In this patch, we teach X86_64TargetMachine that it has a ILP32

(defined by the x32 ABI) mode, in which case its pointers are 32-bits
in size. This knowledge is also added to X86RegisterInfo that now
returns the appropriate registers in getPointerRegClass.

There are many outcomes to this change. In order to keep the patches
separate and manageable, we start by focusing on some simple testable
cases. The patch adds a test with passing a pointer to a function -
focusing on the difference between the two data models for x86-64.
Another test is added for handling of 'sret' arguments (and
functionality is added in X86ISelLowering to make it work).

A note on naming: the "x32 ABI" document refers to the AMD64
architecture (in LLVM it's distinguished by being is64Bits() in the
x86 subtarget) with two variations: the LP64 (default) data model, and
the ILP32 data model. This patch adds predicates to the subtarget
which are consistent with this naming scheme.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173503 91177308-0d34-0410-b5e6-96231b3b80d8

6 files changed, 82 insertions, 22 deletions

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 04758146e5..0c12410361 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1646,10 +1646,10 @@ X86TargetLowering::LowerReturn(SDValue Chain,
     Flag = Chain.getValue(1);
   }
 
-  // The x86-64 ABI for returning structs by value requires that we copy
-  // the sret argument into %rax for the return. We saved the argument into
-  // a virtual register in the entry block, so now we copy the value out
-  // and into %rax.
+  // The x86-64 ABIs require that for returning structs by value we copy
+  // the sret argument into %rax/%eax (depending on ABI) for the return.
+  // We saved the argument into a virtual register in the entry block,
+  // so now we copy the value out and into %rax/%eax.
   if (Subtarget->is64Bit() &&
       DAG.getMachineFunction().getFunction()->hasStructRetAttr()) {
     MachineFunction &MF = DAG.getMachineFunction();
@@ -1659,11 +1659,12 @@ X86TargetLowering::LowerReturn(SDValue Chain,
            "SRetReturnReg should have been set in LowerFormalArguments().");
     SDValue Val = DAG.getCopyFromReg(Chain, dl, Reg, getPointerTy());
 
-    Chain = DAG.getCopyToReg(Chain, dl, X86::RAX, Val, Flag);
+    unsigned RetValReg = Subtarget->isTarget64BitILP32() ? X86::EAX : X86::RAX;
+    Chain = DAG.getCopyToReg(Chain, dl, RetValReg, Val, Flag);
     Flag = Chain.getValue(1);
 
-    // RAX now acts like a return value.
-    MRI.addLiveOut(X86::RAX);
+    // RAX/EAX now acts like a return value.
+    MRI.addLiveOut(RetValReg);
   }
 
   RetOps[0] = Chain;  // Update chain.
@@ -2015,14 +2016,16 @@ X86TargetLowering::LowerFormalArguments(SDValue Chain,
     InVals.push_back(ArgValue);
   }
 
-  // The x86-64 ABI for returning structs by value requires that we copy
-  // the sret argument into %rax for the return. Save the argument into
-  // a virtual register so that we can access it from the return points.
+  // The x86-64 ABIs require that for returning structs by value we copy
+  // the sret argument into %rax/%eax (depending on ABI) for the return.
+  // Save the argument into a virtual register so that we can access it
+  // from the return points.
   if (Is64Bit && MF.getFunction()->hasStructRetAttr()) {
     X86MachineFunctionInfo *FuncInfo = MF.getInfo<X86MachineFunctionInfo>();
     unsigned Reg = FuncInfo->getSRetReturnReg();
     if (!Reg) {
-      Reg = MF.getRegInfo().createVirtualRegister(getRegClassFor(MVT::i64));
+      MVT PtrTy = getPointerTy();
+      Reg = MF.getRegInfo().createVirtualRegister(getRegClassFor(PtrTy));
       FuncInfo->setSRetReturnReg(Reg);
     }
     SDValue Copy = DAG.getCopyToReg(DAG.getEntryNode(), dl, Reg, InVals[0]);
diff --git a/lib/Target/X86/X86RegisterInfo.cpp b/lib/Target/X86/X86RegisterInfo.cpp
index 58064b8695..1dd1e41af8 100644
--- a/lib/Target/X86/X86RegisterInfo.cpp
+++ b/lib/Target/X86/X86RegisterInfo.cpp
@@ -177,20 +177,21 @@ X86RegisterInfo::getLargestLegalSuperClass(const TargetRegisterClass *RC) const{
 const TargetRegisterClass *
 X86RegisterInfo::getPointerRegClass(const MachineFunction &MF, unsigned Kind)
                                                                          const {
+  const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
   switch (Kind) {
   default: llvm_unreachable("Unexpected Kind in getPointerRegClass!");
   case 0: // Normal GPRs.
-    if (TM.getSubtarget<X86Subtarget>().is64Bit())
+    if (Subtarget.isTarget64BitLP64())
       return &X86::GR64RegClass;
     return &X86::GR32RegClass;
   case 1: // Normal GPRs except the stack pointer (for encoding reasons).
-    if (TM.getSubtarget<X86Subtarget>().is64Bit())
+    if (Subtarget.isTarget64BitLP64())
       return &X86::GR64_NOSPRegClass;
     return &X86::GR32_NOSPRegClass;
   case 2: // Available for tailcall (not callee-saved GPRs).
-    if (TM.getSubtarget<X86Subtarget>().isTargetWin64())
+    if (Subtarget.isTargetWin64())
       return &X86::GR64_TCW64RegClass;
-    if (TM.getSubtarget<X86Subtarget>().is64Bit())
+    else if (Subtarget.is64Bit())
       return &X86::GR64_TCRegClass;
 
     const Function *F = MF.getFunction();
diff --git a/lib/Target/X86/X86Subtarget.h b/lib/Target/X86/X86Subtarget.h
index 080f4cfeca..b325f62126 100644
--- a/lib/Target/X86/X86Subtarget.h
+++ b/lib/Target/X86/X86Subtarget.h
@@ -194,7 +194,20 @@ public:
   /// instruction.
   void AutoDetectSubtargetFeatures();
 
