I got the shared-counter benchmark working. The problem was that the source code was wrong. There is a few lines code missing for some reason that I am very intrigued about. But anyway, I got it compiled and running on GEMS simulator.
I need to verify the results of barnes run and write scripts that could run the simulation.
Ok. The following is the Benchmark that is used in LogTM. I need to be able to replicate that.
The first 4 of the benchmarks are the ones shown in TTM
Barnes 512 bodies locks on tree nodes
Ocean contiguous partitions, 258barriers
Raytrace-Basesmall image (teapot) work list & counterlocks
Water NSquared 512 molecules barriers
These three are later added to LogTM
Cholesky 14 task queue locks
Radiosity room task queue & bufferlocks
Raytrace-Opt smallimage (teapot) work list & counter locks
So, I will get the first 4 working and worry about the last three later.
Raytrace is done. I had to modify the rt.H file, placed all _iobuf with FILE, and
changed the followings lines
extern void InitRayTreeStack(long TreeDepth, INT pid);
extern void PushRayTreeStack(struct ray *rmsg, INT pid);
extern long PopRayTreeStack(struct ray *rmsg, INT pid);
All of them are done exception the optimized version of Raytrace.
Wednesday, April 05, 2006
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment