init

COPYING

@@ -0,0 +1,121 @@
+Creative Commons Legal Code
+
+CC0 1.0 Universal
+
+    CREATIVE COMMONS CORPORATION IS NOT A LAW FIRM AND DOES NOT PROVIDE
+    LEGAL SERVICES. DISTRIBUTION OF THIS DOCUMENT DOES NOT CREATE AN
+    ATTORNEY-CLIENT RELATIONSHIP. CREATIVE COMMONS PROVIDES THIS
+    INFORMATION ON AN "AS-IS" BASIS. CREATIVE COMMONS MAKES NO WARRANTIES
+    REGARDING THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS
+    PROVIDED HEREUNDER, AND DISCLAIMS LIABILITY FOR DAMAGES RESULTING FROM
+    THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS PROVIDED
+    HEREUNDER.
+
+Statement of Purpose
+
+The laws of most jurisdictions throughout the world automatically confer
+exclusive Copyright and Related Rights (defined below) upon the creator
+and subsequent owner(s) (each and all, an "owner") of an original work of
+authorship and/or a database (each, a "Work").
+
+Certain owners wish to permanently relinquish those rights to a Work for
+the purpose of contributing to a commons of creative, cultural and
+scientific works ("Commons") that the public can reliably and without fear
+of later claims of infringement build upon, modify, incorporate in other
+works, reuse and redistribute as freely as possible in any form whatsoever
+and for any purposes, including without limitation commercial purposes.
+These owners may contribute to the Commons to promote the ideal of a free
+culture and the further production of creative, cultural and scientific
+works, or to gain reputation or greater distribution for their Work in
+part through the use and efforts of others.
+
+For these and/or other purposes and motivations, and without any
+expectation of additional consideration or compensation, the person
+associating CC0 with a Work (the "Affirmer"), to the extent that he or she
+is an owner of Copyright and Related Rights in the Work, voluntarily
+elects to apply CC0 to the Work and publicly distribute the Work under its
+terms, with knowledge of his or her Copyright and Related Rights in the
+Work and the meaning and intended legal effect of CC0 on those rights.
+
+1. Copyright and Related Rights. A Work made available under CC0 may be
+protected by copyright and related or neighboring rights ("Copyright and
+Related Rights"). Copyright and Related Rights include, but are not
+limited to, the following:
+
+  i. the right to reproduce, adapt, distribute, perform, display,
+     communicate, and translate a Work;
+ ii. moral rights retained by the original author(s) and/or performer(s);
+iii. publicity and privacy rights pertaining to a person's image or
+     likeness depicted in a Work;
+ iv. rights protecting against unfair competition in regards to a Work,
+     subject to the limitations in paragraph 4(a), below;
+  v. rights protecting the extraction, dissemination, use and reuse of data
+     in a Work;
+ vi. database rights (such as those arising under Directive 96/9/EC of the
+     European Parliament and of the Council of 11 March 1996 on the legal
+     protection of databases, and under any national implementation
+     thereof, including any amended or successor version of such
+     directive); and
+vii. other similar, equivalent or corresponding rights throughout the
+     world based on applicable law or treaty, and any national
+     implementations thereof.
+
+2. Waiver. To the greatest extent permitted by, but not in contravention
+of, applicable law, Affirmer hereby overtly, fully, permanently,
+irrevocably and unconditionally waives, abandons, and surrenders all of
+Affirmer's Copyright and Related Rights and associated claims and causes
+of action, whether now known or unknown (including existing as well as
+future claims and causes of action), in the Work (i) in all territories
+worldwide, (ii) for the maximum duration provided by applicable law or
+treaty (including future time extensions), (iii) in any current or future
+medium and for any number of copies, and (iv) for any purpose whatsoever,
+including without limitation commercial, advertising or promotional
+purposes (the "Waiver"). Affirmer makes the Waiver for the benefit of each
+member of the public at large and to the detriment of Affirmer's heirs and
+successors, fully intending that such Waiver shall not be subject to
+revocation, rescission, cancellation, termination, or any other legal or
+equitable action to disrupt the quiet enjoyment of the Work by the public
+as contemplated by Affirmer's express Statement of Purpose.
+
+3. Public License Fallback. Should any part of the Waiver for any reason
+be judged legally invalid or ineffective under applicable law, then the
+Waiver shall be preserved to the maximum extent permitted taking into
+account Affirmer's express Statement of Purpose. In addition, to the
+extent the Waiver is so judged Affirmer hereby grants to each affected
+person a royalty-free, non transferable, non sublicensable, non exclusive,
+irrevocable and unconditional license to exercise Affirmer's Copyright and
+Related Rights in the Work (i) in all territories worldwide, (ii) for the
+maximum duration provided by applicable law or treaty (including future
+time extensions), (iii) in any current or future medium and for any number
+of copies, and (iv) for any purpose whatsoever, including without
+limitation commercial, advertising or promotional purposes (the
+"License"). The License shall be deemed effective as of the date CC0 was
+applied by Affirmer to the Work. Should any part of the License for any
+reason be judged legally invalid or ineffective under applicable law, such
+partial invalidity or ineffectiveness shall not invalidate the remainder
+of the License, and in such case Affirmer hereby affirms that he or she
+will not (i) exercise any of his or her remaining Copyright and Related
+Rights in the Work or (ii) assert any associated claims and causes of
+action with respect to the Work, in either case contrary to Affirmer's
+express Statement of Purpose.
+
+4. Limitations and Disclaimers.
+
+ a. No trademark or patent rights held by Affirmer are waived, abandoned,
+    surrendered, licensed or otherwise affected by this document.
+ b. Affirmer offers the Work as-is and makes no representations or
+    warranties of any kind concerning the Work, express, implied,
+    statutory or otherwise, including without limitation warranties of
+    title, merchantability, fitness for a particular purpose, non
+    infringement, or the absence of latent or other defects, accuracy, or
+    the present or absence of errors, whether or not discoverable, all to
+    the greatest extent permissible under applicable law.
+ c. Affirmer disclaims responsibility for clearing rights of other persons
+    that may apply to the Work or any use thereof, including without
+    limitation any person's Copyright and Related Rights in the Work.
+    Further, Affirmer disclaims responsibility for obtaining any necessary
+    consents, permissions or other rights required for any use of the
+    Work.
+ d. Affirmer understands and acknowledges that Creative Commons is not a
+    party to this document and has no duty or obligation with respect to
+    this CC0 or use of the Work.

