Running all contractions for benchmarking is a waste of time. Just running some of the diagrams does not give the correct result due to re-use across diagrams. Reducing momenta also skews the picture. We can, however, just compute one time slice combination instead of all of them. The code will then just run with a single thread in the contraction phase and we should get a profile much quicker.
Running all contractions for benchmarking is a waste of time. Just running some of the diagrams does not give the correct result due to re-use across diagrams. Reducing momenta also skews the picture. We can, however, just compute one time slice combination instead of all of them. The code will then just run with a single thread in the contraction phase and we should get a profile much quicker.