I looked back at the B3 tests I generated, and I realized that while I used distinct ops (add vs sub, and fmadd vs fmsub), I invoked the same paths by having cases of effective addition in both. There will be a PR in the future fixing this in the test generation, but (if this is possible) it would be nice to have something in coverage that looks to make sure all effective operations are being used instead of all opcodes being used.
I looked back at the B3 tests I generated, and I realized that while I used distinct ops (add vs sub, and fmadd vs fmsub), I invoked the same paths by having cases of effective addition in both. There will be a PR in the future fixing this in the test generation, but (if this is possible) it would be nice to have something in coverage that looks to make sure all effective operations are being used instead of all opcodes being used.