;------------------------------------------------------------------------------ ; ; Fichier : mpega_decode_asm.a ; ; Auteur : Stéphane TAVENARD ; ; $VER: mpega_decode_asm.a 1.0 (23/06/1995) ; ; (C) Copyright 1995 Stéphane TAVENARD ; All Rights Reserved ; ; #Rev| Date | Comment ; ----|----------|-------------------------------------------------------- ; 0 |11/04/1995| Initial revision ST ; 1 |18/06/1995| Use of fast cosine transform for synthesis filter ST ; 2 |23/06/1995| Aminet release ST ; ; ------------------------------------------------------------------------ ; ; Routines pour le decodage MPEG audio (optimisées) ; ;------------------------------------------------------------------------------*/ ; ; OPTIMIZE: to do: ; - put read2_bits where possible (!d2 is affected) ; - ? move.l d0,(a4) .. move.l d0,4*2*SBLIMIT(a4) .. addq.l #4,a4 ; -> move.l d0,(a4)+ .. move.l d0,(4*2*SBLIMIT)-4(a4) ; - reduce some arrays from LONG to BYTE (bit_alloc, alloc, ...) ; IFND INCLUDE_SOURCE incdir "include:" include "exec/funcdef.i" include "exec/exec.i" include "exec/exec_lib.i" ENDC include "mpega_decode_asm.i" include "mpega_decode_asm_tables.i" include "BitStream.i" include "AudioPort.i" SKIP_UPPER_BAND equ 1 ; if skip upper part of sub band (half & quart freq) ; this flag is used for read & deqnor ; !!! incompatible with FILTER_?_SUM because ; if set filter use all of the sub band even in ; half or quart frequency mode READ3_BITS equ 1 ; if @BSTR_read3_bits func used XREF @BSTR_open XREF @BSTR_close XREF @BSTR_rewind XREF @BSTR_end XREF @BSTR_read_bits XREF @BSTR_skip_bits XREF @BSTR_read3_bits XREF @BSTR_read2_bits XREF @BSTR_seek_sync XREF @BSTR_seek_tell XREF @AU_open XREF @AU_write XREF @AU_close XDEF @ASM_init_decode XDEF @ASM_read_header XDEF @ASM_decode_header XDEF @ASM_I_decode_bitalloc XDEF @ASM_I_decode_scale XDEF @ASM_I_buffer_sample XDEF @ASM_I_deqnorm XDEF @ASM_II_decode_bitalloc XDEF @ASM_II_decode_scale XDEF @ASM_II_buffer_sample XDEF @ASM_II_deqnorm XDEF @ASM_sub_band_synthesis XDEF @ASM_I_decode_frame XDEF @ASM_II_decode_frame section ASMCODE,code ; d0: int sub_band_size ; d1: int output_8bits ; d2: int quality (0..2) ; 0 : 1/4 windowing ; 1 : 1/2 windowing ; 2 : full quality ; @ASM_init_decode movem.l d2-d4,-(sp) move.w d1,output_8bits move.w d0,sub_band_size moveq.l #1,d3 ; skip value for full freq output clr.w half_freq clr.w quar_freq cmp.w #SBLIMIT/2,d0 beq init_decode_1 cmp.w #SBLIMIT/4,d0 beq init_decode_2 bra init_decode_5 init_decode_1 move.w #-1,half_freq moveq.l #2,d3 ; skip value for half freq output bra init_decode_5 init_decode_2 move.w #-1,quar_freq moveq.l #4,d3 ; skip value for quarter freq output init_decode_5 move.w d3,window_skip moveq.l #0,d3 ; start offset for windowing moveq.l #8-1,d4 ; #coeff (#-1)/2 used for windowing move.w d2,quality cmp.b #2,d2 ; best quality beq.s init_decode_6 move.w #2*128,d3 moveq.l #4-1,d4 cmp.b #1,d2 beq.s init_decode_6 move.w #3*128,d3 moveq.l #2-1,d4 init_decode_6 move.w d3,window_start move.w d4,window_coeff init_decode_9 moveq.w #64,d0 lea band_offset,a0 move.w d0,(a0)+ move.w d0,(a0)+ lea filter_buffer,a0 move.w #2*2*HAN_SIZE-1,d0 init_decode_91 clr.w (a0)+ dbra d0,init_decode_91 movem.l (sp)+,d2-d4 rts ; Read header information ; ; a0: frame_params *fr_ps ; -> d0 = 0 if end of bitstream ; @ASM_read_header movem.l a2-a3/d2,-(sp) move.l a0,a3 move.l fp_header(a3),a2 ; a2 = header move.l fp_bs(a3),a0 ; a0 = bitstream moveq.l #8,d1 move.b #$FF,d2 ASM_read_hdr1 bsr @BSTR_read_bits cmp.b d0,d2 beq ASM_read_hdr1 ; skip extra $FF sync bfextu d0{24:1},d1 ; Version move.l d1,hdr_version(a2) bfextu d0{25:2},d2 ; 4 - Layer subq.l #4,d2 neg.l d2 ; Layer move.l d2,hdr_lay(a2) bfextu d0{27:1},d1 ; !Error protection bchg #0,d1 move.l d1,hdr_error_protection(a2) bfextu d0{28:4},d1 ; Bit rate index move.l d1,hdr_bitrate_index(a2) subq.l #1,d2 lsl.l #4,d2 ; (lay-1)*16 add.l d1,d2 lea BITRATE_table,a1 move.w (a1,d2.w*2),d1 ; d1 = bitrate ext.l d1 move.l d1,fp_bitstream_rate(a3) moveq.l #12,d1 bsr @BSTR_read_bits ; Read end of header bfextu d0{20:2},d1 ; Sampling frequency move.l d1,hdr_sampling_frequency(a2) lea FREQ_table,a1 move.l (a1,d1.w*4),d1 move.l d1,fp_bitstream_freq(a3) bfextu d0{22:1},d1 ; Padding move.l d1,hdr_padding(a2) bfextu d0{23:1},d1 ; Extension move.l d1,hdr_extension(a2) bfextu d0{24:2},d1 ; Mode move.l d1,hdr_mode(a2) bfextu d0{26:2},d1 ; Mode ext move.l d1,hdr_mode_ext(a2) bfextu d0{28:1},d1 ; Copyright move.l d1,hdr_copyright(a2) bfextu d0{29:1},d1 ; Original move.l d1,hdr_original(a2) bfextu d0{30:2},d1 ; Emphasis move.l d1,hdr_emphasis(a2) bsr @BSTR_end not.l d0 ASM_read_hdr9 movem.l (sp)+,a2-a3/d2 rts ; Get the right alloc table for layer II ; ; a0: frame_params *fr_ps ; -> d0.l = sub band limit ; a0 is preserved ; ASM_II_get_alloc_table movem.l a2,-(sp) move.l fp_header(a0),a1 ; a1 = header move.l hdr_lay(a1),d0 subq.l #1,d0 lsl.l #4,d0 ; (lay-1)*16 move.l hdr_bitrate_index(a1),d1 add.l d1,d0 lea BITRATE_table,a2 move.w (a2,d0.w*2),d1 ; d1 = bitrate move.l fp_stereo(a0),d0 cmp.b #1,d0 beq ASM_II_get_a00 lsr.w #1,d1 ; d1 = bitrate per channel ASM_II_get_a00 move.l hdr_sampling_frequency(a1),d0 cmp.w #56,d1 blt ASM_II_get_a01 cmp.w #1,d0 ; 48kHz beq ASM_II_get_a0 cmp.w #80,d1 bgt ASM_II_get_a01 ASM_II_get_a0 moveq.l #0,d0 ; -> alloc table 0 lea ALLOC_table0,a1 bra ASM_II_get_a4 ASM_II_get_a01 cmp.w #1,d0 ; 48kHz beq ASM_II_get_a02 cmp.w #96,d1 blt ASM_II_get_a02 ASM_II_get_a1 moveq.l #1,d0 ; -> alloc table 1 lea ALLOC_table1,a1 bra ASM_II_get_a4 ASM_II_get_a02 cmp.w #2,d0 ; 32kHz beq ASM_II_get_a3 cmp.w #48,d1 bgt ASM_II_get_a3 ASM_II_get_a2 moveq.l #2,d0 ; -> alloc table 2 lea ALLOC_table2,a1 bra ASM_II_get_a4 ASM_II_get_a3 moveq.l #3,d0 ; -> alloc table 3 lea ALLOC_table3,a1 ASM_II_get_a4 move.l d0,fp_tab_num(a0) move.l a1,fp_alloc(a0) lea ALLOC_sub_band_limit,a1 move.l (a1,d0.l*4),d0 ; move.l d0,fp_sblimit(a0) movem.l (sp)+,a2 rts ; Get the joint sub band ; ; a0: frame_params *fr_ps ; -> d0.l = join sub band ; a0 is preserved ; ASM_joint_sub_band move.l fp_header(a0),a1 ; a1 = header move.l hdr_lay(a1),d0 subq.l #1,d0 lsl.l #2,d0 ; (lay-1)*4 move.l hdr_mode_ext(a1),d1 add.l d1,d0 lea JSB_table,a1 move.l (a1,d0.l*4),d0 rts ; Decode header ; ; a0: frame_params *fr_ps ; @ASM_decode_header movem.l a2,-(sp) move.l fp_header(a0),a2 ; a2 = header move.l hdr_mode(a2),d1 move.l d1,fp_actual_mode(a0) moveq.l #1,d0 cmp.w #MPG_MD_MONO,d1 beq ASM_decode_he1 addq.l #1,d0 ASM_decode_he1 move.l d0,fp_stereo(a0) move.l d0,fp_channels(a0) moveq.l #SBLIMIT,d0 move.l hdr_lay(a2),d1 cmp.w #2,d1 bne ASM_decode_he2 bsr ASM_II_get_alloc_table ASM_decode_he2 move.l d0,fp_sblimit(a0) move.l hdr_mode(a2),d1 cmp.w #MPG_MD_JOINT_STEREO,d1 bne ASM_decode_he3 bsr ASM_joint_sub_band ASM_decode_he3 move.l d0,fp_jsbound(a0) movem.l (sp)+,a2 rts ; Decode bit allocation (MPEG I Lay I) ; ; a0: frame_params *fr_ps @ASM_I_decode_bitalloc movem.l d2-d3/a2-a5,-(sp) move.l a0,a2 ; a2 = frame_params move.l fp_bs(a2),a0 ; a0 = bitstream move.l fp_bit_alloc(a2),a4 ; a4 = &bitalloc[0][0] moveq.l #4,d1 ; 4 bits per bitalloc cmp.l #1,fp_stereo(a2) bgt.s I_dec_bitall5 moveq.l #SBLIMIT-1,d2 I_dec_bitall1 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4)+ dbra d2,I_dec_bitall1 bra.s I_dec_bitall9 I_dec_bitall5 lea.l 4*SBLIMIT(a4),a5 ; a5 = &bit_alloc[1][0] move.l fp_jsbound(a2),d2 subq.l #1,d2 bmi.s I_dec_bitall7 ; jsbound <= 0 ? I_dec_bitall6 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4)+ bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a5)+ dbra d2,I_dec_bitall6 I_dec_bitall7 moveq.l #SBLIMIT,d2 sub.l fp_jsbound(a2),d2 subq.l #1,d2 bmi.s I_dec_bitall9 ; jsbound >= SBLIMIT I_dec_bitall8 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4)+ move.l d0,(a5)+ ; same band for 2 channels dbra d2,I_dec_bitall8 I_dec_bitall9 movem.l (sp)+,d2-d3/a2-a5 rts ; Decode scale (MPEG I Lay I) ; ; a0: frame_params *fr_ps ; ; OPTIMIZATION: bitalloc, scale_index -> BYTE ; @ASM_I_decode_scale movem.l d2-d7/a2-a6,-(sp) move.l a0,a2 ; a2 = frame_params move.l fp_bs(a2),a0 ; a0 = bitstream move.l fp_scale_index(a2),a3 move.l fp_bit_alloc(a2),a4 ; a4 = &bitalloc[0][0] moveq.l #6,d1 ; 6 bits per scale_index moveq.l #SBLIMIT-1,d7 cmp.l #1,fp_stereo(a2) bgt I_dec_scale5 ; stereo I_dec_scale1 moveq.l #SCALE_RANGE-1,d0 tst.l (a4)+ beq.s I_dec_scale11 bsr @BSTR_read_bits ; read d1 bits from bitstream I_dec_scale11 move.l d0,(a3)+ dbra d7,I_dec_scale1 bra.s I_dec_scale9 I_dec_scale5 lea 4*SBLIMIT(a4),a5 I_dec_scale51 moveq.l #SCALE_RANGE-1,d0 tst.l (a4)+ beq.s I_dec_scale52 bsr @BSTR_read_bits ; read d1 bits from bitstream I_dec_scale52 move.l d0,(a3) moveq.l #SCALE_RANGE-1,d0 tst.l (a5)+ beq.s I_dec_scale53 bsr @BSTR_read_bits ; read d1 bits from bitstream I_dec_scale53 move.l d0,4*3*SBLIMIT(a3) addq.l #4,a3 dbra d7,I_dec_scale51 I_dec_scale9 movem.l (sp)+,d2-d7/a2-a6 rts ; Decode bit allocation (MPEG I Lay II) ; ; a0: frame_params *fr_ps @ASM_II_decode_bitalloc movem.l d2-d3/a2-a5,-(sp) move.l a0,a2 ; a2 = frame_params move.l fp_bs(a2),a0 ; a0 = bitstream move.l fp_bit_alloc(a2),a1 ; a1 = &bitalloc[0][0] move.l fp_alloc(a2),a3 ; a3 = &alloc[0][0] move.l a1,a4 ; a4 = &bitalloc[0][0] moveq.l #4*16,d3 ; d3 = next alloc step cmp.l #1,fp_stereo(a2) bgt.s II_dec_bitall5 move.l fp_sblimit(a2),d2 subq.l #1,d2 bmi.s II_dec_bitall2 ; sblimit <= 0 ? II_dec_bitall1 bfextu (a3){ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4)+ add.l d3,a3 ; next -> &alloc[i++][0] dbra d2,II_dec_bitall1 bra.s II_dec_bitall9 II_dec_bitall2 ; moveq.l #SBLIMIT,d2 move.w sub_band_size,d2 ext.l d2 sub.l fp_sblimit(a2),d2 subq.l #1,d2 bmi.s II_dec_bitall9 ; sblimit >= sub_band_size clr.l d0 ; 0 -> upper band II_dec_bitall3 move.l d0,(a4)+ dbra d2,II_dec_bitall3 bra.s II_dec_bitall9 II_dec_bitall5 move.l a4,a5 add.l #4*SBLIMIT,a5 ; a5 = &bit_alloc[1][0] move.l fp_jsbound(a2),d2 subq.l #1,d2 bmi.s II_dec_bitall8 ; jsbound <= 0 ? II_dec_bitall6 bfextu (a3){ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4)+ bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a5)+ add.l d3,a3 ; next -> &alloc[i++][0] dbra d2,II_dec_bitall6 move.l fp_sblimit(a2),d2 sub.l fp_jsbound(a2),d2 subq.l #1,d2 bmi.s II_dec_bitall8 ; jsbound >= sblimit II_dec_bitall7 bfextu (a3){ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4)+ move.l d0,(a5)+ ; same band for 2 channels add.l d3,a3 ; next -> &alloc[i++][0] dbra d2,II_dec_bitall7 II_dec_bitall8 ; moveq.l #SBLIMIT,d2 move.w sub_band_size,d2 ext.l d2 sub.l fp_sblimit(a2),d2 subq.l #1,d2 bmi.s II_dec_bitall9 ; sblimit >= sub_band_size clr.l d0 ; 0 -> upper band II_dec_bitall81 move.l d0,(a4)+ move.l d0,(a5)+ dbra d2,II_dec_bitall81 II_dec_bitall9 movem.l (sp)+,d2-d3/a2-a5 rts ; Decode scale (MPEG I Lay II) ; ; a0: frame_params *fr_ps ; ; OPTIMIZATION: bitalloc, scfsi, scale_index -> BYTE ; @ASM_II_decode_scale movem.l d2-d7/a2-a6,-(sp) move.l a0,a2 ; a2 = frame_params move.l fp_bs(a2),a0 ; a0 = bistream move.l fp_bit_alloc(a2),a1 move.l fp_scfsi(a2),a3 move.l fp_scale_index(a2),a4 move.l a1,a5 ; a5 = &bitalloc[0][0] moveq.l #2,d1 ; 2 bits per scfsi cmp.l #1,fp_stereo(a2) bgt II_dec_scale5 ; stereo move.l fp_sblimit(a2),d7 subq.l #1,d7 bmi II_dec_scale3 ; sblimit <= 0 ! movem.l a3/a5,-(sp) II_dec_scale1 tst.l (a5)+ beq.s II_dec_scale11 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a3) II_dec_scale11 addq.l #4,a3 dbra d7,II_dec_scale1 movem.l (sp)+,a3/a5 II_dec_scale2 move.l fp_sblimit(a2),d7 subq.l #1,d7 moveq.l #6,d1 II_dec_scale21 tst.l (a5)+ beq.s II_dec_scale26 move.l (a3)+,d0 bne.s II_dec_scale22 ; scfsi[0][i] > 0 bsr @BSTR_read3_bits ; read 3*d1 bits from bitstream move.l d4,(a4) ; scale_index[0][0][i] move.l d5,4*SBLIMIT(a4) ; scale_index[0][1][i] move.l d6,8*SBLIMIT(a4) ; scale_index[0][2][i] bra.s II_dec_scale27 II_dec_scale22 subq.b #1,d0 bne.s II_dec_scale23 ; scfsi[0][i] > 1 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,8*SBLIMIT(a4) ; scale_index[0][2][i] bra.s II_dec_scale27 II_dec_scale23 subq.b #1,d0 bne.s II_dec_scale24 ; scfsi[0][i] > 2 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] bra.s II_dec_scale27 II_dec_scale24 ; scfsi[0][i] == 3 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; scale_index[0][0][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*SBLIMIT(a4) ; scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] bra.s II_dec_scale27 II_dec_scale26 moveq.l #SCALE_RANGE-1,d0 move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] addq.l #4,a3 ; next scfsi