On Tuesday, 2 October 2012 at 13:46:58 UTC, Joseph Rushton
Wakeling wrote:
The executables produced are nice and fast -- maybe about
10-15% slower than GDC for the number-crunching code I tested,
[…]
Is any of the code public, in the sense that you could give e.g.
me access for benchmarking purposes? We are currently using a
presumably suboptimal optimization pass schedule, which for
example doesn't include the auto vectorizer, but are going to
switch to a new built-in pass manager soon (which uses the same
pass schedule as Clang), and it would be very interesting to see
its impact on performance-critical code.
David
