llvm dev - Mar 2018 - TargetOpcode::KILL confusion

Hello,


Could someone please explain the semantics of TargetOpcode::KILL?  Specifically,
in this example, which register is killed?  Would it be legal for operands 0 and
1 to refer to different registers?


128B    %R3<def> = KILL %R3, %R3_1<imp-use>, %R3_23<imp-use>


(In my out-of-tree target, %R3 is a <4xi32> register, %R3_1 is an i32
sub-register of %R3, and %R3_23 is a <2xi32> sub-register of %R3).


Thanks,

Nick


-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180322/0267f9a4/attachment.html>

KILL is usually just a placeholder for an operation that has been removed.
We keep it around not to have to update some data structure.
Its “semantic” is similar to a nop but unlike nops they won’t generate any code.

Typically, KILL instructions are created out of identity copies or instructions
with only dead defs.
> On Mar 22, 2018, at 8:20 AM, Johnson, Nicholas Paul via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> 
> Hello,
> 
> Could someone please explain the semantics of TargetOpcode::KILL? 
Specifically, in this example, which register is killed?  Would it be legal for
operands 0 and 1 to refer to different registers?
> 
> 128B    %R3<def> = KILL %R3, %R3_1<imp-use>,
%R3_23<imp-use>
> 
> (In my out-of-tree target, %R3 is a <4xi32> register, %R3_1 is an i32
sub-register of %R3, and %R3_23 is a <2xi32> sub-register of %R3).
> 
> Thanks,
> Nick
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
<http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180322/c2672691/attachment.html>