uint32_t
dequant_mpeg_intra_mxu(int16_t * data,
// const int16_t * coeff,
const uint32_t quant,
const uint32_t dcscalar,
const uint16_t * mpeg_quant_matrices)
{
const uint16_t *intra_matrix = mpeg_quant_matrices;
int32_t i = 0;
/* deal with data[0] then save to xr6 */
S32I2M(xr3,-2048);
S32I2M(xr4,2047);
S32I2M(xr5,quant);
S32MUL(xr0,xr6,(int32_t)data[0],dcscalar);
S32LUI(xr9,1,0);
D16MUL_WW(xr0,xr6,xr9,xr6);
S32MIN(xr6,xr6,xr4);
S32MAX(xr6,xr6,xr3);
data-=2;
intra_matrix-=2;
for (i = 0; i < 32; i++) {
S32LDI(xr1,data,4);
S32LDI(xr2,intra_matrix,4);
D16MUL_LW(xr13,xr9,xr1,xr14); // resave values of data[i] and data[i+1]
D16CPS(xr1,xr1,xr1);
/* abs(level) *( intra_matrix[i]*quant) >> 3 */
D16MUL_LW(xr7,xr5,xr2,xr8);
S32SFL(xr15,xr7,xr8,xr2,3);
D16MUL_WW(xr7,xr1,xr2,xr8);
D32SLR(xr7,xr7,xr8,xr8,3);
/* -2048 < data[i+1] < 2047 */
S32CPS(xr7,xr7,xr13);
S32MAX(xr10,xr7,xr3);
S32MIN(xr10,xr10,xr4);
/* -2048 < data[i] < 2047 */
S32CPS(xr8,xr8,xr14);
S32MAX(xr11,xr8,xr3);
S32MIN(xr11,xr11,xr4);
S32SFL(xr0,xr10,xr11,xr12,3);
S32STD(xr12,data,0);
}
S16STD(xr6,data,-62*2,0);//xr6 to data[0]
return(0);
}
OMX_BOOL AudioDecoder::ALumeDecInit(OMX_AUDIO_CONFIG_EQUALIZERTYPE* aEqualizerType){
S32I2M(xr16,7);
return OMX_TRUE;
}
static void ff_vp3_idct_add_mxu(uint8_t *src, int stride, DCTELEM *input, uint8_t idct_row)
{
int i;
DCTELEM *blk;
int32_t wf = (int32_t)whirl_idct;
S32LDD(xr5, wf, 0x0); // xr5(w7, w3)
S32LDD(xr6, wf, 0x4); // xr6(w9, w8)
S32LDD(xr7, wf, 0x8); // xr7(w11,w10)
S32LDD(xr8, wf, 0xc); // xr8(w13,w12)
S32LDD(xr9, wf, 0x10); // xr9(w6, w0)
S32LDD(xr10,wf, 0x14);
blk = input - 8;
/* Inverse DCT on the rows now */
for (i=0; i<idct_row; i++) {
S32LDI(xr1, blk, 0x10); // xr1 (x4, x0)
S32LDD(xr2, blk, 0x4); // xr2 (x7, x3)
S32LDD(xr3, blk, 0x8); // xr3 (x6, x1)
S32LDD(xr4, blk, 0xc); // xr4 (x5, x2)
S32OR(xr12, xr2,xr3);
S32OR(xr11,xr12,xr4);
S32OR(xr12,xr11,xr1);
if (S32M2I(xr12) == 0) {
continue; //blk[0]= blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=0
}
S32SFL(xr12,xr0,xr1,xr13,ptn3);
S32OR(xr11,xr11,xr12);
if (S32M2I(xr11) == 0 && S32M2I(xr13) != 0) {
D16MUL_HW(xr0,xr5,xr13,xr13);
D32SAR(xr0,xr0,xr13,xr13,15);
S32SFL(xr0,xr13,xr13,xr13,ptn3);
S32STD(xr13,blk, 0x0);
S32STD(xr13,blk, 0x4);
S32STD(xr13,blk, 0x8);
S32STD(xr13,blk, 0xc);
continue; //blk[0]!=0, and blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=0
}
S32SFL(xr1,xr1,xr2,xr2, ptn3); //xr1:s1, s3, xr2: s0, s2
S32SFL(xr3,xr3,xr4,xr4, ptn3); //xr3:s5, s7, xr4: s4, s6
D16MUL_WW(xr11, xr2, xr5, xr12);//xr11: s0*c4, xr12: s2*c2
D16MAC_AA_WW(xr11,xr4,xr6,xr12);//xr11: s0*c4+s4*c4, xr12: s2*c2+s6*c6
D16MUL_WW(xr13, xr2, xr6, xr14);//xr13: s0*c4, xr14: s2*c6
D16MAC_SS_WW(xr13,xr4,xr5,xr14);//xr13: s0*c4 - s4*c4, xr14: s2*c6-s6*c2
D16MUL_HW(xr2, xr1, xr7, xr4); //xr2: s1*c1, xr4: s1*c3
D16MAC_AS_LW(xr2,xr1,xr9,xr4); //xr2: s1*c1+s3*c3, xr4: s1*c3-s3*c7
D16MAC_AS_HW(xr2,xr3,xr10,xr4); //xr2: s1*c1+s3*c3+s5*c5,
// xr4: s1*c3-s3*c7-s5*c1
D16MAC_AS_LW(xr2,xr3,xr8,xr4); //xr2: s1*c1+s3*c3+s5*c5+s7*c7,
//xr4: s1*c3-s3*c7-s5*c1-s7*c5
D32SAR(xr11, xr11,xr13,xr13,15);
S32SFL(xr0, xr11,xr13,xr11,ptn3);
