From 2a46e93aedd2d4f2b3e5049e45e3a358cab1577f Mon Sep 17 00:00:00 2001
From: Hans Wennborg <hans@hanshq.net>
Date: Wed, 1 Feb 2017 22:00:37 +0000
Subject: [PATCH] Merging r293542:
 ------------------------------------------------------------------------
 r293542 | arsenm | 2017-01-30 11:50:17 -0800 (Mon, 30 Jan 2017) | 7 lines

LSR: Don't drop address space when type doesn't match

For targets with different addressing modes in each address space,
if this is dropped querying isLegalAddressingMode later with this
will give a nonsense result, breaking the isLegalUse assertions.

This is a candidate for the 4.0 release branch.
------------------------------------------------------------------------


git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_40@293819 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Transforms/Scalar/LoopStrengthReduce.cpp  | 11 ++--
 .../AMDGPU/preserve-addrspace-assert.ll       | 54 +++++++++++++++++++
 2 files changed, 61 insertions(+), 4 deletions(-)
 create mode 100644 test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll

diff --git a/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/lib/Transforms/Scalar/LoopStrengthReduce.cpp
index 01728ae680d..194587a85e7 100644
--- a/lib/Transforms/Scalar/LoopStrengthReduce.cpp
+++ b/lib/Transforms/Scalar/LoopStrengthReduce.cpp
@@ -158,8 +158,9 @@ struct MemAccessTy {
 
   bool operator!=(MemAccessTy Other) const { return !(*this == Other); }
 
-  static MemAccessTy getUnknown(LLVMContext &Ctx) {
-    return MemAccessTy(Type::getVoidTy(Ctx), UnknownAddressSpace);
+  static MemAccessTy getUnknown(LLVMContext &Ctx,
+                                unsigned AS = UnknownAddressSpace) {
+    return MemAccessTy(Type::getVoidTy(Ctx), AS);
   }
 };
 
@@ -2279,8 +2280,10 @@ bool LSRInstance::reconcileNewOffset(LSRUse &LU, int64_t NewOffset,
   // TODO: Be less conservative when the type is similar and can use the same
   // addressing modes.
   if (Kind == LSRUse::Address) {
-    if (AccessTy != LU.AccessTy)
-      NewAccessTy = MemAccessTy::getUnknown(AccessTy.MemTy->getContext());
+    if (AccessTy.MemTy != LU.AccessTy.MemTy) {
+      NewAccessTy = MemAccessTy::getUnknown(AccessTy.MemTy->getContext(),
+                                            AccessTy.AddrSpace);
+    }
   }
 
   // Conservatively assume HasBaseReg is true for now.
diff --git a/test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll b/test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll
new file mode 100644
index 00000000000..b3b696d42c5
--- /dev/null
+++ b/test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll
@@ -0,0 +1,54 @@
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -loop-reduce %s | FileCheck %s
+
+; Test for assert resulting from inconsistent isLegalAddressingMode
+; answers when the address space was dropped from the query.
+
+target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
+
+%0 = type { i32, double, i32, float }
+
+; CHECK-LABEL: @lsr_crash_preserve_addrspace_unknown_type(
+; CHECK: %tmp4 = bitcast %0 addrspace(3)* %tmp to double addrspace(3)*
+; CHECK: %scevgep5 = getelementptr double, double addrspace(3)* %tmp4, i32 1
+; CHEC: load double, double addrspace(3)* %scevgep5
+
+; CHECK: %scevgep = getelementptr i32, i32 addrspace(3)* %tmp1, i32 4
+; CHECK:%tmp14 = load i32, i32 addrspace(3)* %scevgep
+define void @lsr_crash_preserve_addrspace_unknown_type() #0 {
+bb:
+  br label %bb1
+
+bb1:                                              ; preds = %bb17, %bb
+  %tmp = phi %0 addrspace(3)* [ undef, %bb ], [ %tmp18, %bb17 ]
+  %tmp2 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 0, i32 1
+  %tmp3 = load double, double addrspace(3)* %tmp2, align 8
+  br label %bb4
+
+bb4:                                              ; preds = %bb1
+  br i1 undef, label %bb8, label %bb5
+
+bb5:                                              ; preds = %bb4
+  unreachable
+
+bb8:                                              ; preds = %bb4
+  %tmp9 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 0, i32 0
+  %tmp10 = load i32, i32 addrspace(3)* %tmp9, align 4
+  %tmp11 = icmp eq i32 0, %tmp10
+  br i1 %tmp11, label %bb12, label %bb17
+
+bb12:                                             ; preds = %bb8
+  %tmp13 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 0, i32 2
+  %tmp14 = load i32, i32 addrspace(3)* %tmp13, align 4
+  %tmp15 = icmp eq i32 0, %tmp14
+  br i1 %tmp15, label %bb16, label %bb17
+
+bb16:                                             ; preds = %bb12
+  unreachable
+
+bb17:                                             ; preds = %bb12, %bb8
+  %tmp18 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 2
+  br label %bb1
+}
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
-- 
2.50.1