- Getting the parallel version to work was a pain. Two major gotchas:
	(1) -ltbb had to be added *at the end* of the command line
	(2) I was sometimes trying clang, but not clang++
  of course everything was terrible to decipher with C++ compile errors
  being the mess they are.
- For measuring time I could also use
	clock_t begin = clock();
	...
	clock_t end = clock();
	double time = (dobule)(end - begin) / CLOCKS_PER_SEC;
  but it does not work with the parallel version (it meausres CPU clocks).