D32SAR(xr12,xr12,xr14,xr14,15);
S32SFL(xr0, xr12,xr14,xr12,ptn3);
D32SAR(xr2, xr2,xr4,xr4,15);
S32SFL(xr0, xr2,xr4,xr2,ptn3);
D16MUL_HW(xr4, xr1, xr8, xr15); //xr4: s1*c7, xr15:s1*c5
D16MAC_SS_LW(xr4,xr1,xr10,xr15); //xr4: s1*c7-s3*c5, xr15: s1*c5-s3*c1
D16MAC_AA_HW(xr4,xr3,xr9,xr15); //xr4: s1*c7-s3*c5+s5*c3, xr15: s1*c5-s3*c1+s5*c7
D16MAC_SA_LW(xr4,xr3,xr7,xr15); //xr4: s1*c7-s3*c5+s5*c3-s7*c1
//xr15: s1*c5-s3*c1+s5*c7+s7*c3
Q16ADD_AS_WW(xr11,xr11,xr12,xr12); //xr11: rnd(s0*c4+s4*c4)>>15+rnd(s2*c2+s6*c6)>>15
// rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15
//xr12: rnd(s0*c4+s4*c4)>>15-rnd(s2*c2+s6*c6)>>15
// rnd(s0*c4-s4*c4)>>15-rnd(s2*c6-s6*c2)>>15
D32SAR(xr15,xr15,xr4,xr4,15);
S32SFL(xr0,xr15,xr4,xr15,ptn3);
Q16ADD_AS_WW(xr11, xr11, xr2, xr2);
//xr11: rnd(s0*c4+s4*c4)>>15+rnd(s2*c2+s6*c6)>>15 + rnd(s1*c1+s3*c3+s5*c5+s7*c7)>>15
// : rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15 + rnd(s1*c3-s3*c7-s5*c1-s7*c5)>>15
//xr2: rnd(s0*c4+s4*c4)>>15+rnd(s2*c2+s6*c6)>>15 - rnd(s1*c1+s3*c3+s5*c5+s7*c7)>>15
// : rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15 - rnd(s1*c3-s3*c7-s5*c1-s7*c5)>>15
Q16ADD_AS_XW(xr12, xr12, xr15, xr15);
//xr12: rnd(s0*c4+s4*c4)>>15-rnd(s2*c2+s6*c6)>>15+rnd(s1*c5-s3*c1+s5*c7+s7*c3)>>15
// : rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15+rnd(s1*c7-s3*c5+s5*c3-s7*c1)>>15
//xr15: rnd(s0*c4+s4*c4)>>15-rnd(s2*c2+s6*c6)>>15-rnd(s1*c5-s3*c1+s5*c7+s7*c3)>>15
// : rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15-rnd(s1*c7-s3*c5+s5*c3-s7*c1)>>15
S32SFL(xr11,xr11,xr12,xr12, ptn3);
//xr11: rnd(s0*c4+s4*c4)>>15+rnd(s2*c2+s6*c6)>>15 + rnd(s1*c1+s3*c3+s5*c5+s7*c7)>>15
// : rnd(s0*c4+s4*c4)>>15-rnd(s2*c2+s6*c6)>>15+rnd(s1*c5-s3*c1+s5*c7+s7*c3)>>15
//xr12: rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15 + rnd(s1*c3-s3*c7-s5*c1-s7*c5)>>15
// : rnd(s0*c4-s4*c4)>>15+rnd(s2*c6-s6*c2)>>15+rnd(s1*c7-s3*c5+s5*c3-s7*c1)>>15
S32SFL(xr12,xr12,xr11,xr11, ptn3);
//xr12: rnd(s0*c4-s4*c4)>>16+rnd(s2*c6-s6*c2)>>16 + rnd(s1*c3-s3*c7-s5*c1-s7*c5)>>16
// : rnd(s0*c4+s4*c4)>>16+rnd(s2*c2+s6*c6)>>16 + rnd(s1*c1+s3*c3+s5*c5+s7*c7)>>16
//xr11: rnd(s0*c4-s4*c4)>>16+rnd(s2*c6-s6*c2)>>16+rnd(s1*c7-s3*c5+s5*c3-s7*c1)>>16
// : rnd(s0*c4+s4*c4)>>16-rnd(s2*c2+s6*c6)>>16+rnd(s1*c5-s3*c1+s5*c7+s7*c3)>>16
S32STD(xr12, blk, 0x0);
S32STD(xr11, blk, 0x4);
S32STD(xr15, blk, 0x8);
S32STD(xr2, blk, 0xc);
}
blk = input - 2;
for (i=0; i<4; i++) /* idct columns */
{
S32I2M(xr5,wxr5); //xr5: c4 , c2
S32I2M(xr6,wxr6); //xr5: c4 , c2
S32LDI(xr1, blk, 0x4); //xr1: ss0, s0
S32LDD(xr3, blk, 0x20); //xr3: ss2, s2
S32LDD(xr11, blk, 0x40); //xr11: ss4, s4
S32LDD(xr13, blk, 0x60); //xr13: ss6, s6
D16MUL_HW(xr15, xr5, xr1, xr2); //xr15: ss0*c4, xr9: s0*c4
D16MAC_AA_HW(xr15,xr5,xr11,xr2); //xr15: ss0*c4+ss4*c4, xr9: s0*c4+s4*c4
D16MUL_LW(xr10, xr5, xr3, xr9); //xr10: ss2*c2, xr9: s2*c2
D16MAC_AA_LW(xr10,xr6,xr13,xr9); //xr10: ss2*c2+ss6*c6, xr9: s2*c2+s6*c6
D32SAR(xr15,xr15,xr2,xr2,15);
S32SFL(xr0,xr15,xr2,xr15,ptn3); //xr15: (ss0*c4+ss4*c4)>>15
D32SAR(xr10,xr10,xr9,xr9,15);
S32SFL(xr0,xr10,xr9,xr10,ptn3); //xr10: (ss2*c2+ss6*c6)>>15
S32LDD(xr2, blk, 0x10); //xr2: ss1, s1
S32LDD(xr4, blk, 0x30); //xr4: ss3, s3
Q16ADD_AS_WW(xr15,xr15,xr10,xr9); //xr15: rnd(ss0*c4+ss4*c4)>>15+rnd(ss2*c2+ss6*c6)>>15
// :rnd(s0*c4+s4*c4)>>15 + rnd(s2*c2 + s6*c6)>>15
//xr9: rnd(ss0*c4+ss4*c4)>>15 - rnd(ss2*c2+ss6*c6)>>15
// : rnd(s0*c4+s4*c4)>>15 - rnd(s2*c2 + s6*c6)>>15
D16MUL_HW(xr10, xr5, xr1, xr1); //xr10: ss0*c4, xr1: s0*c4
D16MAC_SS_HW(xr10,xr5,xr11,xr1); //xr10: ss0*c4-ss4*c4, xr1: s0*c4 - s4*c4
D16MUL_LW(xr11, xr6, xr3, xr12); //xr11: ss2*c6, xr1: s2*c6
D16MAC_SS_LW(xr11,xr5,xr13,xr12); //xr11: ss2*c6-ss6*c2, xr1: s2*c6-s6*c2
D32SAR(xr10,xr10,xr1,xr1,15);
S32SFL(xr0,xr10,xr1,xr10,ptn3); //xr10: (ss0*c4-ss4*c4)>>15 // : (s0*c4 - s4*c4)>>15
D32SAR(xr11,xr11,xr12,xr12,15);
S32SFL(xr0,xr11,xr12,xr11,ptn3); //xr11:(ss2*c6-ss6*c2)>>15
// :(s2*c6-s6*c2)>>15
S32LDD(xr12, blk, 0x50); //xr12: ss5, s5
S32LDD(xr14, blk, 0x70); //xr14: ss7, s7
Q16ADD_AS_WW(xr10,xr10,xr11,xr1); //xr10: rnd(ss0*c4-ss4*c4)>>15)+rnd(ss2*c6-ss6*c2)>>15
// : rnd(s0*c4 - s4*c4)>>15 +rnd(s2*c6 - s6*c2)>>15
//xr1 : rnd(ss0*c4-ss4*c4)>>15-rnd(ss2*c6-ss6*c2)>>15
// : rnd(s0*c4 - s4*c4)>>15-rnd(s2*c6 - s6*c2)>>15
D16MUL_HW(xr11, xr7, xr2, xr13); //xr11: ss1*c1, xr13: s1*c1
D16MAC_AA_LW(xr11,xr7,xr4,xr13); //xr11: ss1*c1+ss3*c3, xr13: s1*c1+s3*c3
D16MAC_AA_LW(xr11,xr8,xr12,xr13); //xr11: ss1*c1+ss3*c3+ss5*c5 //xr13: s1*c1+s3*c3+s5*c5
D16MAC_AA_HW(xr11,xr8,xr14,xr13); //xr11: ss1*c1+ss3*c3+ss5*c5+ss7*c7
//xr13: s1*c1+s3*c3+s5*c5+s7*c7
D16MUL_LW(xr3, xr7, xr2, xr5); //xr3: ss1*c3, xr13: s1*c3
D16MAC_SS_HW(xr3,xr8,xr4,xr5); //xr3: ss1*c3-ss3*c7, xr13: s1*c3-s3*c7
D16MAC_SS_HW(xr3,xr7,xr12,xr5); //xr3: ss1*c3-ss3*c7-ss5*c1
//xr13: s1*c3-s3*c7-s5*c1
D16MAC_SS_LW(xr3,xr8,xr14,xr5); //xr3: ss1*c3-ss3*c7-ss5*c1-ss7*c5
//xr13: s1*c3-s3*c7-s7*c5
D32SAR(xr11,xr11,xr13,xr13,15);
S32SFL(xr0,xr11,xr13,xr11,ptn3); //xr11: (ss1*c1+ss3*c3+ss5*c5+ss7*c7)>>15 // : (s1*c1+s3*c3+s5*c5+s7*c7)>>15
D32SAR(xr3,xr3,xr5,xr5,15);
S32SFL(xr0,xr3,xr5,xr3,ptn3); //xr3: (ss1*c3-ss3*c7-ss5*c1-ss7*c5)>>15
// : (s1*c3-s3*c7-s7*c5)>>15
D16MUL_LW(xr5, xr8, xr2, xr13); //xr5: ss1*c5, xr13:s1*c5
D16MAC_SS_HW(xr5,xr7,xr4,xr13); //xr5: ss1*c5-ss3*c1, xr13:s1*c5-s3*c1
D16MAC_AA_HW(xr5,xr8,xr12,xr13); //xr5: ss1*c5-ss3*c1+ss5*c7
// : s1*c5 - s3*c1+ s5*c7
D16MAC_AA_LW(xr5,xr7,xr14,xr13); //xr5: ss1*c5-ss3*c1+ss5*c7+ss7*c1
// : s1*c5 - s3*c1+ s5*c7+ s7*c1
D16MUL_HW(xr2, xr8, xr2, xr6); //xr2: ss1*c7, xr13: s1*c7
D16MAC_SS_LW(xr2,xr8,xr4,xr6); //xr2: ss1*c7-ss3*c5, xr13: s1*c7-s3*c5
D16MAC_AA_LW(xr2,xr7,xr12,xr6); //xr2: ss1*c7-ss3*c5+ss5*c1 //xr13: s1*c7-s3*c5+s5*c1
D16MAC_SS_HW(xr2,xr7,xr14,xr6); //xr2: ss1*c7-ss3*c5+ss5*c1-ss7*c3
//xr13: s1*c7-s3*c5+s5*c1-s7*c3
D32SAR(xr5,xr5,xr13,xr13,15);
S32SFL(xr0,xr5,xr13,xr5,ptn3); //xr5: (ss1*c5-ss3*c1+ss5*c7+ss7*c1)>>15 // :(s1*c5 - s3*c1+ s5*c7+ s7*c1)>>15
D32SAR(xr2,xr2,xr6,xr6,15);
S32SFL(xr0,xr2,xr6,xr2,ptn3); //xr2:(ss1*c7-ss3*c5+ss5*c1-ss7*c3)>>15
// :(s1*c7-s3*c5+s5*c1-s7*c3)>>15
S32I2M(xr4, 0x00080008);//round value 8;
Q16ADD_AS_WW(xr15,xr15,xr11,xr11); //xr15:rnd(ss0*c4+ss4*c4)>>16+rnd(ss2*c2+ss6*c6)>>16+
// rnd(ss1*c1+ss3*c3+ss5*c5+ss7*c7)>>16
// rnd(s0*c4+s4*c4)>>16 + rnd(s2*c2 + s6*c6)>>16+
// rnd(s1*c1+s3*c3+s5*c5+s7*c7)>>16
//xr11:rnd(ss0*c4+ss4*c4)>>16+rnd(ss2*c2+ss6*c6)>>16-
// rnd(ss1*c1+ss3*c3+ss5*c5+ss7*c7)>>16
// rnd(s0*c4+s4*c4)>>16 + rnd(s2*c2 + s6*c6)>>16-
// rnd(s1*c1+s3*c3+s5*c5+s7*c7)>>16
Q16ADD_AS_WW(xr10,xr10,xr3,xr3); //xr10:rnd(ss0*c4-ss4*c4)>>16)+rnd(ss2*c6-ss6*c2)>>16+
// rnd(ss1*c3-ss3*c7-ss5*c1-ss7*c5)>>16
// rnd(s0*c4 - s4*c4)>>16 +rnd(s2*c6 - s6*c2)>>16+
// rnd(s1*c3-s3*c7-s7*c5)>>16
//xr10:rnd(ss0*c4-ss4*c4)>>16)+rnd(ss2*c6-ss6*c2)>>16-
// rnd(ss1*c3-ss3*c7-ss5*c1-ss7*c5)>>16
// rnd(s0*c4 - s4*c4)>>16 +rnd(s2*c6 - s6*c2)>>16-
// rnd(s1*c3-s3*c7-s7*c5)>>16
Q16ADD_AS_WW(xr1,xr1,xr5,xr5); //xr1: rnd(ss0*c4-ss4*c4)>>16-rnd(ss2*c6-ss6*c2)>>16+
// rnd(ss1*c5-ss3*c1+ss5*c7+ss7*c1)>>16
// rnd(s0*c4 - s4*c4)>>16 +rnd(s2*c6 - s6*c2)>>16+
// rnd(s1*c5 - s3*c1+ s5*c7+ s7*c1)>>16
//xr1: rnd(ss0*c4-ss4*c4)>>16-rnd(ss2*c6-ss6*c2)>>16-
// rnd(ss1*c5-ss3*c1+ss5*c7+ss7*c1)>>16
// rnd(s0*c4 - s4*c4)>>16 +rnd(s2*c6 - s6*c2)>>16-
// rnd(s1*c5 - s3*c1+ s5*c7+ s7*c1)>>16
Q16ADD_AS_WW(xr9,xr9,xr2,xr2); //xr9: rnd(ss0*c4+ss4*c4)>>16 - rnd(ss2*c2+ss6*c6)>>16+
// rnd(ss1*c7-ss3*c5+ss5*c1-ss7*c3)>>16
// rnd(s0*c4+s4*c4)>>16 - rnd(s2*c2 + s6*c6)>>16+
// rnd(s1*c7-s3*c5+s5*c1-s7*c3)>>16
//xr9: rnd(ss0*c4+ss4*c4)>>16 - rnd(ss2*c2+ss6*c6)>>16-
// rnd(ss1*c7-ss3*c5+ss5*c1-ss7*c3)>>16
// rnd(s0*c4+s4*c4)>>16 - rnd(s2*c2 + s6*c6)>>16-
// rnd(s1*c7-s3*c5+s5*c1-s7*c3)>>16
Q16ACCM_AA(xr15,xr4,xr4,xr10);
Q16ACCM_AA(xr11,xr4,xr4,xr1);
Q16ACCM_AA(xr9,xr4,xr4,xr2);
Q16ACCM_AA(xr5,xr4,xr4,xr3);
Q16SAR(xr15,xr15,xr10,xr10,4);
Q16SAR(xr11,xr11,xr1,xr1,4);
Q16SAR(xr9,xr9,xr2,xr2,4);
Q16SAR(xr5,xr5,xr3,xr3,4);
S32STD(xr15, blk, 0x00);
S32STD(xr10, blk, 0x10);
S32STD(xr1, blk, 0x20);
S32STD(xr9, blk, 0x30);
S32STD(xr2, blk, 0x40);
S32STD(xr5, blk, 0x50);
S32STD(xr3, blk, 0x60);
S32STD(xr11, blk, 0x70);
}
blk = input - 8;
src -= stride;
for (i=0; i<8; i++) {
S32LDIV(xr1, src, stride, 0x0);
S32LDI(xr3, blk, 0x10);
S32LDD(xr4, blk, 0x4);
Q8ACCE_AA(xr4, xr1, xr0, xr3);
S32LDD(xr2, src, 0x4);
S32LDD(xr5, blk, 0x8);
S32LDD(xr6, blk, 0xc);
Q8ACCE_AA(xr6, xr2, xr0, xr5);
Q16SAT(xr1, xr4, xr3);
S32STD(xr1, src, 0x0);
Q16SAT(xr2, xr6, xr5);
S32STD(xr2, src, 0x4);
}
}
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 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;
}
__tcsm1_main int main() {
int c;
S32I2M(xr16,3);
c = i_la(_gp);
int fifo_rp = 0;
unsigned int XCHGtmp;
H264_Frame_GlbARGs *dFRM = (H264_Frame_GlbARGs *)TCSM1_FRM_ARGS;
H264_MB_DecARGs *dMB_aux = (H264_MB_DecARGs *)TCSM1_MBARGS_BUF;
H264_MB_DecARGs *dMB2 = (H264_MB_DecARGs *)TCSM1_MBARGS_BUF2;
H264_AUX_T *AUX_T = (H264_AUX_T *)TCSM1_AUX_T;
MDMA_DesNode *MDMA1_TRAN = (MDMA_DesNode *)TCSM1_GP1_TRAN_CHAIN;
MDMA_DesNode *MDMA1_ARG = (MDMA_DesNode *)TCSM1_GP1_ARG_CHAIN;
H264_MB_DecARGs *dMBsrc;
fifo_wp = (int *)TCSM1_FIFO_WP;
dMBsrc = (H264_MB_DecARGs *)(dFRM->dMB_baseaddr_aux);
int mb_num;
int mb_start;
int total_mbs;
mb_start = dFRM->first_mb_in_frame;
total_mbs = dFRM->mb_width * dFRM->mb_height;
int i;
uint16_t *intra_pred4x4_top_ptr[2];
intra_pred4x4_top_ptr[0]=intra_pred4x4_top[0];
intra_pred4x4_top_ptr[1]=intra_pred4x4_top[1];
uint16_t *intra_pred4x4_left_ptr[2];
intra_pred4x4_left_ptr[0]=intra_pred4x4_left[0];
intra_pred4x4_left_ptr[1]=intra_pred4x4_left[1];
uint8_t mb_x_d1, mb_y_d1;
uint8_t mb_x_d2, mb_y_d2;
mb_x_d1 = mb_y_d1 = mb_x_d2 = mb_y_d2 =0;
AUX_PMON_CREAT(mc);
AUX_PMON_CREAT(idct);
AUX_PMON_CREAT(dblk);
AUX_PMON_CREAT(intra);
AUX_PMON_CREAT(mdma);
AUX_PMON_CREAT(sync);
do{
}while(fifo_rp == *fifo_wp);
fifo_rp++;
MDMA1_ARG->TSA = get_phy_addr_aux((uint32_t)dMBsrc);
MDMA1_ARG->TDA = (uint32_t)dMB_aux;
MDMA1_ARG->STRD = MDMA_STRD(64, 64);
MDMA1_ARG->UNIT = MDMA_UNIT(1,64,(sizeof(H264_MB_DecARGs)));
SET_MDMA1_DHA((uint32_t)TCSM1_GP1_ARG_CHAIN);
MDMA1_RUN();
dMBsrc++;
POLLING_MDMA1_END();
char bakup_x0 = dFRM->mb_width-1;
char bakup_x1 = 0;
int mb_type_last= 0;
volatile unsigned int *infar = (unsigned int *)TCSM1_H264_DBLK_INFAR;
uint8_t non_zero_count_cache_luma_last[16];
for ( mb_num = mb_start; (mb_num < total_mbs+2); mb_num ++ ) {
AUX_PMON_ON(sync);
do{
}while(fifo_rp == *fifo_wp);
fifo_rp++;
int gp1_tran_start = 0;
int gp1_tran_len = sizeof(H264_MB_DecARGs);
if (dMB_aux->next_mb_no_weight_flag) {
gp1_tran_start = (53 << 2);
gp1_tran_len -= (53 << 2);
}
if (dMB_aux->next_mb_skip_flag) {
gp1_tran_len -= (192 << 2);
}
AUX_PMON_OFF(sync);
AUX_PMON_ON(mdma);
AUX_PMON_OFF(mdma);
AUX_PMON_ON(intra);
MDMA1_ARG->TSA = (get_phy_addr_aux((uint32_t)dMBsrc) + gp1_tran_start);
MDMA1_ARG->TDA = ((uint32_t)dMB2 + gp1_tran_start);
MDMA1_ARG->STRD = MDMA_STRD(64, 64);
MDMA1_ARG->UNIT = MDMA_UNIT(1,64,gp1_tran_len);
dMBsrc++;
POLLING_MDMA1_END(); //ensure curr dblked MB has been trans out
SET_MDMA1_DHA((uint32_t)TCSM1_GP1_ARG_CHAIN);
MDMA1_RUN();
int mb_x= dMB_aux->mb_x;
int mb_y= dMB_aux->mb_y;
int mb_type= dMB_aux->mb_type;
if (dMB_aux->curr_mb_skip_flag)
for(i=0; i<24; i++)
dMB_aux->mb[i*16] = 0;
AUX_T->mc_des_dirty=0;
if(IS_INTRA(mb_type_last)){
// chroma predict
Intra_pred_chroma(dMB_aux->chroma_pred_mode, AUX_T->h264_urecon[1],
AUX_T->h264_urecon[0] + MB_CHROM_WIDTH, TCSM1_BOTTOM_U_13PIX+4);
Intra_pred_chroma(dMB_aux->chroma_pred_mode, AUX_T->h264_vrecon[1],
AUX_T->h264_vrecon[0] + MB_CHROM_WIDTH, TCSM1_BOTTOM_V_13PIX+4);
// luma predict
if(IS_INTRA4x4(mb_type_last)){
if(IS_8x8DCT(mb_type_last)){
for(i=0; i<16; i+=4){
uint8_t * src_left = (uint32_t)(intra_pred4x4_left_ptr[1][i]) | 0xF4000000;
uint8_t * src_top = (uint32_t)(intra_pred4x4_top_ptr[1][i]) | 0xF4000000;
uint8_t * src_topleft;
if ( i==8 )
src_topleft = src_left - RECON_BUF_STRIDE - 1;
else src_topleft = src_top - 1;
uint8_t * ptr= AUX_T->h264_yrecon[1] + MB_offset_recon[i];
int dir= dMB_aux->intra4x4_pred_mode_cache[ i ];
int nnz = non_zero_count_cache_luma_last[i];
Intra_pred_luma_8x8l(dir,ptr,(dMB_aux->topleft_samples_available<<i)&0x8000,
(dMB_aux->topright_samples_available<<i)&0x4000, src_left, src_top, src_topleft, RECON_BUF_STRIDE);
if(nnz){
if(nnz == 1 && dMB_aux->mb[i*16])
ff_h264_idct8_dc_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else
ff_h264_idct8_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}else {
for(i=0; i<16; i++){
