[parsec-users] Fluidanimate implementation

Yungang Bao ybao at CS.Princeton.EDU
Tue May 24 22:59:15 EDT 2011


Hi Gagan,

Here is Christian Bienia's message which I forward to you. 

"The current version of fluidanimate that is publicly available has a limitation that does not allow the program to compute more than one frame, so it's simply computing the first one on and on and on. That's how Intel wrote it. I've improved the program so that it doesn't have this limitation anymore, we'll release it with PARSEC 3.0."

Yungang

----- Original Message -----
From: gagang at cs.wisc.edu
To: parsec-users at lists.cs.princeton.edu
Sent: Tuesday, May 24, 2011 10:20:40 PM
Subject: [parsec-users] Fluidanimate implementation

Hello,

I was studying the serial and pthread versions of the code for
fluidanimate and am a bit puzzled by the implementations. Here is what I
see:

1. cells2 is initialized with data read from the file.

2. cells is populated from cells2 after building the grid.
3. cells is updated after computing the forces, detecting collisions and
advancing particles.
4. Frame is advanced and steps 2 and 3 are repeated.
     Step 2, in my opinion, takes the old data in cells2 and repopulates
cells.
     cells2 does not seem to  be updated after step 3. So in some sense, the
     same computation is being repeated over and over again.

I would appreciate it very much if someone could shed some light and
possibly point out what I am missing.
Many thanks.

Gagan

_______________________________________________
parsec-users mailing list
parsec-users at lists.cs.princeton.edu
https://lists.cs.princeton.edu/mailman/listinfo/parsec-users


More information about the parsec-users mailing list