get_interpolat_uv.asm

adi bf533视频编码程序

#define  IPPVC_MC_APX_FF   0
#define  IPPVC_MC_APX_FH   4
#define  IPPVC_MC_APX_HF   8
#define  IPPVC_MC_APX_HH   12
.global _get_interpolate_uv;
.section L1_code;

/**************************************************************************
void get_interpolate_uv(Ipp8u *pSrc,int srcStep,Ipp8u *pDst,Ipp32s mcType);
***************************************************************************/
_get_interpolate_uv:


             [--sp]=(r7:4,p5:3);
             [--sp]=rets;
             
             l3=0;
             L2=0;
             l1=0;
             L0=0;
             I3=R2;//DST
             i0=r0;//the address of pSrc     
             m3=r0;//the address of pSrc
             b3=r0;
             p5=r1;//the srcCurStep
             r6=[sp+44];//mcType
             m1=r1;
             R7=R1;
             p0=8; 
             //M2=12;// in order to match Step =16;        
/***********************IPPVC_MC_APX_FF**********************/
            cc=r6==IPPVC_MC_APX_FF;
            if !cc jump FH_8x8;
          
              R7+=-8;
              m0=R7;//refstep modify
            
              i1=i0;
             lsetup (ff_8x8_start,ff_8x8_end)lc0=p0;
                disalgnexcpt||r0=[i0++]||r2=[i1++];
      
              ff_8x8_start:
                   disalgnexcpt||r1=[i0++]||r3=[i1++];
                   r6=byteop1p(r1:0,r3:2)||r0=[i0++m0]||r2=[i1++m0];
                   r7=byteop1p(r1:0,r3:2)(r)||[i3++]=r6||r0=[i0++];     
              ff_8x8_end:
                     disalgnexcpt||[i3++]=r7||r2=[i1++]; 
             jump interpolate_8x8_end;
/**********************IPPVC_MC_APX_FH***********************/
           FH_8x8:
               r5=IPPVC_MC_APX_FH;
               cc=r6==r5;
               if !cc jump HF_8x8;
            
               m1=r7;//ref step       
               r7+=-8;
               m0=r7;//refmodify
              
               lsetup (fh_8x8_start,fh_8x8_end)lc0=p0;
                   i1=i0;
                   i1+=m1;
                  disalgnexcpt||r0=[i0++]||r2=[i1++];
               fh_8x8_start:
                   disalgnexcpt||r1=[i0++]||r3=[i1++]; 
                   r6=byteop1p(r1:0,r3:2)||r0=[i0++m0]||r2=[i1++m0];
                   r7=byteop1p(r1:0,r3:2)(r)||[i3++]=r6||r0=[i0++];       
               fh_8x8_end:
                   disalgnexcpt||[i3++]=r7||r2=[i1++];      
                   jump interpolate_8x8_end;
/************************IPPVC_MC_APX_HF************************/
           HF_8x8:
              r5=IPPVC_MC_APX_HF;
              cc=r6==r5;
              if !cc jump HH_8x8;
                 m1=1;//to make the align8    
                 r7+=-8;
                 m0=r7;
                 i1=i0;
                 i1+=m1;
                
                 r6=3;       //这一段主要是为I0的后两位全为1
                 r7=i0;      //考虑的。因为此时I0＋1后，后两位
                 r7=r7&r6;   //全为0，此时所选的寄存器，就不是
                 cc=r7==r6;  //R3,而是R2,故分开考虑。
                 if cc jump byte_align_HF;
                lsetup(hf_8x8_start,hf_8x8_end)lc0=p0;
                i2=i0;
                disalgnexcpt||r0=[i0++]||r2=[i2++];
                hf_8x8_start:
 
                   
                   disalgnexcpt||r1=[i0++]||r3=[i2++];
                   r6=byteop1p(r1:0,r3:2)||r0=[i0++m0]||r2=[i2++m0];
                   r7=byteop1p(r1:0,r3:2)(r)||[i3++]=r6||r0=[i0++];
                   
              hf_8x8_end:     
                   disalgnexcpt||[i3++]=r7||r2=[i2++];                 
                jump interpolate_8x8_end;
                
             byte_align_HF: //特殊情况
                 lsetup(bytealign_start,bytealign_end) lc0=p0;
                 i2=i0;
                 i2+=4;
                  disalgnexcpt||r0=[i0++]||r2=[i2++];
                
                bytealign_start:
                    disalgnexcpt||r1=[i0++]||r3=[i2++];
                    r6=byteop1p(r1:0,r3:2)||r0=[i0++m0]||r2=[i2++m0];       
                    r7=byteop1p(r1:0,r3:2)(r)||[i3++]=r6||r0=[i0++];
                bytealign_end:
                    disalgnexcpt||[i3++]=r7||r2=[i2++];
                    
