;******************************************************************** ; ; add_block ; ; 1.0 - 30.7.98 (bifat) written in assembler, loops unrolled ; ;******************************************************************** section text XDEF _Add_Block XDEF _Clear_Block XREF _picture_structure ; int XREF _current_frame ; unsigned char * [3]; XREF _Coded_Picture_Width ; int XREF _chroma_format ; int XREF _Chroma_Width ; int XREF _ld ; struct layerdata * XREF _Clip ; char[1024] FRAME_PICTURE EQU 3 CHROMA420 EQU 1 CHROMA422 EQU 2 CHROMA444 EQU 3 ;******************************************************************** cnop 0,4 _Add_Block: movem.l d2-d7/a2-a6,-(a7) ; d0 int comp ; d1 int bx ; d2 int by ; d3 int dct_type ; d4 int addflag lea _current_frame,a6 ; /* derive color component index */ ; /* equivalent to ISO/IEC 13818-2 Table 7-1 */ ; cc = (comp<4) ? 0 : (comp&1)+1; /* color component index */ moveq #1,d5 and.l d0,d5 addq.l #1,d5 cmp.w #4,d0 bge .cc_is_not_0 moveq #0,d5 .cc_is_0 cmp.w #FRAME_PICTURE,_picture_structure+2 bne.b .cc0notfp tst.w d3 beq.b .notdcttype ; /* field DCT coding */ ; rfp = current_frame[0] ; + Coded_Picture_Width*(by+((comp&2)>>1)) + bx + ((comp&1)<<3); ; iincr = (Coded_Picture_Width<<1) - 8; move.l _Coded_Picture_Width,d6 move.l (a6),a0 ; currentframe moveq #2,d5 and.l d0,d5 ; comp&2 asr.l #1,d5 ; >> 1 add.l d2,d5 ; +by muls.l d6,d5 ; * codedpicturewidth add.l d5,a0 add.l d1,a0 ; +bx moveq #1,d5 and.l d0,d5 ; comp&1 asl.l #3,d5 add.l d5,a0 ; -> rfp asl.l #1,d6 subq.l #8,d6 ; iincr bra .cont .notdcttype ; /* frame DCT coding */ ; rfp = current_frame[0] ; + Coded_Picture_Width*(by+((comp&2)<<2)) + bx + ((comp&1)<<3); ; iincr = Coded_Picture_Width - 8; move.l _Coded_Picture_Width,d6 move.l (a6),a0 ; currentframe moveq #2,d5 and.l d0,d5 ; comp&2 asl.l #2,d5 ; <<2 add.l d2,d5 ; +by muls.l d6,d5 ; *codedpicturewidth add.l d5,a0 add.l d1,a0 ; +bx moveq #1,d5 and.l d0,d5 ; comp&1 asl.l #3,d5 add.l d5,a0 ; -> rfp subq.l #8,d6 ; iincr bra .cont .cc0notfp ; /* field picture */ ; rfp = current_frame[0] ; + (Coded_Picture_Width<<1)*(by+((comp&2)<<2)) + bx + ((comp&1)<<3); ; iincr = (Coded_Picture_Width<<1) - 8; move.l _Coded_Picture_Width,d6 add.l d6,d6 move.l (a6),a0 ; currentframe moveq #2,d5 and.l d0,d5 ; comp&2 asl.l #2,d5 ; <<2 add.l d2,d5 ; +by muls.l d6,d5 ; *codedpicturewidth add.l d5,a0 add.l d1,a0 ; +bx moveq #1,d5 and.l d0,d5 ; comp&1 asl.l #3,d5 ; <<3 add.l d5,a0 ; -> rfp subq.l #8,d6 ; -> iincr bra .cont .cc_is_not_0 ; /* chrominance */ ; ; /* scale coordinates */ ; if (chroma_format!=CHROMA444) ; bx >>= 1; ; if (chroma_format==CHROMA420) ; by >>= 1; move.w _chroma_format+2,d7 cmp.w #CHROMA444,d7 beq.b .noc1 asr.l #1,d1 .noc1 cmp.w #CHROMA420,d7 bne.b .noc2 asr.l #1,d2 .noc2 ; if (picture_structure==FRAME_PICTURE) ; { cmp.w #FRAME_PICTURE,_picture_structure+2 bne.b .ccnot0fp ; if (dct_type && (chroma_format!