[PATCH] D10208: Add tests for full unroll heuristic: folding CFG, folding IVs.

[Gerolf Hoflehner]

Gerolf added a subscriber: Gerolf.

================
Comment at: test/Transforms/LoopUnroll/full-unroll-heuristics-cfg.ll:6
@@ +5,3 @@
+
+; If a load becomes a constant after loop unrolling, we sometimes can simplify
+; CFG. This test verifies that we handle such cases.
----------------
Please comment in more detail which cases can be simplified. Are there scenarios with constant loads that won't get simplified?

================
Comment at: test/Transforms/LoopUnroll/full-unroll-heuristics-cfg.ll:69
@@ +68,3 @@
+; Induction variables are often casted to another type, and that shouldn't
+; prevent us from folding branches. Tthis test specifically checks if we can
+; handle this.
----------------
Tthis...

================
Comment at: test/Transforms/LoopUnroll/full-unroll-heuristics-cfg.ll:100
@@ +99,3 @@
+
+; In this test %i1024 has a step of 1024, and it's truncated to i8 on every
+; iteration. Obviously, result of this trunc is always 0, and we should be able
----------------
1024 (=2^10). 
it's -> it is


http://reviews.llvm.org/D10208