Makefile.am

@@ -0,0 +1,5 @@
+ACLOCAL_AMFLAGS = -I m4 
+
+EXTRA_DIST = COPYING
+
+SUBDIRS = src

autogen.sh

@@ -0,0 +1,2 @@
+#! /bin/sh
+autoreconf -fvi

clean.sh

@@ -0,0 +1,13 @@
+make maintainer-clean
+rm -rf Makefile.in\
+       aclocal.m4\
+       config.guess\
+       config.sub\
+       configure\
+       depcomp\
+       install-sh\
+       ltmain.sh\
+       missing\
+       src/Makefile.in\
+       src/config.h.in
+

configure.ac

@@ -0,0 +1,76 @@
+AC_PREREQ([2.61])
+AC_INIT([libb2], [0.97], [contact@blake2.net], [libb2], [https://blake2.net])
+AC_CONFIG_SRCDIR([src/blake2b.c])
+AC_CONFIG_HEADERS([src/config.h])
+AM_INIT_AUTOMAKE([foreign 1.9])
+AC_CONFIG_MACRO_DIR([m4])
+
+B2_LIBRARY_VERSION=1:2:0 # interface, revision, age
+AC_SUBST(B2_LIBRARY_VERSION)
+
+AC_LANG_C
+AC_PROG_CC
+AC_PROG_CC_C99
+
+AC_CHECK_HEADERS([stddef.h stdint.h stdlib.h string.h])
+AC_OPENMP
+AX_CHECK_ALIGNED_ACCESS_REQUIRED
+# AX_FORCEINLINE()
+AC_C_BIGENDIAN(
+	[],
+	AC_DEFINE(NATIVE_LITTLE_ENDIAN, 1, [machine is little-endian]),
+	AC_MSG_ERROR(unknown endianness),
+	AC_MSG_ERROR(universal endianness not supported)
+) 
+
+# Checks for typedefs, structures, and compiler characteristics.
+AC_C_INLINE
+AC_TYPE_SIZE_T
+AC_TYPE_UINT32_T
+AC_TYPE_UINT64_T
+AC_TYPE_UINT8_T
+
+# Checks for library functions.
+AC_CHECK_FUNCS([memset])
+
+dnl AM_PROG_AR 
+AC_PROG_LIBTOOL
+dnl LT_INIT
+
+
+AC_ARG_ENABLE(fat,
+AC_HELP_STRING([--enable-fat],
+               [build a fat binary on systems that support it [[default=no]]]),
+               [case $enableval in
+                  yes|no) ;;
+                  *) AC_MSG_ERROR([bad value $enableval for --enable-fat, need yes or no]) ;;
+                esac],
+               [enable_fat=no]
+)
+
+AX_CHECK_COMPILE_FLAG([-O3], [CFLAGS=-O3])
+
+if test $enable_fat == "yes"; then
+  dnl Fat build needs compiler who knows all the possible instruction sets
+  AX_CHECK_COMPILE_FLAG([-msse2],   [], AC_MSG_ERROR([Compiler does not know -msse2.]))
+  AX_CHECK_COMPILE_FLAG([-mssse3],  [], AC_MSG_ERROR([Compiler does not know -mssse3.]))
+  AX_CHECK_COMPILE_FLAG([-msse4.1], [], AC_MSG_ERROR([Compiler does not know -msse4.1.]))
+  AX_CHECK_COMPILE_FLAG([-mavx],    [], AC_MSG_ERROR([Compiler does not know -mavx.]))
+  AX_CHECK_COMPILE_FLAG([-mxop],    [], AC_MSG_ERROR([Compiler does not know -mxop.]))
+else
+  AX_EXT
+  CFLAGS="${CFLAGS} -march=native"
+fi
+
+case $host_os in
+	*mingw*) LDFLAGS="${LDFLAGS} -no-undefined" ;;
+esac
+
+AM_CONDITIONAL([USE_FAT], [test "$enable_fat" = "yes"])
+dnl Only move away from ref with SSSE3; SSE2 is generally slower
+AM_CONDITIONAL([USE_SSE], [test "$ax_cv_have_ssse3_ext" = "yes"])
+
+AC_CONFIG_FILES([Makefile
+                 src/Makefile
+                 ])
+AC_OUTPUT

