libmpeg2-0.2.0 merge

git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@37 b3059339-0415-0410-9bf9-f77b7e298cf2

16 files changed, 791 insertions, 852 deletions

diff --git a/libmpeg2/Makefile b/libmpeg2/Makefile
index f599aecfd2..ccc2e2595a 100644
--- a/libmpeg2/Makefile
+++ b/libmpeg2/Makefile
@@ -3,8 +3,8 @@ LIBNAME = libmpeg2.a
 
 include ../config.mak
 
-SRCS	= decode.c header.c idct.c idct_mmx.c motion_comp.c motion_comp_mmx.c slice.c stats.c
-OBJS	= decode.o header.o idct.o idct_mmx.o motion_comp.o motion_comp_mmx.o slice.o stats.o
+SRCS	= header.c idct.c idct_mmx.c motion_comp.c motion_comp_mmx.c slice.c stats.c decode.c
+OBJS	= header.o idct.o idct_mmx.o motion_comp.o motion_comp_mmx.o slice.o stats.o decode.o
 INCLUDE = -I. -I../libvo -I..
 CFLAGS  = $(OPTFLAGS) $(INCLUDE) -DMPG12PLAY
 
diff --git a/libmpeg2/attributes.h b/libmpeg2/attributes.h
index dfbf129411..3e14cab271 100644
--- a/libmpeg2/attributes.h
+++ b/libmpeg2/attributes.h
@@ -1,6 +1,6 @@
 /*
  * attributes.h
- * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
+ * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -19,11 +19,7 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
-//use gcc attribs to align critical data structures
-
-/* maximum supported data alignment */
-#define ATTRIBUTE_ALIGNED_MAX 64
-
+/* use gcc attribs to align critical data structures */
 #ifdef ATTRIBUTE_ALIGNED_MAX
 #define ATTR_ALIGN(align) __attribute__ ((__aligned__ ((ATTRIBUTE_ALIGNED_MAX < align) ? ATTRIBUTE_ALIGNED_MAX : align)))
 #else
diff --git a/libmpeg2/decode.c b/libmpeg2/decode.c
index e8bbb02112..9426f4ff59 100644
--- a/libmpeg2/decode.c
+++ b/libmpeg2/decode.c
@@ -3,8 +3,7 @@
 
 /* mpeg2dec version: */
 #define PACKAGE "mpeg2dec"
-//#define VERSION "0.1.7-cvs"
-#define VERSION "0.1.8-cvs"
+#define VERSION "0.2.0-release"
 
 #include <stdio.h>
 #include <stdlib.h>
@@ -14,7 +13,8 @@
 
 #include "config.h"
 
-//#include "video_out.h"
+#include "video_out.h"
+#include <inttypes.h>
 
 #include "mpeg2.h"
 #include "mpeg2_internal.h"
@@ -32,6 +32,9 @@
 #include "mmx.h"
 #endif
 
+#include "mm_accel.h"
+
+
 //this is where we keep the state of the decoder
 //picture_t picture_data;
 //picture_t *picture=&picture_data;
@@ -48,7 +51,9 @@ mpeg2_config_t config;
 static int drop_flag = 0;
 static int drop_frame = 0;
 
+#ifdef POSTPROC
 int quant_store[MBR+1][MBC+1]; // [Review]
+#endif
 
 void mpeg2_init (void)
 {
@@ -73,7 +78,7 @@ void mpeg2_init (void)
     picture=shmem_alloc(sizeof(picture_t)); // !!! NEW HACK :) !!!
 
     header_state_init (picture);
-    picture->repeat_count=0;
+//    picture->repeat_count=0;
     
     picture->pp_options=0;
 
@@ -81,10 +86,13 @@ void mpeg2_init (void)
     motion_comp_init ();
 }
 
+static vo_frame_t frames[3];
+
 void mpeg2_allocate_image_buffers (picture_t * picture)
 {
 	int frame_size,buff_size;
         unsigned char *base=NULL;
+	int i;
 
         // height+1 requires for yuv2rgb_mmx code (it reads next line after last)
 	frame_size = picture->coded_picture_width * (1+picture->coded_picture_height);
@@ -92,56 +100,44 @@ void mpeg2_allocate_image_buffers (picture_t * picture)
         buff_size = frame_size + (frame_size/4)*2; // 4Y + 1U + 1V
 
