// MODE 3
static void pred4x4_down_left_mxu(uint8_t *dst, uint8_t *src, uint8_t *topright,
uint8_t *top, uint8_t *topleft){
//load
S32LDD(xr1, top, 0x0); // xr1 <- t3, t2, t1, t0 ;
S32LDD(xr2, topright, 0x0); // xr2 <- t7, t6, t5, t4 ;
S32LDDR(xr15, topright, 0x0); //xr15 <- t4, t5, t6, t7 ;
S32ALNI(xr3, xr2, xr1, ptn2); // xr3: t5, t4, t3, t2 ;
Q8AVG(xr4, xr1, xr3);
S32ALNI(xr5, xr2, xr1, ptn3); // xr5: t4, t3, t2, t1 ;
Q8AVGR(xr6, xr4, xr5);
S32ALNI(xr7, xr2, xr1, ptn1); // xr7: t6, t5, t4, t3 ;
S32ALNI(xr8, xr15, xr2, ptn3);// xr8: t7, t7, t6, t5 ;
Q8AVG(xr9, xr7, xr8);
Q8AVGR(xr10, xr9, xr2);
D32SLL(xr11, xr6, xr0, xr0, 0x8);
S32ALNI(xr12, xr10, xr11, ptn1);
S32ALNI(xr13, xr10, xr11, ptn2);
//store
S32STD(xr6, dst, 0x0);
S32SDIV(xr13, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIV(xr12, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIV(xr10, dst, MB_LUMA_EDGED_WIDTH, 0x0);
}
// MODE 5
static void pred4x4_vertical_right_mxu(uint8_t *dst, uint8_t *src, uint8_t *topright,
uint8_t *top, uint8_t *topleft){
uint8_t *src_left; // left address
src_left = src -0x4;
// load right
S32LDD(xr8, top, 0x0); // xr8: t3, t2, t1, t0 ; high -> low, [31->0];
// load left
S32LDD(xr1, topleft, -0x4); // xr1[31:24] <- src_topleft[3] (lt) ;
S32LDD(xr2, src_left, 0x0); // xr2[31:24] <- src_topleft[stride+3] (l0) ;
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr3[31:24] <- src_topleft[2*stride+3] (l1) ;
S32LDIV(xr4, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr4[31:24] <- src_topleft[3*stride+3] (l2) ;
S32SFL(xr5, xr2, xr1, xr0, ptn2); // xr5[31:16] <- l0, lt ;
S32SFL(xr6, xr4, xr3, xr0, ptn2); // xr6[31:16] <- l2, l1 ;
S32SFL(xr7, xr3, xr2, xr0, ptn2); // xr7[31:16] <- l1, l0 ;
// alni
S32ALNI(xr3, xr8, xr1, ptn1); // xr3: t2, t1, t0, lt ;
S32ALNI(xr4, xr3, xr2, ptn1); // xr4: t1, t0, lt, l0 ;
// cal
Q8AVGR(xr1, xr3, xr8); // xr1:
Q8AVG(xr9, xr4, xr8);
Q8AVGR(xr2, xr9, xr3); // xr2:
Q8AVG(xr10, xr5, xr6);
Q8AVGR(xr11, xr10, xr7); // xr11: src[0,3], src[0,2], ~, ~ ;
// alni
S32ALNI(xr12, xr2, xr11, ptn1);
D32SLL(xr13, xr11, xr0, xr0, 0x8);
S32ALNI(xr14, xr1, xr13, ptn1);
// store
S32STD(xr1, dst, 0x0);
S32SDIV(xr2, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIV(xr14, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIV(xr12, dst, MB_LUMA_EDGED_WIDTH, 0x0);
}
// MODE 6
static void pred4x4_horizontal_down_mxu(uint8_t *dst, uint8_t *src, uint8_t *topright,
uint8_t *top, uint8_t *topleft){
uint8_t *src_left; // left address
src_left = src - 0x4;
// load TOP
S32LDDR(xr8, top, 0x0); // xr8[31:0]: t0, t1, t2, t3 ;
S32LDDR(xr15, topleft, -0x4);// xr15[7:0]: lt ;
S32LDD(xr9, topleft, -0x4); // xr9[31:24]: lt ;
S32ALNI(xr10, xr15, xr8, ptn3); //xr10[31:0]: lt, t0, t1, t2 ;
// load LEFT
S32LDDR(xr1, src_left, 0x0); // xr1[7:0] <- src_left[3] (l0) ;
S32LDIVR(xr2, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr2[7:0] <- src_left[stride+3] (l1) ;
S32LDIVR(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr3[7:0] <- src_left[2*stride+3] (l2) ;
S32LDIVR(xr4, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr4[7:0] <- src_left[3*stride+3] (l3) ;
S32SFL(xr0, xr2, xr1, xr5, ptn2); // xr5[15:0] <- l1, l0 ;
S32SFL(xr0, xr4, xr3, xr6, ptn2); // xr6[15:0] <- l3, l2 ;
S32SFL(xr0, xr6, xr5, xr7, ptn3); // xr7[31:0] <- l3, l2, l1, l0 ;
// ALNI for CAL
S32ALNI(xr11, xr7, xr9, ptn1); // xr11: l2, l1, l0, lt ;
S32ALNI(xr12, xr1, xr10, ptn3); // xr12: l0, lt, t0, t1 ;
D32SLL(xr0, xr0, xr11, xr13, 0x8); // xr13: l1, l0, lt, 0 ;
// CAL
Q8AVGR(xr1, xr11, xr7); // xr1: src[0,3], src[0,2]/src[2,3], src[0,1]/src[2,2], src[0,0]/src[2,1] ;
Q8AVG(xr2, xr12, xr8);
Q8AVGR(xr3, xr2, xr10); // xr3: src[1,0]/src[3,1], src[2,0], src[3,0], ~ ;
Q8AVG(xr4, xr13, xr7);
Q8AVGR(xr5, xr4, xr11); // xr5: src[1,3], src[1,2]/src[3,3], src[1,1]/src[3,2], ~ ;
// ALNI for STORE
S32ALNI(xr8, xr1, xr3, ptn3); // xr8: src[0,0]/src[2,1], src[1,0]/src[3,1], src[2,0], src[3,0] ;
S32SFL(xr9, xr1, xr5, xr10, ptn0); // xr9: src[0,3], src[1,3], src[0,2]/src[2,3], src[1,2]/src[3,3] ;
//xr10: src[0,1]/src[2,2], src[1,1]/src[3,2], src[0,0]/src[2,1], ~ ;