II_dec_scale27 addq.l #4,a4 ; next scale_index dbra d7,II_dec_scale21 II_dec_scale3 ; moveq.l #SBLIMIT,d7 move.w sub_band_size,d7 ext.l d7 sub.l fp_sblimit(a2),d7 subq.l #1,d7 bmi II_dec_scale9 ; sblimit >= sub_band_size moveq.l #SCALE_RANGE-1,d0 II_dec_scale31 move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] addq.l #4,a4 ; next scale_index dbra d7,II_dec_scale31 bra II_dec_scale9 II_dec_scale5 move.l fp_sblimit(a2),d7 subq.l #1,d7 bmi II_dec_scale8 ; sblimit <= 0 ! movem.l a3/a5,-(sp) II_dec_scale51 tst.l (a5) beq.s II_dec_scale52 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a3) II_dec_scale52 tst.l 4*SBLIMIT(a5) beq.s II_dec_scale53 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*SBLIMIT(a3) II_dec_scale53 addq.l #4,a3 addq.l #4,a5 dbra d7,II_dec_scale51 movem.l (sp)+,a3/a5 move.l fp_sblimit(a2),d7 subq.l #1,d7 moveq.l #6,d1 II_dec_scale61 tst.l (a5) beq.s II_dec_scale66 move.l (a3),d0 bne.s II_dec_scale62 ; scfsi[0][i] > 0 bsr @BSTR_read3_bits ; read 3*d1 bits from bitstream move.l d4,(a4) ; scale_index[0][0][i] move.l d5,4*SBLIMIT(a4) ; scale_index[0][1][i] move.l d6,8*SBLIMIT(a4) ; scale_index[0][2][i] bra.s II_dec_scale67 II_dec_scale62 subq.b #1,d0 bne.s II_dec_scale63 ; scfsi[0][i] > 1 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,8*SBLIMIT(a4) ; scale_index[0][2][i] bra.s II_dec_scale67 II_dec_scale63 subq.b #1,d0 bne.s II_dec_scale64 ; scfsi[0][i] > 2 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] bra.s II_dec_scale67 II_dec_scale64 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; scale_index[0][0][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*SBLIMIT(a4) ; scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] bra.s II_dec_scale67 II_dec_scale66 moveq.l #SCALE_RANGE-1,d0 move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] II_dec_scale67 tst.l 4*SBLIMIT(a5) beq.s II_dec_scale76 move.l 4*SBLIMIT(a3),d0 bne.s II_dec_scale72 ; scfsi[0][i] > 0 bsr @BSTR_read3_bits ; read 3*d1 bits from bitstream move.l d4,4*3*SBLIMIT(a4) ; scale_index[1][0][i] move.l d5,4*4*SBLIMIT(a4) ; scale_index[1][1][i] move.l d6,4*5*SBLIMIT(a4) ; scale_index[1][2][i] bra.s II_dec_scale77 II_dec_scale72 subq.b #1,d0 bne.s II_dec_scale73 ; scfsi[0][i] > 1 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*3*SBLIMIT(a4) ; scale_index[1][0][i] move.l d0,4*4*SBLIMIT(a4) ; = scale_index[1][1][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*5*SBLIMIT(a4) ; scale_index[1][2][i] bra.s II_dec_scale77 II_dec_scale73 subq.b #1,d0 bne.s II_dec_scale74 ; scfsi[0][i] > 2 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*3*SBLIMIT(a4) ; scale_index[1][0][i] move.l d0,4*4*SBLIMIT(a4) ; = scale_index[1][1][i] move.l d0,4*5*SBLIMIT(a4) ; = scale_index[1][2][i] bra.s II_dec_scale77 II_dec_scale74 bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*3*SBLIMIT(a4) ; scale_index[1][0][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*4*SBLIMIT(a4) ; = scale_index[1][1][i] move.l d0,4*5*SBLIMIT(a4) ; = scale_index[1][2][i] bra.s II_dec_scale77 II_dec_scale76 moveq.l #SCALE_RANGE-1,d0 move.l d0,4*3*SBLIMIT(a4) ; scale_index[1][0][i] move.l d0,4*4*SBLIMIT(a4) ; = scale_index[1][1][i] move.l d0,4*5*SBLIMIT(a4) ; = scale_index[1][2][i] II_dec_scale77 addq.l #4,a3 ; next scfsi addq.l #4,a4 ; next scale_index addq.l #4,a5 ; next bitalloc dbra d7,II_dec_scale61 II_dec_scale8 ; moveq.l #SBLIMIT,d7 move.w sub_band_size,d7 ext.l d7 sub.l fp_sblimit(a2),d7 subq.l #1,d7 bmi II_dec_scale9 ; sblimit >= sub_band_size moveq.l #SCALE_RANGE-1,d0 II_dec_scale81 move.l d0,(a4) ; scale_index[0][0][i] move.l d0,4*SBLIMIT(a4) ; = scale_index[0][1][i] move.l d0,8*SBLIMIT(a4) ; = scale_index[0][2][i] move.l d0,4*3*SBLIMIT(a4) ; scale_index[1][0][i] move.l d0,4*4*SBLIMIT(a4) ; = scale_index[1][1][i] move.l d0,4*5*SBLIMIT(a4) ; = scale_index[1][2][i] addq.l #4,a4 ; next scale_index dbra d7,II_dec_scale81 ; bra II_dec_scale9 II_dec_scale9 movem.l (sp)+,d2-d7/a2-a6 rts ; Read a sub band sample (MPEG I Lay I) into a buffer ; ; a0: frame_params *fr_ps ; @ASM_I_buffer_sample movem.l d2-d7/a0-a6,-(sp) move.l fp_sblimit(a0),d0 move.l fp_channels(a0),d1 move.l fp_jsbound(a0),d2 move.l fp_alloc(a0),a3 move.l fp_bit_alloc(a0),a2 move.l fp_sample(a0),a1 move.l fp_bs(a0),a0 move.w d0,bsamp_sblimit move.w d1,bsamp_channels move.w d2,bsamp_jsbound move.l a1,a4 ; a4 = &sample[0][0][0] clr.l d7 ; d7=i:0..sblimit ASM_I_bsamp_1 move.l (0,a2,d7.w*4),d1 ; d1 = bit_alloc[0][d7] beq ASM_I_bsamp_3 ; no bit allocated for this sample addq.l #1,d1 ; d1 = bits to read bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; sample[0][0][i] bra.s ASM_I_bsamp_4 ASM_I_bsamp_3 clr.l (a4) ; sample[0][0][i] = 0 ASM_I_bsamp_4 cmp.w #1,bsamp_channels beq ASM_I_bsamp_8 ; only 1 channel cmp.w bsamp_jsbound,d7 bge ASM_I_bsamp_7 ; i>=jsbound & stereo -> L=R move.l (4*SBLIMIT,a2,d7.w*4),d1 ; d1 = bit_alloc[1][d7] beq ASM_I_bsamp_6 ; no bit allocated for this sample addq.l #1,d1 ; d1 = bits to read bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*3*SBLIMIT(a4) ; sample[1][0][i] bra ASM_I_bsamp_8 ASM_I_bsamp_6 clr.l 4*3*SBLIMIT(a4) ; sample[1][0][i] = 0 bra ASM_I_bsamp_8 ASM_I_bsamp_7 move.l (a4),4*3*SBLIMIT(a4) ; sample[1][0][i] = sample[0][0][i] ASM_I_bsamp_8 addq.l #4,a4 ; next sample ptr addq.w #1,d7 ; i++ IFND SKIP_UPPER_BAND cmp.w bsamp_sblimit,d7 blt ASM_I_bsamp_1 ; i= sub_band_size -> skip others cmp.w bsamp_sblimit,d7 blt ASM_I_bsamp_1 bra ASM_I_bsamp_9 ASM_I_bsamp_81 cmp.w bsamp_sblimit,d7 bge ASM_I_bsamp_9 ; i >= sblimit -> end of read move.l (0,a2,d7.w*4),d1 ; d1 = bit_alloc[0][d7] beq ASM_I_bsamp_84 ; no bit allocated for this sample addq.l #1,d1 ; d1 = bits to read bsr @BSTR_skip_bits ; skip d1 bits from bitstream ASM_I_bsamp_84 cmp.w #1,bsamp_channels beq ASM_I_bsamp_88 ; only 1 channel cmp.w bsamp_jsbound,d7 bge ASM_I_bsamp_88 ; i>=jsbound & stereo -> L=R move.l (4*SBLIMIT,a2,d7.w*4),d1 ; d1 = bit_alloc[1][d7] beq.s ASM_I_bsamp_88 ; no bit allocated for this sample addq.l #1,d1 ; d1 = bits to read bsr @BSTR_skip_bits ; skip d1 bits from bitstream ASM_I_bsamp_88 addq.l #4,a4 ; next sample ptr addq.w #1,d7 ; i++ bra ASM_I_bsamp_81 ENDC ASM_I_bsamp_9 IFND SKIP_UPPER_BAND cmp.w #SBLIMIT,d7 ELSE cmp.w sub_band_size,d7 ENDC bge