[4/4][PATCH] Discussing PR83507

[Segher Boessenkool]

On Tue, Apr 16, 2019 at 03:08:23PM +0300, Roman Zhuykov wrote:
> This issue unfortunately was not solved correctly.  In that example we 
> don’t have -fmodulo-sched-allow-regmoves enabled and we should not 
> create any register moves at all.

Yes, but if we do for whatever reason, we should never create register
moves of hard registers.  Because that is incorrect (there may not be
insns to do it).  It's a separate issue.

You're extending Jakub's patch here, not replacing it, so that's fine.

> In pr84524.c we got a loop with an extended inline asm:
> asm volatile ("" : "+r" (v))
> which also gives us a “surprising” situation Alexander predicts.
> 
> For sched-deps scanner such volatile asm is a “true barrier” and we 
> create dependencies to almost all other instructions, while DF scanners 
> don’t give us such information.

There is no such thing as a "true barrier" in extended asm.  The only thing
volatile asm means is that the asm has a side effect unknown to the compiler;
this can *not* be a modification of a register or of memory contents, such
things are known by the compiler, that's what clobbers and "memory" clobber
are about.

> Maybe it is a good idea somehow re-implement modulo scheduler using only 
> one scanner instead of two, but at the moment the best thing to do is to 
> detect the situation earlier and skip such loops.

Or fix the broken code...


Segher