summaryrefslogtreecommitdiff
path: root/readme.txt
diff options
context:
space:
mode:
Diffstat (limited to 'readme.txt')
-rw-r--r--readme.txt19
1 files changed, 2 insertions, 17 deletions
diff --git a/readme.txt b/readme.txt
index fa275a1..89cf3e6 100644
--- a/readme.txt
+++ b/readme.txt
@@ -30,22 +30,7 @@ To open bench, just run bin/bench after building as above.
The "tabulator" program runs every possible pairing of warriors from a selection against each other with every possible separation
and turn order, and then shows the number of wins of each warrior against each other warrior in a table format. The column header
-is the winner, and the row header is the loser. This program uses MPI to run batches of these rounds in different processes, and
-communicate the results back to a head process.
+is the winner, and the row header is the loser. This program uses MPI to run this across multiple processes.
To run all of the included warriors against each other, run
- mpirun bin/tabulator-mpi warriors/*.red
-
-Note that tabulator expects at least two processes (one head process and at least one worker). If you only have one core, you may run
- mpirun -np 2 --oversubscribe bin/tabulator-mpi warriors/*.red
-
-=== evolver ===
-
-The experimental "evolver" program attempts to create warriors who are good against a particular other
-warrior via a process similar to genetic evolution. I may or may not attempt to improve it in the future.
-
-The evolver takes an arbitrary number of input warriors as "seeds". To evolve a warrior against Dwarf, using
-all of the included warriors as seeds, and storing the outputs from each round in a directory named output, run
- mpirun bin/evolver warriors/dwarf.red output warriors/*.red
-
-As with tabulator, evolver requires at least two processes, and you will have to oversubscribe if you only have one core.
+ mpirun bin/tabulator warriors/*.red