[hpct] HPCT: Do we need a call?
Vitali A. Morozov
morozov at anl.gov
Wed Nov 5 11:09:52 CST 2008
Dear David and Kumar;
We were scheduled to have a call tomorrow, Nov 6, 10:00 am CST.
In fact, we would like to have an update on the issues, which we have
left. Luckily, I-Hsin's visit has cleared most of them, and therefore,
this list is very short. So, if you, David and Kumar, think these issues
can be solved by routine fashion, we could save time and cancel the call.
The list of issues:
1) MPI_trace. We have a problem with instrumentation Fortran programs.
Debugging information is unavailable.
2) MPI_trace. We have a problem with instrumentation C++ programs,
compiled with any level of optimization. Debugging information is
unavailable.
"No optimization" flag makes the program instrumented correctly.
3) Xprofiler. We have a problem with Double FPU instructions being not
shown in disassembly window.
4) Xprofiler. We have a problem with assembly-written function being not
disassembled.
5) We want to know the parameters of "-pg" profiling, particularly, we
want to know how to modify the sampling interval.
The test suite is located on surveyor:~morozov/HPCT-TestSuite
Thanks,
Vitali
More information about the hpct
mailing list