[AMDGPU] New tbuffer intrinsics

Summary:
This commit adds new intrinsics
  llvm.amdgcn.raw.tbuffer.load
  llvm.amdgcn.struct.tbuffer.load
  llvm.amdgcn.raw.tbuffer.store
  llvm.amdgcn.struct.tbuffer.store

with the following changes from the llvm.amdgcn.tbuffer.* intrinsics:

* there are separate raw and struct versions: raw does not have an index
  arg and sets idxen=0 in the instruction, and struct always sets
  idxen=1 in the instruction even if the index is 0, to allow for the
  fact that gfx9 does bounds checking differently depending on whether
  idxen is set;

* there is a combined format arg (dfmt+nfmt)

* there is a combined cachepolicy arg (glc+slc)

* there are now only two offset args: one for the offset that is
  included in bounds checking and swizzling, to be split between the
  instruction's voffset and immoffset fields, and one for the offset
  that is excluded from bounds checking and swizzling, to go into the
  instruction's soffset field.

The AMDISD::TBUFFER_* SD nodes always have an index operand, all three
offset operands, combined format operand, combined cachepolicy operand,
and an extra idxen operand.

The tbuffer pseudo- and real instructions now also have a combined
format operand.

The obsolescent llvm.amdgcn.tbuffer.* and llvm.SI.tbuffer.store
intrinsics continue to work.

V2: Separate raw and struct intrinsics.
V3: Moved extract_glc and extract_slc defs to a more sensible place.
V4: Rebased on D49995.
V5: Only two separate offset args instead of three.
V6: Pseudo- and real instructions have joint format operand.
V7: Restored optionality of dfmt and nfmt in assembler.
V8: Addressed minor review comments.

Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, t-tye, llvm-commits

Differential Revision: https://reviews.llvm.org/D49026

Change-Id: If22ad77e349fac3a5d2f72dda53c010377d470d4
llvm-svn: 340268

1 files changed, 55 insertions, 10 deletions

diff --git a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
index 31e2885c833..75deeb7bd67 100644
--- a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
+++ b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -161,8 +161,7 @@ public:
     ImmTyExpTgt,
     ImmTyExpCompr,
     ImmTyExpVM,
-    ImmTyDFMT,
-    ImmTyNFMT,
+    ImmTyFORMAT,
     ImmTyHwreg,
     ImmTyOff,
     ImmTySendMsg,
@@ -312,8 +311,7 @@ public:
   bool isSLC() const { return isImmTy(ImmTySLC); }
   bool isTFE() const { return isImmTy(ImmTyTFE); }
   bool isD16() const { return isImmTy(ImmTyD16); }
-  bool isDFMT() const { return isImmTy(ImmTyDFMT) && isUInt<8>(getImm()); }
-  bool isNFMT() const { return isImmTy(ImmTyNFMT) && isUInt<8>(getImm()); }
+  bool isFORMAT() const { return isImmTy(ImmTyFORMAT) && isUInt<8>(getImm()); }
   bool isBankMask() const { return isImmTy(ImmTyDppBankMask); }
   bool isRowMask() const { return isImmTy(ImmTyDppRowMask); }
   bool isBoundCtrl() const { return isImmTy(ImmTyDppBoundCtrl); }
@@ -666,8 +664,7 @@ public:
     case ImmTySLC: OS << "SLC"; break;
     case ImmTyTFE: OS << "TFE"; break;
     case ImmTyD16: OS << "D16"; break;
-    case ImmTyDFMT: OS << "DFMT"; break;
-    case ImmTyNFMT: OS << "NFMT"; break;
+    case ImmTyFORMAT: OS << "FORMAT"; break;
     case ImmTyClampSI: OS << "ClampSI"; break;
     case ImmTyOModSI: OS << "OModSI"; break;
     case ImmTyDppCtrl: OS << "DppCtrl"; break;
@@ -1061,6 +1058,7 @@ public:
   OperandMatchResultTy parseRegWithFPInputMods(OperandVector &Operands);
   OperandMatchResultTy parseRegWithIntInputMods(OperandVector &Operands);
   OperandMatchResultTy parseVReg32OrOff(OperandVector &Operands);
+  OperandMatchResultTy parseDfmtNfmt(OperandVector &Operands);
 
   void cvtDSOffset01(MCInst &Inst, const OperandVector &Operands);
   void cvtDS(MCInst &Inst, const OperandVector &Operands) { cvtDSImpl(Inst, Operands, false); }
@@ -3522,6 +3520,53 @@ AMDGPUAsmParser::parseStringWithPrefix(StringRef Prefix, StringRef &Value) {
   return MatchOperand_Success;
 }
 
