2 files changed, 178 insertions, 25 deletions

diff --git a/lib/Target/Sparc/SparcCallingConv.td b/lib/Target/Sparc/SparcCallingConv.td
index b38ac616dc..54784e0188 100644
--- a/lib/Target/Sparc/SparcCallingConv.td
+++ b/lib/Target/Sparc/SparcCallingConv.td
@@ -12,25 +12,9 @@
 //===----------------------------------------------------------------------===//
 
 //===----------------------------------------------------------------------===//
-// Return Value Calling Conventions
+// SPARC v8 32-bit.
 //===----------------------------------------------------------------------===//
 
-// Sparc 32-bit C return-value convention.
-def RetCC_Sparc32 : CallingConv<[
-  CCIfType<[i32], CCAssignToReg<[I0, I1, I2, I3, I4, I5]>>,
-  CCIfType<[f32], CCAssignToReg<[F0, F1, F2, F3]>>,
-  CCIfType<[f64], CCAssignToReg<[D0, D1]>>
-]>;
-
-// Sparc 64-bit C return-value convention.
-def RetCC_Sparc64 : CallingConv<[
-  CCIfType<[i32], CCPromoteToType<i64>>,
-  CCIfType<[i64], CCAssignToReg<[I0, I1, I2, I3, I4, I5]>>,
-  CCIfType<[f32], CCAssignToReg<[F0, F1, F2, F3]>>,
-  CCIfType<[f64], CCAssignToReg<[D0, D1]>>
-]>;
-
-// Sparc 32-bit C Calling convention.
 def CC_Sparc32 : CallingConv<[
   //Custom assign SRet to [sp+64].
   CCIfSRet<CCCustom<"CC_Sparc_Assign_SRet">>,
@@ -43,14 +27,93 @@ def CC_Sparc32 : CallingConv<[
   CCAssignToStack<4, 4>
 ]>;
 
-// Sparc 64-bit C Calling convention.
+def RetCC_Sparc32 : CallingConv<[
+  CCIfType<[i32], CCAssignToReg<[I0, I1, I2, I3, I4, I5]>>,
+  CCIfType<[f32], CCAssignToReg<[F0, F1, F2, F3]>>,
+  CCIfType<[f64], CCAssignToReg<[D0, D1]>>
+]>;
+
+
+//===----------------------------------------------------------------------===//
+// SPARC v9 64-bit.
+//===----------------------------------------------------------------------===//
+//
+// The 64-bit ABI conceptually assigns all function arguments to a parameter
+// array starting at [%fp+BIAS+128] in the callee's stack frame. All arguments
+// occupy a multiple of 8 bytes in the array. Integer arguments are extended to
+// 64 bits by the caller. Floats are right-aligned in their 8-byte slot, the
+// first 4 bytes in the slot are undefined.
+//
+// The integer registers %i0 to %i5 shadow the first 48 bytes of the parameter
+// array at fixed offsets. Integer arguments are promoted to registers when
+// possible.
+//
+// The floating point registers %f0 to %f31 shadow the first 128 bytes of the
+// parameter array at fixed offsets. Float and double parameters are promoted
+// to these registers when possible.
+//
+// Structs up to 16 bytes in size are passed by value. They are right-aligned
+// in one or two 8-byte slots in the parameter array. Struct members are
+// promoted to both floating point and integer registers when possible. A
+// struct containing two floats would thus be passed in %f0 and %f1, while two
+// float function arguments would occupy 8 bytes each, and be passed in %f1 and
+// %f3.
+//
+// When a struct { int, float } is passed by value, the int goes in the high
+// bits of an integer register while the float goes in a floating point
+// register.
+//
+// The difference is encoded in LLVM IR using the inreg atttribute on function
+// arguments:
+//
+//   C:   void f(float, float);
+//   IR:  declare void f(float %f1, float %f3)
+//
+//   C:   void f(struct { float f0, f1; });
+//   IR:  declare void f(float inreg %f0, float inreg %f1)
+//
+//   C:   void f(int, float);
+//   IR:  declare void f(int signext %i0, float %f3)
+//
+//   C:   void f(struct { int i0high; float f1; });
+//   IR:  declare void f(i32 inreg %i0high, float inreg %f1)
+//
+// Two ints in a struct are simply coerced to i64:
+//
+//   C:   void f(struct { int i0high, i0low; });
+//   IR:  declare void f(i64 %i0.coerced)
+//
+// The frontend and backend divide the task of producing ABI compliant code for
+// C functions. The C frontend will:
+//
+//  - Annotate integer arguments with zeroext or signext attributes.
+//
+//  - Split structs into one or two 64-bit sized chunks, or 32-bit chunks with
+//    inreg attributes.
+//
+//  - Pass structs larger than 16 bytes indirectly with an explicit pointer
+//    argument. The byval attribute is not used.
+//
+// The backend will:
+//
+//  - Assign all arguments to 64-bit aligned stack slots, 32-bits for inreg.
+//
+//  - Promote to integer or floating point registers depending on type.
+//
+// Function return values are passed exactly like function arguments, except a
+// struct up to 32 bytes in size can be returned in registers.
+
+// Function arguments AND return values.
 def CC_Sparc64 : CallingConv<[
+  // The frontend uses the inreg flag to indicate i32 and float arguments from
+  // structs. These arguments are not promoted to 64 bits, but they can still
+  // be assigned to integer and float registers.
+  CCIfInReg<CCIfType<[i32, f32], CCCustom<"CC_Sparc64_Half">>>,
+
   // All integers are promoted to i64 by the caller.
   CCIfType<[i32], CCPromoteToType<i64>>,
-  // Integer arguments get passed in integer registers if there is space.
-  CCIfType<[i64], CCAssignToReg<[I0, I1, I2, I3, I4, I5]>>,
-  // FIXME: Floating point arguments.
 
