[PATCH] D145535: [RISCV] Store fli min/nan/inf in index form in RISCVOperand.

[Craig Topper via Phabricator via llvm-commits]

craig.topper created this revision.
craig.topper added reviewers: reames, joshua-arch1, asb, zixuan-wu.
Herald added subscribers: luke, VincentWu, vkmr, frasercrmck, evandro, luismarques, apazos, sameer.abuasal, s.egerton, Jim, benna, psnobl, jocewei, PkmX, arphaman, the_o, brucehoult, MartinMosbeck, rogfer01, edward-jones, zzheng, jrtc27, shiva0217, kito-cheng, niosHD, sabuasal, simoncook, johnrusso, rbar, hiraditya, arichardson.
Herald added a project: All.
craig.topper requested review of this revision.
Herald added subscribers: pcwang-thead, eopXD, MaskRay.
Herald added a project: LLVM.

Instead of converting to FP value, store them as Index using an
immediate operand.

Do the same for the explicit index form.

This avoids using the FP32 version of these special values as the
representation for fli.h and fli.d. inf/nan aren't so bad, but
"min" is problematic as the current implementation allows fli.d and
fli.h to accept the decimal version of the fp32 minimum value. I
will submit another patch to fix that.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D145535

Files:
  llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp


Index: llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
===================================================================
--- llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
+++ llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
@@ -492,6 +492,8 @@
 
   /// Return true if the operand is a valid fli.s floating-point immediate.
   bool isLoadFPImm() const {
+    if (isImm())
+      return isUImm5();
     return Kind == KindTy::FPImmediate &&
            RISCVLoadFPImm::getLoadFP32Imm(APInt(32, getFPConst())) != -1;
   }
@@ -973,6 +975,11 @@
 
   void addFPImmOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
+    if (isImm()) {
+      addExpr(Inst, getImm(), isRV64Imm());
+      return;
+    }
+
     int Imm = RISCVLoadFPImm::getLoadFP32Imm(APInt(32, getFPConst()));
     Inst.addOperand(MCOperand::createImm(Imm));
   }
@@ -1567,16 +1574,17 @@
   if (getTok().is(AsmToken::Identifier)) {
     StringRef Identifier = getTok().getIdentifier();
     if (Identifier.compare_insensitive("inf") == 0) {
-      APFloat SpecialVal = APFloat::getInf(APFloat::IEEEsingle());
-      Operands.push_back(RISCVOperand::createFPImm(
-          SpecialVal.bitcastToAPInt().getZExtValue(), S));
+      Operands.push_back(
+          RISCVOperand::createImm(MCConstantExpr::create(30, getContext()), S,
+                                  getTok().getEndLoc(), isRV64()));
     } else if (Identifier.compare_insensitive("nan") == 0) {
-      APFloat SpecialVal = APFloat::getNaN(APFloat::IEEEsingle());
-      Operands.push_back(RISCVOperand::createFPImm(
-          SpecialVal.bitcastToAPInt().getZExtValue(), S));
+      Operands.push_back(
+          RISCVOperand::createImm(MCConstantExpr::create(31, getContext()), S,
+                                  getTok().getEndLoc(), isRV64()));
     } else if (Identifier.compare_insensitive("min") == 0) {
-      unsigned SpecialVal = RISCVLoadFPImm::getFPImm(1);
-      Operands.push_back(RISCVOperand::createFPImm(SpecialVal, S));
+      Operands.push_back(
+          RISCVOperand::createImm(MCConstantExpr::create(1, getContext()), S,
+                                  getTok().getEndLoc(), isRV64()));
     } else {
       TokError("invalid floating point literal");
       return MatchOperand_ParseFail;
@@ -1602,8 +1610,9 @@
       TokError("encoded floating point value out of range");
       return MatchOperand_ParseFail;
     }
-    unsigned F = RISCVLoadFPImm::getFPImm(Tok.getIntVal());
-    Operands.push_back(RISCVOperand::createFPImm(F, S));
+    Operands.push_back(RISCVOperand::createImm(
+        MCConstantExpr::create(Tok.getIntVal(), getContext()), S,
+        Tok.getEndLoc(), isRV64()));
   } else {
     // Parse FP representation.
     APFloat RealVal(APFloat::IEEEsingle());


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D145535.503201.patch
Type: text/x-patch
Size: 2794 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230308/40cb9af2/attachment.bin>