Optimize JPEG stack usage

This commit is contained in:
iabdalkader 2016-02-18 17:55:45 +02:00
parent c771793ac9
commit 04a2e89fd5

View File

@ -215,16 +215,17 @@ static void jpeg_calcBits(int val, uint16_t bits[2]) {
bits[0] = val & ((1<<bits[1])-1);
}
static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, const uint16_t (*HTDC)[2], const uint16_t (*HTAC)[2])
static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int8_t *CDU, int *fdtbl, int DC, const uint16_t (*HTDC)[2], const uint16_t (*HTAC)[2])
{
int DU[64], DUQ[64];
int t0, t1, t2, t3, t4, t5, t6, t7, t10, t11, t12, t13;
const uint16_t EOB[2] = { HTAC[0x00][0], HTAC[0x00][1] };
const uint16_t M16zeroes[2] = { HTAC[0xF0][0], HTAC[0xF0][1] };
int t0, t1, t2, t3, t4, t5, t6, t7, t10, t11, t12, t13;
// BinDCT-a1
// DCT rows
for (int i=8, *p=CDU; i>0; i--, p+=8) {
int8_t *p=CDU;
for (int i=8, *du=DU; i>0; i--, p+=8, du+=8) {
t0 = p[0] + p[7];
t1 = p[1] + p[6];
t2 = p[2] + p[5];
@ -241,14 +242,14 @@ static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, co
t11 = t1 + t2 ;
t12 = t1 - t2;
p[0] = (t10 + t11); /* phase 3 */
p[4] = ((p[0] ) >> 1) - t11; /* Jie 05/18/00 */
du[0] = (t10 + t11); /* phase 3 */
du[4] = ((du[0] ) >> 1) - t11; /* jie 05/18/00 */
/*1/2, -1/2: alter the sign to get positive scaling factor */
p[6] = (( t13 ) >> 1) - t12;
p[2] = t13 - ((p[6] ) >> 1);
du[6] = (( t13 ) >> 1) - t12;
du[2] = t13 - ((du[6] ) >> 1);
/* Odd part */
/* odd part */
/* pi/4 = -1/2u 3/4d -1/2u*/
t10 = t5 - (( t6 ) >> 1);
t6 = t6 + t10 - ((t10 ) >> 2);
@ -260,27 +261,26 @@ static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, co
t13 = t7 + t6;
/* 7pi/16 = 1/4u -1/4d: alter the sign to get positive scaling factor */
p[7] = ((t13 ) >> 2) - t10;
p[1] = t13 - ((p[7] ) >> 2);
du[7] = ((t13 ) >> 2) - t10;
du[1] = t13 - ((du[7] ) >> 2);
/* 3pi/16 = */
/* new version: 1, -1/2 */
p[5] = t11 + t12;
p[3] = t12 - ((p[5] ) >> 1);
du[5] = t11 + t12;
du[3] = t12 - ((du[5] ) >> 1);
}
// DCT columns
for (int i=8, *p=CDU; i>0; i--, p++) {
t0 = (p[8*0] + p[8*7]);
t1 = (p[8*1] + p[8*6]);
t2 = (p[8*2] + p[8*5]);
t3 = (p[8*3] + p[8*4]);
for (int i=8, *du=DU; i>0; i--, du++) {
t0 = (du[8*0] + du[8*7]);
t1 = (du[8*1] + du[8*6]);
t2 = (du[8*2] + du[8*5]);
t3 = (du[8*3] + du[8*4]);
t7 = (p[8*0] - p[8*7]);
t6 = (p[8*1] - p[8*6]);
t5 = (p[8*2] - p[8*5]);
t4 = (p[8*3] - p[8*4]);
t7 = (du[8*0] - du[8*7]);
t6 = (du[8*1] - du[8*6]);
t5 = (du[8*2] - du[8*5]);
t4 = (du[8*3] - du[8*4]);
/* Even part */
t10 = t0 + t3; /* phase 2 */
@ -288,12 +288,12 @@ static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, co
t11 = t1 + t2;
t12 = t1 - t2;
p[8*0] = (t10 + t11); /* phase 3 */
p[8*4] = ((p[8*0] ) >> 1) - t11; /* Jie 05/18/00 */
