Splash Benchmark

Today, I still have trouble with the shared-counter benchmark from Kevin. The problem seems like to do with a function call of sched.h file -- sched_yield(); The library could not resolve the symbol for that file. It is not used so I removed that function call and re-compiled the library and then recompiled the benchmark. It now compiles but it would produce a core dump whenever it runs. I had an initial look at the core dump with the dbx ultility that Solaris provides. It doesn't make any sense to me at the moment.

Given the sluggish progress of porting Wisconsin version of shared-counter benchmake, I started to compile the Splash2 suite with gcc compiler and pthread. With this set of arrangement, I actually can compile the Splash2 suite with minor modification to the make file of the programs and turned off sched_yield() method

So far at least the Splash2+gcc+pthread combination is working with sched_yield() commneted out, but we are still need to make sure sched_yield() can be fixed.
Big plan for tomorrow. 1) contact Kevin about the shared-counter benchmark 2) get ruby magic calls into the Splash2+gcc+pthread 3) if 2) succeed, trying to run it in GMES 4) fix the sched_yield() part if possible 5) contact wisconsin GEMS team about the broken Opal simulator 6) ask lab support people to fix my desktop
Task 1) 2) 4) 5) 6)are done as 4/5/2006
Task 3) simulation runs but really slow, libm.so.2 copied from /lib directory from Solaris machine