PATCH RFC: Change some tree-ssa-operands fields to unsigned
Mark Mitchell
mark@codesourcery.com
Tue Jan 23 19:02:00 GMT 2007
Andrew MacLeod wrote:
> Probably should have been unsigned in the first place, but I never
> bothered going back and changing it because it never seemed to matter.
> I'm guilty of that from time to time :-)
I am fully aware of concerns about GCC's compile-time performance, but I
bet that the signed-overflow optimization has approximately zero impact
on GCC's performance, which is driven primarily by memory usage,
algorithmic choices, file I/O performance, etc. GCC isn't a scientific
program.
Therefore, my feeling is that we should use unsigned types wherever the
quantities really are never negative. That's good documentation. So,
I'm all for Ian's patch, and I'm all for defining variables as unsigned
in future, when the values assigned are always going to be non-negative.
--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713
More information about the Gcc-patches
mailing list