-  bool is64Bit() const { return In64BitMode; }
+  /// Is this x86_64? (disregarding specific ABI / programming model)
+  bool is64Bit() const {
+    return In64BitMode;
+  }
+
+  /// Is this x86_64 with the ILP32 programming model (x32 ABI)?
+  bool isTarget64BitILP32() const {
+    return In64BitMode && (TargetTriple.getEnvironment() == Triple::GNUX32);
+  }
+
+  /// Is this x86_64 with the LP64 programming model (standard AMD64, no x32)?
+  bool isTarget64BitLP64() const {
+    return In64BitMode && (TargetTriple.getEnvironment() != Triple::GNUX32);
+  }
 
   PICStyles::Style getPICStyle() const { return PICStyle; }
   void setPICStyle(PICStyles::Style Style)  { PICStyle = Style; }
diff --git a/lib/Target/X86/X86TargetMachine.cpp b/lib/Target/X86/X86TargetMachine.cpp
index 799f140407..8aa58a2042 100644
--- a/lib/Target/X86/X86TargetMachine.cpp
+++ b/lib/Target/X86/X86TargetMachine.cpp
@@ -59,8 +59,12 @@ X86_64TargetMachine::X86_64TargetMachine(const Target &T, StringRef TT,
                                          Reloc::Model RM, CodeModel::Model CM,
                                          CodeGenOpt::Level OL)
   : X86TargetMachine(T, TT, CPU, FS, Options, RM, CM, OL, true),
-    DL("e-p:64:64-s:64-f64:64:64-i64:64:64-f80:128:128-f128:128:128-"
-               "n8:16:32:64-S128"),
+    // The x32 ABI dictates the ILP32 programming model for x64.
+    DL(getSubtargetImpl()->isTarget64BitILP32() ?
+        "e-p:32:32-s:64-f64:64:64-i64:64:64-f80:128:128-f128:128:128-"
+        "n8:16:32:64-S128" :
+        "e-p:64:64-s:64-f64:64:64-i64:64:64-f80:128:128-f128:128:128-"
+        "n8:16:32:64-S128"),
     InstrInfo(*this),
     TLInfo(*this),
     TSInfo(*this),
diff --git a/test/CodeGen/X86/x86-64-ptr-arg-simple.ll b/test/CodeGen/X86/x86-64-ptr-arg-simple.ll
new file mode 100644
index 0000000000..6d46663989
--- /dev/null
+++ b/test/CodeGen/X86/x86-64-ptr-arg-simple.ll
@@ -0,0 +1,29 @@
+; RUN: llc -mtriple=x86_64-pc-linux < %s | FileCheck %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnux32 < %s | FileCheck -check-prefix=X32ABI %s
+
+; %in is kept in %esi for both ABIs. But the pointer will be passed in %edi
+; for x32, not %rdi
+
+; CHECK: movl %esi, (%rdi)
+; X32ABI: movl %esi, (%edi)
+
+define void @foo(i32* nocapture %out, i32 %in) nounwind {
+entry:
+  store i32 %in, i32* %out, align 4
+  ret void
+}
+
+; CHECK: bar
+; CHECK: movl (%rsi), %eax
+
+; Similarly here, but for loading
+; X32ABI: bar
+; X32ABI: movl (%esi), %eax
+
+define void @bar(i32* nocapture %pOut, i32* nocapture %pIn) nounwind {
+entry:
+  %0 = load i32* %pIn, align 4
+  store i32 %0, i32* %pOut, align 4
+  ret void
+}
+
diff --git a/test/CodeGen/X86/x86-64-sret-return.ll b/test/CodeGen/X86/x86-64-sret-return.ll
index 7b5f189faa..bc8a543465 100644
--- a/test/CodeGen/X86/x86-64-sret-return.ll
+++ b/test/CodeGen/X86/x86-64-sret-return.ll
@@ -1,11 +1,16 @@
-; RUN: llc < %s | FileCheck %s
+; RUN: llc -mtriple=x86_64-apple-darwin8 < %s | FileCheck %s
+; RUN: llc -mtriple=x86_64-pc-linux < %s | FileCheck %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnux32 < %s | FileCheck -check-prefix=X32ABI %s
 
-target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
-target triple = "x86_64-apple-darwin8"
-	%struct.foo = type { [4 x i64] }
+%struct.foo = type { [4 x i64] }
 
 ; CHECK: bar:
 ; CHECK: movq %rdi, %rax
+
+; For the x32 ABI, pointers are 32-bit so 32-bit instructions will be used
+; X32ABI: bar:
+; X32ABI: movl %edi, %eax
+
 define void @bar(%struct.foo* noalias sret  %agg.result, %struct.foo* %d) nounwind  {
 entry:
 	%d_addr = alloca %struct.foo*		; <%struct.foo**> [#uses=2]
@@ -57,6 +62,11 @@ return:		; preds = %entry
 
 ; CHECK: foo:
 ; CHECK: movq %rdi, %rax
+
+; For the x32 ABI, pointers are 32-bit so 32-bit instructions will be used
+; X32ABI: foo:
+; X32ABI: movl %edi, %eax
+
 define void @foo({ i64 }* noalias nocapture sret %agg.result) nounwind {
   store { i64 } { i64 0 }, { i64 }* %agg.result
   ret void