AMDGPU: Cleanup SMRD buffer selection

[arsenm]

Summary

The usage of the Imm out argument from SelectSMRDOffset is pretty
confusing. Stop trying to reject CI immediates in the case where the
offset field can be used. It's not an illegal way to encode the
immediate, so just prefer the better encoding pattern with
AddedComplexity.

We probably don't even really need the different opcodes for the
different offset types anymore, but that will be more work to cleanup.

The SMRD non-buffer load patterns could also use a cleanup to be done
separately.

[rampitec]

LGTM

[arsenm]

75fcdfa1fcf647d89f2e82fcfc7e8f2f310aa7d0

llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp

bool AMDGPUDAGToDAGISel::SelectSMRDSgpr(SDValue Addr, SDValue &SBase,
                                         SDValue &Offset) const {
   bool Imm = false;
   return SelectSMRD(Addr, SBase, Offset, Imm) && !Imm &&
          !isa<ConstantSDNode>(Offset);
 }
 
 bool AMDGPUDAGToDAGISel::SelectSMRDBufferImm(SDValue Addr,
                                              SDValue &Offset) const {
-  bool Imm = false;
-  return SelectSMRDOffset(Addr, Offset, Imm) && Imm;
+  if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Addr)) {
+    if (auto Imm = AMDGPU::getSMRDEncodedOffset(*Subtarget,
+                                                C->getZExtValue())) {
+      Offset = CurDAG->getTargetConstant(*Imm, SDLoc(Addr), MVT::i32);
+      return true;
+    }
+  }
+
+  return false;
 }
 
 bool AMDGPUDAGToDAGISel::SelectSMRDBufferImm32(SDValue Addr,
                                                SDValue &Offset) const {
   assert(Subtarget->getGeneration() == AMDGPUSubtarget::SEA_ISLANDS);
 
-  bool Imm = false;
-  if (!SelectSMRDOffset(Addr, Offset, Imm))
-    return false;
+  if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Addr)) {
+    if (auto Imm = AMDGPU::getSMRDEncodedLiteralOffset32(*Subtarget,
+                                                         C->getZExtValue())) {
+      Offset = CurDAG->getTargetConstant(*Imm, SDLoc(Addr), MVT::i32);
+      return true;
+    }
+  }
 
-  return !Imm && isa<ConstantSDNode>(Offset);
+  return false;
 }
 
 bool AMDGPUDAGToDAGISel::SelectMOVRELOffset(SDValue Index,
                                             SDValue &Base,
                                             SDValue &Offset) const {
   SDLoc DL(Index);
 
   if (CurDAG->isBaseWithConstantOffset(Index)) {

llvm/lib/Target/AMDGPU/SMInstructions.td

multiclass SMRD_Pattern <string Instr, ValueType vt> {
   >;
 }
 
 multiclass SMLoad_Pattern <string Instr, ValueType vt> {
   // 1. Offset as an immediate
   def : GCNPat <
     (SIsbuffer_load v4i32:$sbase, (SMRDBufferImm i32:$offset), timm:$cachepolicy),
     (vt (!cast<SM_Pseudo>(Instr#"_IMM") $sbase, $offset, (extract_glc $cachepolicy),
-                                        (extract_dlc $cachepolicy)))
-  >;
+                                        (extract_dlc $cachepolicy)))> {
+    let AddedComplexity = 2;
+  }
 
   // 2. 32-bit IMM offset on CI
   def : GCNPat <
     (vt (SIsbuffer_load v4i32:$sbase, (SMRDBufferImm32 i32:$offset), timm:$cachepolicy)),
     (!cast<InstSI>(Instr#"_IMM_ci") $sbase, $offset, (extract_glc $cachepolicy), (extract_dlc $cachepolicy))> {
     let OtherPredicates = [isGFX7Only];
+    let AddedComplexity = 1;
   }
 
   // 3. Offset loaded in an 32bit SGPR
   def : GCNPat <
     (SIsbuffer_load v4i32:$sbase, i32:$offset, timm:$cachepolicy),
     (vt (!cast<SM_Pseudo>(Instr#"_SGPR") $sbase, $offset, (extract_glc $cachepolicy),
                                          (extract_dlc $cachepolicy)))
   >;