ac_vld_decode_h.asm

h263,jpeg,mpeg2编解码核心程序(TI DSP C64xx)

*       Steps to be performed after the loop                                * 
*                                                                           * 
*   a)  *num_of_rles = numberval - 1. The number of run-level pairs found   * 
*       is numberval - 1. Remember numberval is 1 more because it           * 
*       increments for EOB as well.                                         * 
*                                                                           * 
*   b)  *bit_pos_vld = bit_pos_val;                                         * 
*       Store off the current bit position in the pointer. Status variable  * 
*       used for chaining VLD's.                                            * 
*                                                                           * 
*   c)  *used_up_bytes= used_bytes_val;                                     * 
*       Store off the number of used up bytes in the pointer. Status        * 
*       variable for chaining VLD's                                         * 
*                                                                           * 
*   d)  *data_word    = data_word_val;                                      * 
*       Store off the present VLD status word to begin resuming this on     * 
*       next pass of the VLD. Status word of the VLD is the most important  * 
*       state variable.                                                     * 
*                                                                           * 
*   e)  *data_ptr     = data_pointer;                                       * 
*       Store off the pointer from which the next set of 5 bytes need       * 
*       to be fetched at the end of the VLD. State variable used for        * 
*       chaining VLD calls.                                                 * 
*                                                                           * 
*   DIAGNOSTICS                                                             * 
*       Return value == 1: Incorrect VLD stream in which K > 63             * 
*       Return value == 0: Correct VLD and valid JPEG stream                * 
*                                                                           * 
*   CYCLES                                                                  * 
*       14 * N + 27                                                         * 
*       where N is the number of symbols in a block                         * 
*                                                                           * 
*       Example: N=15 symbols/block: 237 cycles                             * 
*                                                                           * 
*   CODESIZE                                                                * 
*       484 bytes                                                           * 
*                                                                           * 
*   BIBLIOGRAPHY                                                            * 
*       1) JPEG still image data compression standard by William B          * 
*          Pennabaker and Joan L. Mitchell                                  * 
*       2) CCITT T.81 standard                                              * 
*                                                                           * 
* ------------------------------------------------------------------------- *
*             Copyright (c) 2002 Texas Instruments, Incorporated.           *
*                            All Rights Reserved.                           *
* ========================================================================= *

                .sect ".text:hand"
                .include "ac_vld_decode_h.h64"

_ac_vld_decode_asm:

; ====================== SYMBOLIC REGISTER ASSIGNMENTS =======================
        .asg            A4,         A_data_ptr
        .asg            B4,         B_data_word
        .asg            A6,         A_run_level
        .asg            B6,         B_num_of_rles
        .asg            A8,         A_bit_position
        .asg            B8,         B_used_up_bytes
        .asg            A10,        A_ac_table_luma
        .asg            B21,        B_data_word_cpy
        .asg            B16,        B_EOB_MASK
        .asg            A6,         A_EOB_CODE
        .asg            A20,        A_data_word_val
        .asg            A16,        A_data_pointer
        .asg            B22,        B_numberval
        .asg            B9,         B_run_tbl
        .asg            B7,         B_sze_tbl
        .asg            B5,         B_shf_tbl
        .asg            A5,         A_const_32
        .asg            B2,         B_K_62
        .asg            A21,        A_bit_pos_val
        .asg            A22,        A_shift_new
        .asg            B1,         B_EOB
        .asg            A9,         A_used_bytes_val
        .asg            A7,         A_run
        .asg            B23,        B_level
        .asg            B24,        B_num_coeffs
        .asg            B2,         B_tbl
        .asg            B18,        B_sze_const
        .asg            B16,        B_shf_const
        .asg            B0,         B_eob_ofs_0
        .asg            A0,         A_eob_byte0
        .asg            B16,        B_eob_ofs_1
        .asg            B0,         B_eob_byte1
        .asg            B16,        B_ofs_len
        .asg            B17,        B_EOB_LEN
; ============================================================================
        LDW     .D1T2   *+A_ac_table_luma(4),    B_run_tbl              ;[ 2,0]
        NOP             1
        LDBU    .D1T1   *A_bit_position,         A_bit_pos_val          ;[ 5,0]
        LDW     .D1T2   *+A_ac_table_luma(12),   B_shf_tbl              ;[ 6,0]

        LDW     .D1T2   *+A_ac_table_luma(8),    B_sze_tbl              ;[ 7,0]
||      MVK     .S2     112,                     B_eob_ofs_0            ;[ 7,0]

