Using a table on ARM for unsigned division by small (<=256) integers.

Saves 0.6% for 64 kb/s and 1.8% for 128 kb/s when decoding on arm7tdmi.

1 files changed, 22 insertions, 0 deletions

diff --git a/celt/entcode.h b/celt/entcode.h
index dd13e49e..c10fe8d7 100644
--- a/celt/entcode.h
+++ b/celt/entcode.h
@@ -34,6 +34,12 @@
 # include <stddef.h>
 # include "ecintrin.h"
 
+extern const opus_uint32 SMALL_DIV_TABLE[129];
+
+#ifdef OPUS_ARM_ASM
+#define USE_SMALL_DIV_TABLE
+#endif
+
 /*OPT: ec_window must be at least 32 bits, but if you have fast arithmetic on a
    larger type, you can speed up the decoder by using it here.*/
 typedef opus_uint32           ec_window;
@@ -114,4 +120,20 @@ static OPUS_INLINE int ec_tell(ec_ctx *_this){
            rounding error is in the positive direction).*/
 opus_uint32 ec_tell_frac(ec_ctx *_this);
 
+/* Tested exhaustively for all n and for 1<=d<=256 */
+static OPUS_INLINE opus_uint32 celt_udiv(opus_uint32 n, opus_uint32 d) {
+#ifdef USE_SMALL_DIV_TABLE
+   if (d>256)
+      return n/d;
+   else {
+      opus_uint32 t, q;
+      t = EC_ILOG(d&-d);
+      q = (opus_uint64)SMALL_DIV_TABLE[d>>t]*(n>>(t-1))>>32;
+      return q+(n-q*d >= d);
+   }
+#else
+   return n/d;
+#endif
+}
+
 #endif