AMDGPU: Stop using CCAssignToRegWithShadow

[arsenm]

Summary

This does not do what it is attempting to use it for
and requires working around in LowerFormalArguments.

[arsenm]

Refactor slightly

[arsenm]

ping

[arsenm]

ping

[mareko]

LGTM.

[arsenm]

r299667

lib/Target/AMDGPU/AMDGPUCallingConv.td

def CC_SI : CallingConv<[
   CCIfInReg<CCIfType<[f32, i32, f16] , CCAssignToReg<[
     SGPR0, SGPR1, SGPR2, SGPR3, SGPR4, SGPR5, SGPR6, SGPR7,
     SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15,
     SGPR16, SGPR17, SGPR18, SGPR19, SGPR20, SGPR21, SGPR22, SGPR23,
     SGPR24, SGPR25, SGPR26, SGPR27, SGPR28, SGPR29, SGPR30, SGPR31,
     SGPR32, SGPR33, SGPR34, SGPR35, SGPR36, SGPR37, SGPR38, SGPR39
   ]>>>,
 
-  CCIfInReg<CCIfType<[i64] , CCAssignToRegWithShadow<
-    [ SGPR0, SGPR2, SGPR4, SGPR6, SGPR8, SGPR10, SGPR12, SGPR14,
-      SGPR16, SGPR18, SGPR20, SGPR22, SGPR24, SGPR26, SGPR28, SGPR30,
-      SGPR32, SGPR34, SGPR36, SGPR38 ],
-    [ SGPR1, SGPR3, SGPR5, SGPR7, SGPR9, SGPR11, SGPR13, SGPR15,
-      SGPR17, SGPR19, SGPR21, SGPR23, SGPR25, SGPR27, SGPR29, SGPR31,
-      SGPR33, SGPR35, SGPR37, SGPR39 ]
-  >>>,
+  // We have no way of referring to the generated register tuples
+  // here, so use a custom function.
+  CCIfInReg<CCIfType<[i64], CCCustom<"allocateSGPRTuple">>>,
+  CCIfByVal<CCIfType<[i64], CCCustom<"allocateSGPRTuple">>>,
 
   // 32*4 + 4 is the minimum for a fetch shader consumer with 32 inputs.
   CCIfNotInReg<CCIfType<[f32, i32, f16] , CCAssignToReg<[
     VGPR0, VGPR1, VGPR2, VGPR3, VGPR4, VGPR5, VGPR6, VGPR7,
     VGPR8, VGPR9, VGPR10, VGPR11, VGPR12, VGPR13, VGPR14, VGPR15,
     VGPR16, VGPR17, VGPR18, VGPR19, VGPR20, VGPR21, VGPR22, VGPR23,
     VGPR24, VGPR25, VGPR26, VGPR27, VGPR28, VGPR29, VGPR30, VGPR31,
     VGPR32, VGPR33, VGPR34, VGPR35, VGPR36, VGPR37, VGPR38, VGPR39,
     VGPR40, VGPR41, VGPR42, VGPR43, VGPR44, VGPR45, VGPR46, VGPR47,
     VGPR48, VGPR49, VGPR50, VGPR51, VGPR52, VGPR53, VGPR54, VGPR55,
     VGPR56, VGPR57, VGPR58, VGPR59, VGPR60, VGPR61, VGPR62, VGPR63,
     VGPR64, VGPR65, VGPR66, VGPR67, VGPR68, VGPR69, VGPR70, VGPR71,
     VGPR72, VGPR73, VGPR74, VGPR75, VGPR76, VGPR77, VGPR78, VGPR79,
     VGPR80, VGPR81, VGPR82, VGPR83, VGPR84, VGPR85, VGPR86, VGPR87,
     VGPR88, VGPR89, VGPR90, VGPR91, VGPR92, VGPR93, VGPR94, VGPR95,
     VGPR96, VGPR97, VGPR98, VGPR99, VGPR100, VGPR101, VGPR102, VGPR103,
     VGPR104, VGPR105, VGPR106, VGPR107, VGPR108, VGPR109, VGPR110, VGPR111,
     VGPR112, VGPR113, VGPR114, VGPR115, VGPR116, VGPR117, VGPR118, VGPR119,
     VGPR120, VGPR121, VGPR122, VGPR123, VGPR124, VGPR125, VGPR126, VGPR127,
     VGPR128, VGPR129, VGPR130, VGPR131, VGPR132, VGPR133, VGPR134, VGPR135
-  ]>>>,
-
-  CCIfByVal<CCIfType<[i64] , CCAssignToRegWithShadow<
-    [ SGPR0, SGPR2, SGPR4, SGPR6, SGPR8, SGPR10, SGPR12, SGPR14,
-      SGPR16, SGPR18, SGPR20, SGPR22, SGPR24, SGPR26, SGPR28, SGPR30,
-      SGPR32, SGPR34, SGPR36, SGPR38 ],
-    [ SGPR1, SGPR3, SGPR5, SGPR7, SGPR9, SGPR11, SGPR13, SGPR15,
-      SGPR17, SGPR19, SGPR21, SGPR23, SGPR25, SGPR27, SGPR29, SGPR31,
-      SGPR33, SGPR35, SGPR37, SGPR39 ]
-  >>>
-
+  ]>>>
 ]>;
 
