[parsec-users] Fluidanimate double buffering bug

Elliott Slaughter slaughter at cs.stanford.edu
Thu Mar 15 21:53:46 EDT 2012

I believe I have found a bug in Parsec 2.1's fluidanimate benchmark.
Specifically, nowhere in fluidanimate are pointers swapped for double
buffering. This can be demonstrated by running fluidanimate for 1 step and
for 2 steps and showing that the result files are identical. I have created
a script which does exactly that (attached). The output is copied below.



We ran fluidanimate 3 times:
  in_5k.fluid --> out1.fluid --> out2.fluid  (one step each time)
  in_5k.fluid -----------------> out2b.fluid (two steps at once)

Comparing out2.fluid and out2b.fluid...
Note: These ought to match, but they don't.

Tolerance: 1e-05
0: Values differ (-0.055712 vs. -0.0564135).

Comparing out1.fluid and out2b.fluid...
Note: These shouldn't match, but they do.

Tolerance: 1e-05
Files match.

Elliott Slaughter

"Don't worry about what anybody else is going to do. The best way to
predict the future is to invent it." - Alan Kay
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20120315/21fe99c0/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: breakme.sh
Type: application/x-sh
Size: 1152 bytes
Desc: not available
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20120315/21fe99c0/attachment.sh>

More information about the parsec-users mailing list