply with gem5 coding guidelines. Academic guys never seem to follow 80 columns and no whitespace guideline :(.
51 lines
2.4 KiB
Text
51 lines
2.4 KiB
Text
GENERAL INFORMATION:
|
|
|
|
This code is an improvement over the original Water code in SPLASH,
|
|
but is mostly the same. The best source of descriptive information,
|
|
therefore is the original SPLASH report. The main change is that
|
|
the locking strategy around the updates to the water accelerations
|
|
(in interf.C) is improved: a process updates a local copy of the
|
|
relevant particle accelerations, and then accumulates into the shared
|
|
copy once at the end.
|
|
|
|
RUNNING THE PROGRAM:
|
|
|
|
To see how to run the program, please see the comment at the top
|
|
of the water.C file or run it as "WATER-NSQUARED -h". The input file
|
|
has 10 parameters, of which the ones you would normally change
|
|
are the number of molecules and the number of processors. The other
|
|
parameters should be left at their values in the supplied input file
|
|
in the normal case. Please do not set the CUTOFF value (the last
|
|
parameter) to a nonzero number in the normal case.
|
|
|
|
The only compile-time option (ifdef) is one that says to change the
|
|
input distribution. The default input distribution of molecules
|
|
arranges them on a cubical lattice. For this, the number of molecules
|
|
must be an integer cube (8, 27, 64, 343, 512 ...). If one wants
|
|
to use a non-cube number of molecules, one can ignore the lattice
|
|
and use a random distribution of particles in a cubical space
|
|
by invoking the -DRANDOM compile-time option (see file initia.C).
|
|
Note that a random distribution does not make too much physical sense,
|
|
since it does not preserve chemical intermolecular distance ranges.
|
|
If you do not use the lattice but use -DRANDOM, please say so explicitly
|
|
in any results you report.
|
|
|
|
The program reads random numbers, to compute initial velocities, from
|
|
a file called random.in in the current working directory. It does
|
|
this rather than generate random numbers to facilitate repeatability
|
|
and comparability of experiments. The supplied file random.in
|
|
has enough numbers for about 512 molecules. If you need more, add
|
|
more random numbers between -4.0 and +4.0 to the file.
|
|
|
|
BASE PROBLEM SIZE:
|
|
|
|
The base problem size for an upto-64 processor machine is 512 molecules.
|
|
For this number of molecules, you can use the input file provided (and
|
|
change only the number of processors).
|
|
|
|
DATA DISTRIBUTION:
|
|
|
|
Our "POSSIBLE ENHANCEMENT" comments in the source code tell where one
|
|
might want to distribute data and how. Data distribution, however,
|
|
does not make much difference to performance on the Stanford DASH
|
|
multiprocessor.
|