du[8*0] = (t10 + t11); /* phase 3 */
du[8*4] = ((du[8*0] ) >> 1) - t11; /* Jie 05/18/00 */
// 1/2, 1/2
p[8*6] = ((t13 ) >> 1) - t12;
p[8*2] = t13 - ((p[8*6] ) >> 1);
du[8*6] = ((t13 ) >> 1) - t12;
du[8*2] = t13 - ((du[8*6] ) >> 1);
/* Odd part */
/* pi/4 = -1/2u 3/4d -1/2u*/
@ -307,29 +307,27 @@ static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, co
t13 = t7 + t6;
/* 7pi/16 = 1/4u -1/4d: alter sign to get positive scaling factor */
p[8*7] = ((t13 ) >> 2) - t10;
p[8*1] = t13 - ((p[8*7] ) >> 2);
du[8*7] = ((t13 ) >> 2) - t10;
du[8*1] = t13 - ((du[8*7] ) >> 2);
/* 3pi/16 = */
/* new : 1 and -1/2 */
p[8*5] = t11 + t12 ;
p[8*3] = t12 - ((p[8*5] ) >> 1);
du[8*5] = t11 + t12 ;
du[8*3] = t12 - ((du[8*5] ) >> 1);
}
// first non-zero element in reverse order
int end0pos = 0;
// Quantize/descale/zigzag the coefficients
int DU[64];
for(int i=0; i<64; ++i) {
DU[s_jpeg_ZigZag[i]] = (CDU[i]*fdtbl[i])/4096;
if (s_jpeg_ZigZag[i] > end0pos && DU[s_jpeg_ZigZag[i]]) {
DUQ[s_jpeg_ZigZag[i]] = (DU[i]*fdtbl[i])/4096;
if (s_jpeg_ZigZag[i] > end0pos && DUQ[s_jpeg_ZigZag[i]]) {
end0pos = s_jpeg_ZigZag[i];
}
}
// Encode DC
int diff = DU[0] - DC;
int diff = DUQ[0] - DC;
if (diff == 0) {
jpeg_writeBits(jpeg_buf, HTDC[0]);
} else {
@ -342,12 +340,12 @@ static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, co
// Encode ACs
if(end0pos == 0) {
jpeg_writeBits(jpeg_buf, EOB);
return DU[0];
return DUQ[0];
}
for(int i = 1; i <= end0pos; ++i) {
int startpos = i;
for (; DU[i]==0 && i<=end0pos ; ++i) {
for (; DUQ[i]==0 && i<=end0pos ; ++i) {
}
int nrzeroes = i-startpos;
if ( nrzeroes >= 16 ) {
@ -357,14 +355,14 @@ static int jpeg_processDU(jpeg_buf_t *jpeg_buf, int *CDU, int *fdtbl, int DC, co
nrzeroes &= 15;
}
uint16_t bits[2];
jpeg_calcBits(DU[i], bits);
jpeg_calcBits(DUQ[i], bits);
jpeg_writeBits(jpeg_buf, HTAC[(nrzeroes<<4)+bits[1]]);
jpeg_writeBits(jpeg_buf, bits);
}
if(end0pos != 63) {
jpeg_writeBits(jpeg_buf, EOB);
}
return DU[0];
return DUQ[0];
}
static void jpeg_init(int quality)
@ -523,7 +521,7 @@ void jpeg_compress(image_t *src, image_t *dst, int quality)
// Encode 8x8 macroblocks
if (src->bpp == 1) {
int YDU[64];
int8_t YDU[64];
uint8_t *pixels = (uint8_t *)src->pixels;
// Copy 8x8 MCUs
@ -544,7 +542,7 @@ void jpeg_compress(image_t *src, image_t *dst, int quality)
}
}
} else if (src->bpp == 2) {// TODO assuming RGB565
int YDU[256], UDU[64], VDU[64];
int8_t YDU[256], UDU[64], VDU[64];
uint16_t *pixels = (uint16_t *)src->pixels;
for (int y=0; y<src->h; y+=16) {
@ -627,7 +625,6 @@ void jpeg_compress(image_t *src, image_t *dst, int quality)
jpeg_put_char(&jpeg_buf, 0xFF);
jpeg_put_char(&jpeg_buf, 0xD9);
dst->bpp = jpeg_buf.idx;
dst->data = jpeg_buf.buf;