On Friday, 23 December 2016 at 18:33:52 UTC, Seb wrote:
So the asm is about 40% faster than the best of the
alternatives. The point being that it's the generated code not
the source that's the problem.
Has anyone considered using ldc for DMD release compilations?
We considered using gdc (or ldc) to compensate for the slowdown
after the ddmd transition, but nobody complained enough for that
to happen ;). Also Windows users were stuck with dmd at that
point.
Using a proper optimizer, profiles, and LTO it was fairly simple
to speedup dmd by some 20+% (IIRC).
Unfortunately wasn't considered before 2.069.0, but that's one of
the main reasons why we setup Travis-CI to test gdc/ldc builds.
https://trello.com/c/OT6jlFNa/85-rebench-ddmd-vs-dmd-compiler-speed
