Re: Improving reload inheritance code generation and predictability

[Jeff Law <law at redhat dot com>]

On 11/18/10 12:34, Bernd Schmidt wrote:
> On 11/18/2010 03:46 PM, Jeff Law wrote:
> > Basically we start searching forward in the stream from the current insn
> > needing a reload noting uses of spill regs as we go.
> Isn't this quadratic in the number of insns?
Potentially, yes.  If we want to go forward, then we probably would need 
a PARAM to limit the number of insns scanned.

> > The further away
> > from the insn needing reloads the spill reg is used the more desirable
> > that spill reg becomes.  If we can't find all the spill regs before
> > reaching the end of the block, the remaining spill regs are considered
> > the most desirable and we allocate these stragglers in the round-robin
> > fashion.
> Ideally we'd allocate the valuable ones only to reloads which we know
> we'll want to inherit in the future.
True.  But that means knowing the set of reloads for future insns.

jeff