Clemson University -- CPSC 231 -- Fall 2009


Instruction formats

  formats are basically rules on how bit patterns are interpreted/decoded

  a format specifies the positions of fields such as opcodes, register
      identifiers, and constants

  format 1: call      +--+----------------------------------+
      bit encoding is |01|.......30-bit displacement........|
                      +--+----------------------------------+

  format 2:           +--+-----+---+------------------------+
      sethi is        |00|rdest|100|..22-bit displacement...|
                      +--+-+---+---+------------------------+
      branch is       |00|a|cnd|op2|..22-bit displacement...|
                      +--+-+---+---+------------------------+

                        "a" is the annul bit, if set then the instruction in the
                        delay slot of the branch is executed only if the branch
                        is taken (thus you can move the first instruction from
                        the branch target into the delay slot without causing
                        incorrect values when the branch is untaken)

  format 3:           +--+-----+------+-----+-+--------+-----+
      3-register is   |1x|rdest|.op3..|rsrc1|0|asi/fpop|rsrc2|
                      +--+-----+------+-----+-+--------+-----+
      immediate is    |1x|rdest|.op3..|rsrc1|1|signed 13-bit |
                      +--+-----+------+-----+-+--------------+

  note that calls and branches are pc-relative, this makes it easier to
      link and load programs


Loading a 32-bit constant/address

  sethi instruction takes 22-bit constant and loads it into a register after
      shifting to the left by 10 places, thus you can get the rest of a
      32-bit constant by or-ing in the low ten bits

  pseudo-ops

    %hi(a)                 ==   a >> 10
    %lo(a)                 ==   a & 0x3ff

    set  0x12345678, %o0   ==   sethi %hi(0x12345678),      %o0
                                or    %o0, %lo(0x12345678), %o0

  loading a non-stack value from memory given a label

    set   x_m, %o0        ! expands into two instructions: sethi and or
    ld    [%o0], %o1

  alternatively, the or can be included as part of the addressing mode

    sethi %hi(x_m), %o0
    ld    [%o0 + %lo(x_m)], %o1

  since set, %hi(), and %lo() often deal with absolute addresses, rather than
      pc-relative, the argument is limited to a symbol +/- small constant