S32SFL(xr11, xr10, xr8, xr0, ptn3); // xr11: src[0,1]/src[2,2], src[1,1]/src[3,2],
// src[0,0]/src[2,1], src[1,0]/src[3,1] ;
S32ALNI(xr12, xr9, xr10, ptn2); // xr12: src[0,2]/src[2,3], src[1,2]/src[3,3],
// src[0,1]/src[2,2], src[1,1]/src[3,2] ;
// STORE
S32STDR(xr8, dst, 0x0);
S32SDIVR(xr11, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIVR(xr12, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIVR(xr9, dst, MB_LUMA_EDGED_WIDTH, 0x0);
}
// MODE 4
static void pred4x4_down_right_mxu(uint8_t *dst, uint8_t *src, uint8_t *topright,
uint8_t *top, uint8_t *topleft){
uint8_t *src_left; // left address
src_left = src - 0x4;
// load right
S32LDDR(xr8, top, 0x0); // xr8: t0, t1, t2, t3 ; high -> low, [31->0];
S32LDDR(xr9, topleft, -0x4); // xr9[7:0]: lt ;
// load left
S32LDD(xr7, topleft, -0x4); // xr7[31:24]: lt ;
S32LDD(xr1, src_left, 0x0); // xr1[31:24] <- src_left[3] (l0) ;
S32LDIV(xr2, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr2[31:24] <- src_left[stride+3] (l1) ;
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr3[31:24] <- src_left[2*stride+3] (l2) ;
S32LDIV(xr4, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr4[31:24] <- src_left[3*stride+3] (l3) ;
S32SFL(xr5, xr2, xr1, xr0, ptn2); // xr5[31:16] <- l1, l0 ;
S32SFL(xr6, xr4, xr3, xr0, ptn2); // xr6[31:16] <- l3, l2 ;
S32SFL(xr1, xr6, xr5, xr0, ptn3); // xr1[31: 0] <- l3, l2, l1, l0 ;
// alni
S32ALNI(xr10, xr9, xr8, ptn3); // xr10: lt, t0, t1, t2 ;
S32ALNI(xr11, xr1, xr7, ptn1); // xr11: l2, l1, l0, lt ;
S32ALNI(xr12, xr11, xr8, ptn2);// xr12: l0, lt, t0, t1 ;
S32ALNI(xr13, xr1, xr10, ptn2);// xr13: l1, l0, lt, t0 ;
// cal
Q8AVG(xr3, xr1, xr13);
Q8AVGR(xr4, xr3, xr11); // xr4: src[0,3], src[0,2]/src[1,3], src[0,1]/src[1,2]/src[2,3],
// src[0,0]/src[1,1]/src[2,2]/src[3,3] ;
Q8AVG(xr5, xr8, xr12);
Q8AVGR(xr6, xr5, xr10); // xr6: src[0,0]/src[1,1]/src[2,2]/src[3,3],
// src[1,0]/src[2,1]/src[3,2], src[2,0]/src[3,1], src[3,0] ;
// alni for store
D32SLL(xr7, xr6, xr0, xr0, 0x8); // xr7: src[1,0]/src[2,1]/src[3,2], src[2,0]/src[3,1], src[3,0] ;
S32ALNI(xr8, xr4, xr7, ptn1);
S32ALNI(xr9, xr4, xr7, ptn2);
//store
S32STDR(xr6, dst, 0x0);
S32SDIVR(xr9, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIVR(xr8, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIVR(xr4, dst, MB_LUMA_EDGED_WIDTH, 0x0);
}
// MODE 8
static void pred4x4_horizontal_up_mxu(uint8_t *dst, uint8_t *src, uint8_t *topright,
uint8_t *top, uint8_t *topleft){
uint8_t *src_left; // left address
src_left = src - 0x4;
//load
S32LDD(xr1, src_left, 0x0); // xr1[31:24] <- src_left[3] (l0) ;
S32LDIV(xr2, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr2[31:24] <- src_left[stride+3] (l1) ;
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr3[31:24] <- src_left[2*stride+3] (l2) ;
S32LDIV(xr4, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr4[31:24] <- src_left[3*stride+3] (l3) ;
S32SFL(xr5, xr2, xr1, xr0, ptn2); // xr5[31:16] <- l1, l0 ;
S32SFL(xr6, xr4, xr3, xr0, ptn2); // xr6[31:16] <- l3, l2 ;
S32SFL(xr1, xr6, xr5, xr0, ptn3); // xr1[31: 0] <- l3, l2, l1, l0 ;
D32SLL(xr2, xr1, xr0, xr0, 0x8); // xr2: l2, l1, l0, 0 ;
S32SFL(xr3, xr1, xr1, xr0, ptn0); // xr3: l3, l3, l2, l2;
Q8AVGR(xr4, xr1, xr2); // xr4: src[2,1]/src[0,2], src[2,0]/src[0,1], src[0,0], ~ ;
Q8AVG(xr5, xr2, xr3);
Q8AVGR(xr6, xr5, xr1); // xr6: src[3,1]/src[1,2], src[3,0]/src[1,1], src[1,0], ~ ;
S32SFL(xr7, xr6, xr4, xr0, ptn0); // xr7: src[3,1]/src[1,2], src[2,1]/src[0,2],
// src[3,0]/src[1,1], src[2,0]/src[0,1];
D32SLR(xr8, xr4, xr6, xr9, 0x8); // xr8: 0, src[2,1]/src[0,2], src[2,0]/src[0,1], src[0,0] ;
// xr9: 0, src[3,1]/src[1,2], src[3,0]/src[1,1], src[1,0] ;
S32SFL(xr0, xr9, xr8, xr10, ptn0); // xr10: src[3,0], src[2,0], src[1,0], src[0,0] ;
S32SFL(xr11, xr3, xr7, xr0, ptn3); // xr11: l3, l3, src[3,1]/src[1,2], src[2,1]/src[0,2] ;
S32SFL(xr12, xr3, xr3, xr0, ptn3); // xr12: l3, l3, l3, l3 ;
//store
S32STD(xr10, dst, 0x0);
S32SDIV(xr7, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIV(xr11, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32SDIV(xr12, dst, MB_LUMA_EDGED_WIDTH, 0x0);
}
void imdct_half_fix_c(MDCTContext_fix *s, FFTSample_fix *output,
const FFTSample_fix *input)
{
//PMON_ON(qmf);
int k, n8, n4, n2, n, j,j1;
const FFTSample_fix *in1, *in2;
const unsigned short *revtab = s->fft.revtab;
const FFTSample_fix *tcos = s->tcos;
const FFTSample_fix *tsin = s->tsin;
FFTComplex_fix *z = (FFTComplex_fix *)output;
n = 1 << s->nbits;//64
n2 = n >> 1;//32
n4 = n >> 2;//16
n8 = n >> 3;//8
/* pre rotation */
in1 = input; //head
in2 = input + n2 - 1;//tail
for(k = 0; k < n8; k++) {
#if 0
j=revtab[k];
FFT_CMUL_fix(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]);
in1 += 2;
in2 -= 2;
#else
FFTSample_fix _are,_bre,_aim,_bim,are,aim;
_are = *in2;
_bre = tcos[k];
_aim = *in1;
_bim = tsin[k];
j=revtab[k];
n=n4-k-1;
j1=revtab[n];
S32MUL(xr1,xr2, _are, _bre);
S32MUL(xr3, xr4, _are, _bim);
in2--;
are = *in2;
S32MUL(xr7,xr8, are, _bre);
S32MUL(xr9, xr10, are, _bim);
S32MSUB(xr1, xr2, _aim, _bim);
S32MADD(xr3, xr4, _aim, _bre); ;
in1++;
aim = *in1;
D32SLL(xr5,xr1,xr3,xr6,1);
S32MSUB(xr7, xr8, aim, _bim);
S32MADD(xr9, xr10, aim, _bre);
z[j].re=S32M2I(xr5);
D32SLL(xr11,xr7,xr9,xr12,1);
z[j].im=S32M2I(xr6);
in1++;
in2--;
z[j1].re=S32M2I(xr11);
z[j1].im=S32M2I(xr12);
#endif
}
s->fft.fft_calc(&s->fft, z);
/* post rotation + reordering */
/* XXX: optimize */
for(k = 0; k < n8; k++) {
FFTSample_fix r0, i0, r1, i1;
FFT_CMUL_fix(r0, i1, z[n8-k-1].im, z[n8-k-1].re, tsin[n8-k-1], tcos[n8-k-1]);
FFT_CMUL_fix(r1, i0, z[n8+k ].im, z[n8+k ].re, tsin[n8+k ], tcos[n8+k ]);
z[n8-k-1].re = r0;
z[n8-k-1].im = i0;
z[n8+k ].re = r1;
z[n8+k ].im = i1;
}
//PMON_OFF(qmf);
}
void Predict_16x16_C(const NEW_GMC_DATA * const This,
uint8_t *dst,
const uint8_t *src,
int dststride,
int srcstride,
int x,
int y,
int rounding)
{
const int W = This->sW;
const int H = This->sH;
const int rho = 3 - This->accuracy;
const int Rounder = ( (1<<7) - (rounding<<(2*rho)) ) << 16;
const int dUx = This->dU[0];
const int dVx = This->dV[0];
const int dUy = This->dU[1];
const int dVy = This->dV[1];
int Uo = This->Uo + 16*(dUy*y + dUx*x);
int Vo = This->Vo + 16*(dVy*y + dVx*x);
int i, j;
dst += 16;
{
unsigned int ri = 16;
unsigned int rj = 16;
int Offset;
int u,v;
uint8_t *srctmp;
uint32_t tmpf = 0;
S32I2M(xr15,dUx);
S32I2M(xr14,dVx);
S32I2M(xr13,dUy);
S32I2M(xr12,dVy);
S32I2M(xr11,Uo); // Uo 11
S32I2M(xr10,Vo); // Vo 10
S32I2M(xr5, Rounder);
for (j = 16; j>0; --j)
{
D32SLL(xr9,xr11,xr10,xr8, 0x0); // U 9 ,V 8
D32ASUM_AA(xr11,xr13,xr12,xr10); // += dUy; +=dVy;
for (i = -16; i<0; ++i)
{
ri = 16;
rj = 16;
// ( U >> 16 ) ,( V >> 16 )
D32SAR(xr7,xr9,xr8,xr6, 0x8);
D32SAR(xr7,xr7,xr6,xr6, 0x8);
D32SLLV(xr7,xr6, rho); // << rho
u = S32M2I(xr7);
v = S32M2I(xr6);
D32ASUM_AA(xr9,xr15,xr14,xr8); // U += dUx; V += dVx;
if (u > 0 && u <= W)
{
ri = MTab[u&15];
Offset = u>>4;
}
else
{
if (u > W)
Offset = W>>4;
else
Offset = 0;
ri = MTab[0];
}
if (v > 0 && v <= H)
{
rj = MTab[v&15];
Offset += (v>>4)*srcstride;
}
uint32_t
dequant_h263_intra_mxu(int16_t * data, uint8_t yuv_len,
const uint32_t quant,
const uint32_t dcscalar,
const uint16_t * mpeg_quant_matrices)
{
uint32_t i = 0;
S32LUI(xr9,1,0);
S32I2M(xr1,quant);
D32SLL(xr5,xr1,xr0,xr0,1);// quant_m_2
/* quant_add */
S32AND(xr15,xr1,xr9);
S32MOVN(xr2,xr15,xr1);
D32ADD_SS(xr1,xr1,xr9,xr3);
S32MOVZ(xr2,xr15,xr1);
S32I2M(xr3,-2048);
S32I2M(xr4,2047);
/* part1 */
//S32MUL(xr4,xr6,*data,dcscalar);
S32MUL(xr0,xr6,(int32_t)data[0],dcscalar);
D16MUL_WW(xr0,xr6,xr9,xr6);
S32MIN(xr6,xr6,xr4);
S32MAX(xr6,xr6,xr3);
/* part2 */
yuv_len = ((yuv_len&~1)+3)>>1;
data-=2;
for (i = 0; i < yuv_len; i++) {
S32LDI(xr1,data,4);
D16MUL_LW(xr13,xr9,xr1,xr14);// resave sign of data[i] and data[i+1]
D16CPS(xr1,xr1,xr1);
/* quant_m_2 * acLevel + quant_add */
D16MUL_LW(xr7,xr5,xr1,xr8);
D32ADD_AA(xr7,xr7,xr2,xr0);
D32ADD_AA(xr8,xr8,xr2,xr0);
#if 0
/* -2048 < data[i+1] <2047 */
S32CPS(xr7,xr7,xr13);
S32MAX(xr10,xr7,xr3);
S32MIN(xr10,xr10,xr4);
S32MOVZ(xr10,xr13,xr13);
/* -2048 < data[i] <2047 */
S32CPS(xr8,xr8,xr14);
S32MAX(xr11,xr8,xr3);
S32MIN(xr11,xr11,xr4);
S32MOVZ(xr11,xr14,xr14);
#else
/* -2048 < data[i+1] <2047 */
S32AND(xr7,xr7,xr4);
S32CPS(xr10,xr7,xr13);
S32MOVZ(xr10,xr13,xr13);
/* -2048 < data[i] <2047 */
S32AND(xr8,xr8,xr4);
S32CPS(xr11,xr8,xr14);
S32MOVZ(xr11,xr14,xr14);
#endif
S32SFL(xr0,xr10,xr11,xr12,3);
S32STD(xr12,data,0);
}
S16STD(xr6,data-(yuv_len*2-2),0,0);// data[0]
return(0);
}
void fft_calc_fix_inverse(FFTContext_fix *s, FFTComplex_fix *z)
{
int ln = s->nbits;
int j, np, np2;
int nblocks, nloops;
register FFTComplex_fix *p, *q;
FFTComplex_fix *exptab = s->exptab;
int l;
FFTSample_fix tmp_re, tmp_im;
np = 1 << ln;
/* function is :butterfly all 4 step ,N=16 */
/* pass 0 */
#if 0
p=&z[0];
j=(np >> 1);
do {
/*
X(k) = G(k)+H(k)*W (= e j*0)
*/
FFT_BF_fix(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
p+=2;
} while (--j);
#endif
/* pass 1 */
p=&z[0];
j=np >> 2;
do {
#if 1
S32LDD(xr1,p,0);
S32LDD(xr2,p,4);
S32LDD(xr3,p,8);
S32LDD(xr4,p,12);
S32LDD(xr5,p,16);
S32LDD(xr6,p,20);
S32LDD(xr7,p,24);
S32LDD(xr8,p,28);
D32ADD_AS(xr1,xr1,xr3,xr3);
D32ADD_AS(xr2,xr2,xr4,xr4);
D32ADD_AS(xr5,xr5,xr7,xr7);
D32ADD_AS(xr6,xr6,xr8,xr8);
D32ADD_AS(xr1,xr1,xr5,xr5);
D32ADD_AS(xr2,xr2,xr6,xr6);
D32ADD_SA(xr3,xr3,xr8,xr9);
D32ADD_AS(xr4,xr4,xr7,xr8);
S32STD(xr1,p,0);
S32STD(xr2,p,4);
S32STD(xr3,p,8);
S32STD(xr4,p,12);
S32STD(xr5,p,16);
S32STD(xr6,p,20);
S32STD(xr9,p,24);
S32STD(xr8,p,28);
#else
FFT_BF_fix(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
FFT_BF_fix(p[2].re, p[2].im, p[3].re, p[3].im,
p[2].re, p[2].im, p[3].re, p[3].im);
FFT_BF_fix(p[0].re, p[0].im, p[2].re, p[2].im,
p[0].re, p[0].im, p[2].re, p[2].im);
FFT_BF_fix(p[1].re, p[1].im, p[3].re, p[3].im,
p[1].re, p[1].im, -p[3].im, p[3].re);
#endif
p+=4;
} while (--j);
/* pass 2 .. ln-1 */
nblocks = np >> 3;
nloops = 1 << 2;
np2 = np >> 1;
do {
p = z;
q = z + nloops;
for (j = 0; j < nblocks; ++j) {
#if 1
S32LDD(xr1,p,0);
S32LDD(xr2,p,4);
S32LDD(xr3,q,0);
S32LDD(xr4,q,4);
D32ADD_AS(xr1,xr1,xr3,xr3);
D32ADD_AS(xr2,xr2,xr4,xr4);
S32STD(xr1,p,0);
S32STD(xr2,p,4);
S32STD(xr3,q,0);
S32STD(xr4,q,4);
#else
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, q->re, q->im);
#endif
p++;
q++;
for(l = nblocks; l < np2; l += nblocks) {
/* FFT_CMUL_fix( ) fuction is :
(-j 2*PI/N *km)
H(i) * E
*/
#if 1
FFTSample_fix _are = exptab[l].re;
FFTSample_fix _bre = q->re;
FFTSample_fix _aim = exptab[l].im;
FFTSample_fix _bim = q->im;
S32MUL(xr1, xr2, _are, _bre);
S32MUL(xr5, xr6, _are, _bim);
S32LDD(xr7,p,0);
S32MSUB(xr1, xr2, _aim, _bim);
S32MADD(xr5, xr6, _aim, _bre);
S32LDD(xr8,p,4);
D32SLL(xr1, xr1, xr5, xr5, 1);
D32ADD_AS(xr7,xr7,xr1,xr1);
D32ADD_AS(xr8,xr8,xr5,xr5);
S32STD(xr7,p,0);
S32STD(xr8,p,4);
S32STD(xr1,q,0);
S32STD(xr5,q,4);
#else
FFT_CMUL_fix(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, tmp_re, tmp_im);
#endif
p++;
q++;
}
p += nloops;
q += nloops;
}
nblocks = nblocks >> 1;
nloops = nloops << 1;
} while (nblocks);
}
/*
* NAME: layer->II()
* DESCRIPTION: decode a single Layer II frame
*/
int mad_layer_II(struct mad_stream *stream, struct mad_frame *frame)
{
struct mad_header *header = &frame->header;
struct mad_bitptr start;
unsigned int index, sblimit, nbal, nch, bound, gr, ch, s, sb;
unsigned char const *offsets;
unsigned char allocation[2][32], scfsi[2][32], scalefactor[2][32][3];
mad_fixed_t samples[3];
nch = MAD_NCHANNELS(header);
if (header->flags & MAD_FLAG_LSF_EXT)
index = 4;
else if (header->flags & MAD_FLAG_FREEFORMAT)
goto freeformat;
else {
unsigned long bitrate_per_channel;
bitrate_per_channel = header->bitrate;
if (nch == 2) {
bitrate_per_channel /= 2;
# if defined(OPT_STRICT)
/*
* ISO/IEC 11172-3 allows only single channel mode for 32, 48, 56, and
* 80 kbps bitrates in Layer II, but some encoders ignore this
* restriction. We enforce it if OPT_STRICT is defined.
*/
if (bitrate_per_channel <= 28000 || bitrate_per_channel == 40000) {
stream->error = MAD_ERROR_BADMODE;
return -1;
}
# endif
}
else { /* nch == 1 */
if (bitrate_per_channel > 192000 && bitrate_per_channel != 320000) {
/*
* ISO/IEC 11172-3 does not allow single channel mode for 224, 256,
* 320, or 384 kbps bitrates in Layer II.