ASM_I_bsamp_end addq.w #1,d7 clr.l (a4) ; sample[0][0][i] = 0 cmp.w #1,bsamp_channels beq ASM_I_bsamp_91 ; only 1 channel clr.l 4*3*SBLIMIT(a4) ; sample[1][0][i] = 0 ASM_I_bsamp_91 addq.l #4,a4 ; next sample ptr bra ASM_I_bsamp_9 ASM_I_bsamp_end movem.l (sp)+,d2-d7/a0-a6 rts ; Read a sub band sample (MPEG I Lay II) into a buffer ; ; a0: frame_params *fr_ps ; @ASM_II_buffer_sample movem.l d2-d7/a0-a6,-(sp) move.l fp_sblimit(a0),d0 move.l fp_channels(a0),d1 move.l fp_jsbound(a0),d2 move.l fp_alloc(a0),a3 move.l fp_bit_alloc(a0),a2 move.l fp_sample(a0),a1 move.l fp_bs(a0),a0 move.w d0,bsamp_sblimit move.w d1,bsamp_channels move.w d2,bsamp_jsbound move.l a1,a4 ; a4 = &sample[0][0][0] clr.l d7 ; d7=i:0..sblimit ASM_II_bsamp_1 move.l (0,a2,d7.w*4),d3 ; d3 = bit_alloc[0][d7] beq ASM_II_bsamp_3 ; no bit allocated for this sample move.l (a3,d3.w*4),d3 ; d3 = alloc_table[i][d3] bfextu d3{ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits btst #ALLOC_B_GROUP,d3 beq ASM_II_bsamp_2 ; group <> 3 ; read each sub band sample separately IFND READ3_BITS bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,(a4) ; sample[0][0][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*SBLIMIT(a4) ; sample[0][1][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,8*SBLIMIT(a4) ; sample[0][2][i] ELSE bsr @BSTR_read3_bits ; read 3*d1 bits from bitstream move.l d4,(a4) ; sample[0][0][i] move.l d5,4*SBLIMIT(a4) ; sample[0][1][i] move.l d6,8*SBLIMIT(a4) ; sample[0][2][i] ENDC bra ASM_II_bsamp_4 ASM_II_bsamp_2 ; bit_alloc.steps = 3, 5, 9 bfextu d3{ALLOC_O_STEPS:ALLOC_W_STEPS},d4 ; d4 = steps bsr @BSTR_read_bits ; read d1 bits from bitstream move.w d0,d1 divs.w d4,d1 ; d1/d4 -> d1.h=r d1.l=q move.l d1,d0 swap d0 ext.l d0 move.l d0,(a4) ; r -> sample[0][0][i] ext.l d1 divs.w d4,d1 move.l d1,d0 swap d0 ext.l d0 move.l d0,4*SBLIMIT(a4) ; r -> sample[0][1][i] ext.l d1 move.l d1,8*SBLIMIT(a4) ; q -> sample[0][2][i] bra ASM_II_bsamp_4 ASM_II_bsamp_3 clr.l (a4) ; sample[0][0][i] = 0 clr.l 4*SBLIMIT(a4) ; sample[0][1][i] = 0 clr.l 8*SBLIMIT(a4) ; sample[0][2][i] = 0 ASM_II_bsamp_4 cmp.w #1,bsamp_channels beq ASM_II_bsamp_8 ; only 1 channel cmp.w bsamp_jsbound,d7 bge ASM_II_bsamp_7 ; i>=jsbound & stereo -> L=R move.l (4*SBLIMIT,a2,d7.w*4),d3 ; d3 = bit_alloc[1][d7] beq ASM_II_bsamp_6 ; no bit allocated for this sample move.l (a3,d3.w*4),d3 ; d3 = alloc_table[i][d3] bfextu d3{ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits btst #ALLOC_B_GROUP,d3 beq ASM_II_bsamp_5 ; group <> 3 IFND READ3_BITS bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*3*SBLIMIT(a4) ; sample[1][0][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*4*SBLIMIT(a4) ; sample[1][1][i] bsr @BSTR_read_bits ; read d1 bits from bitstream move.l d0,4*5*SBLIMIT(a4) ; sample[1][2][i] ELSE bsr @BSTR_read3_bits ; read 3*d1 bits from bitstream move.l d4,4*3*SBLIMIT(a4) ; sample[1][0][i] move.l d5,4*4*SBLIMIT(a4) ; sample[1][1][i] move.l d6,4*5*SBLIMIT(a4) ; sample[1][2][i] ENDC bra ASM_II_bsamp_8 ASM_II_bsamp_5 ; bit_alloc.steps = 3, 5, 9 bfextu d3{ALLOC_O_STEPS:ALLOC_W_STEPS},d4 ; d4 = steps bsr @BSTR_read_bits ; read d1 bits from bitstream tst.w d4 beq.s ASM_II_bsamp_6 ; if d4=0 (should not happend !) move.w d0,d1 divs.w d4,d1 ; d1/d4 -> d1.h=r d1.l=q move.l d1,d0 swap d0 ext.l d0 move.l d0,4*3*SBLIMIT(a4) ; r -> sample[1][0][i] ext.l d1 divs.w d4,d1 move.l d1,d0 swap d0 ext.l d0 move.l d0,4*4*SBLIMIT(a4) ; r -> sample[1][1][i] ext.l d1 move.l d1,4*5*SBLIMIT(a4) ; q -> sample[1][2][i] bra ASM_II_bsamp_8 ASM_II_bsamp_6 clr.l 4*3*SBLIMIT(a4) ; sample[1][0][i] = 0 clr.l 4*4*SBLIMIT(a4) ; sample[1][1][i] = 0 clr.l 4*5*SBLIMIT(a4) ; sample[1][2][i] = 0 bra ASM_II_bsamp_8 ASM_II_bsamp_7 move.l (a4),4*3*SBLIMIT(a4) ; sample[1][0][i] = sample[0][0][i] move.l 4*SBLIMIT(a4),4*4*SBLIMIT(a4) ; sample[1][1][i] = sample[0][1][i] move.l 8*SBLIMIT(a4),4*5*SBLIMIT(a4) ; sample[1][2][i] = sample[0][2][i] ASM_II_bsamp_8 addq.l #4,a4 ; next sample ptr add.l #4*16,a3 ; next alloc_table ptr addq.w #1,d7 ; i++ IFND SKIP_UPPER_BAND cmp.w bsamp_sblimit,d7 blt ASM_II_bsamp_1 ; i= sub_band_size -> skip others cmp.w bsamp_sblimit,d7 blt ASM_II_bsamp_1 bra ASM_II_bsamp_9 ASM_II_bsamp_81 cmp.w bsamp_sblimit,d7 bge ASM_II_bsamp_9 ; i >= sblimit -> end of read move.l (0,a2,d7.w*4),d3 ; d3 = bit_alloc[0][d7] beq ASM_II_bsamp_84 ; no bit allocated for this sample move.l (a3,d3.w*4),d3 ; d3 = alloc_table[i][d3] bfextu d3{ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits btst #ALLOC_B_GROUP,d3 beq ASM_II_bsamp_82 ; group <> 3 ; read each sub band sample separately move.l d1,d2 add.l d1,d1 add.l d2,d1 bsr @BSTR_skip_bits ; skip 3*d1 bits from bitstream bra ASM_II_bsamp_84 ASM_II_bsamp_82 bsr @BSTR_skip_bits ; skip d1 bits from bitstream ASM_II_bsamp_84 cmp.w #1,bsamp_channels beq ASM_II_bsamp_88 ; only 1 channel cmp.w bsamp_jsbound,d7 bge ASM_II_bsamp_88 ; i>=jsbound & stereo -> L=R move.l (4*SBLIMIT,a2,d7.w*4),d3 ; d3 = bit_alloc[1][d7] beq.s ASM_II_bsamp_88 ; no bit allocated for this sample move.l (a3,d3.w*4),d3 ; d3 = alloc_table[i][d3] bfextu d3{ALLOC_O_BITS:ALLOC_W_BITS},d1 ; d1 = bits btst #ALLOC_B_GROUP,d3 beq ASM_II_bsamp_85 ; group <> 3 ; read each sub band sample separately move.l d1,d2 add.l d1,d1 add.l d2,d1 bsr @BSTR_skip_bits ; skip 3*d1 bits from bitstream bra ASM_II_bsamp_88 ASM_II_bsamp_85 bsr @BSTR_skip_bits ; skip d1 bits from bitstream ASM_II_bsamp_88 addq.l #4,a4 ; next sample ptr add.l #4*16,a3 ; next alloc_table ptr addq.w #1,d7 ; i++ bra ASM_II_bsamp_81 ENDC ASM_II_bsamp_9 IFND SKIP_UPPER_BAND cmp.w #SBLIMIT,d7 ELSE cmp.w sub_band_size,d7 ENDC bge ASM_II_bsamp_end addq.w #1,d7 clr.l (a4) ; sample[0][0][i] = 0 clr.l 4*SBLIMIT(a4) ; sample[0][1][i] = 0 clr.l 8*SBLIMIT(a4) ; sample[0][2][i] = 0 cmp.w #1,bsamp_channels beq ASM_II_bsamp_91 ; only 1 channel clr.l 4*3*SBLIMIT(a4) ; sample[1][0][i] = 0 clr.l 4*4*SBLIMIT(a4) ; sample[1][1][i] = 0 clr.l 4*5*SBLIMIT(a4) ; sample[1][2][i] = 0 ASM_II_bsamp_91 addq.l #4,a4 bra ASM_II_bsamp_9 ASM_II_bsamp_end movem.l (sp)+,d2-d7/a0-a6 rts ; Dequantize & denormalize samples (layer I) ; ; a0: frame_params *fr_ps @ASM_I_deqnorm movem.l d2-d7/a2-a6,-(sp) move.l fp_sblimit(a0),d0 move.l fp_stereo(a0),d1 move.l fp_scale_index(a0),a3 move.l fp_fraction(a0),a2 