From: Andy Polyakov <appro@openssl.org>
Date: Mon, 13 May 2013 20:49:58 +0000 (+0200)
Subject: Add AES-SHA256 stitch.
X-Git-Tag: master-post-reformat~1317
X-Git-Url: https://git.openssl.org/gitweb/?p=openssl.git;a=commitdiff_plain;h=8a97a33063d93be1130b762daefd729346af4d29;hp=22de0e6583b7f8764e383cf818db45863281eeec

Add AES-SHA256 stitch.
---

diff --git a/Configure b/Configure
index e2497a4f99..601f8931a6 100755
--- a/Configure
+++ b/Configure
@@ -128,7 +128,7 @@ my $x86_asm="x86cpuid.o:bn-586.o co-586.o x86-mont.o x86-gf2m.o:des-586.o crypt5
 
 my $x86_elf_asm="$x86_asm:elf";
 
-my $x86_64_asm="x86_64cpuid.o:x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o::aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o::md5-x86_64.o:sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o::rc4-x86_64.o rc4-md5-x86_64.o:::wp-x86_64.o:cmll-x86_64.o cmll_misc.o:ghash-x86_64.o aesni-gcm-x86_64.o:e_padlock-x86_64.o";
+my $x86_64_asm="x86_64cpuid.o:x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o::aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o::md5-x86_64.o:sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o::rc4-x86_64.o rc4-md5-x86_64.o:::wp-x86_64.o:cmll-x86_64.o cmll_misc.o:ghash-x86_64.o aesni-gcm-x86_64.o:e_padlock-x86_64.o";
 my $ia64_asm="ia64cpuid.o:bn-ia64.o ia64-mont.o::aes_core.o aes_cbc.o aes-ia64.o::md5-ia64.o:sha1-ia64.o sha256-ia64.o sha512-ia64.o::rc4-ia64.o rc4_skey.o:::::ghash-ia64.o::void";
 my $sparcv9_asm="sparcv9cap.o sparccpuid.o:bn-sparcv9.o sparcv9-mont.o sparcv9a-mont.o vis3-mont.o sparct4-mont.o sparcv9-gf2m.o:des_enc-sparc.o fcrypt_b.o dest4-sparcv9.o:aes_core.o aes_cbc.o aes-sparcv9.o aest4-sparcv9.o::md5-sparcv9.o:sha1-sparcv9.o sha256-sparcv9.o sha512-sparcv9.o::::::camellia.o cmll_misc.o cmll_cbc.o cmllt4-sparcv9.o:ghash-sparcv9.o::void";
 my $sparcv8_asm=":sparcv8.o:des_enc-sparc.o fcrypt_b.o:::::::::::::void";
diff --git a/TABLE b/TABLE
index 46d148fb4f..1e90ee37d5 100644
--- a/TABLE
+++ b/TABLE
@@ -308,7 +308,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -803,7 +803,7 @@ $bn_ops       = SIXTY_FOUR_BIT RC4_CHUNK_LL DES_INT EXPORT_VAR_AS_FN
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = bn_asm.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -1496,7 +1496,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -1661,7 +1661,7 @@ $bn_ops       = SIXTY_FOUR_BIT RC4_CHUNK_LL DES_INT EXPORT_VAR_AS_FN
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = bn_asm.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -1760,7 +1760,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -1826,7 +1826,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -2024,7 +2024,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -2552,7 +2552,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -2750,7 +2750,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -2816,7 +2816,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -4466,7 +4466,7 @@ $bn_ops       = SIXTY_FOUR_BIT RC4_CHUNK_LL DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -4499,7 +4499,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -4532,7 +4532,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -4730,7 +4730,7 @@ $bn_ops       = SIXTY_FOUR_BIT RC4_CHUNK_LL DES_INT EXPORT_VAR_AS_FN
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -5720,7 +5720,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
@@ -5753,7 +5753,7 @@ $bn_ops       = SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL
 $cpuid_obj    = x86_64cpuid.o
 $bn_obj       = x86_64-gcc.o x86_64-mont.o x86_64-mont5.o x86_64-gf2m.o modexp512-x86_64.o
 $des_obj      = 
-$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o
+$aes_obj      = aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o aesni-x86_64.o aesni-sha1-x86_64.o aesni-sha256-x86_64.o
 $bf_obj       = 
 $md5_obj      = md5-x86_64.o
 $sha1_obj     = sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o
diff --git a/crypto/aes/Makefile b/crypto/aes/Makefile
index e185ae75db..d2069e46d3 100644
--- a/crypto/aes/Makefile
+++ b/crypto/aes/Makefile
@@ -65,6 +65,8 @@ aesni-x86_64.s: asm/aesni-x86_64.pl
 	$(PERL) asm/aesni-x86_64.pl $(PERLASM_SCHEME) > $@
 aesni-sha1-x86_64.s:	asm/aesni-sha1-x86_64.pl
 	$(PERL) asm/aesni-sha1-x86_64.pl $(PERLASM_SCHEME) > $@
+aesni-sha256-x86_64.s:	asm/aesni-sha256-x86_64.pl
+	$(PERL) asm/aesni-sha256-x86_64.pl $(PERLASM_SCHEME) > $@
 