=CHROMA420)) ; { tst.w d3 beq.b .ccnot0notdct cmp.w #CHROMA420,d7 beq.b .ccnot0notdct ; /* field DCT coding */ ; rfp = current_frame[cc] ; + Chroma_Width*(by+((comp&2)>>1)) + bx + (comp&8); ; iincr = (Chroma_Width<<1) - 8; move.l _Chroma_Width,d6 move.l (a6,d5.w*4),a0 ; currentframe[cc] moveq #2,d5 and.l d0,d5 ; comp&2 asr.l #1,d5 ; >>1 add.l d2,d5 ; +by muls.l d6,d5 ; *chromawidth add.l d5,a0 add.l d1,a0 ; +bx moveq #8,d5 and.l d0,d5 ; comp&8 add.l d5,a0 add.l d6,d6 subq.l #8,d6 ; (chromawidth<<1)-8 bra.b .cont .ccnot0notdct ; /* frame DCT coding */ ; rfp = current_frame[cc] ; + Chroma_Width*(by+((comp&2)<<2)) + bx + (comp&8); ; iincr = Chroma_Width - 8; move.l _Chroma_Width,d6 move.l (a6,d5.w*4),a0 ; currentframe[cc] moveq #2,d5 and.l d0,d5 ; comp&2 asl.l #2,d5 ; <<2 add.l d2,d5 ; +by muls.l d6,d5 ; *chromawidth add.l d5,a0 add.l d1,a0 ; +bx moveq #8,d5 and.l d0,d5 ; comp&8 add.l d5,a0 subq.l #8,d6 ; chromawidth-8 bra.b .cont .ccnot0fp ; /* field picture */ ; rfp = current_frame[cc] ; + (Chroma_Width<<1)*(by+((comp&2)<<2)) + bx + (comp&8); ; iincr = (Chroma_Width<<1) - 8; move.l _Chroma_Width,d6 add.l d6,d6 move.l (a6,d5.w*4),a0 ; currentframe[cc] moveq #2,d5 and.l d0,d5 ; comp&2 asl.l #2,d5 ; <<2 add.l d2,d5 ; +by muls.l d6,d5 ; *chromawidth<<1 add.l d5,a0 add.l d1,a0 ; +bx moveq #8,d5 and.l d0,d5 add.l d5,a0 ; +comp&8 subq.l #8,d6 ; chromawidth<<1 - 8 .cont ; bp = ld->block[comp]; move.l _ld,a1 asl.l #7,d0 ; *128 add.l d0,a1 move.l _Clip,a2 move.l d6,a3 move.w #8,a4 tst.l d4 beq.b .noaddf .ilop1 movem.w (a1)+,d0-d6/a5 moveq #0,d7 move.b (a0)+,d7 add.w d7,d0 move.b (a0)+,d7 add.w d7,d1 move.b (a0)+,d7 add.w d7,d2 move.b (a0)+,d7 add.w d7,d3 move.b (a0)+,d7 add.w d7,d4 move.b (a0)+,d7 add.w d7,d5 move.b (a0)+,d7 add.w d7,d6 move.b (a0)+,d7 add.w a5,d7 subq.w #8,a0 move.b (a2,d0.w),(a0)+ move.b (a2,d1.w),(a0)+ move.b (a2,d2.w),(a0)+ move.b (a2,d3.w),(a0)+ move.b (a2,d4.w),(a0)+ move.b (a2,d5.w),(a0)+ move.b (a2,d6.w),(a0)+ subq.w #1,a4 move.b (a2,d7.w),(a0)+ ; move.b d0,(a0)+ ; move.b d1,(a0)+ ; move.b d2,(a0)+ ; move.b d3,(a0)+ ; move.b d4,(a0)+ ; move.b d5,(a0)+ ; move.b d6,(a0)+ ; move.b d7,(a0)+ add.l a3,a0 move.w a4,d0 bne.b .ilop1 bra.b .raus cnop 0,4 .noaddf add.w #128,a2 .ilop2 movem.w (a1)+,d0-d7 move.b (a2,d0.w),(a0)+ move.b (a2,d1.w),(a0)+ move.b (a2,d2.w),(a0)+ move.b (a2,d3.w),(a0)+ move.b (a2,d4.w),(a0)+ move.b (a2,d5.w),(a0)+ move.b (a2,d6.w),(a0)+ move.b (a2,d7.w),(a0)+ ; move.b d0, ; move.b d1, ; move.b d2, ; move.b d3, ; move.b d4, ; move.b d5, ; move.b d6, ; move.b d7, subq.w #1,a4 add.l a3,a0 move.w a4,d0 bne.b .ilop2 .raus