Modest improvement coefficients mean that code already had some
parallelism and there was not very much room for improvement. Special
thanks to Ted Krovetz for benchmarking the code with such patience.

Submitted by: Antony, Benoy <bantony@ebay.com>
Approved by: steve@openssl.org

Fix bug in AES wrap code when t > 0xff.

PR: 2194

add ghash benchmark to apps/speed.c.

AES_ctr32_encrypt to aes-s390x.pl.

ARM assembler modules.

PR: 2296

opcodes detected in executable segment" error.

PR: 2284

improvement.

Submitted By: Mattias Ellert <mattias.ellert@fysast.uu.se>

Fix type checking macro SKM_ASN1_SET_OF_i2d

Submitted By: Victor Wagner <vitus@cryptocom.ru>

Fix error reporting in load_key function.

Submitted By: De Rudder, Stephen L." <s_derudder@tditx.com>

Workaround for newer Windows headers which define EADDRINUSE but not to the
same value as WSAEADDRINUSE.

Submitted By: Ger Hobbelt <ger@hobbelt.com>

Base64 BIO fixes:

Use OPENSSL_assert() instead of assert().
Use memmove() as buffers overlap.
Fix write retry logic.

Submitted By: Jonathan Gray <jsg@goblin.cx>

Correct ioctl definitions.

in gcm_ghash_4bit. Taking the idea a step further I've added extra
256+16 bytes of per-key storage, so that one can speak about 3rd variant
in addition to "256B" and "4KB": "528B" one. Commonly it should be
~50% faster than "256B" implementation or ~25% slower than "4KB" one.