Fix a bug in DemoteRegToStack where a reload instruction was inserted into the

wrong basic block.

This would happen when the result of an invoke was used by a phi instruction
in the invoke's normal destination block. An instruction to reload the invoke's
value would get inserted before the critical edge was split and a new basic
block (which is the correct insertion point for the reload) was created. This
commit fixes the bug by splitting the critical edge before all the reload
instructions are inserted.

Also, hoist up the code which computes the insertion point to the only place
that need that computation.

rdar://problem/15978721

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228566 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Utils/DemoteRegToStack.cpp b/lib/Transforms/Utils/DemoteRegToStack.cpp
index 9972b22a07e400fdd0b1ffeaa8038e8f99df6464..003da58ee79831e6fdec3704f5f2a36876876294 100644 (file)
--- a/lib/Transforms/Utils/DemoteRegToStack.cpp
+++ b/lib/Transforms/Utils/DemoteRegToStack.cpp
@@ -39,6 +39,19 @@ AllocaInst *llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
                           F->getEntryBlock().begin());
   }
 
+  // We cannot demote invoke instructions to the stack if their normal edge
+  // is critical. Therefore, split the critical edge and create a basic block
+  // into which the store can be inserted.
+  if (InvokeInst *II = dyn_cast<InvokeInst>(&I)) {
+    if (!II->getNormalDest()->getSinglePredecessor()) {
+      unsigned SuccNum = GetSuccessorNumber(II->getParent(), II->getNormalDest());
+      assert(isCriticalEdge(II, SuccNum) && "Expected a critical edge!");
+      BasicBlock *BB = SplitCriticalEdge(II, SuccNum);
+      assert(BB && "Unable to split critical edge.");
+      (void)BB;
+    }
+  }
+
   // Change all of the users of the instruction to read from the stack slot.
   while (!I.use_empty()) {
     Instruction *U = cast<Instruction>(I.user_back());
@@ -71,7 +84,6 @@ AllocaInst *llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
     }
   }
 
-
   // Insert stores of the computed value into the stack slot. We have to be
   // careful if I is an invoke instruction, because we can't insert the store
   // AFTER the terminator instruction.
@@ -79,27 +91,13 @@ AllocaInst *llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
   if (!isa<TerminatorInst>(I)) {
     InsertPt = &I;
     ++InsertPt;
+    for (; isa<PHINode>(InsertPt) || isa<LandingPadInst>(InsertPt); ++InsertPt)
+      /* empty */;   // Don't insert before PHI nodes or landingpad instrs.
   } else {
     InvokeInst &II = cast<InvokeInst>(I);
-    if (II.getNormalDest()->getSinglePredecessor())
-      InsertPt = II.getNormalDest()->getFirstInsertionPt();
-    else {
-      // We cannot demote invoke instructions to the stack if their normal edge
-      // is critical.  Therefore, split the critical edge and insert the store
-      // in the newly created basic block.
-      unsigned SuccNum = GetSuccessorNumber(I.getParent(), II.getNormalDest());
-      TerminatorInst *TI = &cast<TerminatorInst>(I);
-      assert (isCriticalEdge(TI, SuccNum) &&
-              "Expected a critical edge!");
-      BasicBlock *BB = SplitCriticalEdge(TI, SuccNum);
-      assert (BB && "Unable to split critical edge.");
-      InsertPt = BB->getFirstInsertionPt();
-    }
+    InsertPt = II.getNormalDest()->getFirstInsertionPt();
   }
 
-  for (; isa<PHINode>(InsertPt) || isa<LandingPadInst>(InsertPt); ++InsertPt)
-    /* empty */;   // Don't insert before PHI nodes or landingpad instrs.
-
   new StoreInst(&I, Slot, InsertPt);
   return Slot;
 }

diff --git a/test/CodeGen/ARM/sjlj-prepare-critical-edge.ll b/test/CodeGen/ARM/sjlj-prepare-critical-edge.ll
index d8241d0dc380b64b8b9fbff8d3e8d01f63c7db67..a7bc22fe9683bba614ef6cb1c7d2d8ae6705b60c 100644 (file)
--- a/test/CodeGen/ARM/sjlj-prepare-critical-edge.ll
+++ b/test/CodeGen/ARM/sjlj-prepare-critical-edge.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -O1 -mtriple thumbv7-apple-ios6
+; RUN: llc < %s -O1 -mtriple thumbv7-apple-ios6 | FileCheck %s
 ; Just make sure no one tries to make the assumption that the normal edge of an
 ; invoke is never a critical edge.  Previously, this code would assert.
 
@@ -65,3 +65,129 @@ declare i32 @__gxx_personality_sj0(...)
 declare void @release(i8*)
 
 declare void @terminatev()
+
+; Make sure that the instruction DemoteRegToStack inserts to reload
+; %call.i.i.i14.i.i follows the instruction that saves the value to the stack in
+; basic block %entry.do.body.i.i.i_crit_edge.
+; Previously, DemoteRegToStack would insert a load instruction into the entry
+; block to reload %call.i.i.i14.i.i before the phi instruction (%0) in block
+; %do.body.i.i.i.
+
+; CHECK-LABEL: __Z4foo1c:
+; CHECK: blx __Znwm
+; CHECK: {{.*}}@ %entry.do.body.i.i.i_crit_edge
+; CHECK: str r0, [sp, [[OFFSET:#[0-9]+]]]
+; CHECK: ldr [[R0:r[0-9]+]], [sp, [[OFFSET]]]
+; CHECK: {{.*}}@ %do.body.i.i.i
+; CHECK: cmp [[R0]], #0
+
+%"class.std::__1::basic_string" = type { %"class.std::__1::__compressed_pair" }
+%"class.std::__1::__compressed_pair" = type { %"class.std::__1::__libcpp_compressed_pair_imp" }
+%"class.std::__1::__libcpp_compressed_pair_imp" = type { %"struct.std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >::__rep" }
+%"struct.std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >::__rep" = type { %union.anon }
+%union.anon = type { %"struct.std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >::__long" }
+%"struct.std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >::__long" = type { i32, i32, i8* }
+
+@.str = private unnamed_addr constant [12 x i8] c"some_string\00", align 1
+
+define void @_Z4foo1c(i8 signext %a) {
+entry:
+  %s1 = alloca %"class.std::__1::basic_string", align 4
+  call void @_ZNSt3__112basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEE6__initEPKcm(%"class.std::__1::basic_string"* %s1, i8* getelementptr inbounds ([12 x i8]* @.str, i32 0, i32 0), i32 11)
+  %call.i.i.i14.i.i = invoke noalias i8* @_Znwm(i32 1024)
+          to label %do.body.i.i.i unwind label %lpad.body
+
+do.body.i.i.i:                                    ; preds = %entry, %_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i
+  %lsr.iv = phi i32 [ %lsr.iv.next, %_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i ], [ -1024, %entry ]
+  %0 = phi i8* [ %incdec.ptr.i.i.i, %_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i ], [ %call.i.i.i14.i.i, %entry ]
+  %new.isnull.i.i.i.i = icmp eq i8* %0, null
+  br i1 %new.isnull.i.i.i.i, label %_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i, label %new.notnull.i.i.i.i
+
+new.notnull.i.i.i.i:                              ; preds = %do.body.i.i.i
+  store i8 %a, i8* %0, align 1
+  br label %_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i
+
+_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i: ; preds = %new.notnull.i.i.i.i, %do.body.i.i.i
+  %1 = phi i8* [ null, %do.body.i.i.i ], [ %0, %new.notnull.i.i.i.i ]
+  %incdec.ptr.i.i.i = getelementptr inbounds i8* %1, i32 1
+  %lsr.iv.next = add i32 %lsr.iv, 1
+  %cmp.i16.i.i = icmp eq i32 %lsr.iv.next, 0
+  br i1 %cmp.i16.i.i, label %invoke.cont, label %do.body.i.i.i
+
+invoke.cont:                                      ; preds = %_ZNSt3__116allocator_traitsINS_9allocatorIcEEE9constructIccEEvRS2_PT_RKT0_.exit.i.i.i
+  invoke void @_Z4foo2Pci(i8* %call.i.i.i14.i.i, i32 1024)
+          to label %invoke.cont5 unwind label %lpad2
+
+invoke.cont5:                                     ; preds = %invoke.cont
+  %cmp.i.i.i15 = icmp eq i8* %call.i.i.i14.i.i, null
+  br i1 %cmp.i.i.i15, label %invoke.cont6, label %_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i19
+
