ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)

Originally committed as revision 996 to svn://svn.ffmpeg.org/ffmpeg/trunk

1 files changed, 100 insertions, 0 deletions

diff --git a/libavcodec/ps2/mpegvideo_mmi.c b/libavcodec/ps2/mpegvideo_mmi.c
new file mode 100644
index 0000000000..f12640cb8b
--- /dev/null
+++ b/libavcodec/ps2/mpegvideo_mmi.c
@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 2000,2001 Fabrice Bellard.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ *
+ * MMI optimization by Leon van Stuivenberg <leonvs@iae.nl>
+ */
+ 
+#include "../dsputil.h"
+#include "../mpegvideo.h"
+#include "../avcodec.h"
+
+void ff_mmi_idct_put(UINT8 *dest, int line_size, DCTELEM *block);
+void ff_mmi_idct_add(UINT8 *dest, int line_size, DCTELEM *block);
+
+
+static void dct_unquantize_h263_mmi(MpegEncContext *s, 
+                                  DCTELEM *block, int n, int qscale)
+{
+    int level=0, qmul, qadd;
+    int nCoeffs;
+    
+    assert(s->block_last_index[n]>=0);
+    
+    qadd = (qscale - 1) | 1;
+    qmul = qscale << 1;
+    
+    if (s->mb_intra) {
+        if (!s->h263_aic) {
+            if (n < 4) 
+                level = block[0] * s->y_dc_scale;
+            else
+                level = block[0] * s->c_dc_scale;
+        }else {
+            qadd = 0;
+	    level = block[0];
+        }
+        nCoeffs= 63; //does not allways use zigzag table 
+    } else {
+        nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
+    }
+
+    asm volatile(
+        "add    $14, $0, %3	\n\t"
+        "pcpyld $8, %0, %0	\n\t"	
+        "pcpyh  $8, $8		\n\t"   //r8 = qmul
+        "pcpyld $9, %1, %1	\n\t"	
+        "pcpyh  $9, $9		\n\t"   //r9 = qadd
+        ".p2align 2             \n\t"
+        "1:			\n\t"
+        "lq     $10, 0($14)	\n\t"   //r10 = level
+        "addi   $14, $14, 16	\n\t"	//block+=8
+        "addi   %2, %2, -8	\n\t"
+        "pcgth  $11, $0, $10	\n\t"   //r11 = level < 0 ? -1 : 0
+        "pcgth  $12, $10, $0	\n\t"   //r12 = level > 0 ? -1 : 0
+        "por    $12, $11, $12	\n\t"
+        "pmulth $10, $10, $8	\n\t"	
+        "paddh  $13, $9, $11	\n\t"
+        "pxor   $13, $13, $11   \n\t"   //r13 = level < 0 ? -qadd : qadd
+        "pmfhl.uw $11		\n\t"
+        "pinteh $10, $11, $10	\n\t"   //r10 = level * qmul
+        "paddh  $10, $10, $13	\n\t"
+        "pand   $10, $10, $12   \n\t"
+        "sq     $10, -16($14)	\n\t"
+        "bgez   %2, 1b		\n\t"
+	:: "r"(qmul), "r" (qadd), "r" (nCoeffs), "r" (block) : "$8", "$9", "$10", "$11", "$12", "$13", "$14", "memory" );
+
+    if(s->mb_intra)
+        block[0]= level;
+}
+
+
+void MPV_common_init_mmi(MpegEncContext *s)
+{
+    int i;
+//  const int dct_algo = s->avctx->dct_algo;
+    const int idct_algo= s->avctx->idct_algo;
+       
+    if(idct_algo==FF_IDCT_AUTO){
+        s->idct_put= ff_mmi_idct_put;
+        s->idct_add= ff_mmi_idct_add;
+        for(i=0; i<64; i++)
+            s->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
+    }
+    s->dct_unquantize_h263 = dct_unquantize_h263_mmi;
+}
+
+