On 16.01.2017 20:28, Gregory P. Smith wrote: > > Is there a good reason not to detect single expression multiply adds > and just emit a new FMA bytecode? > Same question here.