I have a new set of files for the BoW program.

I have changed the following:

compile.sh              # It now compiles all of the programs including the new ones
cherry.f90              # Fixed the memory leak
shade_module.f90        # Fixed the memory leak

I have also changed the files to Unix format.  If the content is the same as last time, then the date and time is the same.  However it will have linux end of line, rather than dos end of line.  This does not make any difference to the g95 compiler on cygwin.

I have added the following:

generate.f90            # This is a program to build a set of load cases - usually 1527 of them
home.f90                # This is a program to write a home escape sequence for a ANSI terminal
bulk.sh                 # This is a procedure to run the 1527 cases
display.sh              # This is procedure to build poor man's graphics

What you do is the following:

./compile.sh            # Seconds
./home.exe > home       # No time it just writes an escape sequence to the file home
./proc.sh small         # Seconds
./proc.sh normal        # Seconds
./generate.exe          # Seconds
./bulk.sh               # Hours - this runs 1560 cases - this could be done on a grid
./display.sh xxx        # Minutes - this builds the graphics file from all of the results
./cat xxx               # Seconds - this does a poor man's display

Why is this interesting?

If the bulk.sh script caused the jobs to run on a grid, we do not care if all of the jobs finish, or if they complete more than once.  All that matters is that their output file name has the meta data about the case they are dealing with.  As we are only building an animated graphic, we can loose the odd case here and there, and no one will notice.

I think that this is small enough to be simple, and big enough to show that a grid can change a computational task like this from being too long to contemplate, to being something which we may do routinely quite soon!

If you want it to run much more quickly, then do not use the nrmal case in line 23.  This way there are only 196 cases and they are all much smaller.

Comments, suggestions etc welcome.

Alistair

19 October 2004


