[AMDGPU] Added MI bit IsDOT

[rampitec]

Summary

NFC, needed for future commit.

[kzhuravl]

LGTM

llvm/trunk/lib/Target/AMDGPU/SIDefines.h

// TODO: Should this be spilt into VOP3 a and b?
 
   // Uses floating point double precision rounding mode
   FPDPRounding = UINT64_C(1) << 52,
 
   // Instruction is FP atomic.
   FPAtomic = UINT64_C(1) << 53,
 
   // Is a MFMA instruction.
-  IsMAI = UINT64_C(1) << 54
+  IsMAI = UINT64_C(1) << 54,
+
+  // Is a DOT instruction.
+  IsDOT = UINT64_C(1) << 55
 };
 
 // v_cmp_class_* etc. use a 10-bit mask for what operation is checked.
 // The result is true if any of these tests are true.
 enum ClassFlags : unsigned {
   S_NAN = 1 << 0,        // Signaling NaN
   Q_NAN = 1 << 1,        // Quiet NaN
   N_INFINITY = 1 << 2,   // Negative infinity

llvm/trunk/lib/Target/AMDGPU/SIInstrFormats.td

class InstSI <dag outs, dag ins, string asm = "",
   field bit FPDPRounding = 0;
 
   // Instruction is FP atomic.
   field bit FPAtomic = 0;
 
   // This bit indicates that this is one of MFMA instructions.
   field bit IsMAI = 0;
 
+  // This bit indicates that this is one of DOT instructions.
+  field bit IsDOT = 0;
+
   // These need to be kept in sync with the enum in SIInstrFlags.
   let TSFlags{0} = SALU;
   let TSFlags{1} = VALU;
 
   let TSFlags{2} = SOP1;
   let TSFlags{3} = SOP2;
   let TSFlags{4} = SOPC;
   let TSFlags{5} = SOPK;
class InstSI <dag outs, dag ins, string asm = "",
   let TSFlags{51} = IsNonFlatSeg;
 
   let TSFlags{52} = FPDPRounding;
 
   let TSFlags{53} = FPAtomic;
 
   let TSFlags{54} = IsMAI;
 
+  let TSFlags{55} = IsDOT;
+
   let SchedRW = [Write32Bit];
 
   field bits<1> DisableSIDecoder = 0;
   field bits<1> DisableVIDecoder = 0;
   field bits<1> DisableDecoder = 0;
 
   let isAsmParserOnly = !if(!eq(DisableDecoder{0}, {0}), 0, 1);
   let AsmVariantName = AMDGPUAsmVariants.Default;

llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.h

public:
   static bool isMAI(const MachineInstr &MI) {
     return MI.getDesc().TSFlags & SIInstrFlags::IsMAI;
   }
 
   bool isMAI(uint16_t Opcode) const {
     return get(Opcode).TSFlags & SIInstrFlags::IsMAI;
   }
 
+  static bool isDOT(const MachineInstr &MI) {
+    return MI.getDesc().TSFlags & SIInstrFlags::IsDOT;
+  }
+
+  bool isDOT(uint16_t Opcode) const {
+    return get(Opcode).TSFlags & SIInstrFlags::IsDOT;
+  }
+
   static bool isScalarUnit(const MachineInstr &MI) {
     return MI.getDesc().TSFlags & (SIInstrFlags::SALU | SIInstrFlags::SMRD);
   }
 
   static bool usesVM_CNT(const MachineInstr &MI) {
     return MI.getDesc().TSFlags & SIInstrFlags::VM_CNT;
   }
 

llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.td

class VOPProfile <list<ValueType> _ArgVT, bit _EnableF32SrcMods = 0,
 
   field bit HasExt = getHasExt<NumSrcArgs, DstVT, Src0VT, Src1VT>.ret;
   field bit HasExtDPP = getHasDPP<NumSrcArgs, DstVT, Src0VT, Src1VT>.ret;
   field bit HasExtSDWA = HasExt;
   field bit HasExtSDWA9 = HasExt;
   field int NeedPatGen = PatGenMode.NoPattern;
 
   field bit IsMAI = 0;
+  field bit IsDOT = 0;
 
   field Operand Src0PackedMod = !if(HasSrc0FloatMods, PackedF16InputMods, PackedI16InputMods);
   field Operand Src1PackedMod = !if(HasSrc1FloatMods, PackedF16InputMods, PackedI16InputMods);
   field Operand Src2PackedMod = !if(HasSrc2FloatMods, PackedF16InputMods, PackedI16InputMods);
 
   field dag Outs = !if(HasDst,(outs DstRC:$vdst),(outs));
 