                    
              jump interpolate_8x8_end;
/************************IPPVC_MC_APX_HH************************/
           HH_8x8:
              
             
              /*
               m1=1;
               m2=r7;//src step
               r7+=-8;
               m0=r7;
               i1=i0;
                 r6=3;       //这一段主要是为I0的后两位全为1
                 r7=i0;      //考虑的。因为此时I0＋1后，后两位
                 r7=r7&r6;   //全为0，此时所选的寄存器，就不是
                 cc=r7==r6;  //R3,而是R2,故分开考虑。
               if cc jump byte_align_HH;

              // p1=2;
               lsetup(hh_8x8_start,hh_8x8_end)lc0=p0;          
                    i2=i0;
                    i2+=m2;
                    disalgnexcpt||r0=[i0++]||r2=[i2++];
               hh_8x8_start:
                    disalgnexcpt||r1=[i0++]||r3=[i2++];
                    r6=byteop2p(r1:0,r3:2)(rndl);
                    i0+=m1;
                    i1+=m1;
                    r7=byteop2p(r1:0,r3:2)(rndh);
                    i0-=m1;
                    i1-=m1;
                    r7=r7+r6;
                    disalgnexcpt||r0=[i0++m0];
                    disalgnexcpt||[i3++]=r7||r2=[i2++m0];
                    r6=byteop2p(r1:0,r3:2)(rndl,r);
                    i1+=m1;
                    i0+=m1;
                    r7=byteop2p(r1:0,r3:2)(rndh,r);
                    i1-=m1;
                    i0-=m1;                    
                    disalgnexcpt||r0=[i0++];
                    r7=r7+r6;
              hh_8x8_end:
                   disalgnexcpt||[i3++]=r7||r2=[i2++];
                   jump  interpolate_8x8_end;
             byte_align_HH://特殊情况
                   lsetup(bytealign_hh_start,bytealign_hh_end)lc0=p0;
                    i2=i0;
                    i2+=m2;
                    disalgnexcpt||r0=[i0++]||r2=[i2++];
                  bytealign_hh_start:
                     disalgnexcpt||r1=[i0++]||r3=[i2++];  
                     r6=byteop2p(r1:0,r3:2)(rndl);
                     i0+=m1;
                     i1+=m1;
                     
                     r7=byteop2p(r1:0,r3:2)(rndh,r);
                     i0-=m1;
                     i1-=m1;
                     r7=r7+r6;
                     disalgnexcpt||r0=[i0++m0];
                     disalgnexcpt||[i3++]=r7||r2=[i2++m0];
                     r6=byteop2p(r1:0,r3:2)(rndl,r);
                     i0+=m1;
                     i1+=m1;
                     r7=byteop2p(r1:0,r3:2)(rndh);
                     i0-=m1;
                     i1-=m1;
                     disalgnexcpt||r0=[i0++];
                     r7=r7+r6;
                  bytealign_hh_end:
                   disalgnexcpt||[i3++]=r7||r2=[i2++];  
                   //jump  interpolate_8x8_end;   
                   */
                   
                     
    M0 = 7;
    M3 = -3 (X);  
    I2 = R0;                // address of the top-left pel in best match reference block
    R0 = R0 + R1 (S);       //next row
    I1 = R0;                // Address of best match - stride
    I0 = R0;
    R1 += -5;        //-9
    M1 = R1;
    P2 = 8;   //9    
    
    
    LSETUP(AVGHV_RND_ST, AVGHV_RND_END) LC0 = P2;
    DISALGNEXCPT || R0 = [I1++] || R2 = [I2++];
    DISALGNEXCPT || R1 = [I1++M3] || R3 = [I2++M3];
    R7 = BYTEOP2P(R1:0,R3:2) (RNDL) || R0 = [I1++M0];
AVGHV_RND_ST:    
        DISALGNEXCPT || I0 += M3 || R2 = [I2++M0];
        R6 = BYTEOP2P(R1:0,R3:2) (RNDH) || R0 = [I1++M3] || R2 = [I2++M3];
        R7 = R6 + R7 (NS) || I0 -= M3;
        R7 = BYTEOP2P(R1:0,R3:2) (RNDL, R) || [I3++] = R7 || R1 = [I1++M1];
        DISALGNEXCPT || I0 += M3 || R3 = [I2++M1];
        R6 = BYTEOP2P(R1:0,R3:2) (RNDH, R) || R0 = [I1++] || R2 = [I2++];
        R7 = R6 + R7 (NS) || I0 -= M3;        
        DISALGNEXCPT || R1 = [I1++M3] || R3 = [I2++M3];
AVGHV_RND_END:
        R7 = BYTEOP2P(R1:0,R3:2) (RNDL) || [I3++] = R7 || R0 = [I1++M0];
               
        
      interpolate_8x8_end:      
      
      RETS=[SP++];
      (R7:4,P5:3)=[SP++];
_get_interpolate_uv.end:
rts;