[FFmpeg-cvslog] dca: ARMv6 optimised decode_blockcode()

[Mans Rullgard]

ffmpeg | branch: master | Mans Rullgard <mans at mansr.com> | Thu Sep 29 22:03:40 2011 +0100| [d4999e0a798cc82db212be67d9edb22f0a7e5847] | committer: Michael Niedermayer

dca: ARMv6 optimised decode_blockcode()

Signed-off-by: Mans Rullgard <mans at mansr.com>
(cherry picked from commit 08e3dea3f7f69309574dafc0af6671615e909720)

Signed-off-by: Michael Niedermayer <michaelni at gmx.at>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d4999e0a798cc82db212be67d9edb22f0a7e5847
---

 libavcodec/arm/dca.h |   39 +++++++++++++++++++++++++++++++++++++++
 libavcodec/dca.c     |    2 ++
 2 files changed, 41 insertions(+), 0 deletions(-)

diff --git a/libavcodec/arm/dca.h b/libavcodec/arm/dca.h
index 38c8d1f..e85b82c 100644
--- a/libavcodec/arm/dca.h
+++ b/libavcodec/arm/dca.h
@@ -23,6 +23,45 @@
 
 #include <stdint.h>
 #include "config.h"
+#include "libavutil/intmath.h"
+
+#if HAVE_ARMV6 && HAVE_INLINE_ASM
+
+#define decode_blockcode decode_blockcode
+static inline int decode_blockcode(int code, int levels, int *values)
+{
+    int v0, v1, v2, v3;
+
+    __asm__ ("smmul   %4, %8, %11           \n"
+             "smlabb  %8, %4, %10, %8       \n"
+             "smmul   %5, %4, %11           \n"
+             "sub     %8, %8, %9, lsr #1    \n"
+             "smlabb  %4, %5, %10, %4       \n"
+             "smmul   %6, %5, %11           \n"
+             "str     %8, %0                \n"
+             "sub     %4, %4, %9, lsr #1    \n"
+             "smlabb  %5, %6, %10, %5       \n"
+             "smmul   %7, %6, %11           \n"
+             "str     %4, %1                \n"
+             "sub     %5, %5, %9, lsr #1    \n"
+             "smlabb  %6, %7, %10, %6       \n"
+             "cmp     %7, #0                \n"
+             "str     %5, %2                \n"
+             "sub     %6, %6, %9, lsr #1    \n"
+             "it      eq                    \n"
+             "mvneq   %7, #0                \n"
+             "str     %6, %3                \n"
+             : "=m"(values[0]), "=m"(values[1]),
+               "=m"(values[2]), "=m"(values[3]),
+               "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3),
+               "+&r"(code)
+             : "r"(levels - 1), "r"(-levels), "r"(ff_inverse[levels])
+             : "cc");
+
+    return v3;
+}
+
+#endif
 
 #if HAVE_NEON && HAVE_INLINE_ASM && HAVE_ASM_MOD_Y
 
diff --git a/libavcodec/dca.c b/libavcodec/dca.c
index d900d88..cdd9161 100644
--- a/libavcodec/dca.c
+++ b/libavcodec/dca.c
@@ -1038,6 +1038,7 @@ static void dca_downmix(float *samples, int srcfmt,
 }
 
 
+#ifndef decode_blockcode
 /* Very compact version of the block code decoder that does not use table
  * look-up but is slightly slower */
 static int decode_blockcode(int code, int levels, int *values)
@@ -1058,6 +1059,7 @@ static int decode_blockcode(int code, int levels, int *values)
         return -1;
     }
 }
+#endif
 
 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };