Re: Need advice on bounds checking approaches

[Greg McGary <gkm at eng dot ascend dot com>]

Jeffrey A Law <law@cygnus.com> writes:

>   > Perhaps I could do that in a limited way, by adding a BP-specific
>   > optimization pass after loop optimization that eliminates redundant
>   > checks and expands "check_bounds" into primitive RTL for targets that
>   > don't HAVE_check_bounds.  Could that pass muster?
> 
> Possibly.  However, I'm not a big fan of having feature specific
> optimization passes, even if we've had some in the past (addressof & 
> constant_p).
> 
> I'd be much happier if we could find a clean way to get the optimizations
> you need using our existing optimizers.

We can do without a special pass.  CSE seems the best place to
eliminate redundant checks, so we'd just need a little bit of code to
handle the new check_bounds_rtx.  We can just forget about doing
default translation of check_bounds_rtx into primitive RTL, and
require that targets supporting bounds checking handle it in the
machine description.  So far, all of the targets I want to support
initially (i960, PowerPC, i386, MIPS) will benefit from special
handling in the MD file anyway.

Does that sound reasonable?

>   > It's a fine thought, but I can't do that today, can I?  I only see
>   > whole function mode being used for C++ inlines.
> 
> That's not my understanding -- you'd probably need to talk to Mark -- I
> was under the impression that we always did function at a time stuff
> for C++.

You're right.  I was careless in my initial reading of the condition
surrounding the setting of cfun->x_whole_function_mode_p.  Even so,
this doesn't help me for C or ObjC, unless those will be converted
soon.