I have seen it used several times in benchmarks comparing it to 'gdc -O3' and 'dmd -O -inline' now, so people apparently expect it to yield the highest available optimization level.