+++ /dev/null
-; RUN: llc < %s -march=arm -mtriple=arm-apple-darwin | \
-; RUN: grep cmpne | count 1
-; RUN: llc < %s -march=arm -mtriple=arm-apple-darwin | \
-; RUN: grep bls | count 1
-; Here, tail call wins over eliminating branches. It is 1 fewer instruction
-; and removes all stack accesses, so seems like a win.
-
-define void @foo(i32 %X, i32 %Y) {
-entry:
- %tmp1 = icmp ult i32 %X, 4 ; <i1> [#uses=1]
- %tmp4 = icmp eq i32 %Y, 0 ; <i1> [#uses=1]
- %tmp7 = or i1 %tmp4, %tmp1 ; <i1> [#uses=1]
- br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock
-
-cond_true: ; preds = %entry
- %tmp10 = tail call i32 (...)* @bar( ) ; <i32> [#uses=0]
- ret void
-
-UnifiedReturnBlock: ; preds = %entry
- ret void
-}
-
-declare i32 @bar(...)
--- /dev/null
+; RUN: llc < %s -march=arm -mtriple=arm-apple-darwin | \
+; RUN: grep cmpne | count 1
+; RUN: llc < %s -march=arm -mtriple=arm-apple-darwin | \
+; RUN: grep ldmiahi | count 1
+
+define void @foo(i32 %X, i32 %Y) {
+entry:
+ %tmp1 = icmp ult i32 %X, 4 ; <i1> [#uses=1]
+ %tmp4 = icmp eq i32 %Y, 0 ; <i1> [#uses=1]
+ %tmp7 = or i1 %tmp4, %tmp1 ; <i1> [#uses=1]
+ br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock
+
+cond_true: ; preds = %entry
+ %tmp10 = call i32 (...)* @bar( ) ; <i32> [#uses=0]
+ ret void
+
+UnifiedReturnBlock: ; preds = %entry
+ ret void
+}
+
+declare i32 @bar(...)