This is the mail archive of the gcc-patches@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: [24/27] Put gcse's can_copy hash into target structures



On 11 Jul 2010, at 21:03, Andrew Pinski wrote:

On Jul 11, 2010, at 12:54 PM, Mike Stump <mikestump@comcast.net> wrote:


On Jul 10, 2010, at 6:59 AM, Richard Sandiford wrote:
For the most part I think *most* of the "char" variables that
you're moving should really be "bool".  However, I think that
should be done in separate changes.

Yeah, agreed that bool would be nicer. I vaguely recall a discussion
about whether that was a good idea though, given that hosts like Darwin
define bool to be wider than char. More fool them, I suppose.

:-) Hey, for -fwhole-program, is there a narrowing pass that can change 4 byte variables into 1 byte variables? I'd not object to using the arguably better bool for these things. bool is only 4 bytes on ppc, not on any of the x86 targets.


Actually bool is only 4 bytes on ppc Darwin. No other host or target. From what I remember this was done not to break the abi. I would say let's not worry about ppc Darwin as a host any more as there are no new versions of the os being released.

Fair enough that there should be little if any priority for future development.


Conversely FSF gcc is the *only* compiler with any modern development for the remaining useful ppc systems (one of which I'm typing this on).
Despite the ideals of Hacking Heaven .. most of us cannot afford to toss away good quality working hardware that is < 5 years old :-)
Please don't break it unless it's actually necessary.


(and yes, I do have other systems, including an 8-core intel system, but this quad g5 _is_ still useful).
Iain



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