 	// allocate images in YV12 format
-        base = shmem_alloc(buff_size);
-	picture->throwaway_frame[0] = base;
-	picture->throwaway_frame[1] = base + frame_size * 5 / 4;
-	picture->throwaway_frame[2] = base + frame_size;
-
-        base = shmem_alloc(buff_size);
-	picture->backward_reference_frame[0] = base;
-	picture->backward_reference_frame[1] = base + frame_size * 5 / 4;
-	picture->backward_reference_frame[2] = base + frame_size;
-
-        base = shmem_alloc(buff_size);
-	picture->forward_reference_frame[0] = base;
-	picture->forward_reference_frame[1] = base + frame_size * 5 / 4;
-	picture->forward_reference_frame[2] = base + frame_size;
+	for(i=0;i<3;i++){
+            base = shmem_alloc(buff_size);
+	    frames[i].base[0] = base;
+	    frames[i].base[1] = base + frame_size * 5 / 4;
+	    frames[i].base[2] = base + frame_size;
+	    frames[i].copy = NULL;
+	    frames[i].vo = NULL;
+	    frames[i].slice=0;
+	}
+	
+	picture->forward_reference_frame=&frames[0];
+	picture->backward_reference_frame=&frames[1];
+	picture->current_frame=&frames[2];
 
+#ifdef POSTPROC
         base = shmem_alloc(buff_size);
 	picture->pp_frame[0] = base;
 	picture->pp_frame[1] = base + frame_size * 5 / 4;
 	picture->pp_frame[2] = base + frame_size;
+#endif
 
 }
 
-static void decode_reorder_frames (void)
-{
-    if (picture->picture_coding_type != B_TYPE) {
-
-	//reuse the soon to be outdated forward reference frame
-	picture->current_frame[0] = picture->forward_reference_frame[0];
-	picture->current_frame[1] = picture->forward_reference_frame[1];
-	picture->current_frame[2] = picture->forward_reference_frame[2];
+static void copy_slice (vo_frame_t * frame, uint8_t ** src){
+    vo_functions_t * output = frame->vo;
+    int stride[3];
+    int y=frame->slice*16;
 
-	//make the backward reference frame the new forward reference frame
-	picture->forward_reference_frame[0] =
-	    picture->backward_reference_frame[0];
-	picture->forward_reference_frame[1] =
-	    picture->backward_reference_frame[1];
-	picture->forward_reference_frame[2] =
-	    picture->backward_reference_frame[2];
+    stride[0]=picture->coded_picture_width;
+    stride[1]=stride[2]=stride[0]/2;
 
-	picture->backward_reference_frame[0] = picture->current_frame[0];
-	picture->backward_reference_frame[1] = picture->current_frame[1];
-	picture->backward_reference_frame[2] = picture->current_frame[2];
+    output->draw_slice (src, stride, 
+                picture->display_picture_width,
+		(y+16<=picture->display_picture_height) ? 16 :
+		    picture->display_picture_height-y,
+		0, y);
 
-    } else {
-
-	picture->current_frame[0] = picture->throwaway_frame[0];
-	picture->current_frame[1] = picture->throwaway_frame[1];
-	picture->current_frame[2] = picture->throwaway_frame[2];
-
-    }
+    ++frame->slice;
 }
 
 static int in_slice_flag=0;
@@ -156,43 +152,27 @@ static int parse_chunk (vo_functions_t * output, int code, uint8_t * buffer)
     if (is_frame_done) {
 	in_slice_flag = 0;
         
-        if(picture->picture_structure != FRAME_PICTURE) printf("Field! %d  \n",picture->second_field);
+//        if(picture->picture_structure != FRAME_PICTURE) printf("Field! %d  \n",picture->second_field);
         
-	    if ( ((HACK_MODE == 2) || (picture->mpeg1))
-                && ((picture->picture_structure == FRAME_PICTURE) ||
+	if (((picture->picture_structure == FRAME_PICTURE) ||
 		 (picture->second_field))
-            ) {
-	        uint8_t ** bar;
-                int stride[3];
-
-		if (picture->picture_coding_type == B_TYPE)
-		    bar = picture->throwaway_frame;
-		else
-		    bar = picture->forward_reference_frame;
-                
-                stride[0]=picture->coded_picture_width;
-                stride[1]=stride[2]=stride[0]/2;
-
-                if(picture->pp_options){
-                    // apply OpenDivX postprocess filter
-                    postprocess(bar, stride[0],
-                        picture->pp_frame, stride[0],
-                        picture->coded_picture_width, picture->coded_picture_height, 
-                        &quant_store[1][1], (MBC+1), picture->pp_options);
-		    output->draw_slice (picture->pp_frame, stride, 
-                        picture->display_picture_width,
-                        picture->display_picture_height, 0, 0);
-                } else {
-		    output->draw_slice (bar, stride, 
+           ) {
+#if 1
+		if (picture->picture_coding_type != B_TYPE) {
+            	    int stride[3];
+            	    stride[0]=picture->coded_picture_width;
+            	    stride[1]=stride[2]=stride[0]/2;
+		    output->draw_slice (picture->forward_reference_frame->base,
+			stride, 
                         picture->display_picture_width,
                         picture->display_picture_height, 0, 0);
                 }
-                
-	    }
+#endif
+	}
 #ifdef ARCH_X86
-	    if (config.flags & MM_ACCEL_X86_MMX) emms ();
+	if (config.flags & MM_ACCEL_X86_MMX) emms();
 #endif
-	    output->flip_page ();
+	output->flip_page();
     }
 
