[RFC] PR c/12245 (not really C frontend issue)

Mark Mitchell mark@codesourcery.com
Mon Feb 23 03:29:00 GMT 2009


Jan Hubicka wrote:

> Yep, the flat replrezentation of initializers is one part of problem,
> not needing a lot of temporary trees is other.
> Can we possibly break out the HOST_WIDE_INT bits for temporary counters?

I don't think there's an inherent reason we need to use INT_CST for
internal counters.  We do however need to make sure that we're matching
host vs. target overflow, etc., correctly; we can't use 32-bit counters
for a 64-bit target, or vice versa.  Unless we have a host integer type
that is documented as being as wide as target size_t for all possible
targets, we have to be careful; in general, we might have to invent some
other bignum type to use.

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713



More information about the Gcc-patches mailing list