Is there any support planned for autovectorization of SIMD instructions, as in
Xcode 2 (GCC 4.0)?
What about autoparallelization of MIMD? Both of these could be done by directly
compiling source
code into optimized assembly code or by creating, optionally, an intermediate D
source code file with
the correct instructions added automatically, so that you can inspect the
modified file and further
optimize it before compilation.
What do you think?

Noam Lenz