     switch (code) {
@@ -227,40 +207,32 @@ static int parse_chunk (vo_functions_t * output, int code, uint8_t * buffer)
 	if (!(in_slice_flag)) {
 	    in_slice_flag = 1;
 
-	    if(!(picture->second_field)) decode_reorder_frames ();
+//	    if(!(picture->second_field)) decode_reorder_frames ();
+
+	    // set current_frame pointer:
+	    if (picture->second_field){
+//		vo_field (picture->current_frame, picture->picture_structure);
+	    } else {
+		if (picture->picture_coding_type == B_TYPE){
+		    picture->current_frame = &frames[2];
+		    picture->current_frame->copy=copy_slice;
+		} else {
+		    picture->current_frame = picture->forward_reference_frame;
+		    picture->forward_reference_frame = picture->backward_reference_frame;
+		    picture->backward_reference_frame = picture->current_frame;
+		    picture->current_frame->copy=NULL;
+		}
+	    }
+	    
+	    picture->current_frame->vo=output;
+	    picture->current_frame->slice=0;
+
 	}
 
 	if (!drop_frame) {
-	    uint8_t ** bar;
 
 	    slice_process (picture, code, buffer);
 
-	    if ((HACK_MODE < 2) && (!(picture->mpeg1))) {
-		uint8_t * foo[3];
-	        uint8_t ** bar;
-		//frame_t * bar;
-                int stride[3];
-		int offset;
-
-		if (picture->picture_coding_type == B_TYPE)
-		    bar = picture->throwaway_frame;
-		else
-		    bar = picture->forward_reference_frame;
-
-		offset = (code-1) * 4 * picture->coded_picture_width;
-		if ((! HACK_MODE) && (picture->picture_coding_type == B_TYPE))
-		    offset = 0;
-
-		foo[0] = bar[0] + 4 * offset;
-		foo[1] = bar[1] + offset;
-		foo[2] = bar[2] + offset;
-                
-                stride[0]=picture->coded_picture_width;
-                stride[1]=stride[2]=stride[0]/2;
-
-		output->draw_slice (foo, stride, 
-                    picture->display_picture_width, 16, 0, (code-1)*16);
-	    }
 #ifdef ARCH_X86
 	    if (config.flags & MM_ACCEL_X86_MMX) emms ();
 #endif
diff --git a/libmpeg2/header.c b/libmpeg2/header.c
index 8f5b34359e..852ff541bd 100644
--- a/libmpeg2/header.c
+++ b/libmpeg2/header.c
@@ -1,6 +1,6 @@
 /*
  * slice.c
- * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
+ * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -26,7 +26,7 @@
 #include "mpeg2_internal.h"
 #include "attributes.h"
 
-// default intra quant matrix, in zig-zag order
+/* default intra quant matrix, in zig-zag order */
 static uint8_t default_intra_quantizer_matrix[64] ATTR_ALIGN(16) = {
     8,
     16, 16,
@@ -47,7 +47,7 @@ static uint8_t default_intra_quantizer_matrix[64] ATTR_ALIGN(16) = {
 
 uint8_t scan_norm[64] ATTR_ALIGN(16) =
 {
-    // Zig-Zag scan pattern
+    /* Zig-Zag scan pattern */
      0, 1, 8,16, 9, 2, 3,10,
     17,24,32,25,18,11, 4, 5,
     12,19,26,33,40,48,41,34,
@@ -60,7 +60,7 @@ uint8_t scan_norm[64] ATTR_ALIGN(16) =
 
 uint8_t scan_alt[64] ATTR_ALIGN(16) =
 {
-    // Alternate scan pattern
+    /* Alternate scan pattern */
     0,8,16,24,1,9,2,10,17,25,32,40,48,56,57,49,
     41,33,26,18,3,11,4,12,19,27,34,42,50,58,35,43,
     51,59,20,28,5,13,6,14,21,29,36,44,52,60,37,45,
@@ -69,47 +69,34 @@ uint8_t scan_alt[64] ATTR_ALIGN(16) =
 
 void header_state_init (picture_t * picture)
 {
-    //FIXME we should set pointers to the real scan matrices here (mmx vs
-    //normal) instead of the ifdefs in header_process_picture_coding_extension
-
     picture->scan = scan_norm;
 }
 
