--- decoder.c	2004/07/03 08:33:16	1.60
+++ decoder.c	2004/07/26 19:32:28	1.64
@@ -20,7 +20,7 @@
  *  along with this program ; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: decoder.c,v 1.60 2004/07/03 08:33:16 syskin Exp $
+ * $Id: decoder.c,v 1.64 2004/07/26 19:32:28 edgomez Exp $
  *
  ****************************************************************************/
 
@@ -48,6 +48,7 @@
 #include "image/interpolate8x8.h"
 #include "image/reduced.h"
 #include "image/font.h"
+#include "image/qpel.h"
 
 #include "bitstream/mbcoding.h"
 #include "prediction/mbprediction.h"
@@ -61,6 +62,11 @@
 #include "image/postprocessing.h"
 #include "utils/mem_align.h"
 
+#ifdef ARCH_IS_IA32
+#define interpolate16x16_quarterpel new_interpolate16x16_quarterpel
+#define interpolate8x8_quarterpel new_interpolate8x8_quarterpel
+#endif 
+
 static int
 decoder_resize(DECODER * dec)
 {
@@ -390,74 +396,78 @@
 				uint8_t * pY_Cur,
 				uint8_t * pU_Cur,
 				uint8_t * pV_Cur,
-				const int reduced_resolution,
+				int reduced_resolution,
 				const MACROBLOCK * pMB)
 {
-	DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE);
-	DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
+	DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE);
 
 	int stride = dec->edged_width;
 	int next_block = stride * (reduced_resolution ? 16 : 8);
-	const int stride2 = stride/2;
 	int i;
 	const uint32_t iQuant = pMB->quant;
 	const int direction = dec->alternate_vertical_scan ? 2 : 0;
-	const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter;
+	typedef void (*get_inter_block_function_t)(
+			Bitstream * bs,
+			int16_t * block,
+			int direction,
+			const int quant,
+			const uint16_t *matrix);
+	typedef void (*add_residual_function_t)(
+			uint8_t *predicted_block,
+			const int16_t *residual,
+			int stride);
+
+	const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
+		? (get_inter_block_function_t)get_inter_block_h263
+		: (get_inter_block_function_t)get_inter_block_mpeg;
+
+	const add_residual_function_t add_residual = (reduced_resolution)
+		? (add_residual_function_t)add_upsampled_8x8_16to8
+		: (add_residual_function_t)transfer_16to8add;
 
-	for (i = 0; i < 6; i++) {
+	uint8_t *dst[6];
+	int strides[6];
+	
+
+	if (dec->interlacing && pMB->field_dct) {
+		next_block = stride;
+		stride *= 2;
+	}
 
-		if (cbp & (1 << (5 - i))) {	/* coded */
+	reduced_resolution = !!reduced_resolution;
+	dst[0] = pY_Cur;
+	dst[2] = pY_Cur + next_block;
+	dst[1] = dst[0] + (8<<reduced_resolution);
+	dst[3] = dst[2] + (8<<reduced_resolution);
+	dst[4] = pU_Cur;
+	dst[5] = pV_Cur;
+	strides[0] = strides[1] = strides[2] = strides[3] = stride;
+	strides[4] = stride/2;
+	strides[5] = stride/2;
+	
+	for (i = 0; i < 6; i++) {
+		/* Process only coded blocks */
+		if (cbp & (1 << (5 - i))) {
 
-			memset(block, 0, 64 * sizeof(int16_t));	/* clear */
+			/* Clear the block */
+			memset(&data[0], 0, 64*sizeof(int16_t));
 
+			/* Decode coeffs and dequantize on the fly */
 			start_timer();
-			get_inter_block(bs, block, direction);
+			get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
 			stop_coding_timer();
 
+			/* iDCT */
 			start_timer();
-			dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices);
-			stop_iquant_timer();
+			idct(&data[0]);
+			stop_idct_timer();
 
+			/* Add this residual to the predicted block */
 			start_timer();
-			idct(&data[i * 64]);
-			stop_idct_timer();
+			add_residual(dst[i], &data[0], strides[i]);
+			stop_transfer_timer();
 		}
 	}
-
-	if (dec->interlacing && pMB->field_dct) {
-		next_block = stride;
-		stride *= 2;
-	}
-
-	start_timer();
-	if (reduced_resolution) {
-		if (cbp & 32)
-			add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
-		if (cbp & 16)
-			add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
-		if (cbp & 8)
-			add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
-		if (cbp & 4)
-			add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
-		if (cbp & 2)
-			add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
-		if (cbp & 1)
-			add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
-	} else {
-		if (cbp & 32)
-			transfer_16to8add(pY_Cur, &data[0 * 64], stride);
-		if (cbp & 16)
-			transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
-		if (cbp & 8)
-			transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
-		if (cbp & 4)
-			transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
-		if (cbp & 2)
-			transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
-		if (cbp & 1)
-			transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
-	}
-	stop_transfer_timer();
 }
 
 /* decode an inter macroblock */
@@ -1032,7 +1042,6 @@
 	if (!direct) {
 		uv_dx = pMB->mvs[0].x;
 		uv_dy = pMB->mvs[0].y;
-
 		b_uv_dx = pMB->b_mvs[0].x;
 		b_uv_dy = pMB->b_mvs[0].y;
 
@@ -1045,21 +1054,20 @@
 
 		uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
 		uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
-
 		b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
 		b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
 
 	} else {
-		if(dec->quarterpel) {
-			uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
-			uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
-			b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
-			b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2);
-		} else {
-			uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
-			uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
-			b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
-			b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
+		uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
+		uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+		b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
+		b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
+
+		if (dec->quarterpel) {
+			uv_dx /= 2;
+			uv_dy /= 2;
+			b_uv_dx /= 2;
+			b_uv_dy /= 2;
 		}
 
 		uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
@@ -1143,32 +1151,32 @@
 	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
 						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
 						dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
 						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
 						dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
 						dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
 						dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
 						dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
 						dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 1, 8);
+						stride2, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 1, 8);
+						stride2, 0, 8);
 
 	stop_comp_timer();