[NFC][LoopIdiom] Add bcmp loop idiom miscompile test from PR43206.

The transform forgot to check SCEV loop scopes.

https://bugs.llvm.org/show_bug.cgi?id=43206

llvm-svn: 374661

1 files changed, 54 insertions, 3 deletions

diff --git a/llvm/test/Transforms/LoopIdiom/bcmp-negative-tests.ll b/llvm/test/Transforms/LoopIdiom/bcmp-negative-tests.ll
index 9d770adf979..a4d4a3d250a 100644
--- a/llvm/test/Transforms/LoopIdiom/bcmp-negative-tests.ll
+++ b/llvm/test/Transforms/LoopIdiom/bcmp-negative-tests.ll
@@ -1,9 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -loop-idiom < %s -S | FileCheck %s
+; RUN: opt -loop-idiom -verify -verify-each -verify-dom-info -verify-loop-info < %s -S | FileCheck %s --implicit-check-not=bcmp --implicit-check-not=memcmp
 
 ; CHECK: source_filename
-; CHECK-NOT; bcmp
-; CHECK-NOT; memcmp
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
 
@@ -478,3 +476,56 @@ cleanup:
   %res = phi i1 [ false, %for.body ], [ true, %for.inc ]
   ret i1 %res
 }
+
+; See https://bugs.llvm.org/show_bug.cgi?id=43206 for original reduced (but runnable) test:
+;
+; bool do_check(int i_max, int j_max, int**bptr, int* fillch) {
+;   for (int i = 0; i < i_max; i++)
+;       for (int j = 0; j < j_max; j++)
+;         if (bptr[i][j] != fillch[i])
+;           return 1;
+;   return 0;
+; }
+; The loads proceed differently here - fillch[i] changes once per outer loop,
+; while bptr[i][j] changes both in inner loop, and in outer loop.
+define i1 @pr43206_different_loops(i32 %i_max, i32 %j_max, i32** %bptr, i32* %fillch) {
+entry:
+  %cmp31 = icmp sgt i32 %i_max, 0
+  %cmp229 = icmp sgt i32 %j_max, 0
+  %or.cond = and i1 %cmp31, %cmp229
+  br i1 %or.cond, label %for.cond1.preheader.us.preheader, label %cleanup12
+
+for.cond1.preheader.us.preheader:                 ; preds = %entry
+  %wide.trip.count38 = zext i32 %i_max to i64
+  %wide.trip.count = zext i32 %j_max to i64
+  br label %for.cond1.preheader.us
+
+for.cond1.preheader.us:                           ; preds = %for.cond1.for.inc10_crit_edge.us, %for.cond1.preheader.us.preheader
+  %indvars.iv36 = phi i64 [ 0, %for.cond1.preheader.us.preheader ], [ %indvars.iv.next37, %for.cond1.for.inc10_crit_edge.us ]
+  %arrayidx.us = getelementptr inbounds i32*, i32** %bptr, i64 %indvars.iv36
+  %v0 = load i32*, i32** %arrayidx.us, align 8
+  %arrayidx8.us = getelementptr inbounds i32, i32* %fillch, i64 %indvars.iv36
+  %v1 = load i32, i32* %arrayidx8.us, align 4
+  br label %for.body4.us
+
+for.cond1.us:                                     ; preds = %for.body4.us
+  %exitcond = icmp eq i64 %indvars.iv.next, %wide.trip.count
+  br i1 %exitcond, label %for.cond1.for.inc10_crit_edge.us, label %for.body4.us
+
+for.body4.us:                                     ; preds = %for.cond1.us, %for.cond1.preheader.us
+  %indvars.iv = phi i64 [ 0, %for.cond1.preheader.us ], [ %indvars.iv.next, %for.cond1.us ]
+  %arrayidx6.us = getelementptr inbounds i32, i32* %v0, i64 %indvars.iv
+  %v2 = load i32, i32* %arrayidx6.us, align 4
+  %cmp9.us = icmp eq i32 %v2, %v1
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  br i1 %cmp9.us, label %for.cond1.us, label %cleanup12
+
+for.cond1.for.inc10_crit_edge.us:                 ; preds = %for.cond1.us
+  %indvars.iv.next37 = add nuw nsw i64 %indvars.iv36, 1
+  %exitcond39 = icmp eq i64 %indvars.iv.next37, %wide.trip.count38
+  br i1 %exitcond39, label %cleanup12, label %for.cond1.preheader.us
+
+cleanup12:                                        ; preds = %for.cond1.for.inc10_crit_edge.us, %for.body4.us, %entry
+  %v3 = phi i1 [ false, %entry ], [ true, %for.body4.us ], [ false, %for.cond1.for.inc10_crit_edge.us ]
+  ret i1 %v3
+}