uint8_t * src_left = (uint32_t)(intra_pred4x4_left_ptr[1][i]) | 0xf4000000;
uint8_t * src_top = (uint32_t)(intra_pred4x4_top_ptr[1][i]) | 0xf4000000;
uint8_t * src_topleft;
if ( (i==2) || (i==8) || (i==10) )
src_topleft = src_left - RECON_BUF_STRIDE;
else src_topleft = src_top;
uint8_t * ptr= AUX_T->h264_yrecon[1] + MB_offset_recon[i];
uint8_t *topright;
int dir= dMB_aux->intra4x4_pred_mode_cache[ i ];
int nnz, tr;
if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
int topright_avail= (dMB_aux->topright_samples_available<<i)&0x8000;
if(!topright_avail){
tr= src_top[3]*0x01010101;
topright= (uint8_t*) &tr;
}else{
topright= src_top + 4;
}
}
Intra_pred_luma_4x4(dir, ptr, src_left, topright, src_top, src_topleft);
nnz = non_zero_count_cache_luma_last[i];
if(nnz){
if(nnz == 1 && dMB_aux->mb[i*16])
ff_h264_idct_dc_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else
ff_h264_idct_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
}else{
Intra_pred_luma_16x16(dMB_aux->intra16x16_pred_mode, AUX_T->h264_yrecon[1],
AUX_T->h264_yrecon[0] + MB_LUMA_WIDTH, TCSM1_BOTTOM_Y_25PIX+4);
luma_dc_dequant_idct_c(dMB_aux->mb, dMB_aux->dequant4_coeff[0]);
for(i=0; i<16; i++){
if(non_zero_count_cache_luma_last[i])
ff_h264_idct_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else if(dMB_aux->mb[i*16])
ff_h264_idct_dc_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
AUX_PMON_OFF(intra);
if(IS_INTER(mb_type)){
hl_motion_hw(dFRM, dMB_aux, AUX_T);
}
AUX_PMON_ON(mc);
MC_POLLING_END();
AUX_PMON_OFF(mc);
AUX_PMON_ON(dblk);
while ((*infar)!= TCSM1_H264_DBLK_INFDA) {}
AUX_PMON_OFF(dblk);
if(AUX_T->mc_des_dirty){
H264_MC_DesNode *h264_mc = (H264_MC_DesNode *)(AUX_T->h264_mc_des_ptr[0]);
h264_mc[AUX_T->mc_des_dirty - 1].VNodeHead = H264_VHEAD_UNLINK;
SET_MC_DHA((uint32_t)h264_mc);
CLEAR_MC_TTEND();
SET_MC_DCS();
}
if(IS_INTRA(mb_type)){
uint32_t * bakup_src = AUX_T->BackupMBbottom_Y[bakup_x1]-4;
uint32_t * bakup_dst = TCSM1_BOTTOM_Y_25PIX;
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
bakup_dst[4] = bakup_src[4];
bakup_dst[5] = bakup_src[5];
bakup_dst[6] = bakup_src[6];
bakup_src = AUX_T->BackupMBbottom_U[bakup_x1]-4;
bakup_dst = TCSM1_BOTTOM_U_13PIX;
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
bakup_src = AUX_T->BackupMBbottom_V[bakup_x1]-4;
bakup_dst = TCSM1_BOTTOM_V_13PIX;
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
}
if(IS_INTER(mb_type_last))
hl_motion_hw_next(dMB_aux,AUX_T,mb_type_last);
*(uint32_t*)&AUX_T->sub_mb_type[0] = *(uint32_t*)&dMB_aux->sub_mb_type[0];
*(uint32_t*)&AUX_T->sub_mb_type[2] = *(uint32_t*)&dMB_aux->sub_mb_type[2];
AUX_PMON_ON(idct);
AUX_PMON_OFF(idct);
if (IS_INTRA_PCM(mb_type)) {
unsigned int x, y;
for(i=0; i<24; i++) {
for (y=0; y<4; y++) {
for (x=0; x<4; x++) {
*(AUX_T->h264_yrecon[0] + MB_offset_recon[i] + y*RECON_BUF_STRIDE + x) = dMB_aux->mb[i*16+y*4+x];
}
}
}
}
{
int linesize, uvlinesize;
uint8_t *dest_y_d2, *dest_u_d2, *dest_v_d2;
linesize = dFRM->linesize;
uvlinesize = dFRM->uvlinesize;
dest_y_d2 = dFRM->current_picture.y_ptr + (mb_y_d2 * 16* linesize ) + mb_x_d2 * 16;