   // VOP3b instructions are a special case with a second explicit

llvm/trunk/lib/Target/AMDGPU/VOP2Instructions.td

 defm V_FMAC_F32 : VOP2Inst <"v_fmac_f32", VOP_MAC_F32>;
 }
 
 } // End SubtargetPredicate = HasDLInsts
 
 let Constraints = "$vdst = $src2",
       DisableEncoding="$src2",
       isConvertibleToThreeAddress = 1,
-      isCommutable = 1 in {
+      isCommutable = 1,
+      IsDOT = 1 in {
   let SubtargetPredicate = HasDot5Insts in
     defm V_DOT2C_F32_F16 : VOP2Inst_e32<"v_dot2c_f32_f16", VOP_DOT_ACC_F32_V2F16>;
   let SubtargetPredicate = HasDot6Insts in
     defm V_DOT4C_I32_I8  : VOP2Inst_e32<"v_dot4c_i32_i8",  VOP_DOT_ACC_I32_I32>;
 
   let SubtargetPredicate = HasDot4Insts in
     defm V_DOT2C_I32_I16 : VOP2Inst_e32<"v_dot2c_i32_i16", VOP_DOT_ACC_I32_I32>;
   let SubtargetPredicate = HasDot3Insts in

llvm/trunk/lib/Target/AMDGPU/VOP3PInstructions.td

class SDot2Pat<Instruction Inst> : GCNPat <
   (add (add_oneuse (AMDGPUmul_i24_oneuse (sra i32:$src0, (i32 16)),
                                          (sra i32:$src1, (i32 16))), i32:$src2),
        (AMDGPUmul_i24_oneuse (sext_inreg i32:$src0, i16),
                              (sext_inreg i32:$src1, i16))),
   (Inst (i32 8), $src0, (i32 8), $src1, (i32 8), $src2, (i1 0))> {
   let SubtargetPredicate = !cast<VOP_Pseudo>(Inst).SubtargetPredicate;
 }
 
+let IsDOT = 1 in {
 let SubtargetPredicate = HasDot2Insts in {
 
 def V_DOT2_F32_F16 : VOP3PInst<"v_dot2_f32_f16", VOP3_Profile<VOP_F32_V2F16_V2F16_F32>>;
 def V_DOT2_I32_I16 : VOP3PInst<"v_dot2_i32_i16", VOP3_Profile<VOP_I32_V2I16_V2I16_I32>>;
 def V_DOT2_U32_U16 : VOP3PInst<"v_dot2_u32_u16", VOP3_Profile<VOP_I32_V2I16_V2I16_I32>>;
 def V_DOT4_U32_U8  : VOP3PInst<"v_dot4_u32_u8", VOP3_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>>;
 def V_DOT8_U32_U4  : VOP3PInst<"v_dot8_u32_u4", VOP3_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>>;
 
 } // End SubtargetPredicate = HasDot2Insts
 
 let SubtargetPredicate = HasDot1Insts in {
 
 def V_DOT4_I32_I8  : VOP3PInst<"v_dot4_i32_i8", VOP3_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>>;
 def V_DOT8_I32_I4  : VOP3PInst<"v_dot8_i32_i4", VOP3_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>>;
 
 } // End SubtargetPredicate = HasDot1Insts
+} // End let IsDOT = 1
 
 multiclass DotPats<SDPatternOperator dot_op,
                    VOP3PInst dot_inst> {
   let SubtargetPredicate = dot_inst.SubtargetPredicate in
   def : GCNPat <
     (dot_op (dot_inst.Pfl.Src0VT (VOP3PMods0 dot_inst.Pfl.Src0VT:$src0, i32:$src0_modifiers)),
             (dot_inst.Pfl.Src1VT (VOP3PMods dot_inst.Pfl.Src1VT:$src1, i32:$src1_modifiers)),
             (dot_inst.Pfl.Src2VT (VOP3PMods dot_inst.Pfl.Src2VT:$src2, i32:$src2_modifiers)), i1:$clamp),