[llvm] [NFC][SLP] Precommit tests for ordered fadd-reduction FMA-fusion cost (PR #210835)

[Dmitry Sidorov via llvm-commits]

================
@@ -0,0 +1,46 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -passes=slp-vectorizer -S -mtriple=nvptx64-nvidia-cuda -mcpu=sm_80 -slp-threshold=-16 < %s | FileCheck %s
----------------
MrSidims wrote:

@alexey-bataev this is a very-unnatural test and I'm not sure expresses real NVPTX backend behaviour. It just leaves this code scalar and I had to force vectorization by -slp-threshold=-16, so then could test a fix from https://github.com/llvm/llvm-project/pull/210399 and another follow-up patch. I'm happy to remove it if you find it useless. 

https://github.com/llvm/llvm-project/pull/210835