-static const int frameratecode2framerate[16] = {
-   0, 24000*10000/1001, 24*10000,25*10000, 30000*10000/1001, 30*10000,50*10000,60000*10000/1001,
-  60*10000, 0,0,0,0,0,0,0
-};
-
 int header_process_sequence_header (picture_t * picture, uint8_t * buffer)
 {
-    unsigned int h_size;
-    unsigned int v_size;
+    int width, height;
     int i;
 
     if ((buffer[6] & 0x20) != 0x20)
-	return 1;	// missing marker_bit
+	return 1;	/* missing marker_bit */
 
-    v_size = (buffer[0] << 16) | (buffer[1] << 8) | buffer[2];
+    height = (buffer[0] << 16) | (buffer[1] << 8) | buffer[2];
 
-    picture->display_picture_width = (v_size >> 12);
-    picture->display_picture_height = (v_size & 0xfff);
+    picture->display_picture_width = (height >> 12);
+    picture->display_picture_height = (height & 0xfff);
 
-    h_size = ((v_size >> 12) + 15) & ~15;
-    v_size = ((v_size & 0xfff) + 15) & ~15;
+    width = ((height >> 12) + 15) & ~15;
+    height = ((height & 0xfff) + 15) & ~15;
 
-    if ((h_size > 768) || (v_size > 576))
-	return 1;	// size restrictions for MP@ML or MPEG1
+    if ((width > 768) || (height > 576))
+	return 1;	/* size restrictions for MP@ML or MPEG1 */
 
-    //XXX this needs field fixups
-    picture->coded_picture_width = h_size;
-    picture->coded_picture_height = v_size;
-    picture->last_mba = ((h_size * v_size) >> 8) - 1;
+    picture->coded_picture_width = width;
+    picture->coded_picture_height = height;
 
-    // this is not used by the decoder
+    /* this is not used by the decoder */
     picture->aspect_ratio_information = buffer[3] >> 4;
     picture->frame_rate_code = buffer[3] & 15;
-    picture->frame_rate = frameratecode2framerate[picture->frame_rate_code];
-
     picture->bitrate = (buffer[4]<<10)|(buffer[5]<<2)|(buffer[6]>>6);
 
     if (buffer[7] & 2) {
@@ -132,15 +119,15 @@ int header_process_sequence_header (picture_t * picture, uint8_t * buffer)
 	    picture->non_intra_quantizer_matrix[i] = 16;
     }
 
-    // MPEG1 - for testing only
+    /* MPEG1 - for testing only */
     picture->mpeg1 = 1;
     picture->intra_dc_precision = 0;
     picture->frame_pred_frame_dct = 1;
     picture->q_scale_type = 0;
     picture->concealment_motion_vectors = 0;
-    //picture->alternate_scan = 0;
+    /* picture->alternate_scan = 0; */
     picture->picture_structure = FRAME_PICTURE;
-    //picture->second_field = 0;
+    /* picture->second_field = 0; */
 
     return 0;
 }
@@ -148,28 +135,20 @@ int header_process_sequence_header (picture_t * picture, uint8_t * buffer)
 static int header_process_sequence_extension (picture_t * picture,
 					      uint8_t * buffer)
 {
-    // MPEG1 - for testing only
-    picture->mpeg1 = 0;
+    /* check chroma format, size extensions, marker bit */
+    if (((buffer[1] & 0x07) != 0x02) || (buffer[2] & 0xe0) ||
+	((buffer[3] & 0x01) != 0x01))
+	return 1;
 
-    // check chroma format, size extensions, marker bit
-    if(((buffer[1]>>1)&3)!=1){
-      printf("This CHROMA format not yet supported :(\n");
-      return 1;
-    }
-    if ((buffer[1] & 1) || (buffer[2] & 0xe0)){
-      printf("Big resolution video not yet supported :(\n");
-      return 1;
-    }
-    if((buffer[3] & 0x01) != 0x01) return 1; // marker bit
-
-
-    // this is not used by the decoder
+    /* this is not used by the decoder */
     picture->progressive_sequence = (buffer[1] >> 3) & 1;
 
     if (picture->progressive_sequence)
 	picture->coded_picture_height =
 	    (picture->coded_picture_height + 31) & ~31;
-    picture->bitrate>>=1; // hack
+
+    /* MPEG1 - for testing only */
+    picture->mpeg1 = 0;
 
     return 0;
 }
@@ -197,7 +176,7 @@ static int header_process_quant_matrix_extension (picture_t * picture,
 
 static int header_process_picture_coding_extension (picture_t * picture, uint8_t * buffer)
 {
-    //pre subtract 1 for use later in compute_motion_vector
+    /* pre subtract 1 for use later in compute_motion_vector */
     picture->f_code[0][0] = (buffer[0] & 15) - 1;
     picture->f_code[0][1] = (buffer[1] >> 4) - 1;
     picture->f_code[1][0] = (buffer[1] & 15) - 1;
@@ -210,12 +189,12 @@ static int header_process_picture_coding_extension (picture_t * picture, uint8_t
     picture->q_scale_type = (buffer[3] >> 4) & 1;
     picture->intra_vlc_format = (buffer[3] >> 3) & 1;
 
