-
Notifications
You must be signed in to change notification settings - Fork 2
Notes 2020 12 17
Marc-Andre Hermanns edited this page Jan 14, 2021
·
2 revisions
- Wesley Bland
- Chris Chambreau
- Michael Knobloch
- Joachim Protze
- Marc-Andre Hermanns
- Martin Schulz
-
QMPI_Pcontrol
- Argonne customer wants an interface to communicate different phases of an applications
- Caliper (LLNL Tool) should be the solution to that
- Argonne customer wants an interface to communicate different phases of an applications
-
Pcontrol semantics are terrible to begin with
-
var_list
arguments are impossible to intercept -
MPI_Pvcontrol(const int level, va_list ap)
- Can easily be intercepted and passed to the next level of QMPI
int MPI_Pcontrol(const int level, ...){ va_list args; va_start(args, level); int r = PMPI_Pvcontrol(level, args); va_end(args); return r; }
- Can be different for every compiler
- Fortran does not have the
...
- Maybe a key-value store (
MPI_Info
?, binary values?) could solve this
-
-
Non-blocking version?
- How to make sure completion can be handled by the library?
- Something along the lines of the
MPI_Pready
call to indicate that a tool that interceptsQcontrol
is done manipulating the info object
- Something along the lines of the
- How to make sure completion can be handled by the library?
-
Should a tool be required to pass on to next layer?
- Is this info for me?
- Exclusively?
- Is this info for me?