|
|
|
@ -54,8 +54,6 @@ static const float odd_table[] __attribute__ ((aligned(8))) = { |
|
|
|
|
#undef c6 |
|
|
|
|
#undef c7 |
|
|
|
|
|
|
|
|
|
#if 1 |
|
|
|
|
|
|
|
|
|
#define load_matrix(table) \ |
|
|
|
|
do { \
|
|
|
|
|
const float *t = table; \
|
|
|
|
@ -84,52 +82,11 @@ static const float odd_table[] __attribute__ ((aligned(8))) = { |
|
|
|
|
register float fr2 __asm__("fr2"); \
|
|
|
|
|
register float fr3 __asm__("fr3") |
|
|
|
|
|
|
|
|
|
#else |
|
|
|
|
|
|
|
|
|
/* generic C code for check */ |
|
|
|
|
|
|
|
|
|
static void ftrv_(const float xf[],float fv[]) |
|
|
|
|
{ |
|
|
|
|
float f0,f1,f2,f3; |
|
|
|
|
f0 = fv[0]; |
|
|
|
|
f1 = fv[1]; |
|
|
|
|
f2 = fv[2]; |
|
|
|
|
f3 = fv[3]; |
|
|
|
|
fv[0] = xf[0]*f0 + xf[4]*f1 + xf[ 8]*f2 + xf[12]*f3; |
|
|
|
|
fv[1] = xf[1]*f0 + xf[5]*f1 + xf[ 9]*f2 + xf[13]*f3; |
|
|
|
|
fv[2] = xf[2]*f0 + xf[6]*f1 + xf[10]*f2 + xf[14]*f3; |
|
|
|
|
fv[3] = xf[3]*f0 + xf[7]*f1 + xf[11]*f2 + xf[15]*f3; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static void load_matrix_(float xf[],const float table[]) |
|
|
|
|
{ |
|
|
|
|
int i; |
|
|
|
|
for(i=0;i<16;i++) xf[i]=table[i]; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
#define ftrv() ftrv_(xf,fv) |
|
|
|
|
#define load_matrix(table) load_matrix_(xf,table) |
|
|
|
|
|
|
|
|
|
#define DEFREG \ |
|
|
|
|
float fv[4],xf[16] |
|
|
|
|
|
|
|
|
|
#define fr0 fv[0] |
|
|
|
|
#define fr1 fv[1] |
|
|
|
|
#define fr2 fv[2] |
|
|
|
|
#define fr3 fv[3] |
|
|
|
|
|
|
|
|
|
#endif |
|
|
|
|
|
|
|
|
|
#if 1 |
|
|
|
|
#define DESCALE(x,n) (x)*(1.0f/(1<<(n))) |
|
|
|
|
#else |
|
|
|
|
#define DESCALE(x,n) (((int)(x)+(1<<(n-1)))>>(n)) |
|
|
|
|
#endif |
|
|
|
|
|
|
|
|
|
/* this code work worse on gcc cvs. 3.2.3 work fine */ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#if 1 |
|
|
|
|
//optimized
|
|
|
|
|
|
|
|
|
|
void idct_sh4(DCTELEM *block) |
|
|
|
@ -252,111 +209,3 @@ void idct_sh4(DCTELEM *block) |
|
|
|
|
|
|
|
|
|
fp_single_leave(fpscr); |
|
|
|
|
} |
|
|
|
|
#else |
|
|
|
|
void idct_sh4(DCTELEM *block) |
|
|
|
|
{ |
|
|
|
|
DEFREG; |
|
|
|
|
|
|
|
|
|
int i; |
|
|
|
|
float tblock[8*8],*fblock; |
|
|
|
|
|
|
|
|
|
/* row */ |
|
|
|
|
|
|
|
|
|
/* even part */ |
|
|
|
|
load_matrix(even_table); |
|
|
|
|
|
|
|
|
|
fblock = tblock; |
|
|
|
|
i = 8; |
|
|
|
|
do { |
|
|
|
|