+_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i19: ; preds = %invoke.cont5
+  call void @_ZdlPv(i8* %call.i.i.i14.i.i)
+  br label %invoke.cont6
+
+invoke.cont6:                                     ; preds = %_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i19, %invoke.cont5
+  %call10 = call %"class.std::__1::basic_string"* @_ZNSt3__112basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEED1Ev(%"class.std::__1::basic_string"* %s1)
+  ret void
+
+lpad.body:                                        ; preds = %entry
+  %2 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_sj0 to i8*)
+          cleanup
+  %3 = extractvalue { i8*, i32 } %2, 0
+  %4 = extractvalue { i8*, i32 } %2, 1
+  br label %ehcleanup
+
+lpad2:                                            ; preds = %invoke.cont
+  %5 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_sj0 to i8*)
+          cleanup
+  %6 = extractvalue { i8*, i32 } %5, 0
+  %7 = extractvalue { i8*, i32 } %5, 1
+  %cmp.i.i.i21 = icmp eq i8* %call.i.i.i14.i.i, null
+  br i1 %cmp.i.i.i21, label %ehcleanup, label %_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i26
+
+_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i26: ; preds = %lpad2
+  call void @_ZdlPv(i8* %call.i.i.i14.i.i)
+  br label %ehcleanup
+
+ehcleanup:                                        ; preds = %_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i26, %lpad2, %lpad.body
+  %exn.slot.0 = phi i8* [ %3, %lpad.body ], [ %6, %lpad2 ], [ %6, %_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i26 ]
+  %ehselector.slot.0 = phi i32 [ %4, %lpad.body ], [ %7, %lpad2 ], [ %7, %_ZNSt3__113__vector_baseIcNS_9allocatorIcEEE5clearEv.exit.i.i.i26 ]
+  %call12 = invoke %"class.std::__1::basic_string"* @_ZNSt3__112basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEED1Ev(%"class.std::__1::basic_string"* %s1)
+          to label %eh.resume unwind label %terminate.lpad
+
+eh.resume:                                        ; preds = %ehcleanup
+  %lpad.val = insertvalue { i8*, i32 } undef, i8* %exn.slot.0, 0
+  %lpad.val13 = insertvalue { i8*, i32 } %lpad.val, i32 %ehselector.slot.0, 1
+  resume { i8*, i32 } %lpad.val13
+
+terminate.lpad:                                   ; preds = %ehcleanup
+  %8 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_sj0 to i8*)
+          catch i8* null
+  %9 = extractvalue { i8*, i32 } %8, 0
+  call void @__clang_call_terminate(i8* %9)
+  unreachable
+}
+
+declare void @_Z4foo2Pci(i8*, i32)
+
+define linkonce_odr hidden void @__clang_call_terminate(i8*) {
+  %2 = tail call i8* @__cxa_begin_catch(i8* %0)
+  tail call void @_ZSt9terminatev()
+  unreachable
+}
+
+declare i8* @__cxa_begin_catch(i8*)
+declare void @_ZSt9terminatev()
+declare %"class.std::__1::basic_string"* @_ZNSt3__112basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEED1Ev(%"class.std::__1::basic_string"* returned)
+declare void @_ZdlPv(i8*) #3
+declare noalias i8* @_Znwm(i32)
+declare void @_ZNSt3__112basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEE6__initEPKcm(%"class.std::__1::basic_string"*, i8*, i32)
+declare void @_Unwind_SjLj_Register({ i8*, i32, [4 x i32], i8*, i8*, [5 x i8*] }*)
+declare void @_Unwind_SjLj_Unregister({ i8*, i32, [4 x i32], i8*, i8*, [5 x i8*] }*)
+declare i8* @llvm.frameaddress(i32)
+declare i8* @llvm.stacksave()
+declare void @llvm.stackrestore(i8*)
+declare i32 @llvm.eh.sjlj.setjmp(i8*)
+declare i8* @llvm.eh.sjlj.lsda()
+declare void @llvm.eh.sjlj.callsite(i32)
+declare void @llvm.eh.sjlj.functioncontext(i8*)