1 files changed, 103 insertions, 8 deletions

diff --git a/libavcodec/ppc/pixblockdsp.c b/libavcodec/ppc/pixblockdsp.c
index 9cac70e2ef..84aa562bb6 100644
--- a/libavcodec/ppc/pixblockdsp.c
+++ b/libavcodec/ppc/pixblockdsp.c
@@ -3,20 +3,20 @@
  * Copyright (c) 2002 Dieter Shirley
  * Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
  *
- * This file is part of Libav.
+ * This file is part of FFmpeg.
  *
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg 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 2.1 of the License, or (at your option) any later version.
  *
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg 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 Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
@@ -33,10 +33,38 @@
 #include "libavcodec/avcodec.h"
 #include "libavcodec/pixblockdsp.h"
 
-#if HAVE_ALTIVEC && HAVE_BIGENDIAN
+#if HAVE_ALTIVEC
 
+#if HAVE_VSX
 static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels,
-                               int line_size)
+                               ptrdiff_t line_size)
+{
+    int i;
+    vector unsigned char perm =
+        (vector unsigned char) {0x00,0x10, 0x01,0x11,0x02,0x12,0x03,0x13,\
+            0x04,0x14,0x05,0x15,0x06,0x16,0x07,0x17};
+    const vector unsigned char zero =
+        (const vector unsigned char) vec_splat_u8(0);
+
+    for (i = 0; i < 8; i++) {
+        /* Read potentially unaligned pixels.
+         * We're reading 16 pixels, and actually only want 8,
+         * but we simply ignore the extras. */
+        vector unsigned char bytes = vec_vsx_ld(0, pixels);
+
+        // Convert the bytes into shorts.
+        //vector signed short shorts = (vector signed short) vec_perm(zero, bytes, perm);
+        vector signed short shorts = (vector signed short) vec_perm(bytes, zero, perm);
+
+        // Save the data to the block, we assume the block is 16-byte aligned.
+        vec_vsx_st(shorts, i * 16, (vector signed short *) block);
+
+        pixels += line_size;
+    }
+}
+#else
+static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels,
+                               ptrdiff_t line_size)
 {
     int i;
     vec_u8 perm = vec_lvsl(0, pixels);
@@ -60,6 +88,71 @@ static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels,
     }
 }
 
+#endif /* HAVE_VSX */
+
+#if HAVE_VSX
+static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1,
+                                const uint8_t *s2, int stride)
+{
+  int i;
+  const vector unsigned char zero =
+    (const vector unsigned char) vec_splat_u8(0);
+  vector signed short shorts1, shorts2;
+
+  for (i = 0; i < 4; i++) {
+    /* Read potentially unaligned pixels.
+     * We're reading 16 pixels, and actually only want 8,
+     * but we simply ignore the extras. */
+    vector unsigned char bytes = vec_vsx_ld(0,  s1);
+
+    // Convert the bytes into shorts.
+    shorts1 = (vector signed short) vec_mergeh(bytes, zero);
+
+    // Do the same for the second block of pixels.
+    bytes =vec_vsx_ld(0,  s2);
+
+    // Convert the bytes into shorts.
+    shorts2 = (vector signed short) vec_mergeh(bytes, zero);
+
+    // Do the subtraction.
+    shorts1 = vec_sub(shorts1, shorts2);
+
+    // Save the data to the block, we assume the block is 16-byte aligned.
+    vec_vsx_st(shorts1, 0, (vector signed short *) block);
+
+    s1    += stride;
+    s2    += stride;
+    block += 8;
+
+    /* The code below is a copy of the code above...
+     * This is a manual unroll. */
+
+    /* Read potentially unaligned pixels.
+     * We're reading 16 pixels, and actually only want 8,
+     * but we simply ignore the extras. */
+    bytes = vec_vsx_ld(0,  s1);
+
+    // Convert the bytes into shorts.
+    shorts1 = (vector signed short) vec_mergeh(bytes, zero);
+
+    // Do the same for the second block of pixels.
+    bytes = vec_vsx_ld(0,  s2);
+
+    // Convert the bytes into shorts.
+    shorts2 = (vector signed short) vec_mergeh(bytes, zero);
+
+    // Do the subtraction.
+    shorts1 = vec_sub(shorts1, shorts2);
+
+    // Save the data to the block, we assume the block is 16-byte aligned.
+    vec_vsx_st(shorts1, 0, (vector signed short *) block);
+
+    s1    += stride;
+    s2    += stride;
+    block += 8;
+  }
+}
+#else
 static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1,
                                 const uint8_t *s2, int stride)
 {
@@ -131,11 +224,13 @@ static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1,
     }
 }
 
+#endif /* HAVE_VSX */
+
 #endif /* HAVE_ALTIVEC */
 
 #if HAVE_VSX
 static void get_pixels_vsx(int16_t *restrict block, const uint8_t *pixels,
-                           int line_size)
+                           ptrdiff_t line_size)
 {
     int i;
     for (i = 0; i < 8; i++) {
@@ -171,7 +266,7 @@ av_cold void ff_pixblockdsp_init_ppc(PixblockDSPContext *c,
                                      AVCodecContext *avctx,
                                      unsigned high_bit_depth)
 {
-#if HAVE_ALTIVEC && HAVE_BIGENDIAN
+#if HAVE_ALTIVEC
     if (!PPC_ALTIVEC(av_get_cpu_flags()))
         return;