 aes-sparcv9.s: asm/aes-sparcv9.pl
 	$(PERL) asm/aes-sparcv9.pl $(CFLAGS) > $@
diff --git a/crypto/aes/asm/aesni-sha256-x86_64.pl b/crypto/aes/asm/aesni-sha256-x86_64.pl
new file mode 100644
index 0000000000..ebce021452
--- /dev/null
+++ b/crypto/aes/asm/aesni-sha256-x86_64.pl
@@ -0,0 +1,1358 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# January 2013
+#
+# This is AESNI-CBC+SHA256 stitch implementation. The idea, as spelled
+# in http://download.intel.com/design/intarch/papers/323686.pdf, is
+# that since AESNI-CBC encrypt exhibit *very* low instruction-level
+# parallelism, interleaving it with another algorithm would allow to
+# utilize processor resources better and achieve better performance.
+# SHA256 instruction sequences(*) are taken from sha512-x86_64.pl and
+# AESNI code is weaved into it. As SHA256 dominates execution time,
+# stitch performance does not depend on AES key length. Below are
+# performance numbers in cycles per processed byte, less is better,
+# for standalone AESNI-CBC encrypt, standalone SHA256, and stitched
+# subroutine:
+#
+#		 AES-128/-192/-256+SHA256	this(**)gain
+# Sandy Bridge	    5.05/6.05/7.05+11.6		13.0	+28%/36%/43%
+# Ivy Bridge	    5.05/6.05/7.05+10.3		11.6	+32%/41%/50%
+# Bulldozer	    5.77/6.89/8.00+13.7		13.7	+42%/50%/58%
+#
+# (*)	there are XOP, AVX1 and AVX2 code pathes, meaning that
+#	Westmere is omitted from loop, this is because gain was not
+#	estimated high enough to justify the effort;
+# (**)	these are EVP-free results, results obtained with 'speed
+#	-evp aes-256-cbc-hmac-sha256' will vary by percent or two;
+
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+		=~ /GNU assembler version ([2-9]\.[0-9]+)/) {
+	$avx = ($1>=2.19) + ($1>=2.22);
+}
+
+if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
+	   `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
+	$avx = ($1>=2.09) + ($1>=2.10);
+}
+
+if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
+	   `ml64 2>&1` =~ /Version ([0-9]+)\./) {
+	$avx = ($1>=10) + ($1>=11);
+}
+
+open OUT,"| \"$^X\" $xlate $flavour $output";
+*STDOUT=*OUT;
+
+$func="aesni_cbc_sha256_enc";
+$TABLE="K256";
+$SZ=4;
+@ROT=($A,$B,$C,$D,$E,$F,$G,$H)=("%eax","%ebx","%ecx","%edx",
+				"%r8d","%r9d","%r10d","%r11d");
+($T1,$a0,$a1,$a2,$a3)=("%r12d","%r13d","%r14d","%r15d","%esi");
+@Sigma0=( 2,13,22);
+@Sigma1=( 6,11,25);
+@sigma0=( 7,18, 3);
+@sigma1=(17,19,10);
+$rounds=64;
+
+########################################################################
+# void aesni_cbc_sha256_enc(const void *inp,
+#			void *out,
+#			size_t length,
+#			const AES_KEY *key,
+#			unsigned char *iv,
+#			SHA256_CTX *ctx,
+#			const void *in0);
+($inp,  $out,  $len,  $key,  $ivp, $ctx, $in0) =
+("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10");
+
+$Tbl="%rbp";
+
+$_inp="16*$SZ+0*8(%rsp)";
+$_out="16*$SZ+1*8(%rsp)";
+$_end="16*$SZ+2*8(%rsp)";
+$_key="16*$SZ+3*8(%rsp)";
+$_ivp="16*$SZ+4*8(%rsp)";
+$_ctx="16*$SZ+5*8(%rsp)";
+$_in0="16*$SZ+6*8(%rsp)";
+$_rsp="16*$SZ+7*8(%rsp)";
+$framesz=16*$SZ+8*8;
+
+$code=<<___;
+.text
+
+.extern	OPENSSL_ia32cap_P
+.globl	$func
+.type	$func,\@abi-omnipotent
+.align	16
+$func:
+___
+$code.=<<___ if ($avx);
+	lea	OPENSSL_ia32cap_P(%rip),%r11
+	mov	\$1,%eax
+	cmp	\$0,`$win64?"%rcx":"%rdi"`
+	je	.Lprobe
+	mov	0(%r11),%eax
+	mov	4(%r11),%r10d
+	mov	8(%r11),%r11d
+
+	test	\$`1<<11`,%r10d			# check for XOP
+	jnz	${func}_xop
+___
+$code.=<<___ if ($avx>1);
+	and	\$`1<<8|1<<5|1<<3`,%r11d	# check for BMI2+AVX2+BMI1
+	cmp	\$`1<<8|1<<5|1<<3`,%r11d
+	je	${func}_avx2
+___
+$code.=<<___ if ($avx);
+	and	\$`1<<30`,%eax			# mask "Intel CPU" bit
+	and	\$`1<<28|1<<9`,%r10d		# mask AVX+SSSE3 bits
+	or	%eax,%r10d
+	cmp	\$`1<<28|1<<9|1<<30`,%r10d
+	je	${func}_avx
+	ud2
+___
+$code.=<<___;
+	xor	%eax,%eax
+	cmp	\$0,`$win64?"%rcx":"%rdi"`
+	je	.Lprobe
+	ud2
+.Lprobe:
+	ret
+.size	$func,.-$func
+
+.align	64
+.type	$TABLE,\@object
+$TABLE:
+	.long	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
+	.long	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
+	.long	0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
+	.long	0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
+	.long	0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
+	.long	0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
+	.long	0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
+	.long	0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
+	.long	0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
+	.long	0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
+	.long	0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
+	.long	0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
+	.long	0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
+	.long	0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
+	.long	0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
+	.long	0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
+	.long	0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
+	.long	0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
+	.long	0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
+	.long	0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
+	.long	0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
+	.long	0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
+	.long	0xd192e819,0xd6990624,0xf40e3585,0x106aa070
+	.long	0xd192e819,0xd6990624,0xf40e3585,0x106aa070
+	.long	0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
+	.long	0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
+	.long	0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
+	.long	0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
+	.long	0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
+	.long	0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
+	.long	0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
+	.long	0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
+
+	.long	0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f
+	.long	0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f
+	.long	0,0,0,0,   0,0,0,0,   -1,-1,-1,-1
+	.long	0,0,0,0,   0,0,0,0
+	.asciz	"AESNI-CBC+SHA256 stitch for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align	64
+___
+
+######################################################################
+# SIMD code paths
+#
+{{{
+($iv,$inout,$roundkey,$temp,
+ $mask10,$mask12,$mask14,$offload)=map("%xmm$_",(8..15));
+
+$aesni_cbc_idx=0;
+@aesni_cbc_block = (
+##	&vmovdqu	($roundkey,"0x00-0x80($inp)");'
+##	&vmovdqu	($inout,($inp));
+##	&mov		($_inp,$inp);
+
+	'&vpxor		($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x10-0x80($inp)");',
+
+	'&vpxor		($inout,$inout,$iv);',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x20-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x30-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x40-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x50-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x60-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x70-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x80-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x90-0x80($inp)");',
+
+	'&vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0xa0-0x80($inp)");',
+
+	'&vaesenclast	($temp,$inout,$roundkey);'.
+	' &vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0xb0-0x80($inp)");',
+
+	'&vpand		($iv,$temp,$mask10);'.
+	' &vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0xc0-0x80($inp)");',
+
+	'&vaesenclast	($temp,$inout,$roundkey);'.
+	' &vaesenc	($inout,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0xd0-0x80($inp)");',
+
+	'&vpand		($temp,$temp,$mask12);'.
+	' &vaesenc	($inout,$inout,$roundkey);'.
+	 '&vmovdqu	($roundkey,"0xe0-0x80($inp)");',
+
+	'&vpor		($iv,$iv,$temp);'.
+	' &vaesenclast	($temp,$inout,$roundkey);'.
+	' &vmovdqu	($roundkey,"0x00-0x80($inp)");'
+
+##	&mov		($inp,$_inp);
+##	&mov		($out,$_out);
+##	&vpand		($temp,$temp,$mask14);
+##	&vpor		($iv,$iv,$temp);
+##	&vmovdqu	($iv,($out,$inp);
+##	&lea		(inp,16($inp));
+);
+
+my $a4=$T1;
+my ($a,$b,$c,$d,$e,$f,$g,$h);
+
+sub AUTOLOAD()		# thunk [simplified] 32-bit style perlasm
+{ my $opcode = $AUTOLOAD; $opcode =~ s/.*:://;
+  my $arg = pop;
+    $arg = "\$$arg" if ($arg*1 eq $arg);
+    $code .= "\t$opcode\t".join(',',$arg,reverse @_)."\n";
+}
+
+sub body_00_15 () {
+	(