m4/ax_check_aligned_access_required.m4

@@ -0,0 +1,84 @@
+# ====================================================================================
+#  http://www.gnu.org/software/autoconf-archive/ax_check_aligned_access_required.html
+# ====================================================================================
+#
+# SYNOPSIS
+#
+#   AC_CHECK_ALIGNED_ACCESS_REQUIRED
+#
+# DESCRIPTION
+#
+#   While the x86 CPUs allow access to memory objects to be unaligned it
+#   happens that most of the modern designs require objects to be aligned -
+#   or they will fail with a buserror. That mode is quite known by
+#   big-endian machines (sparc, etc) however the alpha cpu is little-
+#   endian.
+#
+#   The following function will test for aligned access to be required and
+#   set a config.h define HAVE_ALIGNED_ACCESS_REQUIRED (name derived by
+#   standard usage). Structures loaded from a file (or mmapped to memory)
+#   should be accessed per-byte in that case to avoid segfault type errors.
+#
+# LICENSE
+#
+#   Copyright (c) 2008 Guido U. Draheim <guidod@gmx.de>
+#
+#   This program is free software; you can redistribute it and/or modify it
+#   under the terms of the GNU General Public License as published by the
+#   Free Software Foundation; either version 3 of the License, or (at your
+#   option) any later version.
+#
+#   This program is distributed in the hope that it will be useful, but
+#   WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
+#   Public License for more details.
+#
+#   You should have received a copy of the GNU General Public License along
+#   with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+#   As a special exception, the respective Autoconf Macro's copyright owner
+#   gives unlimited permission to copy, distribute and modify the configure
+#   scripts that are the output of Autoconf when processing the Macro. You
+#   need not follow the terms of the GNU General Public License when using
+#   or distributing such scripts, even though portions of the text of the
+#   Macro appear in them. The GNU General Public License (GPL) does govern
+#   all other use of the material that constitutes the Autoconf Macro.
+#
+#   This special exception to the GPL applies to versions of the Autoconf
+#   Macro released by the Autoconf Archive. When you make and distribute a
+#   modified version of the Autoconf Macro, you may extend this special
+#   exception to the GPL to apply to your modified version as well.
+
+#serial 7
+
+AC_DEFUN([AX_CHECK_ALIGNED_ACCESS_REQUIRED],
+[AC_CACHE_CHECK([if pointers to integers require aligned access],
+  [ax_cv_have_aligned_access_required],
+  [AC_TRY_RUN([
+#include <stdio.h>
+#include <stdlib.h>
+
+int main()
+{
+  char* string = malloc(40);
+  int i;
+  for (i=0; i < 40; i++) string[[i]] = i;
+  {
+     void* s = string;
+     int* p = s+1;
+     int* q = s+2;
+
+     if (*p == *q) { return 1; }
+  }
+  return 0;
+}
+              ],
+     [ax_cv_have_aligned_access_required=yes],
+     [ax_cv_have_aligned_access_required=no],
+     [ax_cv_have_aligned_access_required=no])
+  ])
+if test "$ax_cv_have_aligned_access_required" = yes ; then
+  AC_DEFINE([HAVE_ALIGNED_ACCESS_REQUIRED], [1],
+    [Define if pointers to integers require aligned access])
+fi
+])

m4/ax_check_compile_flag.m4

@@ -0,0 +1,73 @@
+# ===========================================================================
+#   http://www.gnu.org/software/autoconf-archive/ax_check_compile_flag.html
+# ===========================================================================
+#
+# SYNOPSIS
+#
+#   AX_CHECK_COMPILE_FLAG(FLAG, [ACTION-SUCCESS], [ACTION-FAILURE], [EXTRA-FLAGS])
+#
+# DESCRIPTION
+#
+#   Check whether the given FLAG works with the current language's compiler
+#   or gives an error.  (Warnings, however, are ignored)
+#
+#   ACTION-SUCCESS/ACTION-FAILURE are shell commands to execute on
+#   success/failure.
+#
+#   If EXTRA-FLAGS is defined, it is added to the current language's default
+#   flags (e.g. CFLAGS) when the check is done.  The check is thus made with
+#   the flags: "CFLAGS EXTRA-FLAGS FLAG".  This can for example be used to
+#   force the compiler to issue an error when a bad flag is given.
+#
+#   NOTE: Implementation based on AX_CFLAGS_GCC_OPTION. Please keep this
+#   macro in sync with AX_CHECK_{PREPROC,LINK}_FLAG.
+#
+# LICENSE
+#
+#   Copyright (c) 2008 Guido U. Draheim <guidod@gmx.de>
+#   Copyright (c) 2011 Maarten Bosmans <mkbosmans@gmail.com>
+#
+#   This program is free software: you can redistribute it and/or modify it
+#   under the terms of the GNU General Public License as published by the
+#   Free Software Foundation, either version 3 of the License, or (at your
+#   option) any later version.
+#
+#   This program is distributed in the hope that it will be useful, but
+#   WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
+#   Public License for more details.
+#
+#   You should have received a copy of the GNU General Public License along
+#   with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+#   As a special exception, the respective Autoconf Macro's copyright owner
+#   gives unlimited permission to copy, distribute and modify the configure
+#   scripts that are the output of Autoconf when processing the Macro. You
+#   need not follow the terms of the GNU General Public License when using
+#   or distributing such scripts, even though portions of the text of the
+#   Macro appear in them. The GNU General Public License (GPL) does govern
+#   all other use of the material that constitutes the Autoconf Macro.
+#
+#   This special exception to the GPL applies to versions of the Autoconf
+#   Macro released by the Autoconf Archive. When you make and distribute a
+#   modified version of the Autoconf Macro, you may extend this special
+#   exception to the GPL to apply to your modified version as well.
+
+#serial 2
+
+AC_DEFUN([AX_CHECK_COMPILE_FLAG],
+[AC_PREREQ(2.59)dnl for _AC_LANG_PREFIX
+AS_VAR_PUSHDEF([CACHEVAR],[ax_cv_check_[]_AC_LANG_ABBREV[]flags_$4_$1])dnl
+AC_CACHE_CHECK([whether _AC_LANG compiler accepts $1], CACHEVAR, [
+  ax_check_save_flags=$[]_AC_LANG_PREFIX[]FLAGS
+  _AC_LANG_PREFIX[]FLAGS="$[]_AC_LANG_PREFIX[]FLAGS $4 $1"
+  AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <stdio.h>]],
+  [[char x[42U], fodder = 0;if (fodder > -1000 && fgets(x,1000,stdin)) puts(x)]])],
+    [AS_VAR_SET(CACHEVAR,[yes])],
+    [AS_VAR_SET(CACHEVAR,[no])])
+  _AC_LANG_PREFIX[]FLAGS=$ax_check_save_flags])
+AS_IF([test x"AS_VAR_GET(CACHEVAR)" = xyes],
+  [m4_default([$2], :)],
+  [m4_default([$3], :)])
+AS_VAR_POPDEF([CACHEVAR])dnl
+])dnl AX_CHECK_COMPILE_FLAGS

