Fix a corner case hit by redundant phi elimination running after LSR.

[oota-llvm.git] / test / Transforms / DeadStoreElimination / simple.ll

diff --git a/test/Transforms/DeadStoreElimination/simple.ll b/test/Transforms/DeadStoreElimination/simple.ll
index 2f45a3e92363a0917a3f440148cb787cf0188824..81eb5a8c705e6dd25146148df3104cd740726c72 100644 (file)
--- a/test/Transforms/DeadStoreElimination/simple.ll
+++ b/test/Transforms/DeadStoreElimination/simple.ll
@@ -1,6 +1,10 @@
 ; RUN: opt < %s -basicaa -dse -S | FileCheck %s
 target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
+declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) nounwind
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
+declare i8* @llvm.init.trampoline(i8*, i8*, i8*)
+
 define void @test1(i32* %Q, i32* %P) {
         %DEAD = load i32* %Q
         store i32 %DEAD, i32* %P
@@ -38,31 +42,28 @@ define i32 @test3(i32* %g_addr) nounwind {
 
 define void @test4(i32* %Q) {
         %a = load i32* %Q
-        volatile store i32 %a, i32* %Q
+        store volatile i32 %a, i32* %Q
         ret void
 ; CHECK: @test4
 ; CHECK-NEXT: load i32
-; CHECK-NEXT: volatile store
+; CHECK-NEXT: store volatile
 ; CHECK-NEXT: ret void
 }
 
 define void @test5(i32* %Q) {
-        %a = volatile load i32* %Q
+        %a = load volatile i32* %Q
         store i32 %a, i32* %Q
         ret void
 ; CHECK: @test5
-; CHECK-NEXT: volatile load
+; CHECK-NEXT: load volatile
 ; CHECK-NEXT: ret void
 }
 
-declare void @llvm.memset.i64(i8*, i8, i64, i32)
-declare void @llvm.memcpy.i64(i8*, i8*, i64, i32)
-
 ; Should delete store of 10 even though memset is a may-store to P (P and Q may
 ; alias).
 define void @test6(i32 *%p, i8 *%q) {
   store i32 10, i32* %p, align 4       ;; dead.
-  call void @llvm.memset.i64(i8* %q, i8 42, i64 900, i32 1)
+  call void @llvm.memset.p0i8.i64(i8* %q, i8 42, i64 900, i32 1, i1 false)
   store i32 30, i32* %p, align 4
   ret void
 ; CHECK: @test6
@@ -73,7 +74,7 @@ define void @test6(i32 *%p, i8 *%q) {
 ; alias).
 define void @test7(i32 *%p, i8 *%q, i8* noalias %r) {
   store i32 10, i32* %p, align 4       ;; dead.
-  call void @llvm.memcpy.i64(i8* %q, i8* %r, i64 900, i32 1)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %q, i8* %r, i64 900, i32 1, i1 false)
   store i32 30, i32* %p, align 4
   ret void
 ; CHECK: @test7
@@ -116,7 +117,6 @@ define double @test10(i8* %X) {
 
 
 ; DSE should delete the dead trampoline.
-declare i8* @llvm.init.trampoline(i8*, i8*, i8*)
 declare void @test11f()
 define void @test11() {
 ; CHECK: @test11
@@ -145,3 +145,116 @@ define void @test12({ i32, i32 }* %x) nounwind  {
 ; CHECK: ret void
 }
 
+
+; %P doesn't escape, the DEAD instructions should be removed.
+declare void @test13f()
+define i32* @test13() {
+        %p = tail call i8* @malloc(i32 4)
+        %P = bitcast i8* %p to i32*
+        %DEAD = load i32* %P
+        %DEAD2 = add i32 %DEAD, 1
+        store i32 %DEAD2, i32* %P
+        call void @test13f( )
+        store i32 0, i32* %P
+        ret i32* %P
+; CHECK: @test13()
+; CHECK-NEXT: malloc
+; CHECK-NEXT: bitcast
+; CHECK-NEXT: call void
+}
+
+declare noalias i8* @malloc(i32)
+
+
+
+define void @test14(i32* %Q) {
+        %P = alloca i32
+        %DEAD = load i32* %Q
+        store i32 %DEAD, i32* %P
+        ret void
+
+; CHECK: @test14
+; CHECK-NEXT: ret void
+}
+
+
+; PR8701
+
+;; Fully dead overwrite of memcpy.
+define void @test15(i8* %P, i8* %Q) nounwind ssp {
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 12, i32 1, i1 false)
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 12, i32 1, i1 false)
+  ret void
+; CHECK: @test15
+; CHECK-NEXT: call void @llvm.memcpy
+; CHECK-NEXT: ret
+}
+
+;; Full overwrite of smaller memcpy.
+define void @test16(i8* %P, i8* %Q) nounwind ssp {
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 8, i32 1, i1 false)
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 12, i32 1, i1 false)
+  ret void
+; CHECK: @test16
+; CHECK-NEXT: call void @llvm.memcpy
+; CHECK-NEXT: ret
+}
+
+;; Overwrite of memset by memcpy.
+define void @test17(i8* %P, i8* noalias %Q) nounwind ssp {
+  tail call void @llvm.memset.p0i8.i64(i8* %P, i8 42, i64 8, i32 1, i1 false)
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 12, i32 1, i1 false)
+  ret void
+; CHECK: @test17
+; CHECK-NEXT: call void @llvm.memcpy
+; CHECK-NEXT: ret
+}
+
+; Should not delete the volatile memset.
+define void @test17v(i8* %P, i8* %Q) nounwind ssp {
+  tail call void @llvm.memset.p0i8.i64(i8* %P, i8 42, i64 8, i32 1, i1 true)
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 12, i32 1, i1 false)
+  ret void
+; CHECK: @test17v
+; CHECK-NEXT: call void @llvm.memset
+; CHECK-NEXT: call void @llvm.memcpy
+; CHECK-NEXT: ret
+}
+
+; PR8728
+; Do not delete instruction where possible situation is:
+; A = B
+; A = A
+define void @test18(i8* %P, i8* %Q, i8* %R) nounwind ssp {
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %Q, i64 12, i32 1, i1 false)
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %R, i64 12, i32 1, i1 false)
+  ret void
+; CHECK: @test18
+; CHECK-NEXT: call void @llvm.memcpy
+; CHECK-NEXT: call void @llvm.memcpy
+; CHECK-NEXT: ret
+}
+
+
+; The store here is not dead because the byval call reads it.
+declare void @test19f({i32}* byval align 4 %P)
+
+define void @test19({i32} * nocapture byval align 4 %arg5) nounwind ssp {
+bb:
+  %tmp7 = getelementptr inbounds {i32}* %arg5, i32 0, i32 0
+  store i32 912, i32* %tmp7
+  call void @test19f({i32}* byval align 4 %arg5)
+  ret void
+
+; CHECK: @test19(
+; CHECK: store i32 912
+; CHECK: call void @test19f
+}
+
+define void @test20() {
+  %m = call i8* @malloc(i32 24)
+  store i8 0, i8* %m
+  ret void
+}
+; CHECK: @test20
+; CHECK-NEXT: ret void