[PATCH] Optimize memcmp(x, y, n)==0 for small n and suitably aligned x/y.

[Chad Rosier]

================
Comment at: test/Transforms/InstCombine/memcmp-1.ll:77
@@ +76,3 @@
+define i1 @test_simplify7(i64 %x, i64 %y) {
+; CHECK-label: @test_simplify7(
+  %x.addr = alloca i64, align 8
----------------
; CHECK-LABEL:

================
Comment at: test/Transforms/InstCombine/memcmp-1.ll:94
@@ +93,3 @@
+define i1 @test_simplify8(i32 %x, i32 %y) {
+; CHECK-label: @test_simplify8(
+  %x.addr = alloca i32, align 4
----------------
; CHECK-LABEL:

================
Comment at: test/Transforms/InstCombine/memcmp-1.ll:111
@@ +110,3 @@
+define i1 @test_simplify9(i16 %x, i16 %y) {
+; CHECK-label: @test_simplify9(
+  %x.addr = alloca i16, align 2
----------------
; CHECK-LABEL:

http://reviews.llvm.org/D6952

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/