Add simple ASMed fixed point multiply functions for ARM to AC3. Gives a 38% speed up on my one sample with a Sansa.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@18570 a1c6a512-1295-4272-9138-f99709370657

1 files changed, 30 insertions, 0 deletions

diff --git a/apps/codecs/liba52/a52_internal.h b/apps/codecs/liba52/a52_internal.h
index cd3286b17a..0db16a8bcf 100644
--- a/apps/codecs/liba52/a52_internal.h
+++ b/apps/codecs/liba52/a52_internal.h
@@ -166,6 +166,36 @@ typedef int16_t quantizer_t;
         : [A] "r" ((a)), [B] "r" ((b))); \
     t; \
 })
+
+#elif defined(CPU_ARM)
+#define MUL(x, y)  \
+    ({ int32_t __hi;  \
+       uint32_t __lo;  \
+       int32_t __result;  \
+       asm ("smull   %0, %1, %3, %4\n\t"  \
+            "movs    %2, %1, lsl #2"  \
+            : "=&r" (__lo), "=&r" (__hi), "=r" (__result)  \
+            : "%r" (x), "r" (y)  \
+            : "cc");  \
+       __result;  \
+    })
+
+
+#define MUL_L(x, y)  \
+    ({ int32_t __hi;  \
+       uint32_t __lo;  \
+       int32_t __result;  \
+       asm ("smull   %0, %1, %3, %4\n\t"  \
+            "movs    %0, %0, lsr %5\n\t"  \
+            "adc    %2, %0, %1, lsl %6"  \
+            : "=&r" (__lo), "=&r" (__hi), "=r" (__result)  \
+            : "%r" (x), "r" (y),  \
+              "M" (26), "M" (32 - 26)  \
+            : "cc");  \
+       __result;  \
+    })
+    
+
 #elif 1
 #define MUL(a,b) \
 ({ int32_t _ta=(a), _tb=(b), _tc; \