(commit by michael)

faster simple_idct in C

Originally committed as revision 265 to svn://svn.ffmpeg.org/ffmpeg/trunk

1 files changed, 315 insertions, 125 deletions

diff --git a/libavcodec/simple_idct.c b/libavcodec/simple_idct.c
index 5459b81b35..ef4ebbbad4 100644
--- a/libavcodec/simple_idct.c
+++ b/libavcodec/simple_idct.c
@@ -45,33 +45,50 @@
 #define ROW_SHIFT 11
 #define COL_SHIFT 20 // 6
 #endif
-#if 1
-static void inline idctRow (int16_t * row)
+
+/* 8x8 Matrix used to do a trivial (slow) 8 point IDCT */
+static int coeff[64]={
+	W4, W4, W4, W4, W4, W4, W4, W4,
+	W1, W3, W5, W7,-W7,-W5,-W3,-W1,
+	W2, W6,-W6,-W2,-W2,-W6, W6, W2,
+	W3,-W7,-W1,-W5, W5, W1, W7,-W3,
+	W4,-W4,-W4, W4, W4,-W4,-W4, W4,
+	W5,-W1, W7, W3,-W3,-W7, W1,-W5,
+	W6,-W2, W2,-W6,-W6, W2,-W2, W6,
+	W7,-W5, W3,-W1, W1,-W3, W5,-W7
+};
+
+static int inline idctRowCondZ (int16_t * row)
 {
 	int a0, a1, a2, a3, b0, b1, b2, b3;
-	const int C1 =W1;
-	const int C2 =W2;
-	const int C3 =W3;
-	const int C4 =W4;
-	const int C5 =W5;
-	const int C6 =W6;
-	const int C7 =W7;
-
-	if( !(row[1] | row[2] |row[3] |row[4] |row[5] |row[6] | row[7])) {
-		row[0] = row[1] = row[2] = row[3] = row[4] =
-			row[5] = row[6] = row[7] = row[0]<<3;
-		return;
+
+	if( !( ((uint32_t*)row)[0]|((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3])) {
+/*		row[0] = row[1] = row[2] = row[3] = row[4] =
+			row[5] = row[6] = row[7] = 0;*/
+		return 0;
 	}
 
-	a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + (1<<(ROW_SHIFT-1));
-	a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + (1<<(ROW_SHIFT-1));
-	a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + (1<<(ROW_SHIFT-1));
-	a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + (1<<(ROW_SHIFT-1));
+	if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){
+		a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1));
+		a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1));
+		a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1));
+		a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1));
 
-	b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7];
-	b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7];
-	b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7];
-	b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7];
+		b0 = W1*row[1] + W3*row[3];
+		b1 = W3*row[1] - W7*row[3];
+		b2 = W5*row[1] - W1*row[3];
+		b3 = W7*row[1] - W5*row[3];
+	}else{
+		a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1));
+		a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1));
+		a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1));
+		a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1));
+
+		b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7];
+		b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7];
+		b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7];
+		b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7];
+	}
 
 	row[0] = (a0 + b0) >> ROW_SHIFT;
 	row[1] = (a1 + b1) >> ROW_SHIFT;
@@ -81,151 +98,324 @@ static void inline idctRow (int16_t * row)
 	row[5] = (a2 - b2) >> ROW_SHIFT;
 	row[6] = (a1 - b1) >> ROW_SHIFT;
 	row[7] = (a0 - b0) >> ROW_SHIFT;
+	
+	return 1;
 }
 
-static void inline idctCol (int16_t * col)
+static int inline idctRowCondDC (int16_t * row)
 {
 	int a0, a1, a2, a3, b0, b1, b2, b3;
-	const int C1 =W1;
-	const int C2 =W2;
-	const int C3 =W3;
-	const int C4 =W4;
-	const int C5 =W5;
-	const int C6 =W6;
-	const int C7 =W7;
+
+	if( !( ((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3]| row[1])) {
+//		row[0] = row[1] = row[2] = row[3] = row[4] = row[5] = row[6] = row[7] = row[0]<<3;
+		uint16_t temp= row[0]<<3;
+		((uint32_t*)row)[0]=((uint32_t*)row)[1]=
+		((uint32_t*)row)[2]=((uint32_t*)row)[3]= temp + (temp<<16);
+		return 0;
+	}
+
+	if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){
+		a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1));
+		a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1));
+		a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1));
+		a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1));
+
+		b0 = W1*row[1] + W3*row[3];
+		b1 = W3*row[1] - W7*row[3];
+		b2 = W5*row[1] - W1*row[3];
+		b3 = W7*row[1] - W5*row[3];
+	}else{
+		a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1));
+		a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1));
+		a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1));
+		a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1));
+
+		b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7];
+		b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7];
+		b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7];
+		b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7];
+	}
+
+	row[0] = (a0 + b0) >> ROW_SHIFT;
+	row[7] = (a0 - b0) >> ROW_SHIFT;
+	row[1] = (a1 + b1) >> ROW_SHIFT;
+	row[6] = (a1 - b1) >> ROW_SHIFT;
+	row[2] = (a2 + b2) >> ROW_SHIFT;
+	row[5] = (a2 - b2) >> ROW_SHIFT;
+	row[3] = (a3 + b3) >> ROW_SHIFT;
+	row[4] = (a3 - b3) >> ROW_SHIFT;
+	
+	return 1;
+}
+
+static void inline idctCol (int16_t * col)
+{
+
 /*
 	if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) {
 		col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] =
 			col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3;
 		return;
 	}*/
+
+	int a0, a1, a2, a3, b0, b1, b2, b3;
 	col[0] += (1<<(COL_SHIFT-1))/W4;
-	a0 = C4*col[8*0] + C2*col[8*2] + C4*col[8*4] + C6*col[8*6];
-	a1 = C4*col[8*0] + C6*col[8*2] - C4*col[8*4] - C2*col[8*6];
-	a2 = C4*col[8*0] - C6*col[8*2] - C4*col[8*4] + C2*col[8*6];
-	a3 = C4*col[8*0] - C2*col[8*2] + C4*col[8*4] - C6*col[8*6];
+	a0 = W4*col[8*0] + W2*col[8*2] + W4*col[8*4] + W6*col[8*6];
+	a1 = W4*col[8*0] + W6*col[8*2] - W4*col[8*4] - W2*col[8*6];
+	a2 = W4*col[8*0] - W6*col[8*2] - W4*col[8*4] + W2*col[8*6];
+	a3 = W4*col[8*0] - W2*col[8*2] + W4*col[8*4] - W6*col[8*6];
 
-	b0 = C1*col[8*1] + C3*col[8*3] + C5*col[8*5] + C7*col[8*7];
-	b1 = C3*col[8*1] - C7*col[8*3] - C1*col[8*5] - C5*col[8*7];
-	b2 = C5*col[8*1] - C1*col[8*3] + C7*col[8*5] + C3*col[8*7];
-	b3 = C7*col[8*1] - C5*col[8*3] + C3*col[8*5] - C1*col[8*7];
+	b0 = W1*col[8*1] + W3*col[8*3] + W5*col[8*5] + W7*col[8*7];
+	b1 = W3*col[8*1] - W7*col[8*3] - W1*col[8*5] - W5*col[8*7];
+	b2 = W5*col[8*1] - W1*col[8*3] + W7*col[8*5] + W3*col[8*7];
+	b3 = W7*col[8*1] - W5*col[8*3] + W3*col[8*5] - W1*col[8*7];
 
 	col[8*0] = (a0 + b0) >> COL_SHIFT;
+	col[8*7] = (a0 - b0) >> COL_SHIFT;
 	col[8*1] = (a1 + b1) >> COL_SHIFT;
+	col[8*6] = (a1 - b1) >> COL_SHIFT;
 	col[8*2] = (a2 + b2) >> COL_SHIFT;
+	col[8*5] = (a2 - b2) >> COL_SHIFT;
 	col[8*3] = (a3 + b3) >> COL_SHIFT;
 	col[8*4] = (a3 - b3) >> COL_SHIFT;
-	col[8*5] = (a2 - b2) >> COL_SHIFT;
-	col[8*6] = (a1 - b1) >> COL_SHIFT;
-	col[8*7] = (a0 - b0) >> COL_SHIFT;
 }
 
