user-guided speculative precomputation? (my wacky ia64 idea)

[Jim Wilson]

On Wed, 2004-04-07 at 17:09, Duraid Madina wrote:
> On that note, does GCC *ever* emit ld8.a / chk.a pairs? This would be a 
> very nice addition for yucky pointer code, and isn't the can of worms 
> for the compiler that ld.s/chk.s is.

No.

Cygnus did try to implement some advanced/speculative load support about
4-5 years ago, but we couldn't get a reliable performance increase.  I
think part of the problem was that the alias info in the RTL wasn't
quite good enough, and as a result we ended up generating too many
advanced loads which hurt performance.  It would be worthwhile trying
again some day, as there have been infrastructure improvements since
then.  Or maybe we didn't find the right set of heuristics to use. 
Unfortunately, there isn't anyone doing serious IA-64 gcc work at this
time.
-- 
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com