From f93099eb1ca6579eaea865c5f67dc0935d78e719 Mon Sep 17 00:00:00 2001
From: Reid Kleckner <reid@kleckner.net>
Date: Tue, 2 Sep 2014 18:42:44 +0000
Subject: [PATCH] CodeGen: Handle va_start in the entry block

Also fix a small copy-paste bug in X86ISelLowering where Chain should
have been used in place of DAG.getEntryToken().

Fixes PR20828.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@216929 91177308-0d34-0410-b5e6-96231b3b80d8
---
 .../SelectionDAG/FunctionLoweringInfo.cpp     | 40 ++++++++-----------
 lib/Target/X86/X86ISelLowering.cpp            |  4 +-
 test/CodeGen/X86/win64_vararg.ll              | 19 +++++++++
 3 files changed, 37 insertions(+), 26 deletions(-)

diff --git a/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp b/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
index ff8d3ad1501..1f58d7c301f 100644
--- a/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
+++ b/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
@@ -75,34 +75,26 @@ void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf,
   // instruction values that are used outside of the block that defines
   // them.
   Function::const_iterator BB = Fn->begin(), EB = Fn->end();
-  for (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I)
-    if (const AllocaInst *AI = dyn_cast<AllocaInst>(I)) {
-      // Don't fold inalloca allocas or other dynamic allocas into the initial
-      // stack frame allocation, even if they are in the entry block.
-      if (!AI->isStaticAlloca())
-        continue;
-
-      if (const ConstantInt *CUI = dyn_cast<ConstantInt>(AI->getArraySize())) {
-        Type *Ty = AI->getAllocatedType();
-        uint64_t TySize = TLI->getDataLayout()->getTypeAllocSize(Ty);
-        unsigned Align =
-          std::max((unsigned)TLI->getDataLayout()->getPrefTypeAlignment(Ty),
-                   AI->getAlignment());
-
-        TySize *= CUI->getZExtValue();   // Get total allocated size.
-        if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
-
-        StaticAllocaMap[AI] =
-          MF->getFrameInfo()->CreateStackObject(TySize, Align, false, AI);
-      }
-    }
-
   for (; BB != EB; ++BB)
     for (BasicBlock::const_iterator I = BB->begin(), E = BB->end();
          I != E; ++I) {
-      // Look for dynamic allocas.
       if (const AllocaInst *AI = dyn_cast<AllocaInst>(I)) {
-        if (!AI->isStaticAlloca()) {
+        // Static allocas can be folded into the initial stack frame adjustment.
+        if (AI->isStaticAlloca()) {
+          const ConstantInt *CUI = cast<ConstantInt>(AI->getArraySize());
+          Type *Ty = AI->getAllocatedType();
+          uint64_t TySize = TLI->getDataLayout()->getTypeAllocSize(Ty);
+          unsigned Align =
+            std::max((unsigned)TLI->getDataLayout()->getPrefTypeAlignment(Ty),
+                     AI->getAlignment());
+
+          TySize *= CUI->getZExtValue();   // Get total allocated size.
+          if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
+
+          StaticAllocaMap[AI] =
+            MF->getFrameInfo()->CreateStackObject(TySize, Align, false, AI);
+
+        } else {
           unsigned Align = std::max(
               (unsigned)TLI->getDataLayout()->getPrefTypeAlignment(
                 AI->getAllocatedType()),
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 1c10691250c..43e4ffeb963 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -2547,11 +2547,11 @@ X86TargetLowering::LowerFormalArguments(SDValue Chain,
     for (MCPhysReg Reg : ArgGPRs.slice(NumIntRegs)) {
       unsigned GPR = MF.addLiveIn(Reg, &X86::GR64RegClass);
       LiveGPRs.push_back(
-          DAG.getCopyFromReg(DAG.getEntryNode(), dl, GPR, MVT::i64));
+          DAG.getCopyFromReg(Chain, dl, GPR, MVT::i64));
     }
     if (!ArgXMMs.empty()) {
       unsigned AL = MF.addLiveIn(X86::AL, &X86::GR8RegClass);
-      ALVal = DAG.getCopyFromReg(DAG.getEntryNode(), dl, AL, MVT::i8);
+      ALVal = DAG.getCopyFromReg(Chain, dl, AL, MVT::i8);
       for (MCPhysReg Reg : ArgXMMs.slice(NumXMMRegs)) {
         unsigned XMMReg = MF.addLiveIn(Reg, &X86::VR128RegClass);
         LiveXMMRegs.push_back(
diff --git a/test/CodeGen/X86/win64_vararg.ll b/test/CodeGen/X86/win64_vararg.ll
index 1a51b2a64a7..8d7f2010a54 100644
--- a/test/CodeGen/X86/win64_vararg.ll
+++ b/test/CodeGen/X86/win64_vararg.ll
@@ -111,3 +111,22 @@ entry:
   %tmp = va_arg i8** %ap, i32
   ret i32 %tmp
 }
+
+define void @sret_arg(i32* sret %agg.result, i8* nocapture readnone %format, ...) {
+entry:
+  %ap = alloca i8*
+  %ap_i8 = bitcast i8** %ap to i8*
+  call void @llvm.va_start(i8* %ap_i8)
+  %tmp = va_arg i8** %ap, i32
+  store i32 %tmp, i32* %agg.result
+  ret void
+}
+; CHECK-LABEL: sret_arg:
+; CHECK: pushq
+; CHECK-DAG: movq %r9, 40(%rsp)
+; CHECK-DAG: movq %r8, 32(%rsp)
+; CHECK: movl 32(%rsp), %[[tmp:[^ ]*]]
+; CHECK: movl %[[tmp]], (%[[sret:[^ ]*]])
+; CHECK: movq %[[sret]], %rax
+; CHECK: popq
+; CHECK: retq
-- 
2.34.1