-void simple_idct (short *block)
+static void inline idctSparseCol (int16_t * col)
 {
-	int i;
-	for(i=0; i<8; i++)
-		idctRow(block + 8*i);
+	int a0, a1, a2, a3, b0, b1, b2, b3;
+	col[0] += (1<<(COL_SHIFT-1))/W4;
+	a0 = W4*col[8*0];
+	a1 = W4*col[8*0];
+	a2 = W4*col[8*0];
+	a3 = W4*col[8*0];
 
-	for(i=0; i<8; i++)
-		idctCol(block + i);
+	if(col[8*2]){
+		a0 +=  + W2*col[8*2];
+		a1 +=  + W6*col[8*2];
+		a2 +=  - W6*col[8*2];
+		a3 +=  - W2*col[8*2];
+	}
 
-}
+	if(col[8*4]){
+		a0 += + W4*col[8*4];
+		a1 += - W4*col[8*4];
+		a2 += - W4*col[8*4];
+		a3 += + W4*col[8*4];
+	}
 
-#else
+	if(col[8*6]){
+		a0 += + W6*col[8*6];
+		a1 += - W2*col[8*6];
+		a2 += + W2*col[8*6];
+		a3 += - W6*col[8*6];
+	}
 
-#define W1  22725  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W2  21407  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W3  19266  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W4  16384  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W5  12873  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W6  8867   //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define W7  4520   //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define COL_SHIFT 31 // 6
+	if(col[8*1]){
+		b0 = W1*col[8*1];
+		b1 = W3*col[8*1];
+		b2 = W5*col[8*1];
+		b3 = W7*col[8*1];
+	}else{
+		b0 = 
+		b1 = 
+		b2 = 
+		b3 = 0;
+	}
 
-static void inline idctRow (int32_t *out, int16_t * row)
-{
-	int a0, a1, a2, a3, b0, b1, b2, b3;
-	const int C1 =W1;
-	const int C2 =W2;
-	const int C3 =W3;
-	const int C4 =W4;
-	const int C5 =W5;
-	const int C6 =W6;
-	const int C7 =W7;
-/*
-	if( !(row[1] | row[2] |row[3] |row[4] |row[5] |row[6] | row[7])) {
-		row[0] = row[1] = row[2] = row[3] = row[4] =
-			row[5] = row[6] = row[7] = row[0]<<14;
-		return;
+	if(col[8*3]){
+		b0 += + W3*col[8*3];
+		b1 += - W7*col[8*3];
+		b2 += - W1*col[8*3];
+		b3 += - W5*col[8*3];
+	}
+
+	if(col[8*5]){
+		b0 += + W5*col[8*5];
+		b1 += - W1*col[8*5];
+		b2 += + W7*col[8*5];
+		b3 += + W3*col[8*5];
+	}
+
+	if(col[8*7]){
+		b0 += + W7*col[8*7];
+		b1 += - W5*col[8*7];
+		b2 += + W3*col[8*7];
+		b3 += - W1*col[8*7];
+	}
+
+	if(!(b0|b1|b2|b3)){
+		col[8*0] = (a0) >> COL_SHIFT;
+		col[8*7] = (a0) >> COL_SHIFT;
+		col[8*1] = (a1) >> COL_SHIFT;
+		col[8*6] = (a1) >> COL_SHIFT;
+		col[8*2] = (a2) >> COL_SHIFT;
+		col[8*5] = (a2) >> COL_SHIFT;
+		col[8*3] = (a3) >> COL_SHIFT;
+		col[8*4] = (a3) >> COL_SHIFT;
+	}else{
+		col[8*0] = (a0 + b0) >> COL_SHIFT;
+		col[8*7] = (a0 - b0) >> COL_SHIFT;
+		col[8*1] = (a1 + b1) >> COL_SHIFT;
+		col[8*6] = (a1 - b1) >> COL_SHIFT;
+		col[8*2] = (a2 + b2) >> COL_SHIFT;
+		col[8*5] = (a2 - b2) >> COL_SHIFT;
+		col[8*3] = (a3 + b3) >> COL_SHIFT;
+		col[8*4] = (a3 - b3) >> COL_SHIFT;
 	}
-*/
-	a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6];
-	a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6];
-	a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6];
-	a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6];
-
-	b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7];
-	b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7];
-	b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7];
-	b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7];
-
-	out[0] = (a0 + b0);
-	out[1] = (a1 + b1);
-	out[2] = (a2 + b2);
-	out[3] = (a3 + b3);
-	out[4] = (a3 - b3);
-	out[5] = (a2 - b2);
-	out[6] = (a1 - b1);
-	out[7] = (a0 - b0);
 }
 
