On 8/2/11 7:34 PM, Robert Clipsham wrote:
On 02/08/2011 00:40, Walter Bright wrote:
Anyone care to examine the assembler output and figure out why?
I was talking to David Nadlinger the other day, and there was some sort
of codegen bug causing things to massively outperform dmd and clang with
equivalent code - it's possible this is the cause, I don't know without
looking though. He may be able to shed some light on it.
Nope, this turned out to be a bug in my program, where some memory chunk
used as test input data was prematurely garbage collected (that only
surfaced with aggressive compiler optimizations, which is why I
suspected a compiler bug).
David