 def RetCC_SI : CallingConv<[
   CCIfType<[i32] , CCAssignToReg<[
     SGPR0, SGPR1, SGPR2, SGPR3, SGPR4, SGPR5, SGPR6, SGPR7,
     SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15,
     SGPR16, SGPR17, SGPR18, SGPR19, SGPR20, SGPR21, SGPR22, SGPR23,
     SGPR24, SGPR25, SGPR26, SGPR27, SGPR28, SGPR29, SGPR30, SGPR31,

lib/Target/AMDGPU/AMDGPUISelLowering.cpp

static bool allocateKernArg(unsigned ValNo, MVT ValVT, MVT LocVT,
   AMDGPUMachineFunction *MFI = MF.getInfo<AMDGPUMachineFunction>();
 
   uint64_t Offset = MFI->allocateKernArg(LocVT.getStoreSize(),
                                          ArgFlags.getOrigAlign());
   State.addLoc(CCValAssign::getCustomMem(ValNo, ValVT, Offset, LocVT, LocInfo));
   return true;
 }
 
+static bool allocateCCRegs(unsigned ValNo, MVT ValVT, MVT LocVT,
+                           CCValAssign::LocInfo LocInfo,
+                           ISD::ArgFlagsTy ArgFlags, CCState &State,
+                           const TargetRegisterClass *RC,
+                           unsigned NumRegs) {
+  ArrayRef<MCPhysReg> RegList = makeArrayRef(RC->begin(), NumRegs);
+  unsigned RegResult = State.AllocateReg(RegList);
+  if (!RegResult)
+    return false;
+
+  State.addLoc(CCValAssign::getReg(ValNo, ValVT, RegResult, LocVT, LocInfo));
+  return true;
+}
+
+static bool allocateSGPRTuple(unsigned ValNo, MVT ValVT, MVT LocVT,
+                              CCValAssign::LocInfo LocInfo,
+                              ISD::ArgFlagsTy ArgFlags, CCState &State) {
+  switch (LocVT.SimpleTy) {
+  case MVT::i64:
+  case MVT::f64:
+  case MVT::v2i32:
+  case MVT::v2f32: {
+    // Up to SGPR0-SGPR39
+    return allocateCCRegs(ValNo, ValVT, LocVT, LocInfo, ArgFlags, State,
+                          &AMDGPU::SGPR_64RegClass, 20);
+  }
+  default:
+    return false;
+  }
+}
+
 #include "AMDGPUGenCallingConv.inc"
 
 // Find a larger type to do a load / store of a vector with.
 EVT AMDGPUTargetLowering::getEquivalentMemType(LLVMContext &Ctx, EVT VT) {
   unsigned StoreSize = VT.getStoreSizeInBits();
   if (StoreSize <= 32)
     return EVT::getIntegerVT(Ctx, StoreSize);
 

lib/Target/AMDGPU/SIISelLowering.cpp

if (VA.isMemLoc()) {
 
       InVals.push_back(Arg);
       Info->setABIArgOffset(Offset + MemVT.getStoreSize());
       continue;
     }
     assert(VA.isRegLoc() && "Parameter must be in a register!");
 
     unsigned Reg = VA.getLocReg();
-
-    if (VT == MVT::i64) {
-      // For now assume it is a pointer
-      Reg = TRI->getMatchingSuperReg(Reg, AMDGPU::sub0,
-                                     &AMDGPU::SGPR_64RegClass);
-      Reg = MF.addLiveIn(Reg, &AMDGPU::SGPR_64RegClass);
-      SDValue Copy = DAG.getCopyFromReg(Chain, DL, Reg, VT);
-      InVals.push_back(Copy);
-      continue;
-    }
-
     const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg, VT);
 
     Reg = MF.addLiveIn(Reg, RC);
     SDValue Val = DAG.getCopyFromReg(Chain, DL, Reg, VT);
 
     if (Arg.VT.isVector()) {
       // Build a vector from the registers
       Type *ParamType = FType->getParamType(Arg.getOrigArgIndex());