llvm dev - Oct 2021 - Can DBG_VALUE instructions use undefined vregs?

One GlobalISel compile time optimization patch (https://reviews.llvm.org/D109750
<https://reviews.llvm.org/D109750>) has generated some debate over whether
it’s semantically allowed/not-an-error for a DBG_VALUE machine instruction to
have a use of a vreg that doesn’t have a corresponding definition.

We talked internally with Adrian and Vedant and didn’t come to a strong
conclusion either way. Does anyone have thoughts here?

Thanks,
Amara
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211005/6b037184/attachment.html>

I can't really be speak about debug instructions in particular. For machine
instructions in general you just mark the use operand with "undef" and
you are good...

- Matthias
> On Oct 5, 2021, at 3:29 PM, Amara Emerson via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
> 
> One GlobalISel compile time optimization patch
(https://reviews.llvm.org/D109750) has generated some debate over whether it’s
semantically allowed/not-an-error for a DBG_VALUE machine instruction to have a
use of a vreg that doesn’t have a corresponding definition.
> 
> We talked internally with Adrian and Vedant and didn’t come to a strong
conclusion either way. Does anyone have thoughts here?
> 
> Thanks,
> Amara
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

Hi,

On Tue, Oct 5, 2021 at 11:29 PM Amara Emerson via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> One GlobalISel compile time optimization patch
(https://reviews.llvm.org/D109750) has generated some debate over whether it’s
semantically allowed/not-an-error for a DBG_VALUE machine instruction to have a
use of a vreg that doesn’t have a corresponding definition.
I don't think there's a semantic reason why DBG_VALUEs of undefined
vregs should be disallowed -- they can be interpreted the same as a
DBG_VALUE $noreg, just in a non-canonical form. They would correctly
become DBG_VALUE $noreg's when LiveDebugVariables runs, as the vreg
wouldn't be live.

In practical terms, it could slightly mislead people reading MIR, and
there might be code out there that expects to find a vreg definition
for anything that DBG_VALUEs refer to, which I imagine would be fairly
easy to work around.

--
Thanks,
Jeremy