[FFmpeg-cvslog] avutil/md5: fix misaligned reads

[James Almer]

ffmpeg | branch: master | James Almer <jamrial at gmail.com> | Fri Mar  3 00:25:54 2017 -0300| [a876958d0fdd8bf10d315175daff12cd7d768053] | committer: Sean McGovern

avutil/md5: fix misaligned reads

This makes ubsan happy and also considerably increases performance on
big endian systems.

Tested on an IBM POWER7 3.55 GHz

Before:

2.24user 0.14system 0:02.39elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k
2.26user 0.11system 0:02.38elapsed 99%CPU (0avgtext+0avgdata 2688maxresident)k
2.23user 0.15system 0:02.38elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k
2.25user 0.12system 0:02.38elapsed 100%CPU (0avgtext+0avgdata 2624maxresident)k
2.20user 0.15system 0:02.36elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k

After:

1.86user 0.13system 0:02.00elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k
1.89user 0.11system 0:02.01elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k
1.85user 0.14system 0:02.00elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k
1.84user 0.15system 0:01.99elapsed 99%CPU (0avgtext+0avgdata 2624maxresident)k
1.89user 0.13system 0:02.02elapsed 99%CPU (0avgtext+0avgdata 2688maxresident)k

Tested-by: Nicolas George <george at nsup.org>
Reviewed-by: Michael Niedermayer <michael at niedermayer.cc>
Signed-off-by: James Almer <jamrial at gmail.com>
Signed-off-by: Sean McGovern <gseanmcg at gmail.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a876958d0fdd8bf10d315175daff12cd7d768053
---

 libavutil/md5.c | 20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)

diff --git a/libavutil/md5.c b/libavutil/md5.c
index 1946d78317..0ddde56b67 100644
--- a/libavutil/md5.c
+++ b/libavutil/md5.c
@@ -83,15 +83,15 @@ static const uint32_t T[64] = { // T[i]= fabs(sin(i+1)<<32)
         a += T[i];                                                      \
                                                                         \
         if (i < 32) {                                                   \
-            if (i < 16)                                                 \
-                a += (d ^ (b & (c ^ d))) + X[i           & 15];         \
-            else                                                        \
-                a += (c ^ (d & (c ^ b))) + X[(1 + 5 * i) & 15];         \
+            if (i < 16)                                                     \
+                a += (d ^ (b & (c ^ d))) + AV_RL32(X + (i          & 15));  \
+            else                                                            \
+                a += (c ^ (d & (c ^ b))) + AV_RL32(X + ((1 + 5 * i) & 15)); \
         } else {                                                        \
             if (i < 48)                                                 \
-                a += (b ^ c ^ d)    + X[(5 + 3 * i) & 15];              \
+                a += (b ^ c ^ d)    + AV_RL32(X + ((5 + 3 * i) & 15));  \
             else                                                        \
-                a += (c ^ (b | ~d)) + X[(7     * i) & 15];              \
+                a += (c ^ (b | ~d)) + AV_RL32(X + ((7     * i) & 15));  \
         }                                                               \
         a = b + (a << t | a >> (32 - t));                               \
     } while (0)
@@ -99,18 +99,14 @@ static const uint32_t T[64] = { // T[i]= fabs(sin(i+1)<<32)
 static void body(uint32_t ABCD[4], uint32_t X[16])
 {
     int t;
-    int i av_unused;
     unsigned int a = ABCD[3];
     unsigned int b = ABCD[2];
     unsigned int c = ABCD[1];
     unsigned int d = ABCD[0];
 
-#if HAVE_BIGENDIAN
-    for (i = 0; i < 16; i++)
-        X[i] = av_bswap32(X[i]);
-#endif
-
 #if CONFIG_SMALL
+    int i;
+
     for (i = 0; i < 64; i++) {
         CORE(i, a, b, c, d);
         t = d;