XCore handling of thread local lowering

Fix XCoreLowerThreadLocal trying to initialise globals
which have no initializer.

Add handling of const expressions containing thread local variables.
These need to be replaced with instructions, as the thread ID is
used to access the thread local variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190300 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 149 insertions, 25 deletions

diff --git a/lib/Target/XCore/XCoreLowerThreadLocal.cpp b/lib/Target/XCore/XCoreLowerThreadLocal.cpp
index 2e328b4..5a83282 100644
--- a/lib/Target/XCore/XCoreLowerThreadLocal.cpp
+++ b/lib/Target/XCore/XCoreLowerThreadLocal.cpp
@@ -22,6 +22,8 @@
 #include "llvm/IR/Module.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/NoFolder.h"
+#include "llvm/Support/ValueHandle.h"
 
 #define DEBUG_TYPE "xcore-lower-thread-local"
 
@@ -71,13 +73,94 @@ createLoweredInitializer(ArrayType *NewType, Constant *OriginalInitializer) {
   return ConstantArray::get(NewType, Elements);
 }
 
-static bool hasNonInstructionUse(GlobalVariable *GV) {
-  for (Value::use_iterator UI = GV->use_begin(), E = GV->use_end(); UI != E;
-       ++UI)
-    if (!isa<Instruction>(*UI))
-      return true;
+static Instruction *
+createReplacementInstr(ConstantExpr *CE, Instruction *Instr) {
+  IRBuilder<true,NoFolder> Builder(Instr);
+  unsigned OpCode = CE->getOpcode();
+  switch (OpCode) {
+    case Instruction::GetElementPtr: {
+      SmallVector<Value *,4> CEOpVec(CE->op_begin(), CE->op_end());
+      ArrayRef<Value *> CEOps(CEOpVec);
+      return dyn_cast<Instruction>(Builder.CreateInBoundsGEP(CEOps[0],
+                                                             CEOps.slice(1)));
+    }
+    case Instruction::Add:
+    case Instruction::Sub:
+    case Instruction::Mul:
+    case Instruction::UDiv:
+    case Instruction::SDiv:
+    case Instruction::FDiv:
+    case Instruction::URem:
+    case Instruction::SRem:
+    case Instruction::FRem:
+    case Instruction::Shl:
+    case Instruction::LShr:
+    case Instruction::AShr:
+    case Instruction::And:
+    case Instruction::Or:
+    case Instruction::Xor:
+      return dyn_cast<Instruction>(
+                  Builder.CreateBinOp((Instruction::BinaryOps)OpCode,
+                                      CE->getOperand(0), CE->getOperand(1),
+                                      CE->getName()));
+    case Instruction::Trunc:
+    case Instruction::ZExt:
+    case Instruction::SExt:
+    case Instruction::FPToUI:
+    case Instruction::FPToSI:
+    case Instruction::UIToFP:
+    case Instruction::SIToFP:
+    case Instruction::FPTrunc:
+    case Instruction::FPExt:
+    case Instruction::PtrToInt:
+    case Instruction::IntToPtr:
+    case Instruction::BitCast:
+      return dyn_cast<Instruction>(
+                  Builder.CreateCast((Instruction::CastOps)OpCode,
+                                     CE->getOperand(0), CE->getType(),
+                                     CE->getName()));
+    default:
+      assert(0 && "Unhandled constant expression!\n");
+  }
+}
+
+static bool replaceConstantExprOp(ConstantExpr *CE) {
+  do {
+    SmallVector<WeakVH,8> WUsers;
+    for (Value::use_iterator I = CE->use_begin(), E = CE->use_end();
+         I != E; ++I)
+      WUsers.push_back(WeakVH(*I));
+    while (!WUsers.empty())
+      if (WeakVH WU = WUsers.pop_back_val()) {
+        if (Instruction *Instr = dyn_cast<Instruction>(WU)) {
+          Instruction *NewInst = createReplacementInstr(CE, Instr);
+          assert(NewInst && "Must build an instruction\n");
+          // If NewInst uses a CE being handled in an earlier recursion the
+          // earlier recursion's do-while-hasNUsesOrMore(1) will run again.
+          Instr->replaceUsesOfWith(CE, NewInst);
+        } else {
+          ConstantExpr *CExpr = dyn_cast<ConstantExpr>(WU);
+          if (!CExpr || !replaceConstantExprOp(CExpr))
+            return false;
+        }
+      }
+  } while (CE->hasNUsesOrMore(1)); // Does a recursion's NewInst use CE?
+  CE->destroyConstant();
+  return true;
+}
 
-  return false;
+static bool rewriteNonInstructionUses(GlobalVariable *GV) {
+  SmallVector<WeakVH,8> WUsers;
+  for (Value::use_iterator I = GV->use_begin(), E = GV->use_end(); I != E; ++I)
+    if (!isa<Instruction>(*I))
+      WUsers.push_back(WeakVH(*I));
+  while (!WUsers.empty())
+    if (WeakVH WU = WUsers.pop_back_val()) {
+      ConstantExpr *CE = dyn_cast<ConstantExpr>(WU);
+      if (!CE || !replaceConstantExprOp(CE))
+        return false;
+    }
+  return true;
 }
 
 static bool isZeroLengthArray(Type *Ty) {
@@ -92,14 +175,16 @@ bool XCoreLowerThreadLocal::lowerGlobal(GlobalVariable *GV) {
     return false;
 
   // Skip globals that we can't lower and leave it for the backend to error.
-  if (hasNonInstructionUse(GV) ||
+  if (!rewriteNonInstructionUses(GV) ||
       !GV->getType()->isSized() || isZeroLengthArray(GV->getType()))
     return false;
 
   // Create replacement global.
   ArrayType *NewType = createLoweredType(GV->getType()->getElementType());
-  Constant *NewInitializer = createLoweredInitializer(NewType,
-                                                      GV->getInitializer());
+  Constant *NewInitializer = 0;
+  if (GV->hasInitializer())
+    NewInitializer = createLoweredInitializer(NewType,
+                                              GV->getInitializer());
   GlobalVariable *NewGV =
     new GlobalVariable(*M, NewType, GV->isConstant(), GV->getLinkage(),
                        NewInitializer, "", 0, GlobalVariable::NotThreadLocal,
diff --git a/test/CodeGen/XCore/threads.ll b/test/CodeGen/XCore/threads.ll
index 5840e77..b56b16b 100644
--- a/test/CodeGen/XCore/threads.ll
+++ b/test/CodeGen/XCore/threads.ll
@@ -13,55 +13,94 @@ declare void @llvm.xcore.initdp.p1i8(i8 addrspace(1)* %r, i8* %value)
 define i8 addrspace(1)* @test_getst(i8 addrspace(1)* %r) {
 ; CHECK-LABEL: test_getst:
 ; CHECK: getst r0, res[r0]
-        %result = call i8 addrspace(1)* @llvm.xcore.getst.p1i8.p1i8(i8 addrspace(1)* %r)
-        ret i8 addrspace(1)* %result
+  %result = call i8 addrspace(1)* @llvm.xcore.getst.p1i8.p1i8(i8 addrspace(1)* %r)
+  ret i8 addrspace(1)* %result
 }
 
 define void @test_ssync() {
 ; CHECK-LABEL: test_ssync:
 ; CHECK: ssync
-	call void @llvm.xcore.ssync()
-	ret void
+  call void @llvm.xcore.ssync()
+  ret void
 }
 
 define void @test_mjoin(i8 addrspace(1)* %r) {
 ; CHECK-LABEL: test_mjoin:
 ; CHECK: mjoin res[r0]
-	call void @llvm.xcore.mjoin.p1i8(i8 addrspace(1)* %r)
-	ret void
+  call void @llvm.xcore.mjoin.p1i8(i8 addrspace(1)* %r)
+  ret void
 }
 
 define void @test_initsp(i8 addrspace(1)* %t, i8* %src) {
 ; CHECK-LABEL: test_initsp:
 ; CHECK: init t[r0]:sp, r1
-        call void @llvm.xcore.initsp.p1i8(i8 addrspace(1)* %t, i8* %src)
-        ret void
+  call void @llvm.xcore.initsp.p1i8(i8 addrspace(1)* %t, i8* %src)
+  ret void
 }
 
 define void @test_initpc(i8 addrspace(1)* %t, i8* %src) {
 ; CHECK-LABEL: test_initpc:
 ; CHECK: init t[r0]:pc, r1
-        call void @llvm.xcore.initpc.p1i8(i8 addrspace(1)* %t, i8* %src)
-        ret void
+  call void @llvm.xcore.initpc.p1i8(i8 addrspace(1)* %t, i8* %src)
+  ret void
 }
 
 define void @test_initlr(i8 addrspace(1)* %t, i8* %src) {
 ; CHECK-LABEL: test_initlr:
 ; CHECK: init t[r0]:lr, r1
-        call void @llvm.xcore.initlr.p1i8(i8 addrspace(1)* %t, i8* %src)
-        ret void
+  call void @llvm.xcore.initlr.p1i8(i8 addrspace(1)* %t, i8* %src)
+  ret void
 }
 
 define void @test_initcp(i8 addrspace(1)* %t, i8* %src) {
 ; CHECK-LABEL: test_initcp:
 ; CHECK: init t[r0]:cp, r1
-        call void @llvm.xcore.initcp.p1i8(i8 addrspace(1)* %t, i8* %src)
-        ret void
+  call void @llvm.xcore.initcp.p1i8(i8 addrspace(1)* %t, i8* %src)
+  ret void
 }
 
 define void @test_initdp(i8 addrspace(1)* %t, i8* %src) {
 ; CHECK-LABEL: test_initdp:
 ; CHECK: init t[r0]:dp, r1
-        call void @llvm.xcore.initdp.p1i8(i8 addrspace(1)* %t, i8* %src)
-        ret void
+  call void @llvm.xcore.initdp.p1i8(i8 addrspace(1)* %t, i8* %src)
+  ret void
 }
+
+@tl = thread_local global [3 x i32] zeroinitializer
+@tle = external thread_local global [2 x i32]
+
+define i32* @f_tl() {
+; CHECK-LABEL: f_tl:
+; CHECK: get r11, id
+; CHECK: ldaw [[R0:r[0-9]]], dp[tl]
+; CHECK: ldc [[R1:r[0-9]]], 8
+; CHECK: ldc [[R2:r[0-9]]], 12
+; r0 = id*12 + 8 + &tl
+; CHECK: lmul {{r[0-9]}}, r0, r11, [[R2]], [[R0]], [[R1]]
+  ret i32* getelementptr inbounds ([3 x i32]* @tl, i32 0, i32 2)
+}
+
+define i32* @f_tle() {
+; CHECK-LABEL: f_tle:
+; CHECK: get r11, id
+; CHECK: shl [[R0:r[0-9]]], r11, 3
+; CHECK: ldaw [[R1:r[0-9]]], dp[tle]
+; r0 = &tl + id*8
+; CHECK: add r0, [[R1]], [[R0]]
+  ret i32* getelementptr inbounds ([2 x i32]* @tle, i32 0, i32 0)
+}
+
+define i32 @f_tlExpr () {
+; CHECK-LABEL: f_tlExpr:
+; CHECK: get r11, id
+; CHECK: shl [[R0:r[0-9]]], r11, 3
+; CHECK: ldaw [[R1:r[0-9]]], dp[tle]
+; CHECK: add [[R2:r[0-9]]], [[R1]], [[R0]]
+; CHECK: add r0, [[R2]], [[R2]]
+  ret i32 add(
+      i32 ptrtoint( i32* getelementptr inbounds ([2 x i32]* @tle, i32 0, i32 0) to i32),
+      i32 ptrtoint( i32* getelementptr inbounds ([2 x i32]* @tle, i32 0, i32 0) to i32))
+}
+
+; CHECK-LABEL: tl:
+; CHECK: .space  96