movem.l (a7)+,d2-d7/a2-a6 rts ;******************************************************************** cnop 0,4 _Clear_Block: ; d0 blocknr movem.l d2-d6,-(a7) move.l _ld,a1 asl.w #7,d0 ; *128 add.w d0,a1 moveq #0,d0 moveq #0,d1 moveq #0,d2 moveq #0,d3 moveq #0,d4 moveq #0,d5 moveq #0,d6 move.l d0,a0 movem.l d0-d6/a0,(a1) movem.l d0-d6/a0,32(a1) movem.l d0-d6/a0,64(a1) movem.l d0-d6/a0,96(a1) movem.l (a7)+,d2-d6 rts END /* IMPLEMENTATION: set scratch pad macroblock to zero */ static void Clear_Block(comp) int comp; { short *Block_Ptr; int i; Block_Ptr = ld->block[comp]; for (i=0; i<64; i++) *Block_Ptr++ = 0; } ;******************************************************************** /* move/add 8x8-Block from block[comp] to backward_reference_frame */ /* copy reconstructed 8x8 block from block[comp] to current_frame[] * ISO/IEC 13818-2 section 7.6.8: Adding prediction and coefficient data * This stage also embodies some of the operations implied by: * - ISO/IEC 13818-2 section 7.6.7: Combining predictions * - ISO/IEC 13818-2 section 6.1.3: Macroblock */ static void Add_Block(comp,bx,by,dct_type,addflag) int comp,bx,by,dct_type,addflag; { int cc,i, j, iincr; unsigned char *rfp; short *bp; /* derive color component index */ /* equivalent to ISO/IEC 13818-2 Table 7-1 */ cc = (comp<4) ? 0 : (comp&1)+1; /* color component index */ if (cc==0) { /* luminance */ if (picture_structure==FRAME_PICTURE) { if (dct_type) { /* field DCT coding */ rfp = current_frame[0] + Coded_Picture_Width*(by+((comp&2)>>1)) + bx + ((comp&1)<<3); iincr = (Coded_Picture_Width<<1) - 8; } else { /* frame DCT coding */ rfp = current_frame[0] + Coded_Picture_Width*(by+((comp&2)<<2)) + bx + ((comp&1)<<3); iincr = Coded_Picture_Width - 8; } } else { /* field picture */ rfp = current_frame[0] + (Coded_Picture_Width<<1)*(by+((comp&2)<<2)) + bx + ((comp&1)<<3); iincr = (Coded_Picture_Width<<1) - 8; } } else { /* chrominance */ /* scale coordinates */ if (chroma_format!=CHROMA444) bx >>= 1; if (chroma_format==CHROMA420) by >>= 1; if (picture_structure==FRAME_PICTURE) { if (dct_type && (chroma_format!=CHROMA420)) { /* field DCT coding */ rfp = current_frame[cc] + Chroma_Width*(by+((comp&2)>>1)) + bx + (comp&8); iincr = (Chroma_Width<<1) - 8; } else { /* frame DCT coding */ rfp = current_frame[cc] + Chroma_Width*(by+((comp&2)<<2)) + bx + (comp&8); iincr = Chroma_Width - 8; } } else { /* field picture */ rfp = current_frame[cc] + (Chroma_Width<<1)*(by+((comp&2)<<2)) + bx + (comp&8); iincr = (Chroma_Width<<1) - 8; } } bp = ld->block[comp]; if (addflag) { for (i=0; i<8; i++) { for (j=0; j<8; j++) { *rfp = Clip[*bp++ + *rfp]; rfp++; } rfp+= iincr; } } else { for (i=0; i<8; i++) { for (j=0; j<8; j++) *rfp++ = Clip[*bp++ + 128]; rfp+= iincr; } } }