Request to cherry-pick "[AMDGPU] Fix for folding v2.16 literals." onto 11.0 release branch

[Quuxplusone]

Bugzilla Link	PR47516
Status	RESOLVED FIXED
Importance	P enhancement
Reported by	Jay Foad (jay.foad@gmail.com)
Reported on	2020-09-14 03:54:40 -0700
Last modified on	2020-09-14 06:27:13 -0700
Version	11.0
Hardware	PC Linux
CC	daniil.fukalov@amd.com, david.stuttard@amd.com, hans@chromium.org, llvm-bugs@lists.llvm.org, maraeo@gmail.com, Matthew.Arsenault@amd.com, tpr.ll@botech.co.uk
Fixed by commit(s)
Attachments	shrink-add-sub-constant.ll (106140 bytes, text/plain)
Blocks	PR46725
Blocked by
See also

Please can you cherry-pick this AMDGPU codegen fix onto the 11.0 release branch?

commit c259d3a061c8fc0f9520208eb265d4352a0ad447
Author: dfukalov <daniil.fukalov@amd.com>
Date:   Fri Sep 4 22:44:01 2020 +0300

    [AMDGPU] Fix for folding v2.16 literals.

    It was found some packed immediate operands (e.g. `<half 1.0, half 2.0>`) are
    incorrectly processed so one of two packed values were lost.

    Introduced new function to check immediate 32-bit operand can be folded.
    Converted condition about current op_sel flags value to fall-through.

    Fixes: SWDEV-247595

    Reviewed By: rampitec

    Differential Revision: https://reviews.llvm.org/D87158

[Quuxplusone]

Attached shrink-add-sub-constant.ll (106140 bytes, text/plain): test case after conflict resolution

[Quuxplusone]

Alternatively here is a cherry-pick with conflicts resolved and tests passing that you can pick from: https://github.com/jayfoad/llvm-project/pull/new/cherry-pick-d87158

[Quuxplusone]

(In reply to Jay Foad from comment #2)
> Alternatively here is a cherry-pick with conflicts resolved and tests
> passing that you can pick from:
> https://github.com/jayfoad/llvm-project/pull/new/cherry-pick-d87158

Thanks! Pushed that as b3fb40b3a3c1fb7ac094eda50762624baad37552.