-static void inline idctCol (int32_t *in, int16_t * col)
+static void inline idctSparse2Col (int16_t * col)
 {
-	int64_t a0, a1, a2, a3, b0, b1, b2, b3;
-	const int64_t C1 =W1;
-	const int64_t C2 =W2;
-	const int64_t C3 =W3;
-	const int64_t C4 =W4;
-	const int64_t C5 =W5;
-	const int64_t C6 =W6;
-	const int64_t C7 =W7;
-/*
-	if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) {
-		col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] =
-			col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3;
-		return;
-	}*/
-	in[0] += (1<<(COL_SHIFT-1))/W4;
-	a0 = C4*in[8*0] + C2*in[8*2] + C4*in[8*4] + C6*in[8*6];
-	a1 = C4*in[8*0] + C6*in[8*2] - C4*in[8*4] - C2*in[8*6];
-	a2 = C4*in[8*0] - C6*in[8*2] - C4*in[8*4] + C2*in[8*6];
-	a3 = C4*in[8*0] - C2*in[8*2] + C4*in[8*4] - C6*in[8*6];
+	int a0, a1, a2, a3, b0, b1, b2, b3;
+	col[0] += (1<<(COL_SHIFT-1))/W4;
+	a0 = W4*col[8*0];
+	a1 = W4*col[8*0];
+	a2 = W4*col[8*0];
+	a3 = W4*col[8*0];
+
+	if(col[8*2]){
+		a0 +=  + W2*col[8*2];
+		a1 +=  + W6*col[8*2];
+		a2 +=  - W6*col[8*2];
+		a3 +=  - W2*col[8*2];
+	}
+
+	if(col[8*4]){
+		a0 += + W4*col[8*4];
+		a1 += - W4*col[8*4];
+		a2 += - W4*col[8*4];
+		a3 += + W4*col[8*4];
+	}
+
+	if(col[8*6]){
+		a0 += + W6*col[8*6];
+		a1 += - W2*col[8*6];
+		a2 += + W2*col[8*6];
+		a3 += - W6*col[8*6];
+	}
 
-	b0 = C1*in[8*1] + C3*in[8*3] + C5*in[8*5] + C7*in[8*7];
-	b1 = C3*in[8*1] - C7*in[8*3] - C1*in[8*5] - C5*in[8*7];
-	b2 = C5*in[8*1] - C1*in[8*3] + C7*in[8*5] + C3*in[8*7];
-	b3 = C7*in[8*1] - C5*in[8*3] + C3*in[8*5] - C1*in[8*7];
+	if(col[8*1] || 1){
+		b0 = W1*col[8*1];
+		b1 = W3*col[8*1];
+		b2 = W5*col[8*1];
+		b3 = W7*col[8*1];
+	}else{
+		b0 = 
+		b1 = 
+		b2 = 
+		b3 = 0;
+	}
+
+	if(col[8*3]){
+		b0 += + W3*col[8*3];
+		b1 += - W7*col[8*3];
+		b2 += - W1*col[8*3];
+		b3 += - W5*col[8*3];
+	}
+
+	if(col[8*5]){
+		b0 += + W5*col[8*5];
+		b1 += - W1*col[8*5];
+		b2 += + W7*col[8*5];
+		b3 += + W3*col[8*5];
+	}
+
+	if(col[8*7]){
+		b0 += + W7*col[8*7];
+		b1 += - W5*col[8*7];
+		b2 += + W3*col[8*7];
+		b3 += - W1*col[8*7];
+	}
 
 	col[8*0] = (a0 + b0) >> COL_SHIFT;
+	col[8*7] = (a0 - b0) >> COL_SHIFT;
 	col[8*1] = (a1 + b1) >> COL_SHIFT;
+	col[8*6] = (a1 - b1) >> COL_SHIFT;
 	col[8*2] = (a2 + b2) >> COL_SHIFT;
+	col[8*5] = (a2 - b2) >> COL_SHIFT;
 	col[8*3] = (a3 + b3) >> COL_SHIFT;
 	col[8*4] = (a3 - b3) >> COL_SHIFT;
-	col[8*5] = (a2 - b2) >> COL_SHIFT;
-	col[8*6] = (a1 - b1) >> COL_SHIFT;
-	col[8*7] = (a0 - b0) >> COL_SHIFT;
 }
 
