Click above to get retro games delivered to your door ever month!

    MMX instruction set summary (P55C and Klamath)

    The table below contains a summary of the MMX instruction set. The
    instruction mnemonics below are the base set of mnemonics; most
    instructions have multiple variations (e.g., packed-byte, -word, and
    -dword variations).
    Complete information on the MMX instructions may be found in the
    Intel Architecture MMX (tm) Technology Programmer's Reference Manual.


    Packed Arithmetic       Wrap Around     Signed Sat      Unsigned Sat
    Addition                PADD            PADDS           PADDUS
    Subtraction             PSUB            PSUBS           PSUBUS
    Multiplication          PMULL/H
    Multiply & add          PMADD
    Shift right Arithmetic  PSRA
    Compare                 PCMPcc

    Conversions             Regular         Signed Sat      Unsigned Sat
    Pack                                    PACKSS          PACKUS
    Unpack                  PUNPCKL/H

    Logical Operations      Packed          Full 64-bit
    And                                     PAND
    And not                                 PANDN
    Or                                      POR
    Exclusive or                            PXOR
    Shift left              PSLL            PSLL
    Shift right             PSRL            PSRL

    Transfers and Memory
    Operations              32-bit          64-bit
    Register-register move  MOVD            MOVQ
    Load from memory        MOVD            MOVQ
    Store to memory         MOVD            MOVQ

    Miscellaneous
    Empty MMX state         EMMS


    The following functions are used in the algorithmic descriptions

    ZeroExtend (value)
                        returns a value zero-extended to the operand-
                        size attribute of the instruction. For example,
                        if OperandSize = 32, ZeroExtend of a byte value
                        of -10 converts the byte from 0F6h to doubleword
                        with hexadecimal value 000000F6h. If the value
                        passed to ZeroExtend and the operand-size
                        attribute are the same size, ZeroExtend returns
                        the value unaltered.

    SignExtend (value)
                        returns a value sign-extended to the operand-size
                        attribute of the instruction. For example, if
                        OperandSize = 32, SignExtend of a byte containing
                        the value -10 converts the byte from 0F6h to
                        doubleword with hexadecimal value 0FFFFFFF6h. If
                        the value passed to SignExtend and the operand-
                        size attribute are the same size, SignExtend
                        returns the value unaltered.

    SaturateSignedWordToSignedByte
                        converts a signed 16-bit value to a signed 8-bit
                        value. If the signed 16-bit value is less than
                        -128, it is represented by the saturated value
                        -128 (80h). If it is greater than 127, it is
                        represented by the saturated value 127 (7Fh).

    SaturateSignedDwordToSignedWord
                        converts a signed 32-bit value to a signed
                        16-bit value. If the signed 32-bit value is less
                        than -32768, it is represented by the saturated
                        value -32768 (8000h). If it is greater than
                        32767, it is represented by the saturated value
                        32767 (7FFFh).

    SaturateSignedWordToUnsignedByte
                        converts a signed 16-bit value to an unsigned
                        8.bit value. If the signed 16-bit value is less
                        than zero it is represented by the saturated
                        value zero (00h). If it is greater than 255
                        it is represented by the saturated value 255
                        (0FFh).

    SaturateToSignedByte
                        represents the result of an operation as a signed
                        8-bit value. If the result is less than -128, it
                        is represented by the saturated value -128 (80h).
                        If it is greater than 127, it is represented by
                        the saturated value 127 (7Fh).

    SaturateToSignedWord
                        represents the result of an operation as a signed
                        16-bit value. If the result is less than -32768,
                        it is represented by the saturated value -32768
                        (8000h). If it is greater than 32767, it is
                        represented by the saturated value 32767 (7FFFh).

    SaturateToUnsignedByte
                        represents the result of an operation as a signed
                        8-bit value. If the result is less than zero it
                        is represented by the saturated value zero (00h).
                        If it is greater than 255, it is represented by
                        the saturated value 255 (0FFh).

    SaturateToUnsignedWord
                        represents the result of an operation as a signed
                        16-bit value. If the result is less than zero it
                        is represented by the saturated value zero (00h).
                        If it is greater than 65535, it is represented
                        by the saturated value 65535 (0FFFFh).