        LDBU    .D2T1   *+B_run_tbl[B_eob_ofs_0],A_eob_byte0            ;[ 8,0]
||      ADD     .L2     B_eob_ofs_0,             1,         B_eob_ofs_1 ;[ 8,0]

        LDBU    .D2T2   *+B_run_tbl[B_eob_ofs_1],B_eob_byte1            ;[ 9,0]
||      ADD     .S2     B_eob_ofs_0,             4,         B_ofs_len   ;[ 9,0]

        LDBU    .D2T2   *+B_run_tbl[B_ofs_len],  B_EOB_LEN              ;[10,0]

        MVKL    .S2     62,                  B_K_62                     ;[11,0]
||      LDW     .D2T1   *B_data_word,        A_data_word_val            ;[11,0]

        MVKH    .S2     62,                  B_K_62                     ;[12,0]
||      LDW     .D2T1   *B_used_up_bytes,    A_used_bytes_val           ;[12,0]

        MVK     .S2     684,                 B_shf_const                ;[13,0]
||      LDW     .D1T1   *A_data_ptr,         A_data_pointer             ;[13,0]

        ZERO    .D2     B_num_coeffs                                    ;[14,0]
||      ADD     .L2X    A_run_level,         2,             B_level     ;[14,0]
||      MVK     .S2     370,                 B_sze_const                ;[14,0]

        ADD     .D2     B_shf_tbl,           B_shf_const,   B_shf_tbl   ;[15,0]
||      SHL     .S2     B_eob_byte1,         16,            B_eob_byte1 ;[15,0]
||      MVK     .L2     -1,                  B_EOB_MASK                 ;[15,0]

        MV      .L1     A_run_level,         A_run                      ;[16,0]
||      ADD     .L2     B_sze_tbl,           B_sze_const,   B_sze_tbl   ;[16,0]
||      ADDAW   .D2     B_run_tbl,           14,            B_run_tbl   ;[16,0]
||      SHL     .S1     A_eob_byte0,         24,            A_eob_byte0 ;[16,0]
||      SHRU    .S2     B_EOB_MASK,          B_EOB_LEN,     B_EOB_MASK  ;[16,0]

        ADD     .D1     A_bit_pos_val,       8,             A_shift_new ;[17,0]
||      MVK     .S1     32,                  A_const_32                 ;[17,0]
||      ZERO    .L2     B_numberval                                     ;[17,0]
||      ADD     .L1X    A_eob_byte0,         B_eob_byte1,   A_EOB_CODE  ;[17,0]
||      NOT     .S2     B_EOB_MASK,          B_EOB_MASK                 ;[17,0]
||      MV      .D2X    A_data_word_val,     B_data_word_cpy            ;[17,0]

        ;==== Branch occurs

; ====================== SYMBOLIC REGISTER ASSIGNMENTS =======================
        .asg            B21,        B_data_word_cpy
        .asg            B16,        B_EOB_MASK
        .asg            A6,         A_EOB_CODE
        .asg            A20,        A_data_word_val
        .asg            A16,        A_data_pointer
        .asg            B22,        B_numberval
        .asg            B9,         B_run_tbl
        .asg            B7,         B_sze_tbl
        .asg            B5,         B_shf_tbl
        .asg            A5,         A_const_32
        .asg            B2,         B_K_62
        .asg            A21,        A_bit_pos_val
        .asg            A22,        A_shift_new
        .asg            B1,         B_EOB
        .asg            A9,         A_used_bytes_val
        .asg            A7,         A_run
        .asg            B23,        B_level
        .asg            B24,        B_num_coeffs
        .asg            B19,        B_EOB_VAL
        .asg            B17,        B_lmbd_val
        .asg            A17,        A_lmbd_val
        .asg            A1,         A_new_hi
        .asg            A3,         A_temp_h
        .asg            A2,         A_temp_l
        .asg            A0,         A_temp_swap_l
        .asg            A0,         A_new_lo
        .asg            B26,        B_t02_ofs
        .asg            B18,        B_ofs
        .asg            B17,        B_t68_ofs
        .asg            B17,        B_t9x_ofs
        .asg            A0,         A_st3
        .asg            B1,         B_st5
        .asg            A0,         A_st8
        .asg            B20,        B_run_val
        .asg            B25,        B_size_val
        .asg            A2,         A_len_cw