+// dfmt and nfmt (in a tbuffer instruction) are parsed as one to allow their
+// values to live in a joint format operand in the MCInst encoding.
+OperandMatchResultTy
+AMDGPUAsmParser::parseDfmtNfmt(OperandVector &Operands) {
+  SMLoc S = Parser.getTok().getLoc();
+  int64_t Dfmt = 0, Nfmt = 0;
+  // dfmt and nfmt can appear in either order, and each is optional.
+  bool GotDfmt = false, GotNfmt = false;
+  while (!GotDfmt || !GotNfmt) {
+    if (!GotDfmt) {
+      auto Res = parseIntWithPrefix("dfmt", Dfmt);
+      if (Res != MatchOperand_NoMatch) {
+        if (Res != MatchOperand_Success)
+          return Res;
+        if (Dfmt >= 16) {
+          Error(Parser.getTok().getLoc(), "out of range dfmt");
+          return MatchOperand_ParseFail;
+        }
+        GotDfmt = true;
+        Parser.Lex();
+        continue;
+      }
+    }
+    if (!GotNfmt) {
+      auto Res = parseIntWithPrefix("nfmt", Nfmt);
+      if (Res != MatchOperand_NoMatch) {
+        if (Res != MatchOperand_Success)
+          return Res;
+        if (Nfmt >= 8) {
+          Error(Parser.getTok().getLoc(), "out of range nfmt");
+          return MatchOperand_ParseFail;
+        }
+        GotNfmt = true;
+        Parser.Lex();
+        continue;
+      }
+    }
+    break;
+  }
+  if (!GotDfmt && !GotNfmt)
+    return MatchOperand_NoMatch;
+  auto Format = Dfmt | Nfmt << 4;
+  Operands.push_back(
+      AMDGPUOperand::CreateImm(this, Format, S, AMDGPUOperand::ImmTyFORMAT));
+  return MatchOperand_Success;
+}
+
 //===----------------------------------------------------------------------===//
 // ds
 //===----------------------------------------------------------------------===//
@@ -4617,8 +4662,7 @@ void AMDGPUAsmParser::cvtMtbuf(MCInst &Inst, const OperandVector &Operands) {
 
   addOptionalImmOperand(Inst, Operands, OptionalIdx,
                         AMDGPUOperand::ImmTyOffset);
-  addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyDFMT);
-  addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyNFMT);
+  addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyFORMAT);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyGLC);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySLC);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyTFE);
@@ -4761,8 +4805,7 @@ static const OptionalOperand AMDGPUOptionalOperandTable[] = {
   {"lds",     AMDGPUOperand::ImmTyLDS, true, nullptr},
   {"offset",  AMDGPUOperand::ImmTyOffset, false, nullptr},
   {"inst_offset", AMDGPUOperand::ImmTyInstOffset, false, nullptr},
-  {"dfmt",    AMDGPUOperand::ImmTyDFMT, false, nullptr},
-  {"nfmt",    AMDGPUOperand::ImmTyNFMT, false, nullptr},
+  {"dfmt",    AMDGPUOperand::ImmTyFORMAT, false, nullptr},
   {"glc",     AMDGPUOperand::ImmTyGLC, true, nullptr},
   {"slc",     AMDGPUOperand::ImmTySLC, true, nullptr},
   {"tfe",     AMDGPUOperand::ImmTyTFE, true, nullptr},
@@ -4844,6 +4887,8 @@ OperandMatchResultTy AMDGPUAsmParser::parseOptionalOpr(OperandVector &Operands)
                Op.Type == AMDGPUOperand::ImmTyNegHi) {
       res = parseOperandArrayWithPrefix(Op.Name, Operands, Op.Type,
                                         Op.ConvertResult);
+    } else if (Op.Type == AMDGPUOperand::ImmTyFORMAT) {
+      res = parseDfmtNfmt(Operands);
     } else {
       res = parseIntWithPrefix(Op.Name, Operands, Op.Type, Op.ConvertResult);
     }