[llvm-dev] Is there a way to correlate operation to machine instruction?

Ryan Taylor via llvm-dev llvm-dev at lists.llvm.org
Fri Apr 14 07:46:53 PDT 2017

Is it possible to insert duplicate SDNodes prior to ISel (preferably during
and directly before selection) that won't be CSE'd out?


On Wed, Apr 12, 2017 at 2:24 PM, Matt Arsenault <Matthew.Arsenault at amd.com>

> On 04/12/2017 11:07 AM, Ryan Taylor wrote:
>> Matt,
>>   so in AMDGPU, the operands are sort of 'generic'? Can you point me to
>> the right places?
>> Thanks.
> Not exactly generic, but they have custom defined constraints. Look at the
> definitions for the VSrc_* operand types in SIRegisterIno.td. These set the
> OperandType on the Operand to a custom enum, and the target
> verifyInstruction checks the custom constraints.
> -Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170414/7a9db398/attachment.html>

More information about the llvm-dev mailing list