move.l fp_bit_alloc(a0),a1 move.l fp_sample(a0),a0 move.w d0,sub_band_limit ; sblimit IFD SKIP_UPPER_BAND cmp.w sub_band_size,d0 ble.s ASM_I_deqnorm_0 ; sub_band_limit <= sub_band_size move.w sub_band_size,sub_band_limit ASM_I_deqnorm_0 ENDC move.w d1,channels ; stereo ? mulu.w #SBLIMIT,d2 ; x*SBLIMIT moveq.l #MULT_FACTOR,d7 moveq.l #DEQUANT_FACTOR,d6 lea.l mult_table,a5 ASM_I_deqnorm_1 movem.l d2/a0-a4,-(sp) move.l a0,a6 ; a6 = sample[k][0][0] ; lea.l (a3,d2.w*4),a3 ; a3 = &scale_index[k][x][0] move.w sub_band_limit,d0 move.w d0,sub_band_index lea.l dequantize_I_table,a0 ASM_I_deqnorm_2 subq.w #1,sub_band_index bmi ASM_I_deqnorm_7 move.l (a3)+,d0 ; d0 = *scale++ move.l (a1)+,d1 ; d1 = *bit_alloc++; beq ASM_I_deqnorm_5 ; no bit allocated move.w (a5,d0.w*2),d5 ; m = d5; move.w (a0,d1.w*4),d3 ; d move.w (2,a0,d1.w*4),d4 ; c muls.w d5,d4 ; c*m asr.l d7,d4 ; >>MULT_FACTOR move.l (a6)+,d0 ; sample[k][0][i] moveq.w #31-16,d2 sub.w d1,d2 asl.l d2,d0 ; left justify sample (normalize) bchg #15,d0 ; Invert Msb of sample asr.w #15-INT_FACTOR,d0 add.w d3,d0 ; d0 + d muls.w d4,d0 ; * c*m asr.l d6,d0 ; >>DEQUANT_FACTOR move.w d0,(a2)+ ; -> fraction[k][0][i] bra ASM_I_deqnorm_2 ; next sub_band ASM_I_deqnorm_5 addq.l #4,a6 ; -> sample ptr ++ (skipped) clr.w (a2)+ ; fraction[k][0][i] = 0 ASM_I_deqnorm_6 bra ASM_I_deqnorm_2 ; next sub_band ASM_I_deqnorm_7 move.w sub_band_limit,d0 ASM_I_deqnorm_8 IFND SKIP_UPPER_BAND ; 27/05/1995 cmp.w #SBLIMIT,d0 ELSE ; !!! incompatible with FILTER_?_SUM cmp.w sub_band_size,d0 ENDC bge ASM_I_deqnorm_9 clr.w (a2)+ ; fraction[k][0][i] = 0 addq.w #1,d0 bra ASM_I_deqnorm_8 ASM_I_deqnorm_9 movem.l (sp)+,d2/a0-a4 subq.w #1,channels ble ASM_I_deqnorm_end add.l #4*3*SBLIMIT,a0 ; next -> &sample[k][0][0] add.l #4*SBLIMIT,a1 ; next -> &bit_alloc[k][0] add.l #2*3*SBLIMIT,a2 ; next -> &fraction[k][0][0] add.l #4*3*SBLIMIT,a3 ; next -> &scale_index[k][0][0] bra ASM_I_deqnorm_1 ASM_I_deqnorm_end movem.l (sp)+,d2-d7/a2-a6 rts ; Dequantize & denormalize samples (layer II) ; ; a0: frame_params *fr_ps ; d0.l: int x (0..2) @ASM_II_deqnorm movem.l d2-d7/a2-a6,-(sp) move.l d0,d2 move.l fp_sblimit(a0),d0 move.l fp_stereo(a0),d1 move.l fp_alloc(a0),a4 move.l fp_scale_index(a0),a3 move.l fp_fraction(a0),a2 move.l fp_bit_alloc(a0),a1 move.l fp_sample(a0),a0 move.w d0,sub_band_limit ; sblimit IFD SKIP_UPPER_BAND cmp.w sub_band_size,d0 ble.s ASM_II_deqnorm_0 ; sub_band_limit <= sub_band_size move.w sub_band_size,sub_band_limit ASM_II_deqnorm_0 ENDC move.w d1,channels ; stereo ? mulu.w #SBLIMIT,d2 ; x*SBLIMIT moveq.l #MULT_FACTOR,d7 moveq.l #DEQUANT_FACTOR,d6 lea.l mult_table,a5 ASM_II_deqnorm_1 movem.l d2/a0-a4,-(sp) move.l a0,a6 ; a6 = sample[k][0][0] lea.l (a3,d2.w*4),a3 ; a3 = &scale_index[k][x][0] move.w sub_band_limit,d0 move.w d0,sub_band_index lea.l dequantize_II_table,a0 ASM_II_deqnorm_2 subq.w #1,sub_band_index bmi ASM_II_deqnorm_7 move.l (a3)+,d0 ; d0 = *scale++ move.l (a1)+,d1 ; d1 = *bit_alloc++; beq ASM_II_deqnorm_5 ; no bit allocated move.w (a5,d0.w*2),d5 ; m = d5; move.l (a4,d1.w*4),d0 ; d0 = alloc_table[i][d3] bfextu d0{ALLOC_O_SHIFT:ALLOC_W_SHIFT},d2 ; d2 = shift bfextu d0{ALLOC_O_QUANT:ALLOC_W_QUANT},d1 ; d1 = quant move.w (a0,d1.w*4),d3 ; d move.w (2,a0,d1.w*4),d4 ; c muls.w d5,d4 ; c*m asr.l d7,d4 ; >>MULT_FACTOR move.l (a6),d0 ; sample[k][0][i] asl.l d2,d0 ; left justify sample (normalize) bchg #15,d0 ; Invert Msb of sample asr.w #15-INT_FACTOR,d0 add.w d3,d0 ; d0 + d muls.w d4,d0 ; * c*m asr.l d6,d0 ; >>DEQUANT_FACTOR move.w d0,(a2) ; -> fraction[k][0][i] move.l SBLIMIT*4(a6),d0 ; sample[k][1][i] asl.l d2,d0 ; left justify sample (normalize) bchg #15,d0 ; Invert Msb of sample asr.w #15-INT_FACTOR,d0 add.w d3,d0 ; d0 + d muls.w d4,d0 ; * c*m asr.l d6,d0 ; >>DEQUANT_FACTOR move.w d0,SBLIMIT*2(a2) ; -> fraction[k][1][i] move.l SBLIMIT*8(a6),d0 ; sample[k][2][i] asl.l d2,d0 ; left justify sample (normalize) bchg #15,d0 ; Invert Msb of sample asr.w #15-INT_FACTOR,d0 add.w d3,d0 ; d0 + d muls.w d4,d0 ; * c*m asr.l d6,d0 ; >>DEQUANT_FACTOR move.w d0,SBLIMIT*4(a2) ; -> fraction[k][2][i] bra ASM_II_deqnorm_6 ASM_II_deqnorm_5 clr.w (a2) ; fraction[k][0][i] = 0 clr.w SBLIMIT*2(a2) ; fraction[k][1][i] = 0 clr.w SBLIMIT*4(a2) ; fraction[k][2][i] = 0 ASM_II_deqnorm_6 addq.l #4,a6 ; -> sample ptr ++ addq.l #2,a2 ; -> fraction ptr ++ add.l #4*16,a4 ; next alloc_table ptr bra ASM_II_deqnorm_2 ; next sub_band ASM_II_deqnorm_7 move.w sub_band_limit,d0 ASM_II_deqnorm_8 IFND SKIP_UPPER_BAND ; 08/05/1995 cmp.w #SBLIMIT,d0 ELSE ; !!! incompatible with FILTER_?_SUM cmp.w sub_band_size,d0 ENDC bge ASM_II_deqnorm_9 clr.w (a2) ; fraction[k][0][i] = 0 clr.w SBLIMIT*2(a2) ; fraction[k][1][i] = 0 clr.w SBLIMIT*4(a2) ; fraction[k][2][i] = 0 addq.l #2,a2 ; -> fraction ptr ++ addq.w #1,d0 bra ASM_II_deqnorm_8 ASM_II_deqnorm_9 movem.l (sp)+,d2/a0-a4 subq.w #1,channels ble ASM_II_deqnorm_end add.l #4*3*SBLIMIT,a0 ; next -> &sample[k][0][0] add.l #4*SBLIMIT,a1 ; next -> &bit_alloc[k][0] add.l #2*3*SBLIMIT,a2 ; next -> &fraction[k][0][0] add.l #4*3*SBLIMIT,a3 ; next -> &scale_index[k][0][0] bra ASM_II_deqnorm_1 ASM_II_deqnorm_end movem.l (sp)+,d2-d7/a2-a6 rts ; Apply the synthesis filter to a sub band ; ; a0: bandPtr (=fraction) ; a1: out_filter_buffer ; @ASM_filter_band tst.w quar_freq bne fast_filter_quart tst.w half_freq bne fast_filter_half bra fast_filter_full ; ; Filter COS values for fast cosine transform ; COS1_64 equ $8028 COS3_64 equ $8167 COS5_64 equ $83F4 COS7_64 equ $87F2 COS9_64 equ $8D98 COS11_64 equ $953B COS13_64 equ $9F5C COS15_64 equ $ACC0 COS17_64 equ $BE9A COS19_64 equ $D6E0 COS21_64 equ $F8FA COS23_64 equ $12B60 COS25_64 equ $17BF2 COS27_64 equ $20ECB COS29_64 equ $36859 COS31_64 equ $A30A4 COS1_32 equ $809F COS3_32 equ $85C2 COS5_32 equ $9123 COS7_32 equ $A596 COS9_32 equ $C9C5 COS11_32 equ $10F89 COS13_32 equ $1B8F2 COS15_32 equ $519E5 COS1_16 equ $8282 COS3_16 equ $99F2 COS5_16 equ $E665 COS7_16 equ $2901B COS1_8 equ $8A8C COS3_8 equ $14E7B COS1_4 equ $B505 FF_MUL MACRO move.l (a2)+,d0 sub.l -(a3),d0 muls.l #\1,d1:d0 move.w d1,d0 swap d0 move.l d0,(a4)+ ENDM FF_ADD MACRO move.l \1*4(a2),d0 add.l \2*4(a2),d0 move.l d0,(a4)+ ENDM FF_SUB_MUL MACRO move.l \1*4(a2),d0 sub.l \2*4(a2),d0 muls.l \3,d4:d0 move.w d4,d0 