[PATCH] Constant Hoisting Pass

[Juergen Ributzka]

Yeah, in this case I would prefer a simpler approach that just uses the debug
location from one of the original instructions where that constant came from.

-Juergen

On Jan 23, 2014, at 3:43 PM, Eric Christopher <echristo at gmail.com> wrote:

> *nod* It's definitely problematic, but at least then it will be
> associated with a line that could have produced the code rather than
> some other random line that doesn't materialize a constant... if I
> could give you an ideal world we might ensure that it's the location
> of the first use that post-dominates the new constant materialization
> block/instruction :) That's probably too much work though.
> 
> -eric
> 
> On Thu, Jan 23, 2014 at 3:38 PM, Juergen Ributzka <juergen at apple.com> wrote:
>> Sure, I can modify the patch to do that instead. I am just wondering how
>> that could affect the debugger,
>> because after hoisting the constant might be in a different basic block than
>> the first use?
>> 
>> -Juergen
>> 
>> On Jan 23, 2014, at 3:31 PM, Eric Christopher <echristo at gmail.com> wrote:
>> 
>> For the hoisted constants I'd prefer you pick the instruction of the
>> earliest such use of the constant in the function if possible. That
>> will a) keep it closer, and b) hopefully match up with some of the
>> uses.
>> 
>> Thoughts?
>> 
>>