+	'($a,$b,$c,$d,$e,$f,$g,$h)=@ROT;'.
+
+	'&ror	($a0,$Sigma1[2]-$Sigma1[1])',
+	'&mov	($a,$a1)',
+	'&mov	($a4,$f)',
+
+	'&xor	($a0,$e)',
+	'&ror	($a1,$Sigma0[2]-$Sigma0[1])',
+	'&xor	($a4,$g)',			# f^g
+
+	'&ror	($a0,$Sigma1[1]-$Sigma1[0])',
+	'&xor	($a1,$a)',
+	'&and	($a4,$e)',			# (f^g)&e
+
+	@aesni_cbc_block[$aesni_cbc_idx++].
+	'&xor	($a0,$e)',
+	'&add	($h,$SZ*($i&15)."(%rsp)")',	# h+=X[i]+K[i]
+	'&mov	($a2,$a)',
+
+	'&ror	($a1,$Sigma0[1]-$Sigma0[0])',
+	'&xor	($a4,$g)',			# Ch(e,f,g)=((f^g)&e)^g
+	'&xor	($a2,$b)',			# a^b, b^c in next round
+
+	'&ror	($a0,$Sigma1[0])',		# Sigma1(e)
+	'&add	($h,$a4)',			# h+=Ch(e,f,g)
+	'&and	($a3,$a2)',			# (b^c)&(a^b)
+
+	'&xor	($a1,$a)',
+	'&add	($h,$a0)',			# h+=Sigma1(e)
+	'&xor	($a3,$b)',			# Maj(a,b,c)=Ch(a^b,c,b)
+
+	'&add	($d,$h)',			# d+=h
+	'&ror	($a1,$Sigma0[0])',		# Sigma0(a)
+	'&add	($h,$a3)',			# h+=Maj(a,b,c)
+
+	'&mov	($a0,$d)',
+	'&add	($a1,$h);'.			# h+=Sigma0(a)
+	'($a2,$a3) = ($a3,$a2); unshift(@ROT,pop(@ROT)); $i++;'
+	);
+}
+
+if ($avx) {{
+######################################################################
+# XOP code path
+#
+$code.=<<___;
+.type	${func}_xop,\@function,6
+.align	64
+${func}_xop:
+.Lxop_shortcut:
+	mov	`($win64?56:8)`(%rsp),$in0	# load 7th parameter
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	mov	%rsp,%r11		# copy %rsp
+	sub	\$`$framesz+$win64*16*10`,%rsp
+	and	\$-64,%rsp		# align stack frame
+
+	shl	\$6,$len
+	sub	$inp,$out		# re-bias
+	sub	$inp,$in0
+	add	$inp,$len		# end of input
+
+	#mov	$inp,$_inp		# saved later
+	mov	$out,$_out
+	mov	$len,$_end
+	#mov	$key,$_key		# remains resident in $inp register
+	mov	$ivp,$_ivp
+	mov	$ctx,$_ctx
+	mov	$in0,$_in0
+	mov	%r11,$_rsp
+___
+$code.=<<___ if ($win64);
+	movaps	%xmm6,`$framesz+16*0`(%rsp)
+	movaps	%xmm7,`$framesz+16*1`(%rsp)
+	movaps	%xmm8,`$framesz+16*2`(%rsp)
+	movaps	%xmm9,`$framesz+16*3`(%rsp)
+	movaps	%xmm10,`$framesz+16*4`(%rsp)
+	movaps	%xmm11,`$framesz+16*5`(%rsp)
+	movaps	%xmm12,`$framesz+16*6`(%rsp)
+	movaps	%xmm13,`$framesz+16*7`(%rsp)
+	movaps	%xmm14,`$framesz+16*8`(%rsp)
+	movaps	%xmm15,`$framesz+16*9`(%rsp)
+___
+$code.=<<___;
+.Lprologue_xop:
+	vzeroall
+
+	mov	$inp,%r12		# borrow $a4
+	lea	0x80($key),$inp		# size optimization, reassign
+	lea	$TABLE+`$SZ*2*$rounds+32`(%rip),%r13	# borrow $a0
+	mov	0xf0-0x80($inp),%r14d	# rounds, borrow $a1
+	mov	$ctx,%r15		# borrow $a2
+	mov	$in0,%rsi		# borrow $a3
+	vmovdqu	($ivp),$iv		# load IV
+	sub	\$9,%r14
+
+	mov	$SZ*0(%r15),$A
+	mov	$SZ*1(%r15),$B
+	mov	$SZ*2(%r15),$C
+	mov	$SZ*3(%r15),$D
+	mov	$SZ*4(%r15),$E
+	mov	$SZ*5(%r15),$F
+	mov	$SZ*6(%r15),$G
+	mov	$SZ*7(%r15),$H
+
+	vmovdqa	0x00(%r13,%r14,8),$mask14
+	vmovdqa	0x10(%r13,%r14,8),$mask12
+	vmovdqa	0x20(%r13,%r14,8),$mask10
+	vmovdqu	0x00-0x80($inp),$roundkey
+	jmp	.Lloop_xop
+___
+					if ($SZ==4) {	# SHA256
+    my @X = map("%xmm$_",(0..3));
+    my ($t0,$t1,$t2,$t3) = map("%xmm$_",(4..7));
+
+$code.=<<___;
+.align	16
+.Lloop_xop:
+	vmovdqa	$TABLE+`$SZ*2*$rounds`(%rip),$t3
+	vmovdqu	0x00(%rsi,%r12),@X[0]
+	vmovdqu	0x10(%rsi,%r12),@X[1]
+	vmovdqu	0x20(%rsi,%r12),@X[2]
+	vmovdqu	0x30(%rsi,%r12),@X[3]
+	vpshufb	$t3,@X[0],@X[0]
+	lea	$TABLE(%rip),$Tbl
+	vpshufb	$t3,@X[1],@X[1]
+	vpshufb	$t3,@X[2],@X[2]
+	vpaddd	0x00($Tbl),@X[0],$t0
+	vpshufb	$t3,@X[3],@X[3]
+	vpaddd	0x20($Tbl),@X[1],$t1
+	vpaddd	0x40($Tbl),@X[2],$t2
+	vpaddd	0x60($Tbl),@X[3],$t3
+	vmovdqa	$t0,0x00(%rsp)
+	mov	$A,$a1
+	vmovdqa	$t1,0x10(%rsp)
+	mov	$B,$a3
+	vmovdqa	$t2,0x20(%rsp)
+	xor	$C,$a3			# magic
+	vmovdqa	$t3,0x30(%rsp)
+	mov	$E,$a0
+	jmp	.Lxop_00_47
+
+.align	16
+.Lxop_00_47:
+	sub	\$-16*2*$SZ,$Tbl	# size optimization
+	vmovdqu	(%r12),$inout		# $a4
+	mov	%r12,$_inp		# $a4
+___
+sub XOP_256_00_47 () {
+my $j = shift;
+my $body = shift;
+my @X = @_;
+my @insns = (&$body,&$body,&$body,&$body);	# 104 instructions
+
+	&vpalignr	($t0,@X[1],@X[0],$SZ);	# X[1..4]
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpalignr	($t3,@X[3],@X[2],$SZ);	# X[9..12]
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vprotd		($t1,$t0,8*$SZ-$sigma0[1]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpsrld		($t0,$t0,$sigma0[2]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpaddd	(@X[0],@X[0],$t3);	# X[0..3] += X[9..12]
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vprotd		($t2,$t1,$sigma0[1]-$sigma0[0]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpxor		($t0,$t0,$t1);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vprotd	($t3,@X[3],8*$SZ-$sigma1[1]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpxor		($t0,$t0,$t2);		# sigma0(X[1..4])
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpsrld	($t2,@X[3],$sigma1[2]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpaddd		(@X[0],@X[0],$t0);	# X[0..3] += sigma0(X[1..4])
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vprotd	($t1,$t3,$sigma1[1]-$sigma1[0]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpxor		($t3,$t3,$t2);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpxor		($t3,$t3,$t1);		# sigma1(X[14..15])
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpsrldq	($t3,$t3,8);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpaddd		(@X[0],@X[0],$t3);	# X[0..1] += sigma1(X[14..15])
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vprotd	($t3,@X[0],8*$SZ-$sigma1[1]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpsrld	($t2,@X[0],$sigma1[2]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vprotd	($t1,$t3,$sigma1[1]-$sigma1[0]);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpxor		($t3,$t3,$t2);
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	 &vpxor		($t3,$t3,$t1);		# sigma1(X[16..17])
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpslldq	($t3,$t3,8);		# 22 instructions
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpaddd		(@X[0],@X[0],$t3);	# X[2..3] += sigma1(X[16..17])
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	  eval(shift(@insns));
+	&vpaddd		($t2,@X[0],16*2*$j."($Tbl)");
+	  foreach (@insns) { eval; }		# remaining instructions
+	&vmovdqa	(16*$j."(%rsp)",$t2);
+}
+
+    $aesni_cbc_idx=0;
+    for ($i=0,$j=0; $j<4; $j++) {
+	&XOP_256_00_47($j,\&body_00_15,@X);
+	push(@X,shift(@X));			# rotate(@X)
+    }
+    	&mov		("%r12",$_inp);		# borrow $a4
+	&vpand		($temp,$temp,$mask14);
+	&mov		("%r15",$_out);		# borrow $a2
+	&vpor		($iv,$iv,$temp);
+	&vmovdqu	("(%r15,%r12)",$iv);	# write output
+	&lea		("%r12","16(%r12)");	# inp++
+
+	&cmpb	($SZ-1+16*2*$SZ."($Tbl)",0);
+	&jne	(".Lxop_00_47");
+
+	&vmovdqu	($inout,"(%r12)");
+	&mov		($_inp,"%r12");
+
+    $aesni_cbc_idx=0;
+    for ($i=0; $i<16; ) {
+	foreach(body_00_15()) { eval; }
+    }
+					}
+$code.=<<___;
+	mov	$_inp,%r12		# borrow $a4
+	mov	$_out,%r13		# borrow $a0
+	mov	$_ctx,%r15		# borrow $a2
+	mov	$_in0,%rsi		# borrow $a3
+
+	vpand	$mask14,$temp,$temp
+	mov	$a1,$A
+	vpor	$temp,$iv,$iv
+	vmovdqu	$iv,(%r13,%r12)		# write output
+	lea	16(%r12),%r12		# inp++
+
+	add	$SZ*0(%r15),$A
+	add	$SZ*1(%r15),$B
+	add	$SZ*2(%r15),$C
+	add	$SZ*3(%r15),$D
+	add	$SZ*4(%r15),$E
+	add	$SZ*5(%r15),$F
+	add	$SZ*6(%r15),$G
+	add	$SZ*7(%r15),$H
+
+	cmp	$_end,%r12
+
+	mov	$A,$SZ*0(%r15)
+	mov	$B,$SZ*1(%r15)
+	mov	$C,$SZ*2(%r15)
+	mov	$D,$SZ*3(%r15)
+	mov	$E,$SZ*4(%r15)
+	mov	$F,$SZ*5(%r15)
+	mov	$G,$SZ*6(%r15)
+	mov	$H,$SZ*7(%r15)
+
+	jb	.Lloop_xop
+
+	mov	$_ivp,$ivp
+	mov	$_rsp,%rsi
+	vmovdqu	$iv,($ivp)		# output IV
+	vzeroall
+___
+$code.=<<___ if ($win64);
+	movaps	`$framesz+16*0`(%rsp),%xmm6
+	movaps	`$framesz+16*1`(%rsp),%xmm7
+	movaps	`$framesz+16*2`(%rsp),%xmm8
+	movaps	`$framesz+16*3`(%rsp),%xmm9
+	movaps	`$framesz+16*4`(%rsp),%xmm10
+	movaps	`$framesz+16*5`(%rsp),%xmm11
+	movaps	`$framesz+16*6`(%rsp),%xmm12
+	movaps	`$framesz+16*7`(%rsp),%xmm13
+	movaps	`$framesz+16*8`(%rsp),%xmm14
+	movaps	`$framesz+16*9`(%rsp),%xmm15
+___
+$code.=<<___;
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lepilogue_xop:
+	ret
+.size	${func}_xop,.-${func}_xop
+___
+######################################################################
+# AVX+shrd code path
+#
+local *ror = sub { &shrd(@_[0],@_) };
+
+$code.=<<___;
+.type	${func}_avx,\@function,6
+.align	64
+${func}_avx:
+.Lavx_shortcut:
+	mov	`($win64?56:8)`(%rsp),$in0	# load 7th parameter
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	mov	%rsp,%r11		# copy %rsp
+	sub	\$`$framesz+$win64*16*10`,%rsp
+	and	\$-64,%rsp		# align stack frame
+
+	shl	\$6,$len
+	sub	$inp,$out		# re-bias
+	sub	$inp,$in0
+	add	$inp,$len		# end of input
+
+	#mov	$inp,$_inp		# saved later
+	mov	$out,$_out
+	mov	$len,$_end
+	#mov	$key,$_key		# remains resident in $inp register
+	mov	$ivp,$_ivp
+	mov	$ctx,$_ctx
+	mov	$in0,$_in0
+	mov	%r11,$_rsp
+___
+$code.=<<___ if ($win64);
+	movaps	%xmm6,`$framesz+16*0`(%rsp)
+	movaps	%xmm7,`$framesz+16*1`(%rsp)
+	movaps	%xmm8,`$framesz+16*2`(%rsp)
+	movaps	%xmm9,`$framesz+16*3`(%rsp)
+	movaps	%xmm10,`$framesz+16*4`(%rsp)
+	movaps	%xmm11,`$framesz+16*5`(%rsp)
+	movaps	%xmm12,`$framesz+16*6`(%rsp)
+	movaps	%xmm13,`$framesz+16*7`(%rsp)
+	movaps	%xmm14,`$framesz+16*8`(%rsp)
+	movaps	%xmm15,`$framesz+16*9`(%rsp)
+___
+$code.=<<___;
+.Lprologue_avx:
+	vzeroall
+
+	mov	$inp,%r12		# borrow $a4
+	lea	0x80($key),$inp		# size optimization, reassign
+	lea	$TABLE+`$SZ*2*$rounds+32`(%rip),%r13	# borrow $a0
+	mov	0xf0-0x80($inp),%r14d	# rounds, borrow $a1
+	mov	$ctx,%r15		# borrow $a2
+	mov	$in0,%rsi		# borrow $a3
+	vmovdqu	($ivp),$iv		# load IV
+	sub	\$9,%r14
+
+	mov	$SZ*0(%r15),$A
+	mov	$SZ*1(%r15),$B
+	mov	$SZ*2(%r15),$C
+	mov	$SZ*3(%r15),$D
+	mov	$SZ*4(%r15),$E
+	mov	$SZ*5(%r15),$F
+	mov	$SZ*6(%r15),$G
+	mov	$SZ*7(%r15),$H
+
+	vmovdqa	0x00(%r13,%r14,8),$mask14
+	vmovdqa	0x10(%r13,%r14,8),$mask12
+	vmovdqa	0x20(%r13,%r14,8),$mask10
+	vmovdqu	0x00-0x80($inp),$roundkey
+___
+					if ($SZ==4) {	# SHA256
+    my @X = map("%xmm$_",(0..3));
+    my ($t0,$t1,$t2,$t3) = map("%xmm$_",(4..7));
+
+$code.=<<___;
+	jmp	.Lloop_avx
+.align	16
+.Lloop_avx:
+	vmovdqa	$TABLE+`$SZ*2*$rounds`(%rip),$t3
+	vmovdqu	0x00(%rsi,%r12),@X[0]
+	vmovdqu	0x10(%rsi,%r12),@X[1]
+	vmovdqu	0x20(%rsi,%r12),@X[2]
+	vmovdqu	0x30(%rsi,%r12),@X[3]
+	vpshufb	$t3,@X[0],@X[0]
+	lea	$TABLE(%rip),$Tbl
+	vpshufb	$t3,@X[1],@X[1]
+	vpshufb	$t3,@X[2],@X[2]
+	vpaddd	0x00($Tbl),@X[0],$t0
+	vpshufb	$t3,@X[3],@X[3]
+	vpaddd	0x20($Tbl),@X[1],$t1
+	vpaddd	0x40($Tbl),@X[2],$t2
+	vpaddd	0x60($Tbl),@X[3],$t3
+	vmovdqa	$t0,0x00(%rsp)
+	mov	$A,$a1
+	vmovdqa	$t1,0x10(%rsp)
+	mov	$B,$a3
+	vmovdqa	$t2,0x20(%rsp)
+	xor	$C,$a3			# magic
+	vmovdqa	$t3,0x30(%rsp)
+	mov	$E,$a0
+	jmp	.Lavx_00_47
+
+.align	16
+.Lavx_00_47:
+	sub	\$-16*2*$SZ,$Tbl	# size optimization
+	vmovdqu	(%r12),$inout		# $a4
+	mov	%r12,$_inp		# $a4
+___
+sub Xupdate_256_AVX () {
+	(
+	'&vpalignr	($t0,@X[1],@X[0],$SZ)',	# X[1..4]
+	 '&vpalignr	($t3,@X[3],@X[2],$SZ)',	# X[9..12]
+	'&vpsrld	($t2,$t0,$sigma0[0]);',
+	 '&vpaddd	(@X[0],@X[0],$t3)',	# X[0..3] += X[9..12]
+	'&vpsrld	($t3,$t0,$sigma0[2])',
+	'&vpslld	($t1,$t0,8*$SZ-$sigma0[1]);',
+	'&vpxor		($t0,$t3,$t2)',
+	 '&vpshufd	($t3,@X[3],0b11111010)',# X[14..15]
+	'&vpsrld	($t2,$t2,$sigma0[1]-$sigma0[0]);',
+	'&vpxor		($t0,$t0,$t1)',
+	'&vpslld	($t1,$t1,$sigma0[1]-$sigma0[0]);',
+	'&vpxor		($t0,$t0,$t2)',
+	 '&vpsrld	($t2,$t3,$sigma1[2]);',
+	'&vpxor		($t0,$t0,$t1)',		# sigma0(X[1..4])
+	 '&vpsrlq	($t3,$t3,$sigma1[0]);',
+	'&vpaddd	(@X[0],@X[0],$t0)',	# X[0..3] += sigma0(X[1..4])
+	 '&vpxor	($t2,$t2,$t3);',
+	 '&vpsrlq	($t3,$t3,$sigma1[1]-$sigma1[0])',
+	 '&vpxor	($t2,$t2,$t3)',		# sigma1(X[14..15])
+	 '&vpshufd	($t2,$t2,0b10000100)',
+	 '&vpsrldq	($t2,$t2,8)',
+	'&vpaddd	(@X[0],@X[0],$t2)',	# X[0..1] += sigma1(X[14..15])
+	 '&vpshufd	($t3,@X[0],0b01010000)',# X[16..17]
+	 '&vpsrld	($t2,$t3,$sigma1[2])',
+	 '&vpsrlq	($t3,$t3,$sigma1[0])',
+	 '&vpxor	($t2,$t2,$t3);',
+	 '&vpsrlq	($t3,$t3,$sigma1[1]-$sigma1[0])',
+	 '&vpxor	($t2,$t2,$t3)',
+	 '&vpshufd	($t2,$t2,0b11101000)',
+	 '&vpslldq	($t2,$t2,8)',
+	'&vpaddd	(@X[0],@X[0],$t2)'	# X[2..3] += sigma1(X[16..17])
+	);
+}
+
+sub AVX_256_00_47 () {
+my $j = shift;
+my $body = shift;
+my @X = @_;
+my @insns = (&$body,&$body,&$body,&$body);	# 104 instructions
+
+	foreach (Xupdate_256_AVX()) {		# 29 instructions
+	    eval;
+	    eval(shift(@insns));
+	    eval(shift(@insns));
+	    eval(shift(@insns));
+	}
+	&vpaddd		($t2,@X[0],16*2*$j."($Tbl)");
+	  foreach (@insns) { eval; }		# remaining instructions
+	&vmovdqa	(16*$j."(%rsp)",$t2);
+}
+
+    $aesni_cbc_idx=0;
+    for ($i=0,$j=0; $j<4; $j++) {
+	&AVX_256_00_47($j,\&body_00_15,@X);
+	push(@X,shift(@X));			# rotate(@X)
+    }
+    	&mov		("%r12",$_inp);		# borrow $a4
+	&vpand		($temp,$temp,$mask14);
+	&mov		("%r15",$_out);		# borrow $a2
+	&vpor		($iv,$iv,$temp);
+	&vmovdqu	("(%r15,%r12)",$iv);	# write output
+	&lea		("%r12","16(%r12)");	# inp++
+
+	&cmpb	($SZ-1+16*2*$SZ."($Tbl)",0);
+	&jne	(".Lavx_00_47");
+
+	&vmovdqu	($inout,"(%r12)");
+	&mov		($_inp,"%r12");
+
+    $aesni_cbc_idx=0;
+    for ($i=0; $i<16; ) {
+	foreach(body_00_15()) { eval; }
+    }
+
+					}
+$code.=<<___;
+	mov	$_inp,%r12		# borrow $a4
+	mov	$_out,%r13		# borrow $a0
+	mov	$_ctx,%r15		# borrow $a2
+	mov	$_in0,%rsi		# borrow $a3
+
+	vpand	$mask14,$temp,$temp
+	mov	$a1,$A
+	vpor	$temp,$iv,$iv
+	vmovdqu	$iv,(%r13,%r12)		# write output
+	lea	16(%r12),%r12		# inp++
+
+	add	$SZ*0(%r15),$A
+	add	$SZ*1(%r15),$B
+	add	$SZ*2(%r15),$C
+	add	$SZ*3(%r15),$D
+	add	$SZ*4(%r15),$E
+	add	$SZ*5(%r15),$F
+	add	$SZ*6(%r15),$G
+	add	$SZ*7(%r15),$H
+
+	cmp	$_end,%r12
+
+	mov	$A,$SZ*0(%r15)
+	mov	$B,$SZ*1(%r15)
+	mov	$C,$SZ*2(%r15)
+	mov	$D,$SZ*3(%r15)
+	mov	$E,$SZ*4(%r15)
+	mov	$F,$SZ*5(%r15)
+	mov	$G,$SZ*6(%r15)
+	mov	$H,$SZ*7(%r15)
+	jb	.Lloop_avx
+
+	mov	$_ivp,$ivp
+	mov	$_rsp,%rsi
+	vmovdqu	$iv,($ivp)		# output IV
+	vzeroall
+___
+$code.=<<___ if ($win64);
+	movaps	`$framesz+16*0`(%rsp),%xmm6
+	movaps	`$framesz+16*1`(%rsp),%xmm7
+	movaps	`$framesz+16*2`(%rsp),%xmm8
+	movaps	`$framesz+16*3`(%rsp),%xmm9
+	movaps	`$framesz+16*4`(%rsp),%xmm10
+	movaps	`$framesz+16*5`(%rsp),%xmm11
+	movaps	`$framesz+16*6`(%rsp),%xmm12
+	movaps	`$framesz+16*7`(%rsp),%xmm13
+	movaps	`$framesz+16*8`(%rsp),%xmm14
+	movaps	`$framesz+16*9`(%rsp),%xmm15
+___
+$code.=<<___;
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lepilogue_avx:
+	ret
+.size	${func}_avx,.-${func}_avx
+___
+
+if ($avx>1) {{
+######################################################################
+# AVX2+BMI code path
+#
+my $a5=$SZ==4?"%esi":"%rsi";	# zap $inp 
+my $PUSH8=8*2*$SZ;
+use integer;
+
+sub bodyx_00_15 () {
+	# at start $a1 should be zero, $a3 - $b^$c and $a4 copy of $f
+	(
+	'($a,$b,$c,$d,$e,$f,$g,$h)=@ROT;'.
+
+	'&add	($h,(32*($i/(16/$SZ))+$SZ*($i%(16/$SZ)))%$PUSH8.$base)',    # h+=X[i]+K[i]
+	'&and	($a4,$e)',		# f&e
+	'&rorx	($a0,$e,$Sigma1[2])',
+	'&rorx	($a2,$e,$Sigma1[1])',
+
+	'&lea	($a,"($a,$a1)")',	# h+=Sigma0(a) from the past
+	'&lea	($h,"($h,$a4)")',
+	'&andn	($a4,$e,$g)',		# ~e&g
+	'&xor	($a0,$a2)',
+
+	'&rorx	($a1,$e,$Sigma1[0])',
+	'&lea	($h,"($h,$a4)")',	# h+=Ch(e,f,g)=(e&f)+(~e&g)
+	'&xor	($a0,$a1)',		# Sigma1(e)
+	'&mov	($a2,$a)',
+
+	'&rorx	($a4,$a,$Sigma0[2])',
+	'&lea	($h,"($h,$a0)")',	# h+=Sigma1(e)
+	'&xor	($a2,$b)',		# a^b, b^c in next round
+	'&rorx	($a1,$a,$Sigma0[1])',
+
+	'&rorx	($a0,$a,$Sigma0[0])',
+	'&lea	($d,"($d,$h)")',	# d+=h
+	'&and	($a3,$a2)',		# (b^c)&(a^b)
+	@aesni_cbc_block[$aesni_cbc_idx++].
+	'&xor	($a1,$a4)',
+
+	'&xor	($a3,$b)',		# Maj(a,b,c)=Ch(a^b,c,b)
+	'&xor	($a1,$a0)',		# Sigma0(a)
+	'&lea	($h,"($h,$a3)");'.	# h+=Maj(a,b,c)
+	'&mov	($a4,$e)',		# copy of f in future
+
+	'($a2,$a3) = ($a3,$a2); unshift(@ROT,pop(@ROT)); $i++;'
+	);
+	# and at the finish one has to $a+=$a1
+}
+
+$code.=<<___;
+.type	${func}_avx2,\@function,6
+.align	64
+${func}_avx2:
+.Lavx2_shortcut:
+	mov	`($win64?56:8)`(%rsp),$in0	# load 7th parameter
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	mov	%rsp,%r11		# copy %rsp
+	sub	\$`2*$SZ*$rounds+8*8+$win64*16*10`,%rsp
+	and	\$-256*$SZ,%rsp		# align stack frame
+	add	\$`2*$SZ*($rounds-8)`,%rsp
+
+	shl	\$6,$len
+	sub	$inp,$out		# re-bias
+	sub	$inp,$in0
+	add	$inp,$len		# end of input
+
+	#mov	$inp,$_inp		# saved later
+	#mov	$out,$_out		# kept in $offload
+	mov	$len,$_end
+	#mov	$key,$_key		# remains resident in $inp register
+	mov	$ivp,$_ivp
+	mov	$ctx,$_ctx
+	mov	$in0,$_in0
+	mov	%r11,$_rsp
+___
+$code.=<<___ if ($win64);
+	movaps	%xmm6,`$framesz+16*0`(%rsp)
+	movaps	%xmm7,`$framesz+16*1`(%rsp)
+	movaps	%xmm8,`$framesz+16*2`(%rsp)
+	movaps	%xmm9,`$framesz+16*3`(%rsp)
+	movaps	%xmm10,`$framesz+16*4`(%rsp)
+	movaps	%xmm11,`$framesz+16*5`(%rsp)
+	movaps	%xmm12,`$framesz+16*6`(%rsp)
+	movaps	%xmm13,`$framesz+16*7`(%rsp)
+	movaps	%xmm14,`$framesz+16*8`(%rsp)
+	movaps	%xmm15,`$framesz+16*9`(%rsp)
+___
+$code.=<<___;
+.Lprologue_avx2:
+	vzeroall
+
+	mov	$inp,%r13		# borrow $a0