swap d0 move.l d0,(a4)+ ENDM FFF_MUL MACRO move.l (a2)+,d0 move.l d0,d1 move.l -(a3),d2 add.l d2,d1 move.l d1,(a4)+ sub.l d2,d0 muls.l #\1,d1:d0 move.w d1,d0 swap d0 move.l d0,\2*4-4(a4) ENDM fast_filter_sub move.l a5,a2 ; @p(0) lea 16*4(a2),a3 ; @p(16) move.l a6,a4 ; @pp(0) FFF_MUL COS1_32,8 ; pp(i=0..7) = p(i) + p(15-i) FFF_MUL COS3_32,8 ; pp(i=8..15) = COSx*[p(i) - p(15-i)] FFF_MUL COS5_32,8 FFF_MUL COS7_32,8 FFF_MUL COS9_32,8 FFF_MUL COS11_32,8 FFF_MUL COS13_32,8 FFF_MUL COS15_32,8 move.l a6,a2 ; @pp(0) lea 8*4(a2),a3 ; @pp(8) move.l a5,a4 ; @p(0) FFF_MUL COS1_16,4 ; p(i=0..3) = pp(i) + pp(7-i) FFF_MUL COS3_16,4 ; p(i=4..7) = COSx*[pp(i) - pp(7-i)] FFF_MUL COS5_16,4 FFF_MUL COS7_16,4 lea 8*4(a6),a2 ; @pp(8) lea 8*4(a2),a3 ; @pp(16) lea 4*4(a4),a4 ; skip p(4)..p(7) FFF_MUL COS1_16,4 FFF_MUL COS3_16,4 FFF_MUL COS5_16,4 FFF_MUL COS7_16,4 move.l a5,a2 ; @p(0) move.l a6,a4 ; @pp(0) move.l #COS1_8,d1 move.l #COS3_8,d2 FF_ADD 0,3 FF_ADD 1,2 FF_SUB_MUL 0,3,d1 FF_SUB_MUL 1,2,d2 FF_ADD 4,7 FF_ADD 5,6 FF_SUB_MUL 4,7,d1 FF_SUB_MUL 5,6,d2 FF_ADD 8,11 FF_ADD 9,10 FF_SUB_MUL 8,11,d1 FF_SUB_MUL 9,10,d2 FF_ADD 12,15 FF_ADD 13,14 FF_SUB_MUL 12,15,d1 FF_SUB_MUL 13,14,d2 lea filter_pp,a2 ; @pp(0) lea filter_p,a4 ; @p(0) move.l #COS1_4,d3 moveq.w #8-1,d7 fast_filter_s5 move.l (a2)+,d0 move.l d0,d2 move.l (a2)+,d1 add.l d1,d0 move.l d0,(a4)+ sub.l d1,d2 muls.l d3,d4:d2 move.w d4,d2 swap d2 move.l d2,(a4)+ dbra d7,fast_filter_s5 rts ; Apply the FAST synthesis filter to a sub band ; Generate full frequency sample ; ; a0: bandPtr (=fraction) ; a1: out_filter_buffer ; fast_filter_full movem.l d2-d7/a2-a6,-(sp) lea filter_p,a5 lea filter_pp,a6 move.l a0,a2 ; fraction(0) lea 32*2(a2),a3 ; fraction(32) move.l a5,a4 ; @p(0) moveq.w #16-1,d7 fast_filter_f1 ; p(i=0..15) = f(i) + f(31-i) move.w (a2)+,d0 ext.l d0 move.w -(a3),d1 ext.l d1 add.l d0,d1 move.l d1,(a4)+ dbra d7,fast_filter_f1 bsr fast_filter_sub move.l a5,a2 ; @p(0) move.l 6*4(a2),d1 add.l 7*4(a2),d1 move.l 5*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,36*2(a1) ; [36-17] => b(36) move.w d0,60*2(a1) ; [36-17] => b(60) move.l 4*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,44*2(a1) ; [44-17] => b(44) move.w d0,52*2(a1) ; [44-17] => b(52) move.l 11*4(a2),d1 add.l 15*4(a2),d1 move.l d1,d0 move.w d0,10*2(a1) ; [10] => b(10) neg.l d0 move.w d0,22*2(a1) ; [10] => -b(22) move.l 13*4(a2),d0 add.l d1,d0 move.w d0,6*2(a1) ; [6] => b(6) neg.l d0 move.w d0,26*2(a1) ; [6] => -b(26) move.l 14*4(a2),d1 add.l 15*4(a2),d1 move.l 8*4(a2),d0 add.l 12*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,46*2(a1) ; [46-17] => b(46) move.w d0,50*2(a1) ; [46-17] => b(50) move.l 9*4(a2),d0 add.l 13*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,34*2(a1) ; [34-17] => b(34) move.w d0,62*2(a1) ; [34-17] => b(62) add.l 10*4(a2),d1 add.l 11*4(a2),d1 move.l 13*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,38*2(a1) ; [38-17] => b(38) move.w d0,58*2(a1) ; [38-17] => b(58) move.l 12*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,42*2(a1) ; [42-17] => b(42) move.w d0,54*2(a1) ; [54-17] => b(54) move.l 9*4(a2),d0 add.l 13*4(a2),d0 add.l 15*4(a2),d0 move.w d0,2*2(a1) ; [2] => b(2) neg.l d0 move.w d0,30*2(a1) ; [2] => -b(30) move.l 5*4(a2),d0 add.l 7*4(a2),d0 move.w d0,4*2(a1) ; [4] => b(4) neg.l d0 move.w d0,28*2(a1) ; [4] => -b(28) move.l 0*4(a2),d0 neg.l d0 move.w d0,48*2(a1) ; [48-17] => b(48) move.l 1*4(a2),d0 move.w d0,0*2(a1) ; [0] => b(0) neg.l d0 move.w d0,32*2(a1) ; [0] => -b(32) move.l 3*4(a2),d0 move.w d0,8*2(a1) ; [8] => b(8) neg.l d0 move.w d0,24*2(a1) ; [8] => -b(24) move.l 7*4(a2),d0 move.w d0,12*2(a1) ; [12] => b(12) neg.l d0 move.w d0,20*2(a1) ; [12] => -b(20) move.l 15*4(a2),d0 move.w d0,14*2(a1) ; [14] => b(14) neg.l d0 move.w d0,18*2(a1) ; [14] => -b(18) move.l 2*4(a2),d0 add.l 3*4(a2),d0 neg.l d0 move.w d0,40*2(a1) ; [40-17] => b(40) move.w d0,56*2(a1) ; [40-17] => b(56) clr.w 16*2(a1) ; 0 => b(16) move.l a0,a2 ; fraction(0) lea 32*2(a2),a3 ; fraction(32) move.l a5,a4 ; @p(0) lea filter_cos64,a0 moveq.w #16-1,d7 fast_filter_f2 ; p(i=0..15) = COSx*[f(i) - f(31-i)] move.w (a2)+,d0 ext.l d0 move.w -(a3),d1 ext.l d1 sub.l d1,d0 muls.l (a0)+,d1:d0 move.w d1,d0 swap d0 move.l d0,(a4)+ dbra d7,fast_filter_f2 bsr fast_filter_sub move.l a5,a2 ; @p(0) move.l 13*4(a2),d1 add.l 15*4(a2),d1 move.l 1*4(a2),d0 add.l 9*4(a2),d0 add.l d1,d0 move.w d0,1*2(a1) ; [1] => b(1) neg.l d0 move.w d0,31*2(a1) ; [1] => -b(31) move.l 5*4(a2),d0 add.l 7*4(a2),d0 add.l 11*4(a2),d0 add.l d1,d0 move.w d0,5*2(a1) ; [5] => b(5) neg.l d0 move.w d0,27*2(a1) ; [5] => -b(27) add.l 9*4(a2),d1 move.l 1*4(a2),d0 add.l 14*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,33*2(a1) ; [33-17] => b(33) move.w d0,63*2(a1) ; [33-17] => b(63) add.l 5*4(a2),d1 add.l 7*4(a2),d1 move.l d1,d0 move.w d0,3*2(a1) ; [3] => b(3) neg.l d0 move.w d0,29*2(a1) ; [3] => -b(29) move.l 6*4(a2),d0 add.l 14*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,35*2(a1) ; [35-17] => b(35) move.w d0,61*2(a1) ; [35-17] => b(61) move.l 10*4(a2),d1 add.l 11*4(a2),d1 add.l 14*4(a2),d1 add.l 15*4(a2),d1 move.l 2*4(a2),d0 add.l 3*4(a2),d0 add.l 13*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,39*2(a1) ; [39-17] => b(39) move.w d0,57*2(a1) ; [39-17] => b(57) move.l 4*4(a2),d0 add.l 6*4(a2),d0 add.l 7*4(a2),d0 add.l 12*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,43*2(a1) ; [43-17] => b(43) move.w d0,53*2(a1) ; [43-17] => b(53) move.l 5*4(a2),d0 add.l 6*4(a2),d0 add.l 7*4(a2),d0 add.l 13*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,37*2(a1) ; [37-17] => b(37) move.w d0,59*2(a1) ; [37-17] => b(59) move.l 2*4(a2),d0 add.l 3*4(a2),d0 add.l 12*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,41*2(a1) ; [41-17] => b(41) move.w d0,55*2(a1) ; [55-17] => b(55) move.l 8*4(a2),d1 add.l 12*4(a2),d1 add.l 14*4(a2),d1 add.l 15*4(a2),d1 move.l 0*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,47*2(a1) ; [47-17] => b(47) move.w d0,49*2(a1) ; [47-17] => b(49) move.l 4*4(a2),d0 add.l 6*4(a2),d0 add.l 7*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,45*2(a1) ; [45-17] => b(45) move.w d0,51*2(a1) ; [45-17] => b(51) move.l 11*4(a2),d1 add.l 15*4(a2),d1 move.l 7*4(a2),d0 add.l d1,d0 move.w d0,11*2(a1) ; [11] => b(11) neg.l