+
 void simple_idct (short *block)
 {
+
 	int i;
-	int32_t temp[64];
+	
+#if 0
+	int nonZero[8];
+	int buffer[64];
+	int nNonZero=0;
+	
+	idctRowCondDC(block);
+	
+	for(i=1; i<8; i++)
+	{
+		nonZero[nNonZero]=i;
+		nNonZero+= idctRowCondZ(block + i*8);
+	}
+	
+	if(nNonZero==0)
+	{
+		for(i=0; i<8; i++)
+		{
+			block[i   ]=
+			block[i+8 ]=
+			block[i+16]=
+			block[i+24]=
+			block[i+32]=
+			block[i+40]=
+			block[i+48]=
+			block[i+56]= (W4*block[i] + (1<<(COL_SHIFT-1))) >> COL_SHIFT;
+		}	
+	}
+	else if(nNonZero==1)
+	{
+		int index= nonZero[0]*8;
+		for(i=0; i<8; i++)
+		{
+			int bias= W4*block[i] + (1<<(COL_SHIFT-1));
+			int c= block[i + index];
+			block[i   ]= (c*coeff[index  ] + bias) >> COL_SHIFT;
+			block[i+8 ]= (c*coeff[index+1] + bias) >> COL_SHIFT;
+			block[i+16]= (c*coeff[index+2] + bias) >> COL_SHIFT;
+			block[i+24]= (c*coeff[index+3] + bias) >> COL_SHIFT;
+			block[i+32]= (c*coeff[index+4] + bias) >> COL_SHIFT;
+			block[i+40]= (c*coeff[index+5] + bias) >> COL_SHIFT;
+			block[i+48]= (c*coeff[index+6] + bias) >> COL_SHIFT;
+			block[i+56]= (c*coeff[index+7] + bias) >> COL_SHIFT;
+		}	
+	}
+/*	else if(nNonZero==2)
+	{
+		int index1= nonZero[0]*8;
+		int index2= nonZero[1]*8;
+		for(i=0; i<8; i++)
+		{
+			int bias= W4*block[i] + (1<<(COL_SHIFT-1));
+			int c1= block[i + index1];
+			int c2= block[i + index2];
+			block[i   ]= (c1*coeff[index1  ] + c2*coeff[index2  ] + bias) >> COL_SHIFT;
+			block[i+8 ]= (c1*coeff[index1+1] + c2*coeff[index2+1] + bias) >> COL_SHIFT;
+			block[i+16]= (c1*coeff[index1+2] + c2*coeff[index2+2] + bias) >> COL_SHIFT;
+			block[i+24]= (c1*coeff[index1+3] + c2*coeff[index2+3] + bias) >> COL_SHIFT;
+			block[i+32]= (c1*coeff[index1+4] + c2*coeff[index2+4] + bias) >> COL_SHIFT;
+			block[i+40]= (c1*coeff[index1+5] + c2*coeff[index2+5] + bias) >> COL_SHIFT;
+			block[i+48]= (c1*coeff[index1+6] + c2*coeff[index2+6] + bias) >> COL_SHIFT;
+			block[i+56]= (c1*coeff[index1+7] + c2*coeff[index2+7] + bias) >> COL_SHIFT;
+		}	
+	}*/
+	else
+	{
+		for(i=0; i<8; i++)
+			idctSparse2Col(block + i);
+	}
+#else		
 	for(i=0; i<8; i++)
-		idctRow(temp+8*i, block + 8*i);
-
+		idctRowCondDC(block + i*8);
+	
 	for(i=0; i<8; i++)
-		idctCol(temp+i, block + i);
-
-}
-
+		idctSparseCol(block + i);
 #endif
+}