Jakub Jelinek writes:
> Hi!
>
> The following testcase is miscompiled by delete_trivially_dead_insns,
> latently since r0-6313, actually since r15-1575.
>
> The problem is in that r0-6313 change, which made count_reg_usage not
> count uses of the pseudo which the containing SET sets. That is need
Jakub Jelinek writes:
> On Fri, Apr 04, 2025 at 01:56:31PM +0100, Richard Sandiford wrote:
>> My suggestion isn't elegant, but: once we've decided not to delete
>> an instruction, we need to keep all definitions of all registers used
>> by the instruction (given the way that the algorithm works).
On Fri, Apr 04, 2025 at 01:56:31PM +0100, Richard Sandiford wrote:
> My suggestion isn't elegant, but: once we've decided not to delete
> an instruction, we need to keep all definitions of all registers used
> by the instruction (given the way that the algorithm works). So how
> about instead just
> The problem is in that r0-6313 change, which made count_reg_usage not
> count uses of the pseudo which the containing SET sets. That is needed
> so we can delete those instructions as trivially dead if they are really
> dead, but has the following problem. After fwprop proper we have:
> (insn 7
> I'm afraid having different behavior for incr > 0 and incr < 0 could be a
> problem, then we would not decrease it even if we've increased it
> previously, so this would make setters of that REGNO effectively
> undeletable.
Yes, the setters that are before it in the RTL stream, which is precisel
On Fri, Apr 04, 2025 at 01:08:31PM +0200, Eric Botcazou wrote:
> I guess the question is: how much do we pessimize if we tweak r0-6313 change?
> For example, we could say that a use of REG in the SET_SRC does not count if
> REG is also set in the SET_DEST *only* if there is no previous use of REG:
