Replace the aes256gcm implementation with Romain Dolbeau's implementation

which is slightly faster than mine.
Reimplement features from the previous implementation: add batch mode and
use two passes in the decryption function in order to check the tag before
decrypting.

AUTHORS

@@ -32,6 +32,9 @@ scrypt                                 Colin Percival
 Implementors
 ============
 
+crypto_aead/aes256gcm/aesni            Romain Dolbeau
+                                       Frank Denis
+
 crypto_aead/chacha20poly1305           Frank Denis
 
 crypto_box/curve25519xsalsa20poly1305  Daniel J. Bernstein

configure.ac

@@ -211,13 +211,8 @@ AX_CHECK_COMPILE_FLAG([-Wwrite-strings], [CFLAGS="$CFLAGS -Wwrite-strings"])
 AX_CHECK_COMPILE_FLAG([-Wdiv-by-zero], [CFLAGS="$CFLAGS -Wdiv-by-zero"])
 AX_CHECK_COMPILE_FLAG([-Wsometimes-uninitialized], [CFLAGS="$CFLAGS -Wsometimes-uninitialized"])
 
-AX_CHECK_COMPILE_FLAG([$CFLAGS -mmmx], [CFLAGS="$CFLAGS -mmmx"])
-AX_CHECK_COMPILE_FLAG([$CFLAGS -msse], [CFLAGS="$CFLAGS -msse"])
-AX_CHECK_COMPILE_FLAG([$CFLAGS -msse2], [CFLAGS="$CFLAGS -msse2"])
-AX_CHECK_COMPILE_FLAG([$CFLAGS -msse3], [CFLAGS="$CFLAGS -msse3"])
-AX_CHECK_COMPILE_FLAG([$CFLAGS -mssse3], [CFLAGS="$CFLAGS -mssse3"])
-AX_CHECK_COMPILE_FLAG([$CFLAGS -maes], [CFLAGS="$CFLAGS -maes"])
-AX_CHECK_COMPILE_FLAG([$CFLAGS -mpclmul], [CFLAGS="$CFLAGS -mpclmul"])
+AC_MSG_CHECKING([Checking if we can compile for westmere])
+AX_CHECK_COMPILE_FLAG([-march=westmere $CFLAGS], [CFLAGS="-march=westmere $CFLAGS"])
 
 AC_ARG_VAR([CWFLAGS], [define to compilation flags for generating extra warnings])
 

src/libsodium/include/sodium/crypto_aead_aes256gcm_aesni.h

@@ -12,11 +12,24 @@ extern "C" {
 #endif
 
 #define crypto_aead_aes256gcm_KEYBYTES  32U
-#define crypto_aead_aes256gcm_NSECBYTES 0U
-#define crypto_aead_aes256gcm_NPUBBYTES 12U
-#define crypto_aead_aes256gcm_ABYTES    16U
+SODIUM_EXPORT
+size_t crypto_aead_aes256gcm_aesni_keybytes(void);
 
-typedef CRYPTO_ALIGN(128) unsigned char crypto_aead_aes256gcm_aesni_state[384];
+#define crypto_aead_aes256gcm_NSECBYTES 0U
+SODIUM_EXPORT
+size_t crypto_aead_aes256gcm_aesni_nsecbytes(void);
+
+#define crypto_aead_aes256gcm_NPUBBYTES 12U
+SODIUM_EXPORT
+size_t crypto_aead_aes256gcm_aesni_npubbytes(void);
+
+#define crypto_aead_aes256gcm_ABYTES    16U
+SODIUM_EXPORT
+size_t crypto_aead_aes256gcm_aesni_abytes(void);
+
+typedef CRYPTO_ALIGN(16) unsigned char crypto_aead_aes256gcm_aesni_state[272];
+SODIUM_EXPORT
+size_t crypto_aead_aes256gcm_aesni_statebytes(void);
 
 SODIUM_EXPORT
 int crypto_aead_aes256gcm_aesni_encrypt(unsigned char *c,

src/libsodium/include/sodium/export.h

@@ -30,7 +30,7 @@
 #endif
 
 #ifndef CRYPTO_ALIGN
-# if defined(_MSC_VER)
+# if defined(__INTEL_COMPILER) || defined(_MSC_VER)
 #  define CRYPTO_ALIGN(x) __declspec(align(x))
 # else
 #  define CRYPTO_ALIGN(x) __attribute__((aligned(x)))