[llvm] [SystemZ] SLP reductions: cost functions of reductions and scalarization (PR #112491)

[Jonas Paulsson via llvm-commits]

================
@@ -0,0 +1,88 @@
+; RUN: opt -S --passes=slp-vectorizer -mtriple=s390x-unknown-linux -mcpu=z16 < %s | FileCheck %s
----------------
JonPsson1 wrote:

There is no difference with cost estimations of the functions in this file - one by one they return the same scalar cost. The interesting thing here is to run SLP and see how it behaves with the improved scalarization costs. So I don't think that running opt with -passes='print<cost-model>' makes much sense. Looking at the SLP output doesn't give me much either, there is just a different number for a particular bundle, but it wouldn't be clear in the test what that means, I think.

This could be precommitted, but I think I would prefer to wait until it is actually ready to be committed.

https://github.com/llvm/llvm-project/pull/112491