+	vpinsrq	\$1,$out,$offload,$offload
+	lea	0x80($key),$inp		# size optimization, reassign
+	lea	$TABLE+`$SZ*2*$rounds+32`(%rip),%r12	# borrow $a4
+	mov	0xf0-0x80($inp),%r14d	# rounds, borrow $a1
+	mov	$ctx,%r15		# borrow $a2
+	mov	$in0,%rsi		# borrow $a3
+	vmovdqu	($ivp),$iv		# load IV
+	lea	-9(%r14),%r14
+
+	vmovdqa	0x00(%r12,%r14,8),$mask14
+	vmovdqa	0x10(%r12,%r14,8),$mask12
+	vmovdqa	0x20(%r12,%r14,8),$mask10
+
+	sub	\$-16*$SZ,%r13		# inp++, size optimization
+	mov	$SZ*0(%r15),$A
+	xor	%r12,%r12		# borrow $a0
+	mov	$SZ*1(%r15),$B
+	cmp	$len,%r13		# $_end
+	mov	$SZ*2(%r15),$C
+	sete	%r12b
+	mov	$SZ*3(%r15),$D
+	mov	$SZ*4(%r15),$E
+	mov	$SZ*5(%r15),$F
+	mov	$SZ*6(%r15),$G
+	mov	$SZ*7(%r15),$H
+	vmovdqu	0x00-0x80($inp),$roundkey
+___
+					if ($SZ==4) {	# SHA256
+    my @X = map("%ymm$_",(0..3));
+    my ($t0,$t1,$t2,$t3) = map("%ymm$_",(4..7));
+
+$code.=<<___;
+	jmp	.Loop_avx2
+.align	16
+.Loop_avx2:
+	shl	\$`log(16*$SZ)/log(2)`,%r12
+	vmovdqa	$TABLE+`$SZ*2*$rounds`(%rip),$t3
+	neg	%r12
+	vmovdqu	-16*$SZ+0(%rsi,%r13),%xmm0
+	add	%rsi,%r12		# next or same input block
+	vmovdqu	-16*$SZ+16(%rsi,%r13),%xmm1
+	vmovdqu	-16*$SZ+32(%rsi,%r13),%xmm2
+	vmovdqu	-16*$SZ+48(%rsi,%r13),%xmm3
+
+	vinserti128	\$1,(%r12,%r13),@X[0],@X[0]
+	vinserti128	\$1,16(%r12,%r13),@X[1],@X[1]
+	 vpshufb	$t3,@X[0],@X[0]
+	vinserti128	\$1,32(%r12,%r13),@X[2],@X[2]
+	 vpshufb	$t3,@X[1],@X[1]
+	vinserti128	\$1,48(%r12,%r13),@X[3],@X[3]
+
+	lea	$TABLE(%rip),$Tbl
+	vpshufb	$t3,@X[2],@X[2]
+	lea	-16*$SZ(%r13),%r13
+	vpaddd	0x00($Tbl),@X[0],$t0
+	vpshufb	$t3,@X[3],@X[3]
+	vpaddd	0x20($Tbl),@X[1],$t1
+	vpaddd	0x40($Tbl),@X[2],$t2
+	vpaddd	0x60($Tbl),@X[3],$t3
+	vmovdqa	$t0,0x00(%rsp)
+	xor	$a1,$a1
+	vmovdqa	$t1,0x20(%rsp)
+	lea	-$PUSH8(%rsp),%rsp
+	mov	$B,$a3
+	vmovdqa	$t2,0x00(%rsp)
+	xor	$C,$a3			# magic
+	vmovdqa	$t3,0x20(%rsp)
+	mov	$F,$a4
+	sub	\$-16*2*$SZ,$Tbl	# size optimization
+	jmp	.Lavx2_00_47
+
+.align	16
+.Lavx2_00_47:
+	vmovdqu	(%r13),$inout
+	vpinsrq	\$0,%r13,$offload,$offload
+___
+
+sub AVX2_256_00_47 () {
+my $j = shift;
+my $body = shift;
+my @X = @_;
+my @insns = (&$body,&$body,&$body,&$body);	# 96 instructions
+my $base = "+2*$PUSH8(%rsp)";
+
+	&lea	("%rsp","-$PUSH8(%rsp)")	if (($j%2)==0);
+	foreach (Xupdate_256_AVX()) {		# 29 instructions
+	    eval;
+	    eval(shift(@insns));
+	    eval(shift(@insns));
+	    eval(shift(@insns));
+	}
+	&vpaddd		($t2,@X[0],16*2*$j."($Tbl)");
+	  foreach (@insns) { eval; }		# remaining instructions
+	&vmovdqa	((32*$j)%$PUSH8."(%rsp)",$t2);
+}
+    $aesni_cbc_idx=0;
+    for ($i=0,$j=0; $j<4; $j++) {
+	&AVX2_256_00_47($j,\&bodyx_00_15,@X);
+	push(@X,shift(@X));			# rotate(@X)
+    }
+	&vmovq		("%r13",$offload);	# borrow $a0
+	&vpextrq	("%r15",$offload,1);	# borrow $a2
+	&vpand		($temp,$temp,$mask14);
+	&vpor		($iv,$iv,$temp);
+	&vmovdqu	("(%r15,%r13)",$iv);	# write output
+	&lea		("%r13","16(%r13)");	# inp++
+
+	&lea	($Tbl,16*2*$SZ."($Tbl)");
+	&cmpb	(($SZ-1)."($Tbl)",0);
+	&jne	(".Lavx2_00_47");
+
+	&vmovdqu	($inout,"(%r13)");
+	&vpinsrq	($offload,$offload,"%r13",0);
+
+    $aesni_cbc_idx=0;
+    for ($i=0; $i<16; ) {
+	my $base=$i<8?"+$PUSH8(%rsp)":"(%rsp)";
+	foreach(bodyx_00_15()) { eval; }
+    }
+					}
+$code.=<<___;
+	vpextrq	\$1,$offload,%r12		# $_out, borrow $a4
+	vmovq	$offload,%r13			# $_inp, borrow $a0
+	mov	`2*$SZ*$rounds+5*8`(%rsp),%r15	# $_ctx, borrow $a2
+	add	$a1,$A
+	lea	`2*$SZ*($rounds-8)`(%rsp),$Tbl
+
+	vpand	$mask14,$temp,$temp
+	vpor	$temp,$iv,$iv
+	vmovdqu	$iv,(%r12,%r13)			# write output
+	lea	16(%r13),%r13
+
+	add	$SZ*0(%r15),$A
+	add	$SZ*1(%r15),$B
+	add	$SZ*2(%r15),$C
+	add	$SZ*3(%r15),$D
+	add	$SZ*4(%r15),$E
+	add	$SZ*5(%r15),$F
+	add	$SZ*6(%r15),$G
+	add	$SZ*7(%r15),$H
+
+	mov	$A,$SZ*0(%r15)
+	mov	$B,$SZ*1(%r15)
+	mov	$C,$SZ*2(%r15)
+	mov	$D,$SZ*3(%r15)
+	mov	$E,$SZ*4(%r15)
+	mov	$F,$SZ*5(%r15)
+	mov	$G,$SZ*6(%r15)
+	mov	$H,$SZ*7(%r15)
+
+	cmp	`$PUSH8+2*8`($Tbl),%r13		# $_end
+	je	.Ldone_avx2
+
+	xor	$a1,$a1
+	mov	$B,$a3
+	mov	$F,$a4
+	xor	$C,$a3			# magic
+	jmp	.Lower_avx2
+.align	16
+.Lower_avx2:
+	vmovdqu	(%r13),$inout
+	vpinsrq	\$0,%r13,$offload,$offload
+___
+    $aesni_cbc_idx=0;
+    for ($i=0; $i<16; ) {
+	my $base="+16($Tbl)";
+	foreach(bodyx_00_15()) { eval; }
+	&lea	($Tbl,"-$PUSH8($Tbl)")	if ($i==8);
+    }
+$code.=<<___;
+	vmovq	$offload,%r13			# borrow $a0
+	vpextrq	\$1,$offload,%r15		# borrow $a2
+	vpand	$mask14,$temp,$temp
+	vpor	$temp,$iv,$iv
+	lea	-$PUSH8($Tbl),$Tbl
+	vmovdqu	$iv,(%r15,%r13)			# write output
+	lea	16(%r13),%r13			# inp++
+	cmp	%rsp,$Tbl
+	jae	.Lower_avx2
+
+	mov	`2*$SZ*$rounds+5*8`(%rsp),%r15	# $_ctx, borrow $a2
+	lea	16*$SZ(%r13),%r13
+	mov	`2*$SZ*$rounds+6*8`(%rsp),%rsi	# $_in0, borrow $a3
+	add	$a1,$A
+	lea	`2*$SZ*($rounds-8)`(%rsp),%rsp
+
+	add	$SZ*0(%r15),$A
+	add	$SZ*1(%r15),$B
+	add	$SZ*2(%r15),$C
+	add	$SZ*3(%r15),$D
+	add	$SZ*4(%r15),$E
+	add	$SZ*5(%r15),$F
+	add	$SZ*6(%r15),$G
+	xor	%r12,%r12
+	add	$SZ*7(%r15),$H
+
+	cmp	$_end,%r13
+
+	mov	$A,$SZ*0(%r15)
+	mov	$B,$SZ*1(%r15)
+	mov	$C,$SZ*2(%r15)
+	mov	$D,$SZ*3(%r15)
+	mov	$E,$SZ*4(%r15)
+	mov	$F,$SZ*5(%r15)
+	mov	$G,$SZ*6(%r15)
+	mov	$H,$SZ*7(%r15)
+
+	sete	%r12b
+	jbe	.Loop_avx2
+	lea	(%rsp),$Tbl
+
+.Ldone_avx2:
+	lea	($Tbl),%rsp
+	mov	$_ivp,$ivp
+	mov	$_rsp,%rsi
+	vmovdqu	$iv,($ivp)		# output IV
+	vzeroall
+___
+$code.=<<___ if ($win64);
+	movaps	`$framesz+16*0`(%rsp),%xmm6
+	movaps	`$framesz+16*1`(%rsp),%xmm7
+	movaps	`$framesz+16*2`(%rsp),%xmm8
+	movaps	`$framesz+16*3`(%rsp),%xmm9
+	movaps	`$framesz+16*4`(%rsp),%xmm10
+	movaps	`$framesz+16*5`(%rsp),%xmm11
+	movaps	`$framesz+16*6`(%rsp),%xmm12
+	movaps	`$framesz+16*7`(%rsp),%xmm13
+	movaps	`$framesz+16*8`(%rsp),%xmm14
+	movaps	`$framesz+16*9`(%rsp),%xmm15
+___
+$code.=<<___;
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lepilogue_avx2:
+	ret
+.size	${func}_avx2,.-${func}_avx2
+___
+}}
+}}}}}
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___ if ($avx);
+.extern	__imp_RtlVirtualUnwind
+.type	se_handler,\@abi-omnipotent
+.align	16
+se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	mov	8($disp),%rsi		# disp->ImageBase
+	mov	56($disp),%r11		# disp->HanderlData
+
+	mov	0(%r11),%r10d		# HandlerData[0]
+	lea	(%rsi,%r10),%r10	# prologue label
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	mov	4(%r11),%r10d		# HandlerData[1]
+	lea	(%rsi,%r10),%r10	# epilogue label
+	cmp	%r10,%rbx		# context->Rip>=epilogue label
+	jae	.Lin_prologue
+___
+$code.=<<___ if ($avx>1);
+	lea	.Lavx2_shortcut(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<avx2_shortcut
+	jb	.Lnot_in_avx2
+
+	and	\$-256*$SZ,%rax
+	add	\$`2*$SZ*($rounds-8)`,%rax
+.Lnot_in_avx2:
+___
+$code.=<<___;
+	mov	%rax,%rsi		# put aside Rsp
+	mov	16*$SZ+7*8(%rax),%rax	# pull $_rsp
+	lea	48(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx
+	jb	.Lin_prologue		# non-AVX code
+
+	lea	16*$SZ+8*8(%rsi),%rsi	# Xmm6- save area
+	lea	512($context),%rdi	# &context.Xmm6
+	mov	\$20,%ecx
+	.long	0xa548f3fc		# cld; rep movsq
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	se_handler,.-se_handler
+
+.section	.pdata
+	.rva	.LSEH_begin_${func}_xop
+	.rva	.LSEH_end_${func}_xop
+	.rva	.LSEH_info_${func}_xop
+
+	.rva	.LSEH_begin_${func}_avx
+	.rva	.LSEH_end_${func}_avx
+	.rva	.LSEH_info_${func}_avx
+___
+$code.=<<___ if ($avx>1);
+	.rva	.LSEH_begin_${func}_avx2
+	.rva	.LSEH_end_${func}_avx2
+	.rva	.LSEH_info_${func}_avx2
+___
+$code.=<<___ if ($avx);
+.section	.xdata
+.align	8
+.LSEH_info_${func}_xop:
+	.byte	9,0,0,0
+	.rva	se_handler
+	.rva	.Lprologue_xop,.Lepilogue_xop		# HandlerData[]
+
+.LSEH_info_${func}_avx:
+	.byte	9,0,0,0
+	.rva	se_handler
+	.rva	.Lprologue_avx,.Lepilogue_avx		# HandlerData[]
+___
+$code.=<<___ if ($avx>1);
+.LSEH_info_${func}_avx2:
+	.byte	9,0,0,0
+	.rva	se_handler
+	.rva	.Lprologue_avx2,.Lepilogue_avx2		# HandlerData[]
+___
+}
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT;
diff --git a/crypto/evp/Makefile b/crypto/evp/Makefile
index d6f8e906dc..98a7e8de56 100644
--- a/crypto/evp/Makefile
+++ b/crypto/evp/Makefile
@@ -29,7 +29,7 @@ LIBSRC= encode.c digest.c evp_enc.c evp_key.c evp_acnf.c evp_cnf.c \
 	c_all.c c_allc.c c_alld.c evp_lib.c bio_ok.c \
 	evp_pkey.c evp_pbe.c p5_crpt.c p5_crpt2.c \
 	e_old.c pmeth_lib.c pmeth_fn.c pmeth_gn.c m_sigver.c \
-	e_aes_cbc_hmac_sha1.c e_rc4_hmac_md5.c
+	e_aes_cbc_hmac_sha1.c e_aes_cbc_hmac_sha256.c e_rc4_hmac_md5.c
 
