On 18/08/15 12:45, Walter Bright via Digitalmars-d wrote:
Martin ran some benchmarks recently that showed that ddmd compiled with dmd was
about 30% slower than when compiled with gdc/ldc. This seems to be fairly
typical.
I'm interested in ways to reduce that gap.
All things considered, looking at the issues involved in the transition, and the
benefits of being able to refactor in D, I can't see that it makes sense to
worry about such things at this stage. Surely the easiest way to handle the
speed gap, in the short term, is to make it easy to use gdc/ldc as the D
compiler to build ddmd, and to do exactly that when building dmd binaries for
distribution.
Admittedly my answer may be Linux-centric (I guess gdc/ldc builds of ddmd are
not feasible for Windows), but to be honest, I don't think a 30% slowdown is
_that_ terrible a thing to have to deal with, in the short term, to manage such
an important transition. Stability at this stage seems _much_ more important.