dest_u_d2 = dFRM->current_picture.u_ptr + (mb_y_d2 * 8 * uvlinesize) + mb_x_d2 * 8;
dest_v_d2 = dFRM->current_picture.v_ptr + (mb_y_d2 * 8 * uvlinesize) + mb_x_d2 * 8;
//move dblked MB out
MDMA1_TRAN[0].TSA = AUX_T->h264_ydblk[1]-4;
MDMA1_TRAN[0].TDA = get_phy_addr_aux((uint32_t)dest_y_d2-4);
MDMA1_TRAN[1].TSA = AUX_T->h264_ydblk[1]-4+DBLK_U_OFST;
MDMA1_TRAN[1].TDA = get_phy_addr_aux((uint32_t)dest_u_d2-4);
//MDMA1_TRAN[1].UNIT = MDMA_UNIT(1,12,96);
MDMA1_TRAN[2].TSA = AUX_T->h264_ydblk[1]-4+DBLK_V_OFST;
MDMA1_TRAN[2].TDA = get_phy_addr_aux((uint32_t)dest_v_d2-4);
POLLING_MDMA1_END(); //ensure next MB's args has been received
SET_MDMA1_DHA((uint32_t)MDMA1_TRAN);
MDMA1_RUN();
}
//---------idct fo inter---------------
if(IS_INTER(mb_type_last)){
void (*idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
int di;
if(IS_8x8DCT(mb_type_last)){
idct_dc_add = ff_h264_idct8_dc_add_c;
idct_add = ff_h264_idct8_add_c;
di = 4;
}else{
idct_dc_add = ff_h264_idct_dc_add_c;
idct_add = ff_h264_idct_add_c;
di = 1;
}
for(i=0; i<16; i+=di){
int nnz = non_zero_count_cache_luma_last[i];
if(nnz){
if(nnz==1 && dMB_aux->mb[i*16])
idct_dc_add(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else
idct_add(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
//------------- chroma idct------------
if(mb_type_last){
chroma_dc_dequant_idct_c(dMB_aux->mb + 16*16, dMB_aux->dequant4_coeff[1]);
chroma_dc_dequant_idct_c(dMB_aux->mb + 16*16+4*16, dMB_aux->dequant4_coeff[2]);
for(i=16; i<16+8; i++){
if(dMB_aux->non_zero_count_cache_chroma[ i - 16 ]) {
ff_h264_idct_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
} else if (dMB_aux->mb[i*16]) {
ff_h264_idct_dc_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
if(!(mb_x==0 && mb_y==0)){
SET_DHA_DBLK((unsigned int)AUX_T->h264_dblk_des_ptr[1]);
*infar = 0;
SET_DCS_DBLK(0x1);
}
filter_mb_dblk(dFRM, dMB_aux, AUX_T);
{
uint32_t * bakup_src = AUX_T->h264_yrecon[1] + 15*RECON_BUF_STRIDE;
uint32_t * bakup_dst = AUX_T->BackupMBbottom_Y[bakup_x0];
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
bakup_src = AUX_T->h264_urecon[1] + 7*RECON_BUF_STRIDE;
bakup_dst = AUX_T->BackupMBbottom_U[bakup_x0];
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_src = AUX_T->h264_vrecon[1] + 7*RECON_BUF_STRIDE;
bakup_dst = AUX_T->BackupMBbottom_V[bakup_x0];
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
}
mb_x_d2 = mb_x_d1;
mb_y_d2 = mb_y_d1;
mb_x_d1 = mb_x;
mb_y_d1 = mb_y;
mb_type_last=mb_type;
for(i=0;i<16;i++)
non_zero_count_cache_luma_last[i]=dMB_aux->non_zero_count_cache_luma[ scan5[i] ];
// abnormal exit
if (dMB_aux->deblocking_filter & 0x8) break;
bakup_x0=(bakup_x0==((dFRM->mb_width)-1))?0:(bakup_x0+1);//hit second line's tail
bakup_x1=(bakup_x1==((dFRM->mb_width)-1))?0:(bakup_x1+1);
XCHG2(AUX_T->h264_yrecon[0],AUX_T->h264_yrecon[1],XCHGtmp);
XCHG2(AUX_T->h264_urecon[0],AUX_T->h264_urecon[1],XCHGtmp);
XCHG2(AUX_T->h264_vrecon[0],AUX_T->h264_vrecon[1],XCHGtmp);
XCHG2(AUX_T->h264_ydblk[0],AUX_T->h264_ydblk[1],XCHGtmp);
XCHG2(AUX_T->h264_dblk_des_ptr[0],AUX_T->h264_dblk_des_ptr[1],XCHGtmp);
XCHG2(AUX_T->h264_mc_des_ptr[0],AUX_T->h264_mc_des_ptr[1],XCHGtmp);