fr0 = block[0]; |
|
|
|
|
fr1 = block[2]; |
|
|
|
|
fr2 = block[4]; |
|
|
|
|
fr3 = block[6]; |
|
|
|
|
block+=8; |
|
|
|
|
ftrv(); |
|
|
|
|
fblock[0] = fr0; |
|
|
|
|
fblock[2] = fr1; |
|
|
|
|
fblock[4] = fr2; |
|
|
|
|
fblock[6] = fr3; |
|
|
|
|
fblock+=8; |
|
|
|
|
} while(--i); |
|
|
|
|
block-=8*8; |
|
|
|
|
fblock-=8*8; |
|
|
|
|
|
|
|
|
|
load_matrix(odd_table); |
|
|
|
|
|
|
|
|
|
i = 8; |
|
|
|
|
|
|
|
|
|
do { |
|
|
|
|
float t0,t1,t2,t3; |
|
|
|
|
fr0 = block[1]; |
|
|
|
|
fr1 = block[3]; |
|
|
|
|
fr2 = block[5]; |
|
|
|
|
fr3 = block[7]; |
|
|
|
|
block+=8; |
|
|
|
|
ftrv(); |
|
|
|
|
t0 = fblock[0]; |
|
|
|
|
t1 = fblock[2]; |
|
|
|
|
t2 = fblock[4]; |
|
|
|
|
t3 = fblock[6]; |
|
|
|
|
fblock[0] = t0 + fr0; |
|
|
|
|
fblock[7] = t0 - fr0; |
|
|
|
|
fblock[1] = t1 + fr1; |
|
|
|
|
fblock[6] = t1 - fr1; |
|
|
|
|
fblock[2] = t2 + fr2; |
|
|
|
|
fblock[5] = t2 - fr2; |
|
|
|
|
fblock[3] = t3 + fr3; |
|
|
|
|
fblock[4] = t3 - fr3; |
|
|
|
|
fblock+=8; |
|
|
|
|
} while(--i); |
|
|
|
|
block-=8*8; |
|
|
|
|
fblock-=8*8; |
|
|
|
|
|
|
|
|
|
/* col */ |
|
|
|
|
|
|
|
|
|
/* even part */ |
|
|
|
|
load_matrix(even_table); |
|
|
|
|
|
|
|
|
|
i = 8; |
|
|
|
|
|
|
|
|
|
do { |
|
|
|
|
fr0 = fblock[8*0]; |
|
|
|
|
fr1 = fblock[8*2]; |
|
|
|
|
fr2 = fblock[8*4]; |
|
|
|
|
fr3 = fblock[8*6]; |
|
|
|
|
ftrv(); |
|
|
|
|
fblock[8*0] = fr0; |
|
|
|
|
fblock[8*2] = fr1; |
|
|
|
|
fblock[8*4] = fr2; |
|
|
|
|
fblock[8*6] = fr3; |
|
|
|
|
fblock++; |
|
|
|
|
} while(--i); |
|
|
|
|
fblock-=8; |
|
|
|
|
|
|
|
|
|
load_matrix(odd_table); |
|
|
|
|
|
|
|
|
|
i=8; |
|
|
|
|
do { |
|
|
|
|
float t0,t1,t2,t3; |
|
|
|
|
fr0 = fblock[8*1]; |
|
|
|
|
fr1 = fblock[8*3]; |
|
|
|
|
fr2 = fblock[8*5]; |
|
|
|
|
fr3 = fblock[8*7]; |
|
|
|
|
ftrv(); |
|
|
|
|
t0 = fblock[8*0]; |
|
|
|
|
t1 = fblock[8*2]; |
|
|
|
|
t2 = fblock[8*4]; |
|
|
|
|
t3 = fblock[8*6]; |
|
|
|
|
fblock++; |
|
|
|
|
block[8*0] = DESCALE(t0 + fr0,3); |
|
|
|
|
block[8*7] = DESCALE(t0 - fr0,3); |
|
|
|
|
block[8*1] = DESCALE(t1 + fr1,3); |
|
|
|
|
block[8*6] = DESCALE(t1 - fr1,3); |
|
|
|
|
block[8*2] = DESCALE(t2 + fr2,3); |
|
|
|
|
block[8*5] = DESCALE(t2 - fr2,3); |
|
|
|
|
block[8*3] = DESCALE(t3 + fr3,3); |
|
|
|
|
block[8*4] = DESCALE(t3 - fr3,3); |
|
|
|
|
block++; |
|
|
|
|
} while(--i); |
|
|
|
|
} |
|
|
|
|
#endif |
|
|
|
|