[RFC, Instruction Scheduler, Stage1] New hook/code to perform fusion of dependent instructions

[Pat Haugen]

On 4/8/20 2:46 AM, Richard Biener wrote:
>> I have coded up a proof of concept that implements our needs via a new
>> target hook. The hook is passed a pair of dependent insns and returns if
>> they are a fusion candidate. It is called while removing the forward
>> dependencies of the just scheduled insn. If a dependent insn becomes
>> available to schedule and it's a fusion candidate with the just
>> scheduled insn, then the new code moves it to the ready list (if
>> necessary) and marks it as SCHED_GROUP (piggy-backing on the existing
>> code used by TARGET_SCHED_MACRO_FUSION) to make sure the fusion
>> candidate will be scheduled next. Following is the scheduling part of
>> the diff. Does this sound like a feasible approach? I welcome any
>> comments/discussion.
> It would be nice test if the i386 use of TARGET_SCHED_MACRO_FUSION hooks
> for compare + branch fusion can be transitioned to that scheme.  I think your
> proposal doesn't improve things there since we can never schedule
> branches earlier
> but we could schedule the compare later.  But your proposal only
> affects scheduling
> of later insns, not the order of insns getting onto the ready list?

Yes, your understanding is correct, my approach only affects later
dependent insns. The first insn of the fusion pair will be scheduled
according to existing priority based scheduling.

-Pat