m4/ax_check_link_flag.m4

@@ -0,0 +1,71 @@
+# ===========================================================================
+#    http://www.gnu.org/software/autoconf-archive/ax_check_link_flag.html
+# ===========================================================================
+#
+# SYNOPSIS
+#
+#   AX_CHECK_LINK_FLAG(FLAG, [ACTION-SUCCESS], [ACTION-FAILURE], [EXTRA-FLAGS])
+#
+# DESCRIPTION
+#
+#   Check whether the given FLAG works with the linker or gives an error.
+#   (Warnings, however, are ignored)
+#
+#   ACTION-SUCCESS/ACTION-FAILURE are shell commands to execute on
+#   success/failure.
+#
+#   If EXTRA-FLAGS is defined, it is added to the linker's default flags
+#   when the check is done.  The check is thus made with the flags: "LDFLAGS
+#   EXTRA-FLAGS FLAG".  This can for example be used to force the linker to
+#   issue an error when a bad flag is given.
+#
+#   NOTE: Implementation based on AX_CFLAGS_GCC_OPTION. Please keep this
+#   macro in sync with AX_CHECK_{PREPROC,COMPILE}_FLAG.
+#
+# LICENSE
+#
+#   Copyright (c) 2008 Guido U. Draheim <guidod@gmx.de>
+#   Copyright (c) 2011 Maarten Bosmans <mkbosmans@gmail.com>
+#
+#   This program is free software: you can redistribute it and/or modify it
+#   under the terms of the GNU General Public License as published by the
+#   Free Software Foundation, either version 3 of the License, or (at your
+#   option) any later version.
+#
+#   This program is distributed in the hope that it will be useful, but
+#   WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
+#   Public License for more details.
+#
+#   You should have received a copy of the GNU General Public License along
+#   with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+#   As a special exception, the respective Autoconf Macro's copyright owner
+#   gives unlimited permission to copy, distribute and modify the configure
+#   scripts that are the output of Autoconf when processing the Macro. You
+#   need not follow the terms of the GNU General Public License when using
+#   or distributing such scripts, even though portions of the text of the
+#   Macro appear in them. The GNU General Public License (GPL) does govern
+#   all other use of the material that constitutes the Autoconf Macro.
+#
+#   This special exception to the GPL applies to versions of the Autoconf
+#   Macro released by the Autoconf Archive. When you make and distribute a
+#   modified version of the Autoconf Macro, you may extend this special
+#   exception to the GPL to apply to your modified version as well.
+
+#serial 2
+
+AC_DEFUN([AX_CHECK_LINK_FLAG],
+[AS_VAR_PUSHDEF([CACHEVAR],[ax_cv_check_ldflags_$4_$1])dnl
+AC_CACHE_CHECK([whether the linker accepts $1], CACHEVAR, [
+  ax_check_save_flags=$LDFLAGS
+  LDFLAGS="$LDFLAGS $4 $1"
+  AC_LINK_IFELSE([AC_LANG_PROGRAM()],
+    [AS_VAR_SET(CACHEVAR,[yes])],
+    [AS_VAR_SET(CACHEVAR,[no])])
+  LDFLAGS=$ax_check_save_flags])
+AS_IF([test x"AS_VAR_GET(CACHEVAR)" = xyes],
+  [m4_default([$2], :)],
+  [m4_default([$3], :)])
+AS_VAR_POPDEF([CACHEVAR])dnl
+])dnl AX_CHECK_LINK_FLAGS

