[PATCH] TargetParser prototype - NFC (ARM only)
meadori at gmail.com
Fri May 1 12:20:10 PDT 2015
On Fri, May 01, 2015 at 05:21:32PM +0000, John Brawn wrote:
> A general question: how far are you planning to go with this,
> in particular are you going to sort out cpu <-> arch mapping
> (which is duplicated in several places)?
> I ask because I'm currently thinking about adjusting the ARM
> -mcpu and -march handling in clang to be more like AArch64,
> in particular making -march=garbage not mean -mcpu=arm7tdmi
> (and instead give an error), and it's be nice to not collide
> with what you're doing.
I am very interested in cleaning up -march as well. I have been
experimenting with some local patches to allow for proper -march
handling where -mcpu isn't magically applied when *only* -march is
specified. I am interested in allowing for architecture level code
generation, which currently is impossible because 1) we always compute
the current CPU (sometimes incorrectly) and 2) we don't currently
support generating '.arch' in the ARM backend.
It looks like Renato's work might make solving that problem a bit
easier. So, +1 from me.
More information about the cfe-commits