Question about the difference between two instruction scheduling passes

Question about the difference between two instruction scheduling passes

[Amker.Cheng]

Hi all:
   I'm currently studying implementation of instruction sched in gcc.

it is possible to schedule insns directly from queue in case
there is nothing better to do and there are still vacant dispatch slots
in the current cycle.

Gcc only does this work in the second pass, but what's the point?
Is it wrong or just not necessary  in the first sched pass?

Thanks.
-- 
Best Regards.

Re: Question about the difference between two instruction scheduling passes

[Ian Lance Taylor]

"Amker.Cheng" <amker.cheng@gmail.com> writes:

>    I'm currently studying implementation of instruction sched in gcc.
>
> it is possible to schedule insns directly from queue in case
> there is nothing better to do and there are still vacant dispatch slots
> in the current cycle.
>
> Gcc only does this work in the second pass, but what's the point?
> Is it wrong or just not necessary  in the first sched pass?

The first scheduling pass runs before register allocation and reload so
it's going to get generally trashed anyhow.  There no point to being
very precise in the first pass, it's just a rough approximation.

Ian

Re: Question about the difference between two instruction scheduling passes

[Alex Turjan]

> Gcc only does this work in the second pass, but what's the
> point? Is it wrong or just not necessary  in the first sched
> pass?

Regardless of the target architecture from the correctness point of view sched1 can be disabled. sched1 has as purpose shortening live ranges.
Short live ranges allow the register allocation to:
1. generate less spills and 
2. also to avoid useless live ranges splitting. 

The second point impacts the register renaming phase which allows sched2 phase to extract more ilp.