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]
Other format: [Raw text]

Re: C constant expressions proposals


Mark Mitchell wrote:

I disagree, but we've had that debate before, so it seems boring to have it again. The bottom line is that optimizer people always think that the middle-end should do this kind of thing; people who do other things don't necessarily agree.

I am confused. Full detection of undefined variables requires full flow analysis and solution of data flow equations. Are you really saying the front end should do this.

Indeed the front end can do some of this (GNAT does a quite reasonable
job, based on a linear flow approximation), but the back end optimizer
can still discover cases that GNAT misses.

Another example are incorrect asm statements: the front end shouldn't be
in the business of validating them.


Perhaps

I am really quite amazed by the "perhaps" here, since it seems obvious to me that it is unreasonable to expect to have a full parser for all possible assemblers in the back end (let alone in the front end).



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