bsd: use number of logical CPUs as automatic thread count

[ffmpeg] / libavcodec / alpha / mpegvideo_alpha.c

diff --git a/libavcodec/alpha/mpegvideo_alpha.c b/libavcodec/alpha/mpegvideo_alpha.c
index d0af5e1d32243df0d429646c69aff034d027b620..add57364a3d0529739064fbda33779ec2e552245 100644 (file)
--- a/libavcodec/alpha/mpegvideo_alpha.c
+++ b/libavcodec/alpha/mpegvideo_alpha.c
@@ -2,87 +2,109 @@
  * Alpha optimized DSP utils
  * Copyright (c) 2002 Falk Hueffner <falk@debian.org>
  *
- * This library is free software; you can redistribute it and/or
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ * version 2.1 of the License, or (at your option) any later version.
  *
- * This library is distributed in the hope that it will be useful,
+ * Libav 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
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavcodec/dsputil.h"
+#include "libavcodec/mpegvideo.h"
 #include "asm.h"
-#include "../dsputil.h"
-#include "../mpegvideo.h"
 
-extern UINT8 zigzag_end[64];
+static void dct_unquantize_h263_axp(DCTELEM *block, int n_coeffs,
+                                    uint64_t qscale, uint64_t qadd)
+{
+    uint64_t qmul = qscale << 1;
+    uint64_t correction = WORD_VEC(qmul * 255 >> 8);
+    int i;
+
+    qadd = WORD_VEC(qadd);
+
+    for(i = 0; i <= n_coeffs; block += 4, i += 4) {
+        uint64_t levels, negmask, zeros, add, sub;
+
+        levels = ldq(block);
+        if (levels == 0)
+            continue;
+
+#ifdef __alpha_max__
+        /* I don't think the speed difference justifies runtime
+           detection.  */
+        negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
+        negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
+#else
+        negmask = cmpbge(WORD_VEC(0x7fff), levels);
+        negmask &= (negmask >> 1) | (1 << 7);
+        negmask = zap(-1, negmask);
+#endif
+
+        zeros = cmpbge(0, levels);
+        zeros &= zeros >> 1;
+        /* zeros |= zeros << 1 is not needed since qadd <= 255, so
+           zapping the lower byte suffices.  */
+
+        levels *= qmul;
+        levels -= correction & (negmask << 16);
+
+        add = qadd & ~negmask;
+        sub = qadd &  negmask;
+        /* Set qadd to 0 for levels == 0.  */
+        add = zap(add, zeros);
+        levels += add;
+        levels -= sub;
+
+        stq(levels, block);
+    }
+}
 
-static void dct_unquantize_h263_axp(MpegEncContext *s, 
-                                   DCTELEM *block, int n, int qscale)
+static void dct_unquantize_h263_intra_axp(MpegEncContext *s, DCTELEM *block,
+                                    int n, int qscale)
 {
-    int i, level;
-    UINT64 qmul, qadd;
-    if (s->mb_intra) {
-        if (n < 4) 
-            block[0] = block[0] * s->y_dc_scale;
+    int n_coeffs;
+    uint64_t qadd;
+    DCTELEM block0 = block[0];
+
+    if (!s->h263_aic) {
+        if (n < 4)
+            block0 *= s->y_dc_scale;
         else
-            block[0] = block[0] * s->c_dc_scale;
-       /* Catch up to aligned point.  */
-       qmul = s->qscale << 1;
-       qadd = (s->qscale - 1) | 1;
-       for (i = 1; i < 4; ++i) {
-           level = block[i];
-           if (level) {
-               if (level < 0) {
-                   level = level * qmul - qadd;
-               } else {
-                   level = level * qmul + qadd;
-               }
-               block[i] = level;
-           }
-       }
-       block += 4;
-       i = 60 / 4;
+            block0 *= s->c_dc_scale;
+        qadd = (qscale - 1) | 1;
     } else {
-        i = zigzag_end[s->block_last_index[n]] / 4;
+        qadd = 0;
     }
-    qmul = s->qscale << 1;
-    qadd = WORD_VEC((qscale - 1) | 1);
-    do {
-       UINT64 levels, negmask, zeromask, corr;
-       levels = ldq(block);
-       if (levels == 0)
-           continue;
-       zeromask = cmpbge(0, levels);
-       zeromask &= zeromask >> 1;
-       /* Negate all negative words.  */
-       negmask = maxsw4(levels, WORD_VEC(0xffff)); /* negative -> ffff (-1) */
-       negmask = minsw4(negmask, 0);               /* positive -> 0000 (0) */
-       corr    = negmask & WORD_VEC(0x0001); /* twos-complement correction */
-       levels ^= negmask;
-       levels += corr;
-
-       levels = levels * qmul;
-       levels += zap(qadd, zeromask);
-
-       /* Re-negate negative words.  */
-       levels -= corr;
-       levels ^= negmask;
-
-       stq(levels, block);
-    } while (block += 4, --i);
+
+    if(s->ac_pred)
+        n_coeffs = 63;
+    else
+        n_coeffs = s->inter_scantable.raster_end[s->block_last_index[n]];
+
+    dct_unquantize_h263_axp(block, n_coeffs, qscale, qadd);
+
+    block[0] = block0;
+}
+
+static void dct_unquantize_h263_inter_axp(MpegEncContext *s, DCTELEM *block,
+                                    int n, int qscale)
+{
+    int n_coeffs = s->inter_scantable.raster_end[s->block_last_index[n]];
+    dct_unquantize_h263_axp(block, n_coeffs, qscale, (qscale - 1) | 1);
 }
 
 void MPV_common_init_axp(MpegEncContext *s)
 {
-    if (amask(AMASK_MVI) == 0) {
-        if (s->out_format == FMT_H263)
-           s->dct_unquantize = dct_unquantize_h263_axp;
-    }
+    s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_axp;
+    s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_axp;
 }