md5: avoid unnecessary copying.

Where necessary use memcpy instead.
Thanks to Giorgio Vazzana [mywing81 gmail] for
spotting this loop as the cause for the bad
performance.
Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>

libavutil/md5.c

@@ -139,20 +139,40 @@ void av_md5_init(AVMD5 *ctx)
     ctx->ABCD[3] = 0x67452301;
 }
 
-void av_md5_update(AVMD5 *ctx, const uint8_t *src, const int len)
+void av_md5_update(AVMD5 *ctx, const uint8_t *src, int len)
 {
-    int i, j;
+    const uint8_t *end;
+    int j;
 
     j = ctx->len & 63;
     ctx->len += len;
 
-    for (i = 0; i < len; i++) {
-        ctx->block[j++] = src[i];
-        if (j == 64) {
-            body(ctx->ABCD, (uint32_t *) ctx->block);
-            j = 0;
+    if (j) {
+        int cnt = FFMIN(len, 64 - j);
+        memcpy(ctx->block + j, src, cnt);
+        src += cnt;
+        len -= cnt;
+        if (j + cnt < 64)
+            return;
+        body(ctx->ABCD, (uint32_t *)ctx->block);
+    }
+
+    end = src + (len & ~63);
+    if (HAVE_BIGENDIAN || (!HAVE_FAST_UNALIGNED && ((intptr_t)src & 3))) {
+       while (src < end) {
+           memcpy(ctx->block, src, 64);
+           body(ctx->ABCD, (uint32_t *) ctx->block);
+           src += 64;
+        }
+    } else {
+        while (src < end) {
+            body(ctx->ABCD, (uint32_t *)src);
+            src += 64;
         }
     }
+    len &= 63;
+    if (len > 0)
+        memcpy(ctx->block, src, len);
 }
 
 void av_md5_final(AVMD5 *ctx, uint8_t *dst)

libavutil/md5.h

@@ -38,7 +38,7 @@ struct AVMD5;
 
 struct AVMD5 *av_md5_alloc(void);
 void av_md5_init(struct AVMD5 *ctx);
-void av_md5_update(struct AVMD5 *ctx, const uint8_t *src, const int len);
+void av_md5_update(struct AVMD5 *ctx, const uint8_t *src, int len);
 void av_md5_final(struct AVMD5 *ctx, uint8_t *dst);
 void av_md5_sum(uint8_t *dst, const uint8_t *src, const int len);