m4/ax_ext.m4

@@ -0,0 +1,213 @@
+# ===========================================================================
+#          http://www.gnu.org/software/autoconf-archive/ax_ext.html
+# ===========================================================================
+#
+# SYNOPSIS
+#
+#   AX_EXT
+#
+# DESCRIPTION
+#
+#   Find supported SIMD extensions by requesting cpuid. When an SIMD
+#   extension is found, the -m"simdextensionname" is added to SIMD_FLAGS if
+#   compilator supports it. For example, if "sse2" is available, then
+#   "-msse2" is added to SIMD_FLAGS.
+#
+#   This macro calls:
+#
+#     AC_SUBST(SIMD_FLAGS)
+#
+#   And defines:
+#
+#     HAVE_MMX / HAVE_SSE / HAVE_SSE2 / HAVE_SSE3 / HAVE_SSSE3 / HAVE_SSE4.1 / HAVE_SSE4.2 / HAVE_AVX
+#
+# LICENSE
+#
+#   Copyright (c) 2007 Christophe Tournayre <turn3r@users.sourceforge.net>
+#
+#   Copying and distribution of this file, with or without modification, are
+#   permitted in any medium without royalty provided the copyright notice
+#   and this notice are preserved. This file is offered as-is, without any
+#   warranty.
+
+#serial 10
+
+AC_DEFUN([AX_EXT],
+[
+  AC_REQUIRE([AC_CANONICAL_HOST])
+
+  case $host_cpu in
+    powerpc*)
+      AC_CACHE_CHECK([whether altivec is supported], [ax_cv_have_altivec_ext],
+          [
+            if test `/usr/sbin/sysctl -a 2>/dev/null| grep -c hw.optional.altivec` != 0; then
+                if test `/usr/sbin/sysctl -n hw.optional.altivec` = 1; then
+                  ax_cv_have_altivec_ext=yes
+                fi
+            fi
+          ])
+
+          if test "$ax_cv_have_altivec_ext" = yes; then
+            AC_DEFINE(HAVE_ALTIVEC,,[Support Altivec instructions])
+            AX_CHECK_COMPILE_FLAG(-faltivec, SIMD_FLAGS="$SIMD_FLAGS -faltivec", [])
+          fi
+    ;;
+
+
+    i[[3456]]86*|x86_64*)
+
+      AC_REQUIRE([AX_GCC_X86_CPUID])
+
+      AX_GCC_X86_CPUID(0x00000001)
+      ecx=`echo $ax_cv_gcc_x86_cpuid_0x00000001 | cut -d ":" -f 3`
+      edx=`echo $ax_cv_gcc_x86_cpuid_0x00000001 | cut -d ":" -f 4`
+
+      AC_CACHE_CHECK([whether mmx is supported], [ax_cv_have_mmx_ext],
+      [
+        ax_cv_have_mmx_ext=no
+        if test "$((0x$edx>>23&0x01))" = 1; then
+          ax_cv_have_mmx_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether sse is supported], [ax_cv_have_sse_ext],
+      [
+        ax_cv_have_sse_ext=no
+        if test "$((0x$edx>>25&0x01))" = 1; then
+          ax_cv_have_sse_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether sse2 is supported], [ax_cv_have_sse2_ext],
+      [
+        ax_cv_have_sse2_ext=no
+        if test "$((0x$edx>>26&0x01))" = 1; then
+          ax_cv_have_sse2_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether sse3 is supported], [ax_cv_have_sse3_ext],
+      [
+        ax_cv_have_sse3_ext=no
+        if test "$((0x$ecx&0x01))" = 1; then
+          ax_cv_have_sse3_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether ssse3 is supported], [ax_cv_have_ssse3_ext],
+      [
+        ax_cv_have_ssse3_ext=no
+        if test "$((0x$ecx>>9&0x01))" = 1; then
+          ax_cv_have_ssse3_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether sse4.1 is supported], [ax_cv_have_sse41_ext],
+      [
+        ax_cv_have_sse41_ext=no
+        if test "$((0x$ecx>>19&0x01))" = 1; then
+          ax_cv_have_sse41_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether sse4.2 is supported], [ax_cv_have_sse42_ext],
+      [
+        ax_cv_have_sse42_ext=no
+        if test "$((0x$ecx>>20&0x01))" = 1; then
+          ax_cv_have_sse42_ext=yes
+        fi
+      ])
+
+      AC_CACHE_CHECK([whether avx is supported], [ax_cv_have_avx_ext],
+      [
+        ax_cv_have_avx_ext=no
+        if test "$((0x$ecx>>28&0x01))" = 1; then
+          ax_cv_have_avx_ext=yes
+        fi
+      ])
+
+      if test "$ax_cv_have_mmx_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-mmmx, ax_cv_support_mmx_ext=yes, [])
+        if test x"$ax_cv_support_mmx_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -mmmx"
+          AC_DEFINE(HAVE_MMX,,[Support mmx instructions])
+        else
+          AC_MSG_WARN([Your processor support mmx instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_sse_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-msse, ax_cv_support_sse_ext=yes, [])
+        if test x"$ax_cv_support_sse_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -msse"
+          AC_DEFINE(HAVE_SSE,,[Support SSE (Streaming SIMD Extensions) instructions])
+        else
+          AC_MSG_WARN([Your processor support sse instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_sse2_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-msse2, ax_cv_support_sse2_ext=yes, [])
+        if test x"$ax_cv_support_sse2_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -msse2"
+          AC_DEFINE(HAVE_SSE2,,[Support SSE2 (Streaming SIMD Extensions 2) instructions])
+        else
+          AC_MSG_WARN([Your processor support sse2 instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_sse3_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-msse3, ax_cv_support_sse3_ext=yes, [])
+        if test x"$ax_cv_support_sse3_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -msse3"
+          AC_DEFINE(HAVE_SSE3,,[Support SSE3 (Streaming SIMD Extensions 3) instructions])
+        else
+          AC_MSG_WARN([Your processor support sse3 instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_ssse3_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-mssse3, ax_cv_support_ssse3_ext=yes, [])
+        if test x"$ax_cv_support_ssse3_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -mssse3"
+          AC_DEFINE(HAVE_SSSE3,,[Support SSSE3 (Supplemental Streaming SIMD Extensions 3) instructions])
+        else
+          AC_MSG_WARN([Your processor support ssse3 instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_sse41_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-msse4.1, ax_cv_support_sse41_ext=yes, [])
+        if test x"$ax_cv_support_sse41_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -msse4.1"
+          AC_DEFINE(HAVE_SSE4_1,,[Support SSSE4.1 (Streaming SIMD Extensions 4.1) instructions])
+        else
+          AC_MSG_WARN([Your processor support sse4.1 instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_sse42_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-msse4.2, ax_cv_support_sse42_ext=yes, [])
+        if test x"$ax_cv_support_sse42_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -msse4.2"
+          AC_DEFINE(HAVE_SSE4_2,,[Support SSSE4.2 (Streaming SIMD Extensions 4.2) instructions])
+        else
+          AC_MSG_WARN([Your processor support sse4.2 instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+      if test "$ax_cv_have_avx_ext" = yes; then
+        AX_CHECK_COMPILE_FLAG(-mavx, ax_cv_support_avx_ext=yes, [])
+        if test x"$ax_cv_support_avx_ext" = x"yes"; then
+          SIMD_FLAGS="$SIMD_FLAGS -mavx"
+          AC_DEFINE(HAVE_AVX,,[Support AVX (Advanced Vector Extensions) instructions])
+        else
+          AC_MSG_WARN([Your processor support avx instructions but not your compilor, can you try another compilor ?])
+        fi
+      fi
+
+  ;;
+  esac
+
+  AC_SUBST(SIMD_FLAGS)
+])