-    if (buffer[3] & 4)	// alternate_scan
+    if (buffer[3] & 4)	/* alternate_scan */
 	picture->scan = scan_alt;
     else
 	picture->scan = scan_norm;
 
-    // these are not used by the decoder
+    /* these are not used by the decoder */
     picture->top_field_first = buffer[3] >> 7;
     picture->repeat_first_field = (buffer[3] >> 1) & 1;
     picture->progressive_frame = buffer[4] >> 7;
@@ -240,13 +219,13 @@ static int header_process_picture_coding_extension (picture_t * picture, uint8_t
 int header_process_extension (picture_t * picture, uint8_t * buffer)
 {
     switch (buffer[0] & 0xf0) {
-    case 0x10:	// sequence extension
+    case 0x10:	/* sequence extension */
 	return header_process_sequence_extension (picture, buffer);
 
-    case 0x30:	// quant matrix extension
+    case 0x30:	/* quant matrix extension */
 	return header_process_quant_matrix_extension (picture, buffer);
 
-    case 0x80:	// picture coding extension
+    case 0x80:	/* picture coding extension */
 	return header_process_picture_coding_extension (picture, buffer);
     }
 
@@ -257,14 +236,14 @@ int header_process_picture_header (picture_t *picture, uint8_t * buffer)
 {
     picture->picture_coding_type = (buffer [1] >> 3) & 7;
 
-    // forward_f_code and backward_f_code - used in mpeg1 only
+    /* forward_f_code and backward_f_code - used in mpeg1 only */
     picture->f_code[0][1] = (buffer[3] >> 2) & 1;
     picture->f_code[0][0] =
 	(((buffer[3] << 1) | (buffer[4] >> 7)) & 7) - 1;
     picture->f_code[1][1] = (buffer[4] >> 6) & 1;
     picture->f_code[1][0] = ((buffer[4] >> 3) & 7) - 1;
 
-    // move in header_process_picture_header
+    /* move in header_process_picture_header */
         picture->second_field =
             (picture->picture_structure != FRAME_PICTURE) &&
             !(picture->second_field);
diff --git a/libmpeg2/idct.c b/libmpeg2/idct.c
index 7411e176dd..d78ad49e8d 100644
--- a/libmpeg2/idct.c
+++ b/libmpeg2/idct.c
@@ -1,6 +1,6 @@
 /*
  * idct.c
- * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
+ * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * Portions of this code are from the MPEG software simulation group
  * idct implementation. This code will be replaced with a new
@@ -52,8 +52,7 @@
 #define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */
 #define W7 565  /* 2048*sqrt (2)*cos (7*pi/16) */
 
-
-// idct main entry point 
+/* idct main entry point  */
 void (*idct_block_copy) (int16_t * block, uint8_t * dest, int stride);
 void (*idct_block_add) (int16_t * block, uint8_t * dest, int stride);
 
diff --git a/libmpeg2/idct_mlib.c b/libmpeg2/idct_mlib.c
index 055ee75fa6..876ab574a4 100644
--- a/libmpeg2/idct_mlib.c
+++ b/libmpeg2/idct_mlib.c
@@ -1,6 +1,6 @@
 /*
  * idct_mlib.c
- * Copyright (C) 1999 Håkan Hjort <d95hjort@dtek.chalmers.se>
+ * Copyright (C) 1999-2001 Håkan Hjort <d95hjort@dtek.chalmers.se>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -38,8 +38,8 @@ void idct_block_copy_mlib (int16_t * block, uint8_t * dest, int stride)
 
 void idct_block_add_mlib (int16_t * block, uint8_t * dest, int stride)
 {
-    // Should we use mlib_VideoIDCT_IEEE_S16_S16 here ??
-    // it's ~30% slower.
+    /* Should we use mlib_VideoIDCT_IEEE_S16_S16 here ?? */
+    /* it's ~30% slower. */
     mlib_VideoIDCT8x8_S16_S16 (block, block);
     mlib_VideoAddBlock_U8_S16 (dest, block, stride);
 }
diff --git a/libmpeg2/idct_mmx.c b/libmpeg2/idct_mmx.c
index 03ea5d7580..6c5d2ed96e 100644
--- a/libmpeg2/idct_mmx.c
+++ b/libmpeg2/idct_mmx.c
@@ -1,6 +1,6 @@
 /*
  * idct_mmx.c
- * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
+ * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -37,7 +37,7 @@
 