 LIBOBJ=	encode.o digest.o evp_enc.o evp_key.o evp_acnf.o evp_cnf.o \
 	e_des.o e_bf.o e_idea.o e_des3.o e_camellia.o\
@@ -42,7 +42,7 @@ LIBOBJ=	encode.o digest.o evp_enc.o evp_key.o evp_acnf.o evp_cnf.o \
 	c_all.o c_allc.o c_alld.o evp_lib.o bio_ok.o \
 	evp_pkey.o evp_pbe.o p5_crpt.o p5_crpt2.o \
 	e_old.o pmeth_lib.o pmeth_fn.o pmeth_gn.o m_sigver.o \
-	e_aes_cbc_hmac_sha1.o e_rc4_hmac_md5.o
+	e_aes_cbc_hmac_sha1.o e_aes_cbc_hmac_sha256.o e_rc4_hmac_md5.o
 
 SRC= $(LIBSRC)
 
diff --git a/crypto/evp/c_allc.c b/crypto/evp/c_allc.c
index 6a39d7a5d0..1e3a89e5a4 100644
--- a/crypto/evp/c_allc.c
+++ b/crypto/evp/c_allc.c
@@ -202,6 +202,10 @@ void OpenSSL_add_all_ciphers(void)
 	EVP_add_cipher(EVP_aes_128_cbc_hmac_sha1());
 	EVP_add_cipher(EVP_aes_256_cbc_hmac_sha1());
 #endif
+#if !defined(OPENSSL_NO_SHA) && !defined(OPENSSL_NO_SHA256)
+	EVP_add_cipher(EVP_aes_128_cbc_hmac_sha256());
+	EVP_add_cipher(EVP_aes_256_cbc_hmac_sha256());
+#endif
 #endif
 
 #ifndef OPENSSL_NO_CAMELLIA
diff --git a/crypto/evp/e_aes_cbc_hmac_sha256.c b/crypto/evp/e_aes_cbc_hmac_sha256.c
new file mode 100644
index 0000000000..53262fa8e0
--- /dev/null
+++ b/crypto/evp/e_aes_cbc_hmac_sha256.c
@@ -0,0 +1,598 @@
+/* ====================================================================
+ * Copyright (c) 2011-2013 The OpenSSL Project.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ *    software must display the following acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ *    endorse or promote products derived from this software without
+ *    prior written permission. For written permission, please contact
+ *    licensing@OpenSSL.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ *    nor may "OpenSSL" appear in their names without prior written
+ *    permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ *    acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * OF THE POSSIBILITY OF SUCH DAMAGE.
+ * ====================================================================
+ */
+
+#include <openssl/opensslconf.h>
+
+#include <stdio.h>
+#include <string.h>
+
+#if !defined(OPENSSL_NO_AES) && !defined(OPENSSL_NO_SHA256)
+
+#include <openssl/evp.h>
+#include <openssl/objects.h>
+#include <openssl/aes.h>
+#include <openssl/sha.h>
+
+#ifndef EVP_CIPH_FLAG_AEAD_CIPHER
+#define EVP_CIPH_FLAG_AEAD_CIPHER	0x200000
+#define EVP_CTRL_AEAD_TLS1_AAD		0x16
+#define EVP_CTRL_AEAD_SET_MAC_KEY	0x17
+#endif
+
+#if !defined(EVP_CIPH_FLAG_DEFAULT_ASN1)
+#define EVP_CIPH_FLAG_DEFAULT_ASN1 0
+#endif
+
+#define TLS1_1_VERSION 0x0302
+
+typedef struct
+    {
+    AES_KEY		ks;
+    SHA256_CTX		head,tail,md;
+    size_t		payload_length;	/* AAD length in decrypt case */
+    union {
+	unsigned int	tls_ver;
+    	unsigned char	tls_aad[16];	/* 13 used */
+    } aux;
+    } EVP_AES_HMAC_SHA256;
+
+#define NO_PAYLOAD_LENGTH	((size_t)-1)
+
+#if	defined(AES_ASM) &&	( \
+	defined(__x86_64)	|| defined(__x86_64__)	|| \
+	defined(_M_AMD64)	|| defined(_M_X64)	|| \
+	defined(__INTEL__)	)
+
+#if defined(__GNUC__) && __GNUC__>=2 && !defined(PEDANTIC)
+# define BSWAP(x) ({ unsigned int r=(x); asm ("bswapl %0":"=r"(r):"0"(r)); r; })
+#endif
+
+extern unsigned int OPENSSL_ia32cap_P[3];
+#define AESNI_AVX_CAPABLE   (1<<(57-32)|1<<(60-32))
+
+int aesni_set_encrypt_key(const unsigned char *userKey, int bits,
+			      AES_KEY *key);
+int aesni_set_decrypt_key(const unsigned char *userKey, int bits,
+			      AES_KEY *key);
+
+void aesni_cbc_encrypt(const unsigned char *in,
+			   unsigned char *out,
+			   size_t length,
+			   const AES_KEY *key,
+			   unsigned char *ivec, int enc);
+
+int aesni_cbc_sha256_enc (const void *inp, void *out, size_t blocks,
+		const AES_KEY *key, unsigned char iv[16],
+		SHA256_CTX *ctx,const void *in0);
+
+#define data(ctx) ((EVP_AES_HMAC_SHA256 *)(ctx)->cipher_data)
+
+static int aesni_cbc_hmac_sha256_init_key(EVP_CIPHER_CTX *ctx,
+			const unsigned char *inkey,
+			const unsigned char *iv, int enc)
+	{
+	EVP_AES_HMAC_SHA256 *key = data(ctx);
+	int ret;
+
+	if (enc)
+		memset(&key->ks,0,sizeof(key->ks.rd_key)),
+		ret=aesni_set_encrypt_key(inkey,ctx->key_len*8,&key->ks);
+	else
+		ret=aesni_set_decrypt_key(inkey,ctx->key_len*8,&key->ks);
+
+	SHA256_Init(&key->head);	/* handy when benchmarking */
+	key->tail = key->head;
+	key->md   = key->head;
+
+	key->payload_length = NO_PAYLOAD_LENGTH;
+
+	return ret<0?0:1;
+	}
+
+#define	STITCHED_CALL
+
+#if !defined(STITCHED_CALL)
+#define	aes_off 0
+#endif
+
+void sha256_block_data_order (void *c,const void *p,size_t len);
+
+static void sha256_update(SHA256_CTX *c,const void *data,size_t len)
+{	const unsigned char *ptr = data;
+	size_t res;
+
+	if ((res = c->num)) {
+		res = SHA256_CBLOCK-res;
+		if (len<res) res=len;
+		SHA256_Update (c,ptr,res);
+		ptr += res;
+		len -= res;
+	}
+
+	res = len % SHA256_CBLOCK;
+	len -= res;
+
+	if (len) {
+		sha256_block_data_order(c,ptr,len/SHA256_CBLOCK);
+
+		ptr += len;
+		c->Nh += len>>29;
+		c->Nl += len<<=3;
+		if (c->Nl<(unsigned int)len) c->Nh++;
+	}
+
+	if (res)
+		SHA256_Update(c,ptr,res);
+}
+
+#ifdef SHA256_Update
+#undef SHA256_Update
+#endif
+#define SHA256_Update sha256_update
+
+static int aesni_cbc_hmac_sha256_cipher(EVP_CIPHER_CTX *ctx, unsigned char *out,
+		      const unsigned char *in, size_t len)
+	{
+	EVP_AES_HMAC_SHA256 *key = data(ctx);
+	unsigned int l;
+	size_t	plen = key->payload_length,
+		iv = 0,		/* explicit IV in TLS 1.1 and later */
+		sha_off = 0;
+#if defined(STITCHED_CALL)
+	size_t	aes_off = 0,
+		blocks;
+
+	sha_off = SHA256_CBLOCK-key->md.num;
+#endif
+
+	key->payload_length = NO_PAYLOAD_LENGTH;
+
+	if (len%AES_BLOCK_SIZE) return 0;
+
+	if (ctx->encrypt) {
+		if (plen==NO_PAYLOAD_LENGTH)
+			plen = len;
+		else if (len!=((plen+SHA256_DIGEST_LENGTH+AES_BLOCK_SIZE)&-AES_BLOCK_SIZE))
+			return 0;
+		else if (key->aux.tls_ver >= TLS1_1_VERSION)
+			iv = AES_BLOCK_SIZE;
+
+#if defined(STITCHED_CALL)
+		if (plen>(sha_off+iv) && (blocks=(plen-(sha_off+iv))/SHA256_CBLOCK)) {
+			SHA256_Update(&key->md,in+iv,sha_off);
+
+			(void)aesni_cbc_sha256_enc(in,out,blocks,&key->ks,
+				ctx->iv,&key->md,in+iv+sha_off);
+			blocks *= SHA256_CBLOCK;
+			aes_off += blocks;
+			sha_off += blocks;
+			key->md.Nh += blocks>>29;
+			key->md.Nl += blocks<<=3;
+			if (key->md.Nl<(unsigned int)blocks) key->md.Nh++;
+		} else {
+			sha_off = 0;
+		}
+#endif
+		sha_off += iv;
+		SHA256_Update(&key->md,in+sha_off,plen-sha_off);
+
+		if (plen!=len)	{	/* "TLS" mode of operation */
+			if (in!=out)
+				memcpy(out+aes_off,in+aes_off,plen-aes_off);
+
+			/* calculate HMAC and append it to payload */
+			SHA256_Final(out+plen,&key->md);
+			key->md = key->tail;
+			SHA256_Update(&key->md,out+plen,SHA256_DIGEST_LENGTH);
+			SHA256_Final(out+plen,&key->md);
+
+			/* pad the payload|hmac */
+			plen += SHA256_DIGEST_LENGTH;
+			for (l=len-plen-1;plen<len;plen++) out[plen]=l;
+			/* encrypt HMAC|padding at once */
+			aesni_cbc_encrypt(out+aes_off,out+aes_off,len-aes_off,
+					&key->ks,ctx->iv,1);
+		} else {
+			aesni_cbc_encrypt(in+aes_off,out+aes_off,len-aes_off,
+					&key->ks,ctx->iv,1);
+		}
+	} else {
+		union { unsigned int  u[SHA256_DIGEST_LENGTH/sizeof(unsigned int)];
+			unsigned char c[64+SHA256_DIGEST_LENGTH]; } mac, *pmac;
+
+		/* arrange cache line alignment */
+		pmac = (void *)(((size_t)mac.c+63)&((size_t)0-64));
+
+		/* decrypt HMAC|padding at once */
+		aesni_cbc_encrypt(in,out,len,
+				&key->ks,ctx->iv,0);
+
+		if (plen) {	/* "TLS" mode of operation */
+			size_t inp_len, mask, j, i;
+			unsigned int res, maxpad, pad, bitlen;
+			int ret = 1;
+			union {	unsigned int  u[SHA_LBLOCK];
+				unsigned char c[SHA256_CBLOCK]; }
+				*data = (void *)key->md.data;
+
+			if ((key->aux.tls_aad[plen-4]<<8|key->aux.tls_aad[plen-3])
+			    >= TLS1_1_VERSION)
+				iv = AES_BLOCK_SIZE;
+
+			if (len<(iv+SHA256_DIGEST_LENGTH+1))
+				return 0;
+
+			/* omit explicit iv */
+			out += iv;
+			len -= iv;
+
+			/* figure out payload length */
+			pad = out[len-1];
+			maxpad = len-(SHA256_DIGEST_LENGTH+1);
+			maxpad |= (255-maxpad)>>(sizeof(maxpad)*8-8);
+			maxpad &= 255;
+
+			inp_len = len - (SHA256_DIGEST_LENGTH+pad+1);
+			mask = (0-((inp_len-len)>>(sizeof(inp_len)*8-1)));
+			inp_len &= mask;
+			ret &= (int)mask;
+
+			key->aux.tls_aad[plen-2] = inp_len>>8;
+			key->aux.tls_aad[plen-1] = inp_len;
+
+			/* calculate HMAC */
+			key->md = key->head;
+			SHA256_Update(&key->md,key->aux.tls_aad,plen);
+
+#if 1
+			len -= SHA256_DIGEST_LENGTH;		/* amend mac */
+			if (len>=(256+SHA256_CBLOCK)) {
+				j = (len-(256+SHA256_CBLOCK))&(0-SHA256_CBLOCK);
+				j += SHA256_CBLOCK-key->md.num;
+				SHA256_Update(&key->md,out,j);
+				out += j;
+				len -= j;
+				inp_len -= j;
+			}
+
+			/* but pretend as if we hashed padded payload */
+			bitlen = key->md.Nl+(inp_len<<3);	/* at most 18 bits */
+#ifdef BSWAP
+			bitlen = BSWAP(bitlen);
+#else
+			mac.c[0] = 0;
+			mac.c[1] = (unsigned char)(bitlen>>16);
+			mac.c[2] = (unsigned char)(bitlen>>8);
+			mac.c[3] = (unsigned char)bitlen;
+			bitlen = mac.u[0];
+#endif
+
+			pmac->u[0]=0;
+			pmac->u[1]=0;
+			pmac->u[2]=0;
+			pmac->u[3]=0;
+			pmac->u[4]=0;
+			pmac->u[5]=0;
+			pmac->u[6]=0;
+			pmac->u[7]=0;
+
+			for (res=key->md.num, j=0;j<len;j++) {
+				size_t c = out[j];
+				mask = (j-inp_len)>>(sizeof(j)*8-8);
+				c &= mask;
+				c |= 0x80&~mask&~((inp_len-j)>>(sizeof(j)*8-8));
+				data->c[res++]=(unsigned char)c;
+
+				if (res!=SHA256_CBLOCK) continue;
+
+				/* j is not incremented yet */
+				mask = 0-((inp_len+7-j)>>(sizeof(j)*8-1));
+				data->u[SHA_LBLOCK-1] |= bitlen&mask;
+				sha256_block_data_order(&key->md,data,1);
+				mask &= 0-((j-inp_len-72)>>(sizeof(j)*8-1));
+				pmac->u[0] |= key->md.h[0] & mask;
+				pmac->u[1] |= key->md.h[1] & mask;
+				pmac->u[2] |= key->md.h[2] & mask;
+				pmac->u[3] |= key->md.h[3] & mask;
+				pmac->u[4] |= key->md.h[4] & mask;
+				pmac->u[5] |= key->md.h[5] & mask;
+				pmac->u[6] |= key->md.h[6] & mask;
+				pmac->u[7] |= key->md.h[7] & mask;
+				res=0;
+			}
+
+			for(i=res;i<SHA256_CBLOCK;i++,j++) data->c[i]=0;
+
+			if (res>SHA256_CBLOCK-8) {
+				mask = 0-((inp_len+8-j)>>(sizeof(j)*8-1));
+				data->u[SHA_LBLOCK-1] |= bitlen&mask;
+				sha256_block_data_order(&key->md,data,1);
+				mask &= 0-((j-inp_len-73)>>(sizeof(j)*8-1));
+				pmac->u[0] |= key->md.h[0] & mask;
+				pmac->u[1] |= key->md.h[1] & mask;
+				pmac->u[2] |= key->md.h[2] & mask;
+				pmac->u[3] |= key->md.h[3] & mask;
+				pmac->u[4] |= key->md.h[4] & mask;
+				pmac->u[5] |= key->md.h[5] & mask;
+				pmac->u[6] |= key->md.h[6] & mask;
+				pmac->u[7] |= key->md.h[7] & mask;
+
+				memset(data,0,SHA256_CBLOCK);
+				j+=64;
+			}
+			data->u[SHA_LBLOCK-1] = bitlen;
+			sha256_block_data_order(&key->md,data,1);
+			mask = 0-((j-inp_len-73)>>(sizeof(j)*8-1));
+			pmac->u[0] |= key->md.h[0] & mask;
+			pmac->u[1] |= key->md.h[1] & mask;
+			pmac->u[2] |= key->md.h[2] & mask;
+			pmac->u[3] |= key->md.h[3] & mask;
+			pmac->u[4] |= key->md.h[4] & mask;
+			pmac->u[5] |= key->md.h[5] & mask;
+			pmac->u[6] |= key->md.h[6] & mask;
+			pmac->u[7] |= key->md.h[7] & mask;
+
+#ifdef BSWAP
+			pmac->u[0] = BSWAP(pmac->u[0]);
+			pmac->u[1] = BSWAP(pmac->u[1]);
+			pmac->u[2] = BSWAP(pmac->u[2]);
+			pmac->u[3] = BSWAP(pmac->u[3]);
+			pmac->u[4] = BSWAP(pmac->u[4]);
+			pmac->u[5] = BSWAP(pmac->u[5]);
+			pmac->u[6] = BSWAP(pmac->u[6]);
+			pmac->u[7] = BSWAP(pmac->u[7]);
+#else
+			for (i=0;i<8;;i++) {
+				res = pmac->u[i];
+				pmac->c[4*i+0]=(unsigned char)(res>>24);
+				pmac->c[4*i+1]=(unsigned char)(res>>16);
+				pmac->c[4*i+2]=(unsigned char)(res>>8);
+				pmac->c[4*i+3]=(unsigned char)res;
+			}
+#endif
+			len += SHA256_DIGEST_LENGTH;
+#else
+			SHA256_Update(&key->md,out,inp_len);
+			res = key->md.num;
+			SHA256_Final(pmac->c,&key->md);
+
+			{
+			unsigned int inp_blocks, pad_blocks;
+
+			/* but pretend as if we hashed padded payload */
+			inp_blocks = 1+((SHA256_CBLOCK-9-res)>>(sizeof(res)*8-1));
+			res += (unsigned int)(len-inp_len);
+			pad_blocks = res / SHA256_CBLOCK;
+			res %= SHA256_CBLOCK;
+			pad_blocks += 1+((SHA256_CBLOCK-9-res)>>(sizeof(res)*8-1));
+			for (;inp_blocks<pad_blocks;inp_blocks++)
+				sha1_block_data_order(&key->md,data,1);
+			}
+#endif
+			key->md = key->tail;
+			SHA256_Update(&key->md,pmac->c,SHA256_DIGEST_LENGTH);
+			SHA256_Final(pmac->c,&key->md);
+
+			/* verify HMAC */
+			out += inp_len;
+			len -= inp_len;
+#if 1
+			{
+			unsigned char *p = out+len-1-maxpad-SHA256_DIGEST_LENGTH;
+			size_t off = out-p;
+			unsigned int c, cmask;
+
+			maxpad += SHA256_DIGEST_LENGTH;
+			for (res=0,i=0,j=0;j<maxpad;j++) {
+				c = p[j];
+				cmask = ((int)(j-off-SHA256_DIGEST_LENGTH))>>(sizeof(int)*8-1);
+				res |= (c^pad)&~cmask;	/* ... and padding */
+				cmask &= ((int)(off-1-j))>>(sizeof(int)*8-1);
+				res |= (c^pmac->c[i])&cmask;
+				i += 1&cmask;
+			}
+			maxpad -= SHA256_DIGEST_LENGTH;
+
+			res = 0-((0-res)>>(sizeof(res)*8-1));
+			ret &= (int)~res;
+			}
+#else
+			for (res=0,i=0;i<SHA256_DIGEST_LENGTH;i++)
+				res |= out[i]^pmac->c[i];
+			res = 0-((0-res)>>(sizeof(res)*8-1));
+			ret &= (int)~res;
+
+			/* verify padding */
+			pad = (pad&~res) | (maxpad&res);
+			out = out+len-1-pad;
+			for (res=0,i=0;i<pad;i++)
+				res |= out[i]^pad;
+
+			res = (0-res)>>(sizeof(res)*8-1);
+			ret &= (int)~res;
+#endif
+			return ret;
+		} else {
+			SHA256_Update(&key->md,out,len);
+		}
+	}
+
+	return 1;
+	}
+
+static int aesni_cbc_hmac_sha256_ctrl(EVP_CIPHER_CTX *ctx, int type, int arg, void *ptr)
+	{
+	EVP_AES_HMAC_SHA256 *key = data(ctx);
+
+	switch (type)
+		{
+	case EVP_CTRL_AEAD_SET_MAC_KEY:
+		{
+		unsigned int  i;
+		unsigned char hmac_key[64];
+
+		memset (hmac_key,0,sizeof(hmac_key));
+
+		if (arg > (int)sizeof(hmac_key)) {
+			SHA256_Init(&key->head);
+			SHA256_Update(&key->head,ptr,arg);
+			SHA256_Final(hmac_key,&key->head);
+		} else {
+			memcpy(hmac_key,ptr,arg);
+		}
+
+		for (i=0;i<sizeof(hmac_key);i++)
+			hmac_key[i] ^= 0x36;		/* ipad */
+		SHA256_Init(&key->head);
+		SHA256_Update(&key->head,hmac_key,sizeof(hmac_key));
+
+		for (i=0;i<sizeof(hmac_key);i++)
+			hmac_key[i] ^= 0x36^0x5c;	/* opad */
+		SHA256_Init(&key->tail);
+		SHA256_Update(&key->tail,hmac_key,sizeof(hmac_key));
+
+		OPENSSL_cleanse(hmac_key,sizeof(hmac_key));
+
+		return 1;
+		}
+	case EVP_CTRL_AEAD_TLS1_AAD:
+		{
+		unsigned char *p=ptr;
+		unsigned int   len=p[arg-2]<<8|p[arg-1];
+
+		if (ctx->encrypt)
+			{
+			key->payload_length = len;
+			if ((key->aux.tls_ver=p[arg-4]<<8|p[arg-3]) >= TLS1_1_VERSION) {
+				len -= AES_BLOCK_SIZE;
+				p[arg-2] = len>>8;
+				p[arg-1] = len;
+			}
+			key->md = key->head;
+			SHA256_Update(&key->md,p,arg);
+
+			return (int)(((len+SHA256_DIGEST_LENGTH+AES_BLOCK_SIZE)&-AES_BLOCK_SIZE)
+				- len);
+			}
+		else
+			{
+			if (arg>13) arg = 13;
+			memcpy(key->aux.tls_aad,ptr,arg);
+			key->payload_length = arg;
+
+			return SHA256_DIGEST_LENGTH;
+			}
+		}
+	default:
+		return -1;
+		}
+	}
+
+static EVP_CIPHER aesni_128_cbc_hmac_sha256_cipher =
+	{
+#ifdef NID_aes_128_cbc_hmac_sha256
+	NID_aes_128_cbc_hmac_sha256,
+#else
+	NID_undef,
+#endif
+	16,16,16,
+	EVP_CIPH_CBC_MODE|EVP_CIPH_FLAG_DEFAULT_ASN1|EVP_CIPH_FLAG_AEAD_CIPHER,
+	aesni_cbc_hmac_sha256_init_key,
+	aesni_cbc_hmac_sha256_cipher,
+	NULL,
+	sizeof(EVP_AES_HMAC_SHA256),
+	EVP_CIPH_FLAG_DEFAULT_ASN1?NULL:EVP_CIPHER_set_asn1_iv,
+	EVP_CIPH_FLAG_DEFAULT_ASN1?NULL:EVP_CIPHER_get_asn1_iv,
+	aesni_cbc_hmac_sha256_ctrl,
+	NULL
+	};
+
+static EVP_CIPHER aesni_256_cbc_hmac_sha256_cipher =
+	{
+#ifdef NID_aes_256_cbc_hmac_sha256
+	NID_aes_256_cbc_hmac_sha256,
+#else
+	NID_undef,
+#endif
+	16,32,16,
+	EVP_CIPH_CBC_MODE|EVP_CIPH_FLAG_DEFAULT_ASN1|EVP_CIPH_FLAG_AEAD_CIPHER,
+	aesni_cbc_hmac_sha256_init_key,
+	aesni_cbc_hmac_sha256_cipher,
+	NULL,
+	sizeof(EVP_AES_HMAC_SHA256),
+	EVP_CIPH_FLAG_DEFAULT_ASN1?NULL:EVP_CIPHER_set_asn1_iv,
+	EVP_CIPH_FLAG_DEFAULT_ASN1?NULL:EVP_CIPHER_get_asn1_iv,
+	aesni_cbc_hmac_sha256_ctrl,
+	NULL
+	};
+
+const EVP_CIPHER *EVP_aes_128_cbc_hmac_sha256(void)
+	{
+	return((OPENSSL_ia32cap_P[1]&AESNI_AVX_CAPABLE)==AESNI_AVX_CAPABLE &&
+		aesni_cbc_sha256_enc(NULL,NULL,0,NULL,NULL,NULL,NULL) ?
+		&aesni_128_cbc_hmac_sha256_cipher:NULL);
+	}
+
+const EVP_CIPHER *EVP_aes_256_cbc_hmac_sha256(void)
+	{
+	return((OPENSSL_ia32cap_P[1]&AESNI_AVX_CAPABLE)==AESNI_AVX_CAPABLE &&
+		aesni_cbc_sha256_enc(NULL,NULL,0,NULL,NULL,NULL,NULL)?
+		&aesni_256_cbc_hmac_sha256_cipher:NULL);
+	}
+#else
+const EVP_CIPHER *EVP_aes_128_cbc_hmac_sha256(void)
+	{
+	return NULL;
+	}
+const EVP_CIPHER *EVP_aes_256_cbc_hmac_sha256(void)
+	{
+	return NULL;
+	}
+#endif
+#endif
diff --git a/crypto/objects/obj_dat.h b/crypto/objects/obj_dat.h
index ad2e1dbda0..241d4e148c 100644
--- a/crypto/objects/obj_dat.h
+++ b/crypto/objects/obj_dat.h
@@ -62,9 +62,9 @@
  * [including the GNU Public Licence.]
  */
 
-#define NUM_NID 935
-#define NUM_SN 928
-#define NUM_LN 928
+#define NUM_NID 938
+#define NUM_SN 931
+#define NUM_LN 931
 #define NUM_OBJ 872
 
 static const unsigned char lvalues[6113]={
@@ -2450,12 +2450,19 @@ static const ASN1_OBJECT nid_objs[NUM_NID]={
 	&(lvalues[6094]),0},
 {"brainpoolP512t1","brainpoolP512t1",NID_brainpoolP512t1,9,
 	&(lvalues[6103]),0},
+{"AES-128-CBC-HMAC-SHA256","aes-128-cbc-hmac-sha256",
+	NID_aes_128_cbc_hmac_sha256,0,NULL,0},
+{"AES-192-CBC-HMAC-SHA256","aes-192-cbc-hmac-sha256",
+	NID_aes_192_cbc_hmac_sha256,0,NULL,0},
+{"AES-256-CBC-HMAC-SHA256","aes-256-cbc-hmac-sha256",
+	NID_aes_256_cbc_hmac_sha256,0,NULL,0},
 };
 
 static const unsigned int sn_objs[NUM_SN]={
 364,	/* "AD_DVCS" */
 419,	/* "AES-128-CBC" */
 916,	/* "AES-128-CBC-HMAC-SHA1" */
+935,	/* "AES-128-CBC-HMAC-SHA256" */
 421,	/* "AES-128-CFB" */
 650,	/* "AES-128-CFB1" */
 653,	/* "AES-128-CFB8" */
@@ -2465,6 +2472,7 @@ static const unsigned int sn_objs[NUM_SN]={
 913,	/* "AES-128-XTS" */
 423,	/* "AES-192-CBC" */
 917,	/* "AES-192-CBC-HMAC-SHA1" */
+936,	/* "AES-192-CBC-HMAC-SHA256" */
 425,	/* "AES-192-CFB" */
 651,	/* "AES-192-CFB1" */
 654,	/* "AES-192-CFB8" */
@@ -2473,6 +2481,7 @@ static const unsigned int sn_objs[NUM_SN]={
 424,	/* "AES-192-OFB" */
 427,	/* "AES-256-CBC" */
 918,	/* "AES-256-CBC-HMAC-SHA1" */
+937,	/* "AES-256-CBC-HMAC-SHA256" */
 429,	/* "AES-256-CFB" */
 652,	/* "AES-256-CFB1" */
 655,	/* "AES-256-CFB8" */
@@ -3530,6 +3539,7 @@ static const unsigned int ln_objs[NUM_LN]={
 606,	/* "additional verification" */
 419,	/* "aes-128-cbc" */
 916,	/* "aes-128-cbc-hmac-sha1" */
+935,	/* "aes-128-cbc-hmac-sha256" */
 896,	/* "aes-128-ccm" */
 421,	/* "aes-128-cfb" */
 650,	/* "aes-128-cfb1" */
@@ -3541,6 +3551,7 @@ static const unsigned int ln_objs[NUM_LN]={
 913,	/* "aes-128-xts" */
 423,	/* "aes-192-cbc" */
 917,	/* "aes-192-cbc-hmac-sha1" */
+936,	/* "aes-192-cbc-hmac-sha256" */
 899,	/* "aes-192-ccm" */
 425,	/* "aes-192-cfb" */
 651,	/* "aes-192-cfb1" */
@@ -3551,6 +3562,7 @@ static const unsigned int ln_objs[NUM_LN]={
 424,	/* "aes-192-ofb" */
 427,	/* "aes-256-cbc" */
 918,	/* "aes-256-cbc-hmac-sha1" */
+937,	/* "aes-256-cbc-hmac-sha256" */
 902,	/* "aes-256-ccm" */
 429,	/* "aes-256-cfb" */
 652,	/* "aes-256-cfb1" */
diff --git a/crypto/objects/obj_mac.h b/crypto/objects/obj_mac.h
index 704697eee0..68063777d8 100644
--- a/crypto/objects/obj_mac.h
+++ b/crypto/objects/obj_mac.h
@@ -4030,6 +4030,18 @@
 #define LN_aes_256_cbc_hmac_sha1		"aes-256-cbc-hmac-sha1"
 #define NID_aes_256_cbc_hmac_sha1		918
 
+#define SN_aes_128_cbc_hmac_sha256		"AES-128-CBC-HMAC-SHA256"
+#define LN_aes_128_cbc_hmac_sha256		"aes-128-cbc-hmac-sha256"
+#define NID_aes_128_cbc_hmac_sha256		935
+
+#define SN_aes_192_cbc_hmac_sha256		"AES-192-CBC-HMAC-SHA256"
+#define LN_aes_192_cbc_hmac_sha256		"aes-192-cbc-hmac-sha256"
+#define NID_aes_192_cbc_hmac_sha256		936
+
+#define SN_aes_256_cbc_hmac_sha256		"AES-256-CBC-HMAC-SHA256"
+#define LN_aes_256_cbc_hmac_sha256		"aes-256-cbc-hmac-sha256"
+#define NID_aes_256_cbc_hmac_sha256		937
+
 #define SN_dhpublicnumber		"dhpublicnumber"
 #define LN_dhpublicnumber		"X9.42 DH"
 #define NID_dhpublicnumber		920
diff --git a/crypto/objects/obj_mac.num b/crypto/objects/obj_mac.num
index 929743572f..181532424d 100644
--- a/crypto/objects/obj_mac.num
+++ b/crypto/objects/obj_mac.num
@@ -932,3 +932,6 @@ brainpoolP384r1		931
 brainpoolP384t1		932
 brainpoolP512r1		933
 brainpoolP512t1		934
+aes_128_cbc_hmac_sha256		935
+aes_192_cbc_hmac_sha256		936
+aes_256_cbc_hmac_sha256		937
diff --git a/crypto/objects/objects.txt b/crypto/objects/objects.txt
index 42514695bd..1868446392 100644
--- a/crypto/objects/objects.txt
+++ b/crypto/objects/objects.txt
@@ -1290,6 +1290,9 @@ kisa 1 6                : SEED-OFB      : seed-ofb
 			: AES-128-CBC-HMAC-SHA1		: aes-128-cbc-hmac-sha1
 			: AES-192-CBC-HMAC-SHA1		: aes-192-cbc-hmac-sha1
 			: AES-256-CBC-HMAC-SHA1		: aes-256-cbc-hmac-sha1
+			: AES-128-CBC-HMAC-SHA256	: aes-128-cbc-hmac-sha256
+			: AES-192-CBC-HMAC-SHA256	: aes-192-cbc-hmac-sha256
+			: AES-256-CBC-HMAC-SHA256	: aes-256-cbc-hmac-sha256
 
 ISO-US 10046 2 1	: dhpublicnumber		: X9.42 DH
 
diff --git a/ssl/ssl_algs.c b/ssl/ssl_algs.c
index d5a7a20e8a..d3646aef97 100644
--- a/ssl/ssl_algs.c
+++ b/ssl/ssl_algs.c
@@ -94,7 +94,10 @@ int SSL_library_init(void)
 	EVP_add_cipher(EVP_aes_128_cbc_hmac_sha1());
 	EVP_add_cipher(EVP_aes_256_cbc_hmac_sha1());
 #endif
-
+#if !defined(OPENSSL_NO_SHA) && !defined(OPENSSL_NO_SHA256)
+	EVP_add_cipher(EVP_aes_128_cbc_hmac_sha256());
+	EVP_add_cipher(EVP_aes_256_cbc_hmac_sha256());
+#endif
 #endif
 #ifndef OPENSSL_NO_CAMELLIA
 	EVP_add_cipher(EVP_camellia_128_cbc());
diff --git a/ssl/ssl_ciph.c b/ssl/ssl_ciph.c
index 1683f7bb1f..afdc8ff5df 100644
--- a/ssl/ssl_ciph.c
+++ b/ssl/ssl_ciph.c
@@ -638,6 +638,14 @@ int ssl_cipher_get_evp(const SSL_SESSION *s, const EVP_CIPHER **enc,
 			 c->algorithm_mac == SSL_SHA1 &&
 			 (evp=EVP_get_cipherbyname("AES-256-CBC-HMAC-SHA1")))
 			*enc = evp, *md = NULL;
+		else if (c->algorithm_enc == SSL_AES128 &&
+			 c->algorithm_mac == SSL_SHA256 &&
+			 (evp=EVP_get_cipherbyname("AES-128-CBC-HMAC-SHA256")))
+			*enc = evp, *md = NULL;
+		else if (c->algorithm_enc == SSL_AES256 &&
+			 c->algorithm_mac == SSL_SHA256 &&
+			 (evp=EVP_get_cipherbyname("AES-256-CBC-HMAC-SHA256")))
+			*enc = evp, *md = NULL;
 		return(1);
 		}
 	else