Introduction: Concurrency

We envision CoreDSL 2 to be suitable for the modelling of ISAX that imply the presence of complex functional units executing independently of the core's main pipeline. The code below presents an implementation of a zero-overhead loop mechanism using the spawn keyword. The spawned region represents logic to periodically check and reset the program counter, and will continue to execute even after the setup_ZOL instruction is completed.

InstructionSet ZeroOverheadLoop extends RISCVBase {
  architectural_state { unsigned<XLEN> count, start_PC, end_PC; }
  instructions {
    setup_ZOL {
      encoding: offset[6:0] :: rs2[4:0] :: rs1[4:0] :: 3'b010 :: rd[4:0] :: 7'b0010111;
      behavior: {
        // set configuration registers
        count    = X[rs1];
        start_PC = (unsigned<XLEN>) PC + 4;
        end_PC   = (unsigned<XLEN>) (start_PC + (signed) offset);

        spawn { // this region will execute independently of setup_ZOL
          for (; count > 0; --count) {
            while (PC != end_PC) /* do nothing */ ;
            PC = start_PC; // jump to loop start
          }
        }
        // execution of setup_ZOL ends
      }
    }
  }
}