 
 #if 0
-// C row IDCT - its just here to document the MMXEXT and MMX versions
+/* C row IDCT - its just here to document the MMXEXT and MMX versions */
 static inline void idct_row (int16_t * row, int offset,
 			     int16_t * table, int32_t * rounder)
 {
@@ -76,7 +76,7 @@ static inline void idct_row (int16_t * row, int offset,
 #endif
 
 
-// MMXEXT row IDCT
+/* MMXEXT row IDCT */
 
 #define mmxext_table(c1,c2,c3,c4,c5,c6,c7)	{  c4,  c2, -c4, -c2,	\
 						   c4,  c6,  c4,  c6,	\
@@ -155,7 +155,7 @@ static inline void mmxext_row_tail (int16_t * row, int store)
     movq_r2m (mm1, *(row+store));	// save y3 y2 y1 y0
     pshufw_r2r (mm4, mm4, 0xb1);	// mm4 = y7 y6 y5 y4
 
-    // slot
+    /* slot */
 
     movq_r2m (mm4, *(row+store+4));	// save y7 y6 y5 y4
 }
@@ -188,7 +188,7 @@ static inline void mmxext_row_mid (int16_t * row, int store,
 }
 
 
-// MMX row IDCT
+/* MMX row IDCT */
 
 #define mmx_table(c1,c2,c3,c4,c5,c6,c7)	{  c4,  c2,  c4,  c6,	\
 					   c4,  c6, -c4, -c2,	\
@@ -276,7 +276,7 @@ static inline void mmx_row_tail (int16_t * row, int store)
 
     por_r2r (mm4, mm7);			// mm7 = y7 y6 y5 y4
 
-    // slot
+    /* slot */
 
     movq_r2m (mm7, *(row+store+4));	// save y7 y6 y5 y4
 }
@@ -320,10 +320,10 @@ static inline void mmx_row_mid (int16_t * row, int store,
 // C column IDCT - its just here to document the MMXEXT and MMX versions
 static inline void idct_col (int16_t * col, int offset)
 {
-// multiplication - as implemented on mmx
+/* multiplication - as implemented on mmx */
 #define F(c,x) (((c) * (x)) >> 16)
 
-// saturation - it helps us handle torture test cases
+/* saturation - it helps us handle torture test cases */
 #define S(x) (((x)>32767) ? 32767 : ((x)<-32768) ? -32768 : (x))
 
     int16_t x0, x1, x2, x3, x4, x5, x6, x7;
@@ -344,25 +344,25 @@ static inline void idct_col (int16_t * col, int offset)
 
     u04 = S (x0 + x4);
     v04 = S (x0 - x4);
-    u26 = S (F (T2, x6) + x2);	// -0.5
-    v26 = S (F (T2, x2) - x6);	// -0.5
+    u26 = S (F (T2, x6) + x2);
+    v26 = S (F (T2, x2) - x6);
 
     a0 = S (u04 + u26);
     a1 = S (v04 + v26);
     a2 = S (v04 - v26);
     a3 = S (u04 - u26);
 
-    u17 = S (F (T1, x7) + x1);	// -0.5
-    v17 = S (F (T1, x1) - x7);	// -0.5
-    u35 = S (F (T3, x5) + x3);	// -0.5
-    v35 = S (F (T3, x3) - x5);	// -0.5
+    u17 = S (F (T1, x7) + x1);
+    v17 = S (F (T1, x1) - x7);
+    u35 = S (F (T3, x5) + x3);
+    v35 = S (F (T3, x3) - x5);
 
     b0 = S (u17 + u35);
     b3 = S (v17 - v35);
     u12 = S (u17 - u35);
     v12 = S (v17 + v35);
-    u12 = S (2 * F (C4, u12));	// -0.5
-    v12 = S (2 * F (C4, v12));	// -0.5
+    u12 = S (2 * F (C4, u12));
+    v12 = S (2 * F (C4, v12));
     b1 = S (u12 + v12);
     b2 = S (u12 - v12);
 
@@ -400,7 +400,6 @@ static inline void idct_col (int16_t * col, int offset)
     static short _T2[] ATTR_ALIGN(8) = {T2,T2,T2,T2};
     static short _T3[] ATTR_ALIGN(8) = {T3,T3,T3,T3};
     static short _C4[] ATTR_ALIGN(8) = {C4,C4,C4,C4};
-    static mmx_t scratch0, scratch1;
 