*/
stream->error = MAD_ERROR_BADMODE;
return -1;
}
}
if (bitrate_per_channel <= 48000)
index = (header->samplerate == 32000) ? 3 : 2;
else if (bitrate_per_channel <= 80000)
index = 0;
else {
freeformat:
index = (header->samplerate == 48000) ? 0 : 1;
}
}
sblimit = sbquant_table[index].sblimit;
offsets = sbquant_table[index].offsets;
bound = 32;
if (header->mode == MAD_MODE_JOINT_STEREO) {
header->flags |= MAD_FLAG_I_STEREO;
bound = 4 + header->mode_extension * 4;
}
if (bound > sblimit)
bound = sblimit;
start = stream->ptr;
/* decode bit allocations */
for (sb = 0; sb < bound; ++sb) {
nbal = bitalloc_table[offsets[sb]].nbal;
for (ch = 0; ch < nch; ++ch)
allocation[ch][sb] = mad_bit_read(&stream->ptr, nbal);
}
for (sb = bound; sb < sblimit; ++sb) {
nbal = bitalloc_table[offsets[sb]].nbal;
allocation[0][sb] =
allocation[1][sb] = mad_bit_read(&stream->ptr, nbal);
}
/* decode scalefactor selection info */
for (sb = 0; sb < sblimit; ++sb) {
for (ch = 0; ch < nch; ++ch) {
if (allocation[ch][sb])
scfsi[ch][sb] = mad_bit_read(&stream->ptr, 2);
}
}
/* check CRC word */
if (header->flags & MAD_FLAG_PROTECTION) {
header->crc_check =
mad_bit_crc(start, mad_bit_length(&start, &stream->ptr),
header->crc_check);
if (header->crc_check != header->crc_target &&
!(frame->options & MAD_OPTION_IGNORECRC)) {
stream->error = MAD_ERROR_BADCRC;
return -1;
}
}
/* decode scalefactors */
for (sb = 0; sb < sblimit; ++sb) {
for (ch = 0; ch < nch; ++ch) {
if (allocation[ch][sb]) {
scalefactor[ch][sb][0] = mad_bit_read(&stream->ptr, 6);
switch (scfsi[ch][sb]) {
case 2:
scalefactor[ch][sb][2] =
scalefactor[ch][sb][1] =
scalefactor[ch][sb][0];
break;
case 0:
scalefactor[ch][sb][1] = mad_bit_read(&stream->ptr, 6);
/* fall through */
case 1:
case 3:
scalefactor[ch][sb][2] = mad_bit_read(&stream->ptr, 6);
}
if (scfsi[ch][sb] & 1)
scalefactor[ch][sb][1] = scalefactor[ch][sb][scfsi[ch][sb] - 1];
# if defined(OPT_STRICT)
/*
* Scalefactor index 63 does not appear in Table B.1 of
* ISO/IEC 11172-3. Nonetheless, other implementations accept it,
* so we only reject it if OPT_STRICT is defined.
*/
if (scalefactor[ch][sb][0] == 63 ||
scalefactor[ch][sb][1] == 63 ||
scalefactor[ch][sb][2] == 63) {
stream->error = MAD_ERROR_BADSCALEFACTOR;
return -1;
}
# endif
}
}
}
/* decode samples */
for (gr = 0; gr < 12; ++gr) {
for (sb = 0; sb < bound; ++sb) {
for (ch = 0; ch < nch; ++ch) {
if ((index = allocation[ch][sb])) {
#ifdef JZ4750_OPT
mad_fixed_t sf_val;
mad_fixed_t *sb_ptr;
sb_ptr = &(frame->sbsample[ch][3*gr-1][sb]);
sf_val = sf_table[scalefactor[ch][sb][gr/4]];
index = offset_table[bitalloc_table[offsets[sb]].offset][index - 1];
II_samples(&stream->ptr, &qc_table[index], samples);
S32MUL(xr1,xr2, samples[0], sf_val);
S32MUL(xr3,xr4, samples[1], sf_val);
S32MUL(xr5,xr6, samples[2], sf_val);
S32EXTR(xr1,xr2,(32 - MAD_F_SCALEBITS), 31);
S32EXTR(xr3,xr4,(32 - MAD_F_SCALEBITS), 31);
S32EXTR(xr5,xr6,(32 - MAD_F_SCALEBITS), 31);
D32SLL(xr1,xr1,xr3,xr3,1);
D32SLL(xr5,xr5,xr0,xr0,1);
S32SDIV(xr1, sb_ptr, 32, 2);
S32SDIV(xr3, sb_ptr, 32, 2);
S32SDIV(xr5, sb_ptr, 32, 2);
#else
index = offset_table[bitalloc_table[offsets[sb]].offset][index - 1];
II_samples(&stream->ptr, &qc_table[index], samples);
for (s = 0; s < 3; ++s) {
frame->sbsample[ch][3 * gr + s][sb] =
mad_f_mul(samples[s], sf_table[scalefactor[ch][sb][gr / 4]]);
}
#endif
}
else {
for (s = 0; s < 3; ++s)
frame->sbsample[ch][3 * gr + s][sb] = 0;
}
}
}
for (sb = bound; sb < sblimit; ++sb) {
if ((index = allocation[0][sb])) {
index = offset_table[bitalloc_table[offsets[sb]].offset][index - 1];
II_samples(&stream->ptr, &qc_table[index], samples);
for (ch = 0; ch < nch; ++ch) {
#ifdef JZ4750_OPT
mad_fixed_t sf_val;
mad_fixed_t *sb_ptr;
sb_ptr = &(frame->sbsample[ch][3*gr-1][sb]);
sf_val = sf_table[scalefactor[ch][sb][gr/4]];