m4/ax_forceinline.m4

@@ -0,0 +1,60 @@
+# ===========================================================================
+#      http://www.gnu.org/software/autoconf-archive/ax_forceinline.html
+# ===========================================================================
+#
+# SYNOPSIS
+#
+#   AX_FORCEINLINE()
+#
+# DESCRIPTION
+#
+#   Provides a test for C compiler support of forced inlining.  If usable,
+#   #define FORCEINLINE to the appropriate force inline keyword.  Otherwise
+#   #define FORCEINLINE to be 'inline'.
+#
+# LICENSE
+#
+#   Copyright (c) 2008 Alan Woodland <ajw05@aber.ac.uk>
+#   Copyright (c) 2009 Rhys Ulerich <rhys.ulerich@gmail.com>
+#
+#   Copying and distribution of this file, with or without modification, are
+#   permitted in any medium without royalty provided the copyright notice
+#   and this notice are preserved. This file is offered as-is, without any
+#   warranty.
+
+#serial 2
+
+AC_DEFUN([AX_FORCEINLINE], [
+  AC_LANG_PUSH([C])
+  AC_MSG_CHECKING(for forced inline keyword)
+  AC_CACHE_VAL(ac_cv_forceinline, [
+    ax_forceinline_keywords="__forceinline inline none"
+    for ax_forceinline_keyword in $ax_forceinline_keywords; do
+       case $ax_forceinline_keyword in
+          none) ac_cv_forceinline=none ; break ;;
+      *)
+             AC_TRY_COMPILE(
+                [#include <stdlib.h>
+                 ] $ax_forceinline_keyword [
+                 static void
+                 foo(void) {
+                 exit(1);
+                 }],
+                 [],
+                 [ac_cv_forceinline=$ax_forceinline_keyword ; break],
+                 ac_cv_forceinline=none
+             )
+      esac
+    done
+])
+
+  if test "$ac_cv_forceinline" = "none"; then
+    ax_forceinline_keyword=
+  else
+    ax_forceinline_keyword=$ac_cv_forceinline
+  fi
+  AC_DEFINE_UNQUOTED([FORCEINLINE],$ax_forceinline_keyword,
+    [The most forceful inline keyword known by the compiler])
+  AC_MSG_RESULT($ac_cv_forceinline)
+  AC_LANG_POP([C])
+])

m4/ax_gcc_x86_cpuid.m4

@@ -0,0 +1,79 @@
+# ===========================================================================
+#     http://www.gnu.org/software/autoconf-archive/ax_gcc_x86_cpuid.html
+# ===========================================================================
+#
+# SYNOPSIS
+#
+#   AX_GCC_X86_CPUID(OP)
+#
+# DESCRIPTION
+#
+#   On Pentium and later x86 processors, with gcc or a compiler that has a
+#   compatible syntax for inline assembly instructions, run a small program
+#   that executes the cpuid instruction with input OP. This can be used to
+#   detect the CPU type.
+#
+#   On output, the values of the eax, ebx, ecx, and edx registers are stored
+#   as hexadecimal strings as "eax:ebx:ecx:edx" in the cache variable
+#   ax_cv_gcc_x86_cpuid_OP.
+#
+#   If the cpuid instruction fails (because you are running a
+#   cross-compiler, or because you are not using gcc, or because you are on
+#   a processor that doesn't have this instruction), ax_cv_gcc_x86_cpuid_OP
+#   is set to the string "unknown".
+#
+#   This macro mainly exists to be used in AX_GCC_ARCHFLAG.
+#
+# LICENSE
+#
+#   Copyright (c) 2008 Steven G. Johnson <stevenj@alum.mit.edu>
+#   Copyright (c) 2008 Matteo Frigo
+#
+#   This program is free software: you can redistribute it and/or modify it
+#   under the terms of the GNU General Public License as published by the
+#   Free Software Foundation, either version 3 of the License, or (at your
+#   option) any later version.
+#
+#   This program is distributed in the hope that it will be useful, but
+#   WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
+#   Public License for more details.
+#
+#   You should have received a copy of the GNU General Public License along
+#   with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+#   As a special exception, the respective Autoconf Macro's copyright owner
+#   gives unlimited permission to copy, distribute and modify the configure
+#   scripts that are the output of Autoconf when processing the Macro. You
+#   need not follow the terms of the GNU General Public License when using
+#   or distributing such scripts, even though portions of the text of the
+#   Macro appear in them. The GNU General Public License (GPL) does govern
+#   all other use of the material that constitutes the Autoconf Macro.
+#
+#   This special exception to the GPL applies to versions of the Autoconf
+#   Macro released by the Autoconf Archive. When you make and distribute a
+#   modified version of the Autoconf Macro, you may extend this special
+#   exception to the GPL to apply to your modified version as well.
+
+#serial 7
+
+AC_DEFUN([AX_GCC_X86_CPUID],
+[AC_REQUIRE([AC_PROG_CC])
+AC_LANG_PUSH([C])
+AC_CACHE_CHECK(for x86 cpuid $1 output, ax_cv_gcc_x86_cpuid_$1,
+ [AC_RUN_IFELSE([AC_LANG_PROGRAM([#include <stdio.h>], [
+     int op = $1, eax, ebx, ecx, edx;
+     FILE *f;
+      __asm__("cpuid"
+        : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx)
+        : "a" (op));
+     f = fopen("conftest_cpuid", "w"); if (!f) return 1;
+     fprintf(f, "%x:%x:%x:%x\n", eax, ebx, ecx, edx);
+     fclose(f);
+     return 0;
+])],
+     [ax_cv_gcc_x86_cpuid_$1=`cat conftest_cpuid`; rm -f conftest_cpuid],
+     [ax_cv_gcc_x86_cpuid_$1=unknown; rm -f conftest_cpuid],
+     [ax_cv_gcc_x86_cpuid_$1=unknown])])
+AC_LANG_POP([C])
+])

m4/ltoptions.m4

@@ -0,0 +1,384 @@
+# Helper functions for option handling.                    -*- Autoconf -*-
+#
+#   Copyright (C) 2004, 2005, 2007, 2008, 2009 Free Software Foundation,
+#   Inc.
+#   Written by Gary V. Vaughan, 2004
+#
+# This file is free software; the Free Software Foundation gives
+# unlimited permission to copy and/or distribute it, with or without
+# modifications, as long as this notice is preserved.
+
+# serial 7 ltoptions.m4
+
+# This is to help aclocal find these macros, as it can't see m4_define.
+AC_DEFUN([LTOPTIONS_VERSION], [m4_if([1])])
+
+
+# _LT_MANGLE_OPTION(MACRO-NAME, OPTION-NAME)
+# ------------------------------------------
+m4_define([_LT_MANGLE_OPTION],
+[[_LT_OPTION_]m4_bpatsubst($1__$2, [[^a-zA-Z0-9_]], [_])])
+
+
+# _LT_SET_OPTION(MACRO-NAME, OPTION-NAME)
+# ---------------------------------------
+# Set option OPTION-NAME for macro MACRO-NAME, and if there is a
+# matching handler defined, dispatch to it.  Other OPTION-NAMEs are
+# saved as a flag.
+m4_define([_LT_SET_OPTION],
+[m4_define(_LT_MANGLE_OPTION([$1], [$2]))dnl
+m4_ifdef(_LT_MANGLE_DEFUN([$1], [$2]),
+        _LT_MANGLE_DEFUN([$1], [$2]),
+    [m4_warning([Unknown $1 option `$2'])])[]dnl
+])
+
+
+# _LT_IF_OPTION(MACRO-NAME, OPTION-NAME, IF-SET, [IF-NOT-SET])
+# ------------------------------------------------------------
+# Execute IF-SET if OPTION is set, IF-NOT-SET otherwise.
+m4_define([_LT_IF_OPTION],
+[m4_ifdef(_LT_MANGLE_OPTION([$1], [$2]), [$3], [$4])])
+
+
+# _LT_UNLESS_OPTIONS(MACRO-NAME, OPTION-LIST, IF-NOT-SET)
+# -------------------------------------------------------
+# Execute IF-NOT-SET unless all options in OPTION-LIST for MACRO-NAME
+# are set.
+m4_define([_LT_UNLESS_OPTIONS],
+[m4_foreach([_LT_Option], m4_split(m4_normalize([$2])),
+	    [m4_ifdef(_LT_MANGLE_OPTION([$1], _LT_Option),
+		      [m4_define([$0_found])])])[]dnl
+m4_ifdef([$0_found], [m4_undefine([$0_found])], [$3
+])[]dnl
+])
+
+
+# _LT_SET_OPTIONS(MACRO-NAME, OPTION-LIST)
+# ----------------------------------------
+# OPTION-LIST is a space-separated list of Libtool options associated
+# with MACRO-NAME.  If any OPTION has a matching handler declared with
+# LT_OPTION_DEFINE, dispatch to that macro; otherwise complain about
+# the unknown option and exit.
+m4_defun([_LT_SET_OPTIONS],
+[# Set options
+m4_foreach([_LT_Option], m4_split(m4_normalize([$2])),
+    [_LT_SET_OPTION([$1], _LT_Option)])
+
+m4_if([$1],[LT_INIT],[
+  dnl
+  dnl Simply set some default values (i.e off) if boolean options were not
+  dnl specified:
+  _LT_UNLESS_OPTIONS([LT_INIT], [dlopen], [enable_dlopen=no
+  ])
+  _LT_UNLESS_OPTIONS([LT_INIT], [win32-dll], [enable_win32_dll=no
+  ])
+  dnl
+  dnl If no reference was made to various pairs of opposing options, then
+  dnl we run the default mode handler for the pair.  For example, if neither
+  dnl `shared' nor `disable-shared' was passed, we enable building of shared
+  dnl archives by default:
+  _LT_UNLESS_OPTIONS([LT_INIT], [shared disable-shared], [_LT_ENABLE_SHARED])
+  _LT_UNLESS_OPTIONS([LT_INIT], [static disable-static], [_LT_ENABLE_STATIC])
+  _LT_UNLESS_OPTIONS([LT_INIT], [pic-only no-pic], [_LT_WITH_PIC])
+  _LT_UNLESS_OPTIONS([LT_INIT], [fast-install disable-fast-install],
+  		   [_LT_ENABLE_FAST_INSTALL])
+  ])
+])# _LT_SET_OPTIONS
+
+
+## --------------------------------- ##
+## Macros to handle LT_INIT options. ##
+## --------------------------------- ##
+
+# _LT_MANGLE_DEFUN(MACRO-NAME, OPTION-NAME)
+# -----------------------------------------
+m4_define([_LT_MANGLE_DEFUN],
+[[_LT_OPTION_DEFUN_]m4_bpatsubst(m4_toupper([$1__$2]), [[^A-Z0-9_]], [_])])
+
+
+# LT_OPTION_DEFINE(MACRO-NAME, OPTION-NAME, CODE)
+# -----------------------------------------------
+m4_define([LT_OPTION_DEFINE],
+[m4_define(_LT_MANGLE_DEFUN([$1], [$2]), [$3])[]dnl
+])# LT_OPTION_DEFINE
+
+
+# dlopen
+# ------
+LT_OPTION_DEFINE([LT_INIT], [dlopen], [enable_dlopen=yes
+])
+
+AU_DEFUN([AC_LIBTOOL_DLOPEN],
+[_LT_SET_OPTION([LT_INIT], [dlopen])
+AC_DIAGNOSE([obsolete],
+[$0: Remove this warning and the call to _LT_SET_OPTION when you
+put the `dlopen' option into LT_INIT's first parameter.])
+])
+
+dnl aclocal-1.4 backwards compatibility:
+dnl AC_DEFUN([AC_LIBTOOL_DLOPEN], [])
+
+
+# win32-dll
+# ---------
+# Declare package support for building win32 dll's.
+LT_OPTION_DEFINE([LT_INIT], [win32-dll],
+[enable_win32_dll=yes
+
+case $host in
+*-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-cegcc*)
+  AC_CHECK_TOOL(AS, as, false)
+  AC_CHECK_TOOL(DLLTOOL, dlltool, false)
+  AC_CHECK_TOOL(OBJDUMP, objdump, false)
+  ;;
+esac
+
+test -z "$AS" && AS=as
+_LT_DECL([], [AS],      [1], [Assembler program])dnl
+
+test -z "$DLLTOOL" && DLLTOOL=dlltool
+_LT_DECL([], [DLLTOOL], [1], [DLL creation program])dnl
+
+test -z "$OBJDUMP" && OBJDUMP=objdump
+_LT_DECL([], [OBJDUMP], [1], [Object dumper program])dnl
+])# win32-dll
+
+AU_DEFUN([AC_LIBTOOL_WIN32_DLL],
+[AC_REQUIRE([AC_CANONICAL_HOST])dnl
+_LT_SET_OPTION([LT_INIT], [win32-dll])
+AC_DIAGNOSE([obsolete],
+[$0: Remove this warning and the call to _LT_SET_OPTION when you
+put the `win32-dll' option into LT_INIT's first parameter.])
+])
+
+dnl aclocal-1.4 backwards compatibility:
+dnl AC_DEFUN([AC_LIBTOOL_WIN32_DLL], [])
+
+
+# _LT_ENABLE_SHARED([DEFAULT])
+# ----------------------------
+# implement the --enable-shared flag, and supports the `shared' and
+# `disable-shared' LT_INIT options.
+# DEFAULT is either `yes' or `no'.  If omitted, it defaults to `yes'.
+m4_define([_LT_ENABLE_SHARED],
+[m4_define([_LT_ENABLE_SHARED_DEFAULT], [m4_if($1, no, no, yes)])dnl
+AC_ARG_ENABLE([shared],
+    [AS_HELP_STRING([--enable-shared@<:@=PKGS@:>@],
+	[build shared libraries @<:@default=]_LT_ENABLE_SHARED_DEFAULT[@:>@])],
+    [p=${PACKAGE-default}
+    case $enableval in
+    yes) enable_shared=yes ;;
+    no) enable_shared=no ;;
+    *)
+      enable_shared=no
+      # Look at the argument we got.  We use all the common list separators.
+      lt_save_ifs="$IFS"; IFS="${IFS}$PATH_SEPARATOR,"
+      for pkg in $enableval; do
+	IFS="$lt_save_ifs"
+	if test "X$pkg" = "X$p"; then
+	  enable_shared=yes
+	fi
+      done
+      IFS="$lt_save_ifs"
+      ;;
+    esac],
+    [enable_shared=]_LT_ENABLE_SHARED_DEFAULT)
+
+    _LT_DECL([build_libtool_libs], [enable_shared], [0],
+	[Whether or not to build shared libraries])
+])# _LT_ENABLE_SHARED
+
+LT_OPTION_DEFINE([LT_INIT], [shared], [_LT_ENABLE_SHARED([yes])])
+LT_OPTION_DEFINE([LT_INIT], [disable-shared], [_LT_ENABLE_SHARED([no])])
+
+# Old names:
+AC_DEFUN([AC_ENABLE_SHARED],
+[_LT_SET_OPTION([LT_INIT], m4_if([$1], [no], [disable-])[shared])
+])
+
+AC_DEFUN([AC_DISABLE_SHARED],
+[_LT_SET_OPTION([LT_INIT], [disable-shared])
+])
+
+AU_DEFUN([AM_ENABLE_SHARED], [AC_ENABLE_SHARED($@)])
+AU_DEFUN([AM_DISABLE_SHARED], [AC_DISABLE_SHARED($@)])
+
+dnl aclocal-1.4 backwards compatibility:
+dnl AC_DEFUN([AM_ENABLE_SHARED], [])
+dnl AC_DEFUN([AM_DISABLE_SHARED], [])
+
+
+
+# _LT_ENABLE_STATIC([DEFAULT])
+# ----------------------------
+# implement the --enable-static flag, and support the `static' and
+# `disable-static' LT_INIT options.
+# DEFAULT is either `yes' or `no'.  If omitted, it defaults to `yes'.
+m4_define([_LT_ENABLE_STATIC],
+[m4_define([_LT_ENABLE_STATIC_DEFAULT], [m4_if($1, no, no, yes)])dnl
+AC_ARG_ENABLE([static],
+    [AS_HELP_STRING([--enable-static@<:@=PKGS@:>@],
+	[build static libraries @<:@default=]_LT_ENABLE_STATIC_DEFAULT[@:>@])],
+    [p=${PACKAGE-default}
+    case $enableval in
+    yes) enable_static=yes ;;
+    no) enable_static=no ;;
+    *)
+     enable_static=no
+      # Look at the argument we got.  We use all the common list separators.
+      lt_save_ifs="$IFS"; IFS="${IFS}$PATH_SEPARATOR,"
+      for pkg in $