-  // Alternatively, they are assigned to the stack in 8-byte aligned units.
-  CCAssignToStack<8, 8>
+  // Custom assignment is required because stack space is reserved for all
+  // arguments whether they are passed in registers or not.
+  CCCustom<"CC_Sparc64_Full">
 ]>;
diff --git a/lib/Target/Sparc/SparcISelLowering.cpp b/lib/Target/Sparc/SparcISelLowering.cpp
index 325f13424b..48c0857f3e 100644
--- a/lib/Target/Sparc/SparcISelLowering.cpp
+++ b/lib/Target/Sparc/SparcISelLowering.cpp
@@ -74,6 +74,78 @@ static bool CC_Sparc_Assign_f64(unsigned &ValNo, MVT &ValVT,
   return true;
 }
 
+// Allocate a full-sized argument for the 64-bit ABI.
+static bool CC_Sparc64_Full(unsigned &ValNo, MVT &ValVT,
+                            MVT &LocVT, CCValAssign::LocInfo &LocInfo,
+                            ISD::ArgFlagsTy &ArgFlags, CCState &State) {
+  assert((LocVT == MVT::f32 || LocVT.getSizeInBits() == 64) &&
+         "Can't handle non-64 bits locations");
+
+  // Stack space is allocated for all arguments starting from [%fp+BIAS+128].
+  unsigned Offset = State.AllocateStack(8, 8);
+  unsigned Reg = 0;
+
+  if (LocVT == MVT::i64 && Offset < 6*8)
+    // Promote integers to %i0-%i5.
+    Reg = SP::I0 + Offset/8;
+  else if (LocVT == MVT::f64 && Offset < 16*8)
+    // Promote doubles to %d0-%d30. (Which LLVM calls D0-D15).
+    Reg = SP::D0 + Offset/8;
+  else if (LocVT == MVT::f32 && Offset < 16*8)
+    // Promote floats to %f1, %f3, ...
+    Reg = SP::F1 + Offset/4;
+
+  // Promote to register when possible, otherwise use the stack slot.
+  if (Reg) {
+    State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
+    return true;
+  }
+
+  // This argument goes on the stack in an 8-byte slot.
+  // When passing floats, LocVT is smaller than 8 bytes. Adjust the offset to
+  // the right-aligned float. The first 4 bytes of the stack slot are undefined.
+  if (LocVT == MVT::f32)
+    Offset += 4;
+
+  State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
+  return true;
+}
+
+// Allocate a half-sized argument for the 64-bit ABI.
+//
+// This is used when passing { float, int } structs by value in registers.
+static bool CC_Sparc64_Half(unsigned &ValNo, MVT &ValVT,
+                            MVT &LocVT, CCValAssign::LocInfo &LocInfo,
+                            ISD::ArgFlagsTy &ArgFlags, CCState &State) {
+  assert(LocVT.getSizeInBits() == 32 && "Can't handle non-32 bits locations");
+  unsigned Offset = State.AllocateStack(4, 4);
+
+  if (LocVT == MVT::f32 && Offset < 16*8) {
+    // Promote floats to %f0-%f31.
+    State.addLoc(CCValAssign::getReg(ValNo, ValVT, SP::F0 + Offset/4,
+                                     LocVT, LocInfo));
+    return true;
+  }
+
+  if (LocVT == MVT::i32 && Offset < 6*8) {
+    // Promote integers to %i0-%i5, using half the register.
+    unsigned Reg = SP::I0 + Offset/8;
+    LocVT = MVT::i64;
+    LocInfo = CCValAssign::AExt;
+
+    // Set the Custom bit if this i32 goes in the high bits of a register.
+    if (Offset % 8 == 0)
+      State.addLoc(CCValAssign::getCustomReg(ValNo, ValVT, Reg,
+                                             LocVT, LocInfo));
+    else
+      State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
+    return true;
+  }
+
+  State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
+  return true;
+}
+
 #include "SparcGenCallingConv.inc"
 
 SDValue
@@ -93,8 +165,7 @@ SparcTargetLowering::LowerReturn(SDValue Chain,
                  DAG.getTarget(), RVLocs, *DAG.getContext());
 
   // Analize return values.
-  CCInfo.AnalyzeReturn(Outs, Subtarget->is64Bit() ?
-                             RetCC_Sparc64 : RetCC_Sparc32);
+  CCInfo.AnalyzeReturn(Outs, Subtarget->is64Bit() ? CC_Sparc64 : RetCC_Sparc32);
 
   SDValue Flag;
   SmallVector<SDValue, 4> RetOps(1, Chain);
@@ -384,6 +455,11 @@ LowerFormalArguments_64(SDValue Chain,
                                    getRegClassFor(VA.getLocVT()));
       SDValue Arg = DAG.getCopyFromReg(Chain, DL, VReg, VA.getLocVT());
 
+      // Get the high bits for i32 struct elements.
+      if (VA.getValVT() == MVT::i32 && VA.needsCustom())
+        Arg = DAG.getNode(ISD::SRL, DL, VA.getLocVT(), Arg,
+                          DAG.getConstant(32, MVT::i32));
+
       // The caller promoted the argument, so insert an Assert?ext SDNode so we
       // won't promote the value again in this function.
       switch (VA.getLocInfo()) {
@@ -409,6 +485,20 @@ LowerFormalArguments_64(SDValue Chain,
 
     // The registers are exhausted. This argument was passed on the stack.
     assert(VA.isMemLoc());
+    // The CC_Sparc64_Full/Half functions compute stack offsets relative to the
+    // beginning of the arguments area at %fp+BIAS+128.
+    unsigned Offset = VA.getLocMemOffset() + 128;
+    unsigned ValSize = VA.getValVT().getSizeInBits() / 8;
+    // Adjust offset for extended arguments, SPARC is big-endian.
+    // The caller will have written the full slot with extended bytes, but we
+    // prefer our own extending loads.
+    if (VA.isExtInLoc())
+      Offset += 8 - ValSize;
+    int FI = MF.getFrameInfo()->CreateFixedObject(ValSize, Offset, true);
+    InVals.push_back(DAG.getLoad(VA.getValVT(), DL, Chain,
+                                 DAG.getFrameIndex(FI, getPointerTy()),
+                                 MachinePointerInfo::getFixedStack(FI),
+                                 false, false, false, 0));
   }
   return Chain;
 }