S32MUL(xr1,xr2, samples[0], sf_val);
S32MUL(xr3,xr4, samples[1], sf_val);
S32MUL(xr5,xr6, samples[2], sf_val);
S32EXTR(xr1,xr2,(32 - MAD_F_SCALEBITS), 31);
S32EXTR(xr3,xr4,(32 - MAD_F_SCALEBITS), 31);
S32EXTR(xr5,xr6,(32 - MAD_F_SCALEBITS), 31);
D32SLL(xr1,xr1,xr3,xr3,1);
D32SLL(xr5,xr5,xr0,xr0,1);
S32SDIV(xr1, sb_ptr, 32, 2);
S32SDIV(xr3, sb_ptr, 32, 2);
S32SDIV(xr5, sb_ptr, 32, 2);
#else
for (s = 0; s < 3; ++s) {
frame->sbsample[ch][3 * gr + s][sb] =
mad_f_mul(samples[s], sf_table[scalefactor[ch][sb][gr / 4]]);
}
#endif
}
}
else {
for (ch = 0; ch < nch; ++ch) {
for (s = 0; s < 3; ++s)
frame->sbsample[ch][3 * gr + s][sb] = 0;
}
}
}
for (ch = 0; ch < nch; ++ch) {
for (s = 0; s < 3; ++s) {
for (sb = sblimit; sb < 32; ++sb)
frame->sbsample[ch][3 * gr + s][sb] = 0;
}
}
}
return 0;
}
// MODE 3
static void pred16x16_plane_mxu(uint8_t *dst, uint8_t *src, uint8_t *top){
int i, j, k, a;
uint8_t *src_top; // top address
uint8_t *src_topleft, *src_left; // left address
src_top = top;
src_topleft = src_top - 0x14;
src_left = src - 0x4;
//----- H, LOAD -----
S32LDD(xr1, src_top, -0x14); // xr1 <- src_top[-4]; xr1: lt, 0, 0, 0 ;
S32LDD(xr5, src_top, 0x0); // xr5 <- src_top[0] ; xr5: t3, t2, t1, t0 ;
S32LDD(xr2, src_top, 0x4); // xr2 <- src_top[4] ; xr2: t7, t6, t5, t4 ;
S32LDDR(xr3, src_top, 0x8); // xr3 <- src_top[8] ; xr3: t8, t9, t10, t11 ;
S32LDDR(xr4, src_top, 0xc); // xr4 <- src_top[12]; xr4: t12, t13, t14, t15 ;
S32ALNI(xr1, xr5, xr1, ptn1); // xr1: t2, t1, t0, lt ;
S32ALNI(xr2, xr2, xr5, ptn1); // xr2: t6, t5, t4, t3 ; ---xr5 is free to use ;
S32I2M(xr9, MUL_12); // xr9 : 0x00010002 ;
S32I2M(xr10, MUL_34); // xr10: 0x00030004 ;
//----- H, SUM -----
Q8ADDE_SS(xr5, xr3, xr2, xr6); // xr5[31:16] <- t8-t6 ; xr5[15:0] <- t9-t5 ;
// xr6[31:16] <- t10-t4; xr6[15:0] <- t11-t3;
S32I2M(xr11, MUL_56); // xr11: 0x00050006 ;
D16MUL_WW(xr13, xr9, xr5, xr14); // xr13 <- 1*(t8-t6) ; xr14 <- 2*(t9-t5) ;
D16MAC_AA_WW(xr13, xr10, xr6, xr14); // xr13 <- 1*(t8-t6)+3*(t10-t4) ; xr14 <- 2*(t9-t5)+4*(t11-t3) ;
Q8ADDE_SS(xr5, xr4, xr1, xr6); // xr5[31:16] <- t12-t2; xr5[15:0] <- t13-t1;
// xr6[31:16] <- t14-t0; xr6[15:0] <- t15-lt;
S32I2M(xr12, MUL_78); // xr12: 0x00070008 ;
D16MAC_AA_WW(xr13, xr11, xr5, xr14); // xr13 <- 1*(t8-t6)+3*(t10-t4)+5*(t12-t2) ;
// xr14 <- 2*(t9-t5)+4*(t11-t3)+6*(t13-t1) ;
D16MAC_AA_WW(xr13, xr12, xr6, xr14); // xr13 <- 1*(t8-t6)+3*(t10-t4)+5*(t12-t2)+7*(t14-t0) ;
// xr14 <- 2*(t9-t5)+4*(t11-t3)+6*(t13-t1)+8*(t15-lt) ;
S32LDD(xr1, src_topleft, 0x0); // xr1[31:24] <- src_topleft[3] (lt) ;
S32LDD(xr2, src_left, 0x0); // xr2[31:24] <- src_topleft[stride+3] (l0) ;
D32ADD_AA(xr15, xr13, xr14, xr0); // xr15 <- 1*(t8-t6)+3*(t10-t4)+5*(t12-t2)+7*(t14-t0)
// + 2*(t9-t5)+4*(t11-t3)+6*(t13-t1)+8*(t15-lt) ;
//----- V, LOAD -----
// S32LDD(xr1, src_topleft, 0x0); // xr1[31:24] <- src_topleft[3] (lt) ;
// S32LDIV(xr2, src_topleft, stride, 0x0); // xr2[31:24] <- src_topleft[stride+3] (l0) ;
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr3[31:24] <- src_topleft[2*stride+3] (l1) ;
S32LDIV(xr8, src_left, MB_LUMA_EDGED_WIDTH, 0x0); // xr9[31:24] <- src_topleft[3*stride+3] (l2) ;
S32SFL(xr5, xr2, xr1, xr0, ptn2); // xr5[31:16] <- l0, lt ;
S32SFL(xr6, xr8, xr3, xr0, ptn2); // xr8[31:16] <- l2, l1 ;
S32SFL(xr7, xr6, xr5, xr0, ptn3); // xr7[31: 0] <- l2, l1, l0, lt ;
S32LDIV(xr1, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr2, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr8, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32SFL(xr5, xr2, xr1, xr0, ptn2);
S32SFL(xr6, xr8, xr3, xr0, ptn2);
S32SFL(xr13, xr6, xr5, xr0, ptn3); // xr13[31:0] <- l6, l5, l4, l3 ;
src_left += MB_LUMA_EDGED_WIDTH;
S32LDIV(xr8, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr2, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr1, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32SFL(xr6, xr8, xr3, xr0, ptn2);
