[llvm] [AMDGPU][Scheduler] Scoring system for rematerialization candidates (PR #153092)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,535 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn -mcpu=gfx908 -run-pass=machine-scheduler -amdgpu-disable-unclustered-high-rp-reschedule -verify-machineinstrs %s -o - | FileCheck %s
+
+# All tests are almost identical, the only differences being that some
+# VGPR-defining instructions are progressively made artificially
+# unrematerializable with an implicit def to test rematerialization
+# priorities. The CFG is the following for all tests in the file.
+#
+#           +---+
+#           | 0 |
+#           +---+
+#             |
+#             v
+#           +---+
+#   +------>| 1 |-----+
+#   |       +---+     |
+#   |         |       v
+#   |         |     +---+
+#   |         |     | 2 |
+#   |         |     +-+-+
+#   |         v       |
+# +---+     +---+     |
+# | 4 |<----| 3 |<----+
+# +---+     +---+
+#             |
+#             v
+#           +---+
+#           | 5 |
+#           +---+
+
+--- |
+  define void @favor_always_benef() {
+    ret void
+  }
+  define void @favor_live_through_in_high_freq_region() {
+    ret void
+  }
+  define void @use_only_region_possible() {
+    ret void
+  }
----------------
arsenm wrote:

Don't need IR section 

https://github.com/llvm/llvm-project/pull/153092