This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Loop unrolling
- To: mrs at wrs dot com (Mike Stump)
- Subject: Re: Loop unrolling
- From: Horst von Brand <vonbrand at sleipnir dot valparaiso dot cl>
- Date: Sat, 13 Jun 1998 08:32:30 -0500
- cc: amylaar at cygnus dot co dot uk, law at cygnus dot com, branko dot cibej at hermes dot si, egcs at cygnus dot com, pfeifer at dbai dot tuwien dot ac dot at
mrs@wrs.com (Mike Stump) said:
[...]
> Well, this wasn't the result of the consersation when last we agreed
> on what to do and what the doc meant and what the compiler should do.
> The decision whether or not a loop could be removed was to be decided
> by the source input tokens to the lexer. Not by the existing rtl
> codes.
That is a horribly broken idea. That way you have to get non-standard
lexical information down to the loop-optimizer. The data paths for that
surely aren't in place, and would just add gratuitous complexity. All just
for supporting a broken feature.
There are non-intrusive solutions, like the asm volatile ("" : : : "r(i)")
idea. If somebody wants to do something like this, they deserve to have to
write it this way, IMHO.
--
Horst von Brand vonbrand@sleipnir.valparaiso.cl
Casilla 9G, Viņa del Mar, Chile +56 32 672616