[AMDGPU] Assembler: fix v_mac_f16 immediates

[artem.tamazov]

Good job, thanks!

lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp

if (type == MVT::f64 || type == MVT::i64) { // Expected 64-bit operand
       return AMDGPU::isInlinableLiteral64(Imm.Val,
                                           AsmParser->hasInv2PiInlineImm());
     }
 
     APFloat FPLiteral(APFloat::IEEEdouble(), APInt(64, Imm.Val));
     if (!canLosslesslyConvertToFPType(FPLiteral, type))
       return false;
 
+    if (type.getScalarSizeInBits() == 16) {
+      return AMDGPU::isInlinableLiteral16(
+        static_cast<int32_t>(FPLiteral.bitcastToAPInt().getZExtValue()),
+        AsmParser->hasInv2PiInlineImm());
+    }
+
     // Check if single precision literal is inlinable
     return AMDGPU::isInlinableLiteral32(
       static_cast<int32_t>(FPLiteral.bitcastToAPInt().getZExtValue()),
       AsmParser->hasInv2PiInlineImm());
   }
 
 
   // We got int literal token.

 }
 
 void AMDGPUAsmParser::cvtSdwaVOPC(MCInst &Inst, const OperandVector &Operands) {
   cvtSDWA(Inst, Operands, SIInstrFlags::VOPC);
 }
 
 void AMDGPUAsmParser::cvtSDWA(MCInst &Inst, const OperandVector &Operands,
                               uint64_t BasicInstType) {
+  using namespace llvm::AMDGPU::SDWA;
   OptionalImmIndexMap OptionalIdx;
 
   unsigned I = 1;
   const MCInstrDesc &Desc = MII.get(Inst.getOpcode());
   for (unsigned J = 0; J < Desc.getNumDefs(); ++J) {
     ((AMDGPUOperand &)*Operands[I++]).addRegOperands(Inst, 1);
   }
 
void AMDGPUAsmParser::cvtSDWA(MCInst &Inst, const OperandVector &Operands,
   }
 
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyClampSI, 0);
 
   if (Inst.getOpcode() != AMDGPU::V_NOP_sdwa_vi) {
     // V_NOP_sdwa_vi has no optional sdwa arguments
     switch (BasicInstType) {
     case SIInstrFlags::VOP1:
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, 2);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, 6);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, DstUnused::UNUSED_PRESERVE);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, SdwaSel::DWORD);
       break;
 
     case SIInstrFlags::VOP2:
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, 2);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, 6);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, DstUnused::UNUSED_PRESERVE);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, SdwaSel::DWORD);
       break;
 
     case SIInstrFlags::VOPC:
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, 6);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, SdwaSel::DWORD);
       break;
 
     default:
       llvm_unreachable("Invalid instruction type. Only VOP1, VOP2 and VOPC allowed");
     }
   }
 
   // special case v_mac_{f16, f32}:

lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp

return (Val == FloatToBits(0.0f)) ||
          (Val == FloatToBits(2.0f)) ||
          (Val == FloatToBits(-2.0f)) ||
          (Val == FloatToBits(4.0f)) ||
          (Val == FloatToBits(-4.0f)) ||
          (Val == 0x3e22f983 && HasInv2Pi);
 }
 
 bool isInlinableLiteral16(int16_t Literal, bool HasInv2Pi) {
-  assert(HasInv2Pi);
+  if (!HasInv2Pi)
+    return false;
 
   if (Literal >= -16 && Literal <= 64)
     return true;
 
   uint16_t Val = static_cast<uint16_t>(Literal);
   return Val == 0x3C00 || // 1.0
          Val == 0xBC00 || // -1.0
          Val == 0x3800 || // 0.5

test/MC/AMDGPU/vop3.s

 // RUN: not llvm-mc -arch=amdgcn -show-encoding %s | FileCheck %s --check-prefix=SICI
-// RUN: llvm-mc -arch=amdgcn -mcpu=hawaii -show-encoding %s | FileCheck %s --check-prefix=CI
+// RUN: not llvm-mc -arch=amdgcn -mcpu=hawaii -show-encoding %s | FileCheck %s --check-prefix=CI
 // RUN: not llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s | FileCheck %s --check-prefix=VI
 
-// RUN: not llvm-mc -arch=amdgcn -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSI
-// RUN: not llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s 2>&1 | FileCheck %s -check-prefix=NOVI
+// RUN: not llvm-mc -arch=amdgcn -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSI --check-prefix=NOSICI
+// RUN: not llvm-mc -arch=amdgcn -mcpu=hawaii -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSICI
+// RUN: not llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOVI
 
 
 //===----------------------------------------------------------------------===//
 // VOPC Instructions
 //===----------------------------------------------------------------------===//
 
 // Test forced e64 encoding
 

 v_mac_f32_e64 v0, v1, v2 mul:2
 // SICI: v_mac_f32_e64 v0, v1, v2 mul:2 ; encoding: [0x00,0x00,0x3e,0xd2,0x01,0x05,0x02,0x08]
 // VI:   v_mac_f32_e64 v0, v1, v2 mul:2 ; encoding: [0x00,0x00,0x16,0xd1,0x01,0x05,0x02,0x08]
 
 v_mac_f32_e64 v0, -v1, |v2|
 // SICI: v_mac_f32_e64 v0, -v1, |v2| ; encoding: [0x00,0x02,0x3e,0xd2,0x01,0x05,0x02,0x20]
 // VI:   v_mac_f32_e64 v0, -v1, |v2| ; encoding: [0x00,0x02,0x16,0xd1,0x01,0x05,0x02,0x20]
 
+v_mac_f16_e64 v0, 0.5, flat_scratch_lo
+// NOSICI: error:
+// VI: v_mac_f16_e64 v0, 0.5, flat_scratch_lo ; encoding: [0x00,0x00,0x23,0xd1,0xf0,0xcc,0x00,0x00]
+
+v_mac_f16_e64 v0, -4.0, flat_scratch_lo
+// NOSICI: error:
+// VI: v_mac_f16_e64 v0, -4.0, flat_scratch_lo ; encoding: [0x00,0x00,0x23,0xd1,0xf6,0xcc,0x00,0x20]
+
+v_mac_f16_e64 v0, flat_scratch_lo, -4.0
+// NOSICI: error:
+// VI: v_mac_f16_e64 v0, flat_scratch_lo, -4.0 ; encoding: [0x00,0x00,0x23,0xd1,0x66,0xec,0x01,0x40]
+
 ///===---------------------------------------------------------------------===//
 // VOP3 Instructions
 ///===---------------------------------------------------------------------===//
 
 // TODO: Modifier tests
 
 v_mad_legacy_f32 v2, v4, v6, v8
 // SICI: v_mad_legacy_f32 v2, v4, v6, v8 ; encoding: [0x02,0x00,0x80,0xd2,0x04,0x0d,0x22,0x04]