XCHG2(intra_pred4x4_top_ptr[0],intra_pred4x4_top_ptr[1],XCHGtmp);
XCHG2(intra_pred4x4_left_ptr[0],intra_pred4x4_left_ptr[1],XCHGtmp);
XCHG2(dMB_aux,dMB2,XCHGtmp);
}
AUX_PMON_TRAN(mc,PMON_MC_BUF);
AUX_PMON_TRAN(idct,PMON_IDCT_BUF);
AUX_PMON_TRAN(dblk,PMON_DBLK_BUF);
AUX_PMON_TRAN(intra,PMON_INTRA_BUF);
AUX_PMON_TRAN(mdma,PMON_MDMA_BUF);
AUX_PMON_TRAN(sync,PMON_SYNC_BUF);
/*
task_done: aux-cpu task done flag, only write by aux-cpu, only read by main-cpu
*/
MDMA1_ARG->TSA = (TCSM1_TASK_DONE);
MDMA1_ARG->TDA = (TCSM0_TASK_DONE);
MDMA1_ARG->STRD = MDMA_STRD(4,4);
MDMA1_ARG->UNIT = MDMA_UNIT(1,4,4);
SET_MDMA1_DHA((uint32_t)TCSM1_GP1_ARG_CHAIN);
MDMA1_RUN();
i_nop;
i_nop;
i_nop;
i_nop;
__asm__ __volatile__ ("wait");
}
// 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);
}
}
/**
*
* @param buf_size size of buf in bits
*/
void ff_init_cabac_decoder_mxu(CABACContext *c, const uint8_t *buf, int buf_size){
c->bytestream_start=
c->bytestream= buf;
c->bytestream_end= buf + buf_size;
#if CABAC_BITS == 16
c->low = (*c->bytestream++)<<18;
c->low+= (*c->bytestream++)<<10;
#else
c->low = (*c->bytestream++)<<10;
#endif
c->low+= ((*c->bytestream++)<<2) + 2;
c->range= 0x1FE;
#ifdef JZC_TCSM_OPT
S32I2M(xr12,ff_h264_lps_range);
S32I2M(xr13,ff_h264_mlps_state+128);
S32I2M(xr14,c->low);
S32I2M(xr15,c->range);
S32I2M(xr11,c->bytestream);
S32I2M(xr10,0x8000);
S32I2M(xr9,0xFFFEFFFF);
S32I2M(xr8,0x8000);
S32I2M(xr7,2);
S32I2M(xr6,0xFFFEFFFF);
S32I2M(xr5,c->bytestream_start);
#endif
}
void ff_init_cabac_states_mxu(CABACContext *c){
uint32_t i = 0;
uint8_t *p_lps_range = lps_range[0]-4;
uint8_t *p_mps_state = mps_state-4;
uint8_t *p_lps_state = lps_state-4;
uint8_t *p_ff_h264_mlps_state1 = &ff_h264_mlps_state[0]+124;
uint8_t *p_ff_h264_mlps_state2 = &ff_h264_mlps_state[0]+128;//
uint8_t *p_ff_h264_lps_range = ff_h264_lps_range-128;
S32I2M(xr15,0x1010101);
for(i=0; i<16; i++){
/* load 32 */
S32LDI(xr1,p_lps_range,4);
S32LDI(xr2,p_lps_range,4);
S32LDI(xr3,p_lps_range,4);
S32LDI(xr4,p_lps_range,4);
S32SFL(xr7,xr2,xr1,xr8,0);
S32SFL(xr9,xr4,xr3,xr10,0);
S32SFL(xr11,xr7,xr7,xr12,0);
S32SFL(xr13,xr8,xr8,xr14,0);
S32SFL(xr1,xr9,xr9,xr2,0);
S32SFL(xr3,xr10,xr10,xr4,0);
S32SDI(xr14,p_ff_h264_lps_range,128);
S32SDI(xr4,p_ff_h264_lps_range,4);
S32SDI(xr13,p_ff_h264_lps_range,124);
S32SDI(xr3,p_ff_h264_lps_range,4);
S32SDI(xr12,p_ff_h264_lps_range,124);
S32SDI(xr2,p_ff_h264_lps_range,4);
S32SDI(xr11,p_ff_h264_lps_range,124);
S32SDI(xr1,p_ff_h264_lps_range,4);
p_ff_h264_lps_range-=128*4-4;
/* part2 */
S32LDI(xr1,p_mps_state,4);
Q8ADD_AA(xr3,xr1,xr1);// 2*mps_state[i]+0
Q8ADD_AA(xr4,xr3,xr15);//2*mps_state[i]+1
S32SFL(xr11,xr4,xr3,xr12,0);
S32SDI(xr12,p_ff_h264_mlps_state1,4);// ff_h264_mlps_state[128+2*i+1]= 2*mps_state[i]+0;
S32SDI(xr11,p_ff_h264_mlps_state1,4);
/*part3 */
S32LDI(xr1,p_lps_state,4);
Q8ADD_AA(xr3,xr1,xr1);// 2*mps_state[i]+0
Q8ADD_AA(xr4,xr3,xr15);//2*mps_state[i]+1
S32SFL(xr11,xr3,xr4,xr12,0);
S32ALN(xr11,xr11,xr11,2);
S32ALN(xr12,xr12,xr12,2);
S32SDI(xr12,p_ff_h264_mlps_state2,-4);
S32SDI(xr11,p_ff_h264_mlps_state2,-4);
}
ff_h264_mlps_state[127]= 1;
ff_h264_mlps_state[126]= 0;
}