     /* column code adapted from peter gubanov */
     /* http://www.elecard.com/peter/idct.shtml */
@@ -428,7 +427,7 @@ static inline void idct_col (int16_t * col, int offset)
     paddsw_r2r (mm2, mm1);		// mm1 = u17
     pmulhw_r2r (mm6, mm7);		// mm7 = (T3-1)*x5
 
-    // slot
+    /* slot */
 
     movq_r2r (mm4, mm2);		// mm2 = T2
     paddsw_r2r (mm3, mm5);		// mm5 = T3*x3
@@ -448,7 +447,7 @@ static inline void idct_col (int16_t * col, int offset)
     psubsw_r2r (mm3, mm4);		// mm4 = v26
     paddsw_r2r (mm6, mm5);		// mm5 = v12
 
-    movq_r2m (mm0, scratch0);		// save b3
+    movq_r2m (mm0, *(col+offset+3*8));	// save b3 in scratch0
     movq_r2r (mm1, mm6);		// mm6 = u17
 
     paddsw_m2r (*(col+offset+2*8), mm2);// mm2 = u26
@@ -463,7 +462,7 @@ static inline void idct_col (int16_t * col, int offset)
     movq_m2r (*_C4, mm0);		// mm0 = C4/2
     psubsw_r2r (mm5, mm7);		// mm7 = u12-v12
 
-    movq_r2m (mm6, scratch1);		// save b0
+    movq_r2m (mm6, *(col+offset+5*8));	// save b0 in scratch1
     pmulhw_r2r (mm0, mm1);		// mm1 = b1/2
 
     movq_r2r (mm4, mm6);		// mm6 = v26
@@ -496,7 +495,7 @@ static inline void idct_col (int16_t * col, int offset)
     psraw_i2r (COL_SHIFT, mm4);		// mm4 = y1
     psubsw_r2r (mm1, mm6);		// mm6 = a1-b1
 
-    movq_m2r (scratch1, mm1);		// mm1 = b0
+    movq_m2r (*(col+offset+5*8), mm1);	// mm1 = b0
     psubsw_r2r (mm7, mm2);		// mm2 = a2-b2
 
     psraw_i2r (COL_SHIFT, mm6);		// mm6 = y6
@@ -508,7 +507,7 @@ static inline void idct_col (int16_t * col, int offset)
     movq_r2m (mm3, *(col+offset+2*8));	// save y2
     paddsw_r2r (mm1, mm5);		// mm5 = a0+b0
 
-    movq_m2r (scratch0, mm4);		// mm4 = b3
+    movq_m2r (*(col+offset+3*8), mm4);	// mm4 = b3
     psubsw_r2r (mm1, mm7);		// mm7 = a0-b0
 
     psraw_i2r (COL_SHIFT, mm5);		// mm5 = y0
@@ -538,17 +537,17 @@ static int32_t rounder0[] ATTR_ALIGN(8) =
     rounder ((1 << (COL_SHIFT - 1)) - 0.5);
 static int32_t rounder4[] ATTR_ALIGN(8) = rounder (0);
 static int32_t rounder1[] ATTR_ALIGN(8) =
-    rounder (1.25683487303);	// C1*(C1/C4+C1+C7)/2
+    rounder (1.25683487303);	/* C1*(C1/C4+C1+C7)/2 */
 static int32_t rounder7[] ATTR_ALIGN(8) =
-    rounder (-0.25);		// C1*(C7/C4+C7-C1)/2
+    rounder (-0.25);		/* C1*(C7/C4+C7-C1)/2 */
 static int32_t rounder2[] ATTR_ALIGN(8) =
-    rounder (0.60355339059);	// C2 * (C6+C2)/2
+    rounder (0.60355339059);	/* C2 * (C6+C2)/2 */
 static int32_t rounder6[] ATTR_ALIGN(8) =
-    rounder (-0.25);		// C2 * (C6-C2)/2
+    rounder (-0.25);		/* C2 * (C6-C2)/2 */
 static int32_t rounder3[] ATTR_ALIGN(8) =
-    rounder (0.087788325588);	// C3*(-C3/C4+C3+C5)/2
+    rounder (0.087788325588);	/* C3*(-C3/C4+C3+C5)/2 */
 static int32_t rounder5[] ATTR_ALIGN(8) =
-    rounder (-0.441341716183);	// C3*(-C5/C4+C5-C3)/2
+    rounder (-0.441341716183);	/* C3*(-C5/C4+C5-C3)/2 */
 
