Hi Sam,
> This seems to suggest that block->loop_end() is the first block inside
the loop, is this true?! If so, I hope it's clear why this is confusing :)
The comment on GetLoopEndRpo explains what loop_end is:
// In Turbofan, the `block->loop_end()` refers to the first after
(outside)
// the loop. In the relevant use cases, we retrieve the backedge block by
// subtracting one from the rpo_number, so for Turboshaft we "fake" this
by
// adding 1 to the backedge block's rpo_number.
So, loop_end is outside of the loop.
> Does the header have exactly one predecessor outside of the loop?
Yes.
> Is loop_end guaranteed to contain a backedge to the header?
You might need a -1, cf the comment above, but yes.
> Are other blocks allowed to have an edge back to the header?
No.
Cheers,
Darius
On Wednesday, June 25, 2025 at 9:40:39 PM UTC+2 sam.p...@arm.com wrote:
> I think it's just the way that it's written, as it was suggested to me
> offline that this annotation is describing a list of blocks: header -> end.
> If this is true then it makes far more sense! Saying 'loop up to' has
> certain connotations for me.
>
> I do have other questions about the properties of loops too though:
> - Does the header have exactly one predecessor outside of the loop?
> - Is loop_end guaranteed to contain a backedge to the header?
> - Are other blocks allowed to have an edge back to the header?
>
> Thanks.
>
>
>
> On Wednesday, June 25, 2025 at 1:47:57 PM UTC+1 in...@bnoordhuis.nl wrote:
>
>> On Wed, Jun 25, 2025 at 12:06 PM Sam Parker-Haynes <sam.p...@arm.com>
>> wrote:
>> >
>> > Hi!
>> >
>> > I'm looking at the backend and the terminology used for loops is really
>> confusing me. AFAICT, in the backend there may be two special blocks for
>> loops but I can't say it makes any sense to me! Here's the code that is
>> boggling my mind:
>> >
>> > if (block->IsLoopHeader()) {
>> > buffer << " (loop up to " << block->loop_end().ToInt() << ")";
>> > }
>> >
>> > This seems to suggest that block->loop_end() is the first block inside
>> the loop, is this true?! If so, I hope it's clear why this is confusing :)
>> >
>> > cheers,
>> > sam
>>
>> Nodes are visited in reverse post-order. Is that the source of the
>> confusion?
>>
>
--
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups
"v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to v8-dev+unsubscr...@googlegroups.com.
To view this discussion visit
https://groups.google.com/d/msgid/v8-dev/b47dfc1d-d39e-4b87-b543-525bc4ccbb31n%40googlegroups.com.
