per process output from --osh-time sorted function is mixed together on stdout

[cwsmith]

Running with more than one rank and --osh-time produces TOP FUNCTIONS output from each rank that gets mixed together on stdout making the results hard to read.

It appears that building with -DOmega_h_DBG=on may resolve this as the TASK_0_cout macro gets defined:

omega_h/src/Omega_h_dbg.hpp

Line 33 in 630e4ed

# define TASK_0_cout if(DBG_COMM && (0 == DBG_COMM->rank())) std::cout

But, for collecting performance data, specifically with a run using GPUs, enabling that build option may not work, and, if it does, it may increase runtime. This needs to be tested.

omega_h/src/Omega_h_profile.cpp

Lines 320 to 334 in 630e4ed

	for (auto i : sorted_result) {
	auto cflags( std::cout.flags() );
	double val = i.second[TOP_AVE];
	if (val*100.0/total_runtime_ave >= h.chop) {
	TASK_0_cout << std::right
	<< std::setw(width) << val*scale << percent
	<< std::setw(width) << i.second[TOP_MIN]*scale << percent
	<< std::setw(width) << i.second[TOP_MAX]*scale << percent
	<< std::setw(width) << i.second[TOP_MAX]/i.second[TOP_MIN] << " ";
	std::cout.flags(cflags);
	TASK_0_cout << i.first << std::endl;
	}
	std::cout.flags(coutflags);
	}
	}