S32SFL(xr5, xr2, xr1, xr0, ptn2);
S32SFL(xr14, xr6, xr5, xr0, ptn3); // xr14[31:0] <- l8, l9, l10, l11 ;
S32LDIV(xr8, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr3, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr2, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32LDIV(xr1, src_left, MB_LUMA_EDGED_WIDTH, 0x0);
S32SFL(xr6, xr8, xr3, xr0, ptn2);
S32SFL(xr5, xr2, xr1, xr0, ptn2);
S32SFL(xr1, xr6, xr5, xr0, ptn3); // xr1[31: 0] <- l12, l13, l14, l15 ;
//----- V, SUM -----
Q8ADDE_SS(xr5, xr14, xr13, xr6);
Q8ADDE_SS(xr2, xr1, xr7, xr3);
D16MUL_WW(xr13, xr9, xr5, xr14);
D16MAC_AA_WW(xr13, xr10, xr6, xr14);
D16MAC_AA_WW(xr13, xr11, xr2, xr14);
D16MAC_AA_WW(xr13, xr12, xr3, xr14);
D32SLR(xr2, xr11, xr12, xr3, 0x8); // xr2: 0x00000500 ; xr3: 0x00000700 ;
D32SLR(xr11, xr2, xr3, xr12, 0x8); //xr11: 0x00000005 ; xr12: 0x00000007 ;
D32ADD_AA(xr14, xr13, xr14, xr0); // xr14 <- 1*(l8-l6)+3*(l10-l4)+5*(l12-l2)+7*(l14-l0)
// + 2*(l9-l5)+4*(l11-l3)+6*(l13-l1)+8*(l15-lt) ;
//----- P, CAL -----
// D32SLR(xr2, xr11, xr12, xr3, 0x8); // xr2: 0x00000500 ; xr3: 0x00000700 ;
// D32SLR(xr11, xr2, xr3, xr12, 0x8); //xr11: 0x00000005 ; xr12: 0x00000007 ;
D16MUL_WW(xr0, xr15, xr11, xr2); // xr2: 5*H ;
D16MUL_WW(xr0, xr14, xr11, xr3); // xr3: 5*V ;
D32SLR(xr8, xr11, xr0, xr0, 0x2); // xr8: 0x00000001 ;
D32SLL(xr13, xr8, xr0, xr0, 0x5); //xr13: 0x00000020 ;
Q8ACCE_AA(xr0, xr1, xr4, xr8); // xr8[15:0]: src1[0] + src2[16] + 1
D32ADD_AA(xr5, xr2, xr13, xr0); // xr5: 5*H+32 ;
D32ADD_AA(xr6, xr3, xr13, xr0); // xr6: 5*V+32 ;
D32SLR(xr2, xr5, xr6, xr3, 0x6); // xr2: ( 5*H+32 ) >> 6 ; xr3: ( 5*V+32 ) >> 6 ;
// Q8ACCE_AA(xr0, xr1, xr4, xr8); // xr8[15:0]: src1[0] + src2[16] + 1
D32SLL(xr5, xr8, xr0, xr0, 0x4); // xr5[15:0]: 16*(src1[0] + src2[16] + 1)
Q16ADD_AA_WW(xr7, xr2, xr3, xr0); // xr7: V+H
// S32NOR(xr0, xr0, xr0); // idle
S32I2M(xr4, MUX_H16); // xr4: 0x0000ffff ;
D16MUL_WW(xr0, xr7, xr12, xr8); // xr8: 7*(V+H)
S32SFL(xr0, xr3, xr3, xr14, ptn3); // xr14[31:16]: V ; xr14[15:0]: V ;
D32SLL(xr7, xr2, xr0, xr0, 0x1);
Q16ADD_SS_WW(xr9, xr5, xr8, xr0); // xr9: 16*(src1[0] + src2[16] + 1) - 7*(V+H)
S32SFL(xr0, xr9, xr9, xr5, ptn3); // xr5[31:16]: a ; xr5[15:0]: a ;
// S32SFL(xr0, xr3, xr3, xr14, ptn3); // xr14[31:16]: V ; xr14[15:0]: V ;
// D32SLL(xr7, xr2, xr0, xr0, 0x1);
S32SFL(xr0, xr7, xr7, xr8, ptn3); // xr8[31:16]: 2H ; xr8[15:0]: 2H ;
S32AND(xr2, xr4, xr2);
Q16ADD_AA_WW(xr15, xr5, xr2, xr0); // xr15[31:16]: a ; xr15[15:0]: a + H ;
dst -= MB_LUMA_EDGED_WIDTH;
//----- SRC, STORE -----
for (i=0; i<16; i++) {
Q16ADD_AA_WW(xr1, xr15, xr8, xr0);
Q16ADD_AA_WW(xr2, xr1, xr8, xr0);
Q16SAR(xr9, xr15, xr1, xr1, 0x5);
Q16ADD_AA_WW(xr3, xr2, xr8, xr0);
Q16SAT(xr10, xr9, xr1);
Q16ADD_AA_WW(xr4, xr3, xr8, xr0);
Q16SAR(xr2, xr2, xr3, xr3, 0x5);
Q16ADD_AA_WW(xr5, xr4, xr8, xr0);
Q16SAT(xr11, xr2, xr3);
Q16ADD_AA_WW(xr6, xr5, xr8, xr0);
Q16SAR(xr4, xr4, xr5, xr5, 0x5);
Q16ADD_AA_WW(xr7, xr6, xr8, xr0);
Q16SAR(xr6, xr6, xr7, xr7, 0x5);
Q16SAT(xr12, xr4, xr5);
Q16SAT(xr13, xr6, xr7);
S32SDIVR(xr10, dst, MB_LUMA_EDGED_WIDTH, 0x0);
S32STDR(xr11, dst, 0x4);
S32STDR(xr12, dst, 0x8);
// S32STDR(xr13, dst, 0xc);
Q16ADD_AA_WW(xr15, xr15, xr14, xr0);
S32STDR(xr13, dst, 0xc);
}
}
// MODE 3
static void pred8x8_plane_mxu(uint8_t *dst, uint8_t *src, uint8_t *top){
unsigned int i;
uint8_t *src_top; // top address
uint8_t *src_topleft, *src_left; // left address
src_top = top;
src_topleft = src_top - 0x1c;
src_left = src - 0x4;
//----- H, LOAD -----
S32LDD(xr1, src_top, -0x1c); // xr1 <- src_top[-4]; xr1: lt, 0, 0, 0 ;
S32LDD(xr3, src_top, 0x0); // xr3 <- src_top[0] ; xr3: t3, t2, t1, t0 ;
S32LDDR(xr2, src_top, 0x4); // xr2 <- src_top[4] ; xr2: t4, t5, t6, t7 ;
S32ALNI(xr1, xr3, xr1, ptn1);// xr1: t2, t1, t0, lt ;