 
 #define declare_idct(idct,table,idct_row_head,idct_row,idct_row_tail,idct_row_mid)	\
@@ -693,7 +692,7 @@ void idct_mmx_init (void)
     extern uint8_t scan_alt[64];
     int i, j;
 
-    // the mmx/mmxext idct uses a reordered input, so we patch scan tables
+    /* the mmx/mmxext idct uses a reordered input, so we patch scan tables */
 
     for (i = 0; i < 64; i++) {
 	j = scan_norm[i];
diff --git a/libmpeg2/mmx.h b/libmpeg2/mmx.h
index bab97b8b1f..ac23866690 100644
--- a/libmpeg2/mmx.h
+++ b/libmpeg2/mmx.h
@@ -1,6 +1,6 @@
 /*
  * mmx.h
- * Copyright (C) 1997-1999 H. Dietz and R. Fisher
+ * Copyright (C) 1997-2001 H. Dietz and R. Fisher
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -41,16 +41,16 @@ typedef	union {
 #define	mmx_i2r(op,imm,reg) \
 	__asm__ __volatile__ (#op " %0, %%" #reg \
 			      : /* nothing */ \
-			      : "X" (imm) )
+			      : "i" (imm) )
 
 #define	mmx_m2r(op,mem,reg) \
 	__asm__ __volatile__ (#op " %0, %%" #reg \
 			      : /* nothing */ \
-			      : "X" (mem))
+			      : "m" (mem))
 
 #define	mmx_r2m(op,reg,mem) \
 	__asm__ __volatile__ (#op " %%" #reg ", %0" \
-			      : "=X" (mem) \
+			      : "=m" (mem) \
 			      : /* nothing */ )
 
 #define	mmx_r2r(op,regs,regd) \
diff --git a/libmpeg2/motion_comp.c b/libmpeg2/motion_comp.c
index 816335c6dc..52e4655362 100644
--- a/libmpeg2/motion_comp.c
+++ b/libmpeg2/motion_comp.c
@@ -1,6 +1,6 @@
 /*
  * motion_comp.c
- * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
+ * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -67,7 +67,7 @@ void motion_comp_init (void)
 #define put(predictor,i) dest[i] = predictor (i)
 #define avg(predictor,i) dest[i] = avg2 (predictor (i), dest[i])
 
-// mc function template
+/* mc function template */
 
 #define MC_FUNC(op,xy)						\
 static void MC_##op##_##xy##16_c (uint8_t * dest, uint8_t * ref,\
@@ -111,7 +111,7 @@ static void MC_##op##_##xy##8_c (uint8_t * dest, uint8_t * ref,	\
     } while (--height);						\
 }
 
-// definitions of the actual mc functions
+/* definitions of the actual mc functions */
 
 MC_FUNC (put,)
 MC_FUNC (avg,)
diff --git a/libmpeg2/motion_comp_mlib.c b/libmpeg2/motion_comp_mlib.c
index e079119eb9..91c0fb5a87 100644
--- a/libmpeg2/motion_comp_mlib.c
+++ b/libmpeg2/motion_comp_mlib.c
@@ -1,6 +1,6 @@
 /*
- * MC_mlib.c
- * Copyright (C) 2000 Håkan Hjort <d95hjort@dtek.chalmers.se>
+ * motion_comp_mlib.c
+ * Copyright (C) 2000-2001 Håkan Hjort <d95hjort@dtek.chalmers.se>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
diff --git a/libmpeg2/motion_comp_mmx.c b/libmpeg2/motion_comp_mmx.c
index f635692045..51b40bac55 100644
--- a/libmpeg2/motion_comp_mmx.c
+++ b/libmpeg2/motion_comp_mmx.c
@@ -1,6 +1,6 @@
 /*
  * motion_comp_mmx.c
- * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
+ * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
  *
@@ -33,7 +33,7 @@
 #define CPU_3DNOW 1
 
 
-//MMX code - needs a rewrite
+/* MMX code - needs a rewrite */
 
 
 
@@ -41,7 +41,7 @@
 
 
 
-// some rounding constants
+/* some rounding constants */
 mmx_t round1 = {0x0001000100010001LL};
 mmx_t round4 = {0x0002000200020002LL};
 
@@ -55,16 +55,14 @@ mmx_t round4 = {0x0002000200020002LL};
 
 static inline void mmx_zero_reg ()
 {
-    // load 0 into mm0
+    /* load 0 into mm0 */
     pxor_r2r (mm0, mm0);
 }
 
 static inline void mmx_average_2_U8 (uint8_t * dest,
 				     uint8_t * src1, uint8_t * src2)
 {
-    //
-    // *dest = (*src1 + *src2 + 1)/ 2;
-    //
+    /* *dest = (*src1 + *src2 + 1)/ 2; */