Re: Pass 4 (non-)optimization speedup

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -

On Tue, Jul 14, 2009 at 07:16:54PM -0700, Josh Stone wrote:
> [...]
> Another approach mentioned on IRC is trying to simplify the code that we
> present to GCC.

Right, and to check again whether we may have some larger
duplicate-elimination opportunities.  We have gone to some
effort to reuse translated functions and probe bodies, but
maybe it's not working well enough.

> [...]  So, is that 2-3% overhead worth 1/2 second compile time?

Not IMO.  Run-time performance is more important.

> Can anyone see ways to improve this patch, or suggest other areas
> that could benefit from focus?

The runtime?
Profile gcc slightly?
(--vp 0009 indicates considerable lex/parse time for some tests.)
Reduce number of #include'd kernel files if possible?
Improving the translate.cxx parts to copy-propagate better to
eliminate more temporary variables?

- FChE