S32I2M(xr8, MUL_12); // xr8: 0x00010002 ;
S32I2M(xr9, MUL_34); // xr9: 0x00030004 ;
//----- H, SUM -----
Q8ADDE_SS(xr3, xr2, xr1, xr4); // xr3[31:16] <- t4-t2 ; xr3[15:0] <- t5-t1 ;
// xr4[31:16] <- t6-t0 ; xr4[15:0] <- t7-lt;
S32LDD(xr1, src_topleft, 0x0); // xr1[31:24] <- src_topleft[3] (lt) ;
D16MUL_WW(xr5, xr8, xr3, xr6); // xr5 <- 1*(t4-t2) ; xr6 <- 2*(t5-t1) ;
D16MAC_AA_WW(xr5, xr9, xr4, xr6); // xr5 <- 1*(t4-t2)+3*(t6-t0) ; xr6 <- 2*(t5-t1)+4*(t7-lt) ;
S32LDD(xr12, src_left, 0x0);//xr12[31:24] <- src_topleft[stride+3] (l0) ;
S32LDIV(xr3, src_left, MB_CHROM_EDGED_WIDTH, 0x0); // xr3[31:24] <- src_topleft[2*stride+3] (l1) ;
D32ADD_AA(xr7, xr5, xr6, xr0); // xr7 <- 1*(t4-t2)+3*(t6-t0)+2*(t5-t1)+4*(t7-lt) ;
//----- V, LOAD -----
// S32LDD(xr1, src_topleft, 0x0); // xr1[31:24] <- src_topleft[3] (lt) ;
// S32LDIV(xr12, src_topleft, stride, 0x0);//xr12[31:24] <- src_topleft[stride+3] (l0) ;
// S32LDIV(xr3, src_topleft, stride, 0x0); // xr3[31:24] <- src_topleft[2*stride+3] (l1) ;
S32LDIV(xr4, src_left, MB_CHROM_EDGED_WIDTH, 0x0); // xr4[31:24] <- src_topleft[3*stride+3] (l2) ;
S32SFL(xr5, xr12, xr1, xr0, ptn2); // xr5[31:16] <- l0, lt ;
S32SFL(xr6, xr4, xr3, xr0, ptn2); // xr8[31:16] <- l2, l1 ;
S32SFL(xr10, xr6, xr5, xr0, ptn3); // xr10[31:0] <- l2, l1, l0, lt ;
src_left += MB_CHROM_EDGED_WIDTH;
S32LDIV(xr4, src_left, MB_CHROM_EDGED_WIDTH, 0x0);
S32LDIV(xr3, src_left, MB_CHROM_EDGED_WIDTH, 0x0);
S32LDIV(xr12, src_left, MB_CHROM_EDGED_WIDTH, 0x0);
S32LDIV(xr1, src_left, MB_CHROM_EDGED_WIDTH, 0x0);
S32SFL(xr6, xr4, xr3, xr0, ptn2);
S32SFL(xr5, xr12, xr1, xr0, ptn2);
S32SFL(xr11, xr6, xr5, xr0, ptn3); // xr11[31:0] <- l4, l5, l6, l7 ;
//----- V, SUM -----
Q8ADDE_SS(xr3, xr11, xr10, xr4);
S32LUI(xr1, 0x1, ptn0); // xr1[31:0]: 0x00000001 ;
D16MUL_WW(xr5, xr8, xr3, xr6);
D16MAC_AA_WW(xr5, xr9, xr4, xr6);
D32ADD_AA(xr13, xr5, xr6, xr0); // xr13 <- 1*(l4-l2)+3*(l6-l0)+2*(l5-l1)+4*(l7-lt) ;
//----- P, CAL ----- useful XRs:xr13, xr7, xr2, xr11;
// S32LUI(xr1, 0x1, ptn0); // xr1[31:0]: 0x00000001 ;
D32SLL(xr5, xr1, xr1, xr6, 0x4); // xr5: 0x00000010; xr6: 0x00000010;
D32SLL(xr3, xr13, xr7, xr4, 0x4);
D32ACC_AA(xr5, xr13, xr3, xr0); // xr5: 17*V+16
D32ACC_AA(xr6, xr7, xr4, xr0); // xr6: 17*H+16
Q8ACCE_AA(xr0, xr2, xr11, xr1); // xr1[15:0]: src1[0] + src2[8] + 1
D32SLR(xr8, xr5, xr6, xr9, 0x5); // xr8: (17*V+16) >> 5 ; xr9: (17*H+16) >> 5 ;
// Q8ACCE_AA(xr0, xr2, xr11, xr1); // xr1[15:0]: src1[0] + src2[8] + 1
D32SLL(xr2, xr1, xr0, xr0, 0x4); // xr2[15:0]: 16*(src1[0] + src2[16] + 1)
Q16ADD_AA_WW(xr7, xr8, xr9, xr0); // xr7: V+H
S32I2M(xr4, MUX_H16); // xr4: 0x0000ffff ;
D32SLL(xr12, xr7, xr0, xr0, 0x1);
D32ADD_AA(xr5, xr12, xr7, xr0); // xr5: 3*(V+H)
// S32LUI(xr12, 0x3, ptn0); // xr12[31:0]: 0x00000003 ;
// D16MUL_WW(xr0, xr7, xr12, xr5); // xr5: 3*(V+H)
// S32I2M(xr4, MUX_H16); // xr4: 0x0000ffff ;
Q16ADD_SS_WW(xr6, xr2, xr5, xr0); // xr6: 16*(src1[0] + src2[16] + 1) - 3*(V+H)
// S32I2M(xr4, MUX_H16); // xr4: 0x0000ffff ;
S32SFL(xr0, xr8, xr8, xr14, ptn3);// xr14[31:16]: V ; xr14[15:0]: V ;
S32SFL(xr0, xr6, xr6, xr5, ptn3); // xr5[31:16]: a ; xr5[15:0]: a ;
D32SLL(xr7, xr9, xr0, xr0, 0x1);
S32SFL(xr0, xr7, xr7, xr8, ptn3); // xr8[31:16]: 2H ; xr8[15:0]: 2H ;
// S32I2M(xr4, MUX_H16); // xr4: 0x0000ffff ;
S32AND(xr9, xr4, xr9);
Q16ADD_AA_WW(xr15, xr5, xr9, xr0); // xr15[31:16]: a ; xr15[15:0]: a + H ;
dst -= MB_CHROM_EDGED_WIDTH;
//----- SRC, STORE -----
for (i=0; i<8; i++) {
Q16ADD_AA_WW(xr1, xr15, xr8, xr0);
Q16ADD_AA_WW(xr2, xr1, xr8, xr0);
Q16SAR(xr9, xr15, xr1, xr1, 0x5);
Q16ADD_AA_WW(xr3, xr2, xr8, xr0);
Q16SAT(xr10, xr9, xr1);
// Q16SAR(xr9, xr15, xr1, xr1, 0x5);
Q16SAR(xr2, xr2, xr3, xr3, 0x5);
// Q16SAT(xr10, xr9, xr1);
Q16SAT(xr11, xr2, xr3);
S32SDIVR(xr10, dst, MB_CHROM_EDGED_WIDTH, 0x0);
Q16ADD_AA_WW(xr15, xr15, xr14, xr0);
S32STDR(xr11, dst, 0x4);
}
}