d0 move.w d0,21*2(a1) ; [11] => -b(21) add.l 3*4(a2),d1 move.l d1,d0 move.w d0,9*2(a1) ; [9] => b(9) neg.l d0 move.w d0,23*2(a1) ; [9] => -b(23) move.l 13*4(a2),d0 add.l d1,d0 move.w d0,7*2(a1) ; [7] => b(7) neg.l d0 move.w d0,25*2(a1) ; [7] => -b(25) move.l 7*4(a2),d0 add.l 15*4(a2),d0 move.w d0,13*2(a1) ; [13] => b(13) neg.l d0 move.w d0,19*2(a1) ; [13] => -b(19) move.l 15*4(a2),d0 move.w d0,15*2(a1) ; [15] => b(15) neg.l d0 move.w d0,17*2(a1) ; [15] => -b(17) movem.l (sp)+,d2-d7/a2-a6 rts FFH_MUL MACRO move.w (a2)+,d0 ext.l d0 move.l d0,d1 move.w -(a3),d2 ext.l d2 add.l d2,d1 move.l d1,(a4)+ sub.l d2,d0 muls.l #\1,d1:d0 move.w d1,d0 swap d0 move.l d0,7*4(a4) ENDM fast_filter_half movem.l d2-d7/a2-a6,-(sp) move.l a0,a2 ; fraction(0) lea 16*2(a2),a3 ; fraction(16) lea filter_pp,a4 ; @pp(0) FFH_MUL COS1_32 ; pp(i=0..7) = f(i) + f(15-i) FFH_MUL COS3_32 ; pp(i=8..15) = COSx*[f(i) - f(15-i)] FFH_MUL COS5_32 FFH_MUL COS7_32 FFH_MUL COS9_32 FFH_MUL COS11_32 FFH_MUL COS13_32 FFH_MUL COS15_32 lea filter_pp,a2 ; @pp(0) lea 8*4(a2),a3 ; @pp(8) lea filter_p,a4 ; @p(0) moveq.w #4-1,d7 fast_filter_h3 ; p(i=0..3) = pp(i) + p(7-i) move.l (a2)+,d0 add.l -(a3),d0 move.l d0,(a4)+ dbra d7,fast_filter_h3 lea filter_pp,a2 ; @pp(0) lea 8*4(a2),a3 ; @pp(8) FF_MUL COS1_16 ; p(i=4..7) = COSx*[p(i) - p(7-i)] FF_MUL COS3_16 FF_MUL COS5_16 FF_MUL COS7_16 lea filter_pp+8*4,a2 ; @pp(8) lea 8*4(a2),a3 ; @pp(16) moveq.w #4-1,d7 fast_filter_h4 ; p(i=8..11) = pp(i) + p(15-i+8) move.l (a2)+,d0 add.l -(a3),d0 move.l d0,(a4)+ dbra d7,fast_filter_h4 lea filter_pp+8*4,a2 ; @pp(8) lea 8*4(a2),a3 ; @pp(16) FF_MUL COS1_16 ; p(i=12..15) = COSx*[p(i) - p(15-i+8)] FF_MUL COS3_16 FF_MUL COS5_16 FF_MUL COS7_16 lea filter_p,a2 ; @p(0) lea filter_pp,a4 ; @pp(0) move.l #COS1_8,d1 move.l #COS3_8,d2 FF_ADD 0,3 FF_ADD 1,2 FF_SUB_MUL 0,3,d1 FF_SUB_MUL 1,2,d2 FF_ADD 4,7 FF_ADD 5,6 FF_SUB_MUL 4,7,d1 FF_SUB_MUL 5,6,d2 FF_ADD 8,11 FF_ADD 9,10 FF_SUB_MUL 8,11,d1 FF_SUB_MUL 9,10,d2 FF_ADD 12,15 FF_ADD 13,14 FF_SUB_MUL 12,15,d1 FF_SUB_MUL 13,14,d2 lea filter_pp,a2 ; @pp(0) lea filter_p,a4 ; @p(0) move.l #COS1_4,d3 moveq.w #8-1,d7 fast_filter_h5 move.l (a2)+,d0 move.l d0,d2 move.l (a2)+,d1 add.l d1,d0 move.l d0,(a4)+ sub.l d1,d2 muls.l d3,d4:d2 move.w d4,d2 swap d2 move.l d2,(a4)+ dbra d7,fast_filter_h5 lea filter_p,a2 ; @p(0) move.l 6*4(a2),d1 add.l 7*4(a2),d1 move.l 5*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,36*2(a1) ; [36-17] => b(36) move.w d0,60*2(a1) ; [36-17] => b(60) move.l 4*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,44*2(a1) ; [44-17] => b(44) move.w d0,52*2(a1) ; [44-17] => b(52) move.l 11*4(a2),d1 add.l 15*4(a2),d1 move.l d1,d0 move.w d0,10*2(a1) ; [10] => b(10) neg.l d0 move.w d0,22*2(a1) ; [10] => -b(22) move.l 13*4(a2),d0 add.l d1,d0 move.w d0,6*2(a1) ; [6] => b(6) neg.l d0 move.w d0,26*2(a1) ; [6] => -b(26) move.l 14*4(a2),d1 add.l 15*4(a2),d1 move.l 8*4(a2),d0 add.l 12*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,46*2(a1) ; [46-17] => b(46) move.w d0,50*2(a1) ; [46-17] => b(50) move.l 9*4(a2),d0 add.l 13*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,34*2(a1) ; [34-17] => b(34) move.w d0,62*2(a1) ; [34-17] => b(62) add.l 10*4(a2),d1 add.l 11*4(a2),d1 move.l 13*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,38*2(a1) ; [38-17] => b(38) move.w d0,58*2(a1) ; [38-17] => b(58) move.l 12*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,42*2(a1) ; [42-17] => b(42) move.w d0,54*2(a1) ; [54-17] => b(54) move.l 9*4(a2),d0 add.l 13*4(a2),d0 add.l 15*4(a2),d0 move.w d0,2*2(a1) ; [2] => b(2) neg.l d0 move.w d0,30*2(a1) ; [2] => -b(30) move.l 5*4(a2),d0 add.l 7*4(a2),d0 move.w d0,4*2(a1) ; [4] => b(4) neg.l d0 move.w d0,28*2(a1) ; [4] => -b(28) move.l 0*4(a2),d0 neg.l d0 move.w d0,48*2(a1) ; [48-17] => b(48) move.l 1*4(a2),d0 move.w d0,0*2(a1) ; [0] => b(0) neg.l d0 move.w d0,32*2(a1) ; [0] => -b(32) move.l 3*4(a2),d0 move.w d0,8*2(a1) ; [8] => b(8) neg.l d0 move.w d0,24*2(a1) ; [8] => -b(24) move.l 7*4(a2),d0 move.w d0,12*2(a1) ; [12] => b(12) neg.l d0 move.w d0,20*2(a1) ; [12] => -b(20) move.l 15*4(a2),d0 move.w d0,14*2(a1) ; [14] => b(14) neg.l d0 move.w d0,18*2(a1) ; [14] => -b(18) move.l 2*4(a2),d0 add.l 3*4(a2),d0 neg.l d0 move.w d0,40*2(a1) ; [40-17] => b(40) move.w d0,56*2(a1) ; [40-17] => b(56) clr.w 16*2(a1) ; 0 => b(16) movem.l (sp)+,d2-d7/a2-a6 rts FFQ_MUL MACRO move.w (a2)+,d0 ext.l d0 move.l d0,d1 move.w -(a3),d2 ext.l d2 add.l d2,d1 move.l d1,(a4)+ sub.l d2,d0 muls.l #\1,d1:d0 move.w d1,d0 swap d0 move.l d0,3*4(a4) ENDM FFQ_MUL2 MACRO movem.l (a2)+,d0-d3 move.l d0,d4 add.l d3,d4 move.l d4,(a4)+ move.l d0,d4 sub.l d3,d4 muls.l \1,d5:d4 move.w d5,d4 swap d4 move.l d4,4(a4) move.l d1,d4 add.l d2,d4 move.l d4,(a4)+ move.l d1,d4 sub.l d2,d4 muls.l \2,d5:d4 move.w d5,d4 swap d4 move.l d4,4(a4) ENDM ; Apply the FAST synthesis filter to a sub band ; Generate 1/4 frequency sample ; ; a0: bandPtr (=fraction) ; a1: out_filter_buffer ; fast_filter_quart movem.l d2-d7/a2-a6,-(sp) lea filter_p,a5 lea filter_pp,a6 move.l a0,a2 ; fraction(0) lea 8*2(a2),a3 ; fraction(8) move.l a5,a4 ; @p(0) ; p(i=0..3) = x(i) + x(7-i) FFQ_MUL COS1_16 ; p(i=4..7) = COSx*[x(i) - x(7-i)] FFQ_MUL COS3_16 FFQ_MUL COS5_16 FFQ_MUL COS7_16 move.l a5,a2 ; @p(0) move.l a6,a4 ; @pp(0) IFD FF_ADD move.l #COS1_8,d1 move.l #COS3_8,d2 FF_ADD 0,3 FF_ADD 1,2 FF_SUB_MUL 0,3,d1 FF_SUB_MUL 1,2,d2 FF_ADD 4,7 FF_ADD 5,6 FF_SUB_MUL 4,7,d1 FF_SUB_MUL 5,6,d2 ELSE move.l #COS1_8,d6 move.l #COS3_8,d7 FFQ_MUL2 d6,d7 addq.l #4*2,a4 FFQ_MUL2 d6,d7 ENDC move.l a6,a2 ; @pp(0) move.l a5,a4 ; @p(0) move.l #COS1_4,d3 moveq.w #4-1,d7 fast_filter_q5 move.l (a2)+,d0 move.l d0,d2 move.l (a2)+,d1 add.l d1,d0 move.l d0,(a4)+ sub.l d1,d2 muls.l d3,d4:d2 move.w d4,d2 swap d2 move.l d2,(a4)+ dbra d7,fast_filter_q5 move.l a5,a2 ; @p(0) move.l 6*4(a2),d1 add.l 7*4(a2),d1 move.l 5*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,36*2(a1) ; [36-17] => b(36) move.w d0,60*2(a1) ; [36-17] => b(60) move.l 4*4(a2),d0 add.l d1,d0 neg.l d0 move.w d0,44*2(a1) ; [44-17] => b(44) move.w d0,52*2(a1) ; [44-17] => b(52) move.l 5*4(a2),d0 add.l 7*4(a2),d0 move.w d0,4*2(a1) ; [4] => b(4) neg.l d0 move.w d0,28*2(a1) ; [4] => -b(28) move.l 0*4(a2),d0 neg.l d0 move.w d0,48*2(a1) ; [48-17] => b(48) move.l 1*4(a2),d0 move.w d0,0*2(a1) ; [0] => b(0) neg.l d0 move.w d0,32*2(a1) ; [0] => -b(32) move.l 3*4(a2),d0 move.w d0,8*2(a1) ; [8] => b(8) neg.l d0 move.w d0,24*2(a1) ; [8] => -b(24) move.l 7*4(a2),d0 move.w d0,12*2(a1) ; [12] => b(12) neg.l d0 move.w d0,20*2(a1) ; [12] => -b(20) move.l 2*4(a2),d0 add.l 3*4(a2),d0 neg.l d0 move.w d0,40*2(a1) ; [40-17] => b(40) move.w d0,56*2(a1) ; [40-17] => b(56) clr.w 16*2(a1) ; 0 => b(16) movem.l (sp)+,d2-d7/a2-a6 rts section ASMDATA,data CNOP 0,4 filter_cos64 dc.l COS1_64, COS3_64, COS5_64, COS7_64 dc.l COS9_64, COS11_64, COS13_64, COS15_64 dc.l COS17_64, COS19_64, COS21_64, COS23_64 dc.l COS25_64, COS27_64, COS29_64, COS31_64 section ASMVARS,bss CNOP 0,4 filter_p ds.l 16 filter_pp ds.l 16 section ASMCODE,code ;WINDOW_CLIP equ 1 ;WINDOW_HARD equ 1 ; ; Window a sub band filtered sample ; ; a0: out_filter_buffer ; a1: out_sample_buffer ; d0: buffer offset ; -> a1 = out_sample_buffer + out_sample_length @ASM_window_band movem.l d2-d7/a2-a6,-(sp) move.l a1,a2 moveq.l #4,d7 ; Band skip for quarter tst.w quar_freq beq.s ASM_window_band0 moveq.l #2,d7 ; Band skip for half tst.w half_freq beq.s ASM_window_band0 moveq.l #1,d7 ; Band skip for normal ASM_window_band0 moveq.l #WINDOW_FACTOR+INT_FACTOR-15,d7 add.w output_8bits,d7 ; add right shift value move.w #$3FF,d2 lea.l window_table,a1 lea.l 32*2*2,a4 lea.l 32,a5 lea.l 96,a6 move.w window_start,d6 ext.l d6 add.w d6,d0 ; add window offset to d0 (buffer offset) add.l d6,a1 ; add window offset to window ptr moveq.l #0,d6 ; d6 = j tst.w output_8bits bne window_band_5 ; 8 bit output selected window_band_1 clr.l d1 ; d1 = sum lea.l (a1,d6.w*2),a3 ; a3 = window[ k(=j) ] move.w d6,d3 add.w d0,d3 ; d3 = l = j + buf_offset move.w window_coeff,d5 ; d5 = (#-1)/2 coeffs window_band_2 and.w d2,d3 ; l &= 0x3FF move.w (a0,d3.w*2),d4 ; bufPtr[ l ] -> d4 muls.w (a3),d4 ; window[ k ] * bufPtr[ l ] -> d4 add.l d4,d1 ; sum += d4 add.w a6,d3 ; l += 96 and.w d2,d3 ; l &= 0x3FF move.w (a0,d3.w*2),d4 ; bufPtr[ l ] -> d4 muls.w 64(a3),d4 ; window[ k ] * bufPtr[ l ] -> d4 add.l a4,a3 ; a3 += 64 add.l d4,d1 ; sum += d4 add.w a5,d3 ; l += 32 dbra d5,window_band_2 asr.l d7,d1 ; sum >>= (WINDOW_FACTOR+INT_FACTOR-15) cmp.l #-32768,d1 blt window_band_4l cmp.l #32767,d1 bgt window_band_4h window_band_4n move.w d1,(a2)+ add.w window_skip,d6 ; next band cmp.w a5,d6 ; equiv cmp.w #32,d6 blt window_band_1 bra window_band_9 window_band_4l move.l #-32768,d1 bra.s window_band_4n window_band_4h move.l #32767,d1 bra.s window_band_4n window_band_5 clr.l d1 ; d1 = sum lea.l (a1,d6.w*2),a3 ; a3 = window[ k(=j) ] move.w d6,d3 add.w d0,d3 ; d3 = l = j + buf_offset move.w window_coeff,d5 ; d5 = (#-1)/2 coeffs window_band_6 and.w d2,d3 ; l &= 0x3FF move.w (a0,d3.w*2),d4 ; bufPtr[ l ] -> d4 muls.w (a3),d4 ; window[ k ] * bufPtr[ l ] -> d4 add.l d4,d1 ; sum += d4 add.w a6,d3 ; l += 96 and.w d2,d3 ; l &= 0x3FF move.w (a0,d3.w*2),d4 ; bufPtr[ l ] -> d4 muls.w 64(a3),d4 ; window[ k ] * bufPtr[ l ] -> d4 add.l a4,a3 ; a3 += 64 add.l d4,d1 ; sum += d4 add.w a5,d3 ; l += 32 dbra d5,window_band_6 asr.l d7,d1 ; sum >>= (WINDOW_FACTOR+INT_FACTOR-15) cmp.w #-128,d1 bge.s window_band_71 moveq.b #-128,d1 bra.s window_band_72 window_band_71 cmp.w #127,d1 ble.s window_band_72 moveq.b #127,d1 window_band_72 move.b d1,(a2)+ add.w window_skip,d6 ; next band cmp.w a5,d6 ; equiv cmp.w #32,d6 blt window_band_5 window_band_9 move.l a2,a1 movem.l (sp)+,d2-d7/a2-a6 rts ; Sub band synthesis ; ; a0: frame_params *fr_ps ; d0: int band_count ; @ASM_sub_band_synthesis movem.l d2-d6/a2-a6,-(sp) move.l a0,a2 ; a2 = frame_params move.l d0,d4 subq.l #1,d4 ; d4 = band_count-1 move.l fp_stereo(a2),d5 subq.l #1,d5 ; d5 = #channels-1 clr.l d6 ; d6 = channel lea fp_out_buffer(a2),a6 ; a6 = &sample[ 0 ] sub_band_synthesis_0 move.l (a6),a5 ; a5 = sample[channel] move.l fp_fraction(a2),a3 move.l d6,d1 mulu.w #2*3*SBLIMIT,d1 add.l d1,a3 ; a3 = fraction[channel] lea.l filter_buffer,a4 move.w d6,d2 mulu.w #2*2*HAN_SIZE,d2 add.l d2,a4 ; -> a4 = &filter_buffer[channel][0] move.w d4,d3 sub_band_synthesis_1 lea.l band_offset,a1 move.w (a1,d6.w*2),d2 sub.w #64,d2 and.w #$3FF,d2 ; d2 = (band_offset[k] - 64) & 0x3ff; move.w d2,(a1,d6.w*2) lea.l (a4,d2.w*2),a1 ; a1 = &filter_buffer[k][d2] move.l a3,a0 ; a0 = &bandPtr[k][j][0] bsr @ASM_filter_band move.l a5,a1 ; a1 = &samples[k][j][0] move.l a4,a0 ; a0 = &filter_buffer[k][0] move.l d2,d0 ; d0 = band_offset[k] bsr @ASM_window_band add.l #2*SBLIMIT,a3 ; -> a3 = &bandPtr[k][j+1][0] move.l a1,a5 ; a5 = next sample ptr dbra d3,sub_band_synthesis_1 ; j++ move.l a5,(a6) ; a5 = sample[channel] addq.l #1,d6 addq.l #4,a6 dbra d5,sub_band_synthesis_0 ; next channel movem.l (sp)+,d2-d6/a2-a6 rts ; ; Decode an audio MPEG I Layer I frame ; ; a0: frame_params *fr_ps ; @ASM_I_decode_frame movem.l d2/a2,-(sp) move.l a0,a2 clr.l d2 ASM_I_decode_frame_1 move.l a2,a0 bsr @ASM_I_buffer_sample move.l a2,a0 bsr @ASM_I_deqnorm move.l a2,a0 moveq.l #1,d0 bsr @ASM_sub_band_synthesis addq.l #1,d2 cmp.w #SCALE_BLOCK,d2 blt ASM_I_decode_frame_1 movem.l (sp)+,d2/a2 rts ; ; Decode an audio MPEG I Layer II frame ; ; a0: frame_params *fr_ps ; @ASM_II_decode_frame movem.l d2/a2,-(sp) move.l a0,a2 clr.l d2 ASM_II_decode_frame_1 move.l a2,a0 bsr @ASM_II_buffer_sample move.l a2,a0 move.l d2,d0 lsr.l #2,d0 bsr @ASM_II_deqnorm move.l a2,a0 moveq.l #3,d0 bsr @ASM_sub_band_synthesis addq.l #1,d2 cmp.w #SCALE_BLOCK,d2 blt ASM_II_decode_frame_1 movem.l (sp)+,d2/a2 rts section ASMDATA,data even band_offset dc.w 64,64 section ASMVARS,bss even bsamp_sblimit ds.w 1 bsamp_jsbound ds.w 1 bsamp_channels ds.w 1 channels ds.w 1 sub_band_limit ds.w 1 sub_band_index ds.w 1 sub_band_size ds.w 1 output_8bits ds.w 1 ; 0 if 16 bits output ; > 0 if 8 bits output = right shift half_freq ds.w 1 ; = 0 if full freq output quar_freq ds.w 1 ; != 0 if quarter freq output window_start ds.w 1 ; start offset for windowing window_coeff ds.w 1 ; #coeff used for windowing window_skip ds.w 1 ; #values skipped (1=full,2=half,4=quart) quality ds.w 1 even band_buffer ds.w 32 h_band_input ds.w 16 h_band_buffer ds.w 16 h_band_buffer2 ds.w 4+8+4+8 q_band_input ds.w 8 q_band_buffer ds.w 8 q_band_buffer2 ds.w 2+4+2+4 filter_buffer ds.w 2*2*HAN_SIZE