int dst_traced = req->dst();
if (req->flags() & MPI_REQ_RECV) { // Is this request a wait for RECV?
if (src_traced == MPI_ANY_SOURCE)
- src_traced = (status != MPI_STATUSES_IGNORE) ? req->comm()->group()->rank(status->MPI_SOURCE) : req->src();
+ src_traced = (status != MPI_STATUS_IGNORE) ? req->comm()->group()->rank(status->MPI_SOURCE) : req->src();
TRACE_smpi_recv(src_traced, dst_traced, req->tag());
}
}
} else if (*request == MPI_REQUEST_NULL) {
retval = MPI_SUCCESS;
} else {
+ //for tracing, save the handle which might get overriden before we can use the helper on it
+ MPI_Request savedreq=*request;
+ if (savedreq!=MPI_REQUEST_NULL && not(savedreq->flags() & MPI_REQ_FINISHED))
+ savedreq->ref();//don't erase te handle in Request::wait, we'll need it later
int my_proc_id = (*request)->comm() != MPI_COMM_NULL
? simgrid::s4u::this_actor::get_pid()
: -1; // TODO: cheinrich: Check if this correct or if it should be MPI_UNDEFINED
//the src may not have been known at the beginning of the recv (MPI_ANY_SOURCE)
TRACE_smpi_comm_out(my_proc_id);
- trace_smpi_recv_helper(request, status);
+ trace_smpi_recv_helper(&savedreq, status);
+ if (savedreq!=MPI_REQUEST_NULL)
+ simgrid::smpi::Request::unref(&savedreq);
}
smpi_bench_begin();
return MPI_SUCCESS;
smpi_bench_end();
+ //for tracing, save the handles which might get overriden before we can use the helper on it
+ MPI_Request* savedreqs= static_cast<MPI_Request*>(xbt_malloc(count * sizeof(MPI_Request)));
+ memcpy(savedreqs, requests, count * sizeof(MPI_Request));
+ for(int i=0; i<count; i++)
+ if (savedreqs[i]!=MPI_REQUEST_NULL && not(savedreqs[i]->flags() & MPI_REQ_FINISHED))
+ savedreqs[i]->ref();
int rank_traced = simgrid::s4u::this_actor::get_pid(); // FIXME: In PMPI_Wait, we check if the comm is null?
TRACE_smpi_comm_in(rank_traced, __func__, new simgrid::instr::CpuTIData("waitAny", static_cast<double>(count)));
*index = simgrid::smpi::Request::waitany(count, requests, status);
if(*index!=MPI_UNDEFINED){
- trace_smpi_recv_helper(&requests[*index], status);
+ trace_smpi_recv_helper(&savedreqs[*index], status);
TRACE_smpi_comm_out(rank_traced);
}
+
+ for(int i=0; i<count; i++)
+ if (savedreqs[i]!=MPI_REQUEST_NULL && not(savedreqs[i]->flags() & MPI_REQ_FINISHED))
+ simgrid::smpi::Request::unref(&(savedreqs[i]));
+ xbt_free(savedreqs);
smpi_bench_begin();
return MPI_SUCCESS;
{
smpi_bench_end();
+ //for tracing, save the handles which might get overriden before we can use the helper on it
+ MPI_Request* savedreqs= static_cast<MPI_Request*>(xbt_malloc(count * sizeof(MPI_Request)));
+ memcpy(savedreqs, requests, count * sizeof(MPI_Request));
+ for(int i=0; i<count; i++)
+ if (savedreqs[i]!=MPI_REQUEST_NULL && not(savedreqs[i]->flags() & MPI_REQ_FINISHED))
+ savedreqs[i]->ref();
+
int rank_traced = simgrid::s4u::this_actor::get_pid(); // FIXME: In PMPI_Wait, we check if the comm is null?
TRACE_smpi_comm_in(rank_traced, __func__, new simgrid::instr::CpuTIData("waitAll", static_cast<double>(count)));
int retval = simgrid::smpi::Request::waitall(count, requests, status);
for (int i = 0; i < count; i++) {
- trace_smpi_recv_helper(&requests[i], &status[i]);
+ trace_smpi_recv_helper(&savedreqs[i], status!=MPI_STATUSES_IGNORE ? &status[i]: MPI_STATUS_IGNORE);
}
TRACE_smpi_comm_out(rank_traced);
+ for(int i=0; i<count; i++)
+ if (savedreqs[i]!=MPI_REQUEST_NULL && not(savedreqs[i]->flags() & MPI_REQ_FINISHED))
+ simgrid::smpi::Request::unref(&(savedreqs[i]));
+ xbt_free(savedreqs);
+
smpi_bench_begin();
return retval;
}