Merge commit 'e6bff23f1e11aefb16a2b5d6ee72bf7469c5a66e'

* commit 'e6bff23f1e11aefb16a2b5d6ee72bf7469c5a66e':
  cpu: add a function for querying maximum required data alignment

Adapted to work with the arbitrary runtime cpuflag changes av_force_cpu_flags()
can generate.

Merged-by: James Almer <jamrial@gmail.com>

doc/APIchanges

@@ -15,6 +15,9 @@ libavutil:     2015-08-28
 
 API changes, most recent first:
 
+2017-09-27 - xxxxxxx - lavu 55.77.100 / lavu 55.31.0 - cpu.h
+  Add av_cpu_max_align() for querying maximum required data alignment.
+
 2017-09-26 - xxxxxxx - lavc 57.106.102 - avcodec.h
   Deprecate AVCodecContext.refcounted_frames. This was useful for deprecated
   API only (avcodec_decode_video2/avcodec_decode_audio4). The new decode APIs

libavutil/cpu.c

@@ -16,9 +16,11 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include <stddef.h>
 #include <stdint.h>
 #include <stdatomic.h>
 
+#include "attributes.h"
 #include "cpu.h"
 #include "cpu_internal.h"
 #include "config.h"
@@ -299,3 +301,40 @@ int av_cpu_count(void)
 
     return nb_cpus;
 }
+
+size_t av_cpu_max_align(void)
+{
+    int av_unused flags = av_get_cpu_flags();
+
+#if ARCH_ARM || ARCH_AARCH64
+    if (flags & AV_CPU_FLAG_NEON)
+        return 16;
+#elif ARCH_PPC
+    if (flags & (AV_CPU_FLAG_ALTIVEC   |
+                 AV_CPU_FLAG_VSX       |
+                 AV_CPU_FLAG_POWER8))
+        return 16;
+#elif ARCH_X86
+    if (flags & (AV_CPU_FLAG_AVX2      |
+                 AV_CPU_FLAG_AVX       |
+                 AV_CPU_FLAG_XOP       |
+                 AV_CPU_FLAG_FMA4      |
+                 AV_CPU_FLAG_FMA3      |
+                 AV_CPU_FLAG_AVXSLOW))
+        return 32;
+    if (flags & (AV_CPU_FLAG_AESNI     |
+                 AV_CPU_FLAG_SSE42     |
+                 AV_CPU_FLAG_SSE4      |
+                 AV_CPU_FLAG_SSSE3     |
+                 AV_CPU_FLAG_SSE3      |
+                 AV_CPU_FLAG_SSE2      |
+                 AV_CPU_FLAG_SSE       |
+                 AV_CPU_FLAG_ATOM      |
+                 AV_CPU_FLAG_SSSE3SLOW |
+                 AV_CPU_FLAG_SSE3SLOW  |
+                 AV_CPU_FLAG_SSE2SLOW))
+        return 16;
+#endif
+
+    return 8;
+}

libavutil/cpu.h

@@ -21,6 +21,8 @@
 #ifndef AVUTIL_CPU_H
 #define AVUTIL_CPU_H
 
+#include <stddef.h>
+
 #include "attributes.h"
 
 #define AV_CPU_FLAG_FORCE    0x80000000 /* force usage of selected flags (OR) */
@@ -113,4 +115,15 @@ int av_parse_cpu_caps(unsigned *flags, const char *s);
  */
 int av_cpu_count(void);
 
+/**
+ * Get the maximum data alignment that may be required by FFmpeg.
+ *
+ * Note that this is affected by the build configuration and the CPU flags mask,
+ * so e.g. if the CPU supports AVX, but libavutil has been built with
+ * --disable-avx or the AV_CPU_FLAG_AVX flag has been disabled through
+ *  av_set_cpu_flags_mask(), then this function will behave as if AVX is not
+ *  present.
+ */
+size_t av_cpu_max_align(void);
+
 #endif /* AVUTIL_CPU_H */

libavutil/version.h

@@ -80,7 +80,7 @@
 
 
 #define LIBAVUTIL_VERSION_MAJOR  55
-#define LIBAVUTIL_VERSION_MINOR  76
+#define LIBAVUTIL_VERSION_MINOR  77
 #define LIBAVUTIL_VERSION_MICRO 100
 
 #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \