This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: basic_block index negative?


Am Mit, 2001-10-17 um 19.46 schrieb 1003340807:

> Dunno.  I guess because they're special and contain no program
> instructions.  The tree SSA code uses it to tell whether it found
> a ghost definition (artificial definitions created by SSA).

I find it quite odd and distracting to have negative indices. This
also forbids using them for arrays.

> If the index is unsigned, then a reverse traversal could get into
> an infinite loop.  Consider:
 
> for (i = n_basic_blocks - 1; i >= 0; i--)

Ah yes, of course, I just had trouble following your terminology. :)

I think it would be really nice to mostly unsigned variables; this
makes it easy to prevent nasty sideeffects when dealing with them
and quite frankly I don't like the just of casts in the code. Maybe
gcc could also create better code but this is just a speculation
for recent versions, it worked on older versions though.

-- 
Servus,
       Daniel


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]