// Copyright 2026 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// Matrix configuration instructions, encoded in the vector-configuration
// (OP-V) major opcode like the vset* family (spec 15.1.1.4). They are
// distinguished from vsetvl (func7 == 0b100'0000) by func7:
//   msetmtype       func7 == 0b100'0001
// msettn/tm/tk    func7 == 0b100'0010, selected by the low rs2 bits.

instruction group RiscVZvtInst32[32] : Inst32Format {
  msetmtype     : VConfig : func7 == 0b100'0001,                 func3 == 0b111, opcode == 0b101'0111;
  msettn        : VConfig : func7 == 0b100'0010, rs2 == 0b00000, func3 == 0b111, opcode == 0b101'0111;
  msettm        : VConfig : func7 == 0b100'0010, rs2 == 0b00001, func3 == 0b111, opcode == 0b101'0111;
  msettk        : VConfig : func7 == 0b100'0010, rs2 == 0b00010, func3 == 0b111, opcode == 0b101'0111;

  // Tile manipulation (OP-V, func3 == 0b110).
  vtmv_v_t      : VArith : func6 == 0b010'000, vs2 == 0b11111, func3 == 0b110, opcode == 0b101'0111;
  vtmv_t_v      : VArith : func6 == 0b010'111,                 func3 == 0b110, opcode == 0b101'0111;
  vtzero        : VArith : func6 == 0b010'000, vs2 == 0b11110, vs1 == 0b00000, func3 == 0b110, opcode == 0b101'0111;
  vtdiscard     : VArith : func6 == 0b010'000, vs2 == 0b11100, vs1 == 0b00000, func3 == 0b110, opcode == 0b101'0111;

  // Tile load/store to memory (LOAD-FP / STORE-FP, mew == 1, width == 0b111;
  // the nf field selects EEW: 0->8, 1->16, 2->32, 3->64).
  vtle8         : VMem : nf == 0b000, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b000'0111;
  vtle16        : VMem : nf == 0b001, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b000'0111;
  vtle32        : VMem : nf == 0b010, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b000'0111;
  vtle64        : VMem : nf == 0b011, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b000'0111;
  vtse8         : VMem : nf == 0b000, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b010'0111;
  vtse16        : VMem : nf == 0b001, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b010'0111;
  vtse32        : VMem : nf == 0b010, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b010'0111;
  vtse64        : VMem : nf == 0b011, mew == 0b1, mop == 0b00, vm == 0b1, width == 0b111, opcode == 0b010'0111;

  // Matrix multiply (OP-VE, func6 == 0b111100). func3 selects int(000)/fp(001);
  // the alt bit selects the alternate datatype variant.
  vtmmu_tvv     : VMatrix : func6 == 0b111'100, func3 == 0b000, alt == 0b0, opcode == 0b111'0111;
  vtmms_tvv     : VMatrix : func6 == 0b111'100, func3 == 0b000, alt == 0b1, opcode == 0b111'0111;
  vtfmm_tvv     : VMatrix : func6 == 0b111'100, func3 == 0b001, alt == 0b0, opcode == 0b111'0111;
  vtfmm_alt_tvv : VMatrix : func6 == 0b111'100, func3 == 0b001, alt == 0b1, opcode == 0b111'0111;
};