HW2 HCL-2 (irmovq + rrmovq + jmp)

Your task

  1. Make an HCL file irrr.hcl that implements a single-cycle processor with
    the following instructions:

    • nop
    • halt
    • irmovq
    • rrmovq
    • unconditional jmp
  2. For instructions not in Y86, make your HCL file set Stat to STAT_INS
    (invalid instruction error). If an instruction is in Y86 and not listed
    above, we do not care how your simulator works on it.
  3. Test your file using make test-irrr.
  4. Submit your result to kytos


Suggested method

  1. We suggest implementing and testing one instruction at a time, in the order, they’re listed above.
  2. Start with an HCL which has a program counter register and increments it based on
    the instruction icode like your pc.hcl from the last lab.

Adding nop/halt

  1. Set Stat based on the icode. It should be STAT_HLT if the icode is halt, STAT_INS for an icode not
    in Y86 (e.g., > 0xb), STAT_AOK otherwise.

Test nop/halt

  1. If you run your simulator on y86/halt.yo, and it should halt after one cycle, and if you run it on
    y86/nophalt.yo, it should halt after 4 cycles.

Adding irmovq

  1. Extract the rB and valC fields of the instruction into a wire.
  2. Send valC to the register file (reg_inputE or reg_inputM) as the new value
    for register number rB (reg_dstE or reg_dstM).
  3. Make sure you do not write to the register file when the icode is not
    an IRMOVQ or RRMOVQ. (Specify REG_NONE for reg_dstE and/or reg_dstM.)

Testing irmovq

  1. If you run your simulator with the -q flag on y86/irmovq.yo, you should see something like:
    +----------------------- halted in state: ------------------------------+
    | RAX:              222   RCX:                2   RDX:               22 |
    | RBX:                0   RSP:                0   RBP:                0 |
    | RSI:                0   RDI:                0   R8:                 0 |
    | R9:                 0   R10:                0   R11:                0 |
    | R12:                0   R13:                0   R14: fedcba9876543210 |
    | register pP(N) { pc=0000000000000033 }                                |
    | used memory:   _0 _1 _2 _3  _4 _5 _6 _7   _8 _9 _a _b  _c _d _e _f    |
    |  0x0000000_:   30 f0 01 00  00 00 00 00   00 00 30 f1  02 00 00 00    |
    |  0x0000001_:   00 00 00 00  30 f2 22 00   00 00 00 00  00 00 30 f0    |
    |  0x0000002_:   22 02 00 00  00 00 00 00   30 fe 10 32  54 76 98 ba    |
    |  0x0000003_:   dc fe                                                  |
    +--------------------- (end of halted state) ---------------------------+
    Cycles run: 6

    (It’s okay if your pc register and register bank have a different name.)

  2. If you see 0 in RAX instead of 222 from y86/irmovq.yo,
    it is likely that you’re writing to the register file for halt as well as irmovq.

Implement rrmovq

  1. Extract the rA field from the instruction.
  2. When the instruction is RRMOVQ, send rA to one of the register file source inputs
    so it reads from this register.
  3. Add rB to one of the register file destination inputs so it writes to that register.
  4. When the instruction is RRMOVQ, send the value just read from rA into the
    register file data input signal (reg_inputE or reg_inputM).
  5. Since you already implemented irmovq, several of the above changes will probably require
    modifying an existing MUX rather than writing new assignments.

Testing rrmovq

  1. If you run your simulator with the -q flag on y86/rrmovq.yo, you should see
    these register values:

    | RAX:               22   RCX:               22   RDX:             162e |
  2. Your simulator’s register values should agree with tools/yis on y86/prog8.yo.

Implement unconditional jmp

  1. Add a MUX for valC. The jXX immediate value is bits 8..72, not 16..80.
  2. Add a MUX for the PC register value (or to an existing PC register value MUX) to
    select valC if the icode is JXX. (You don’t need to handle conditional
    jumps in the lab.)

Testing unconditional jmp

  1. If you run your simulator on y86/jmp.yo, you should end up with
    ace and not bad in rax and nothing in rbx.
  2. If you have trouble with the y86/jmp.yo test case, see the source code
    in y86/jmp.ys and a trace of the expected cycle-by-cycle
    output in testdata/seq-traces/jmp.txt.

More complete testing

  1. Run make test-irrr. This compares your simulators output to our references over the files listed in testdata/irrr-tests.txt.

Debugging Test Failures

  1. You can see the expected cycle-by-cycle of a single-cycle processor on any failing test in
    testdata/seq-traces/TESTNAME.yo. This is our recommendation for debugging test failures
  2. You can run your simulator with the -d flag to see the values of all wires during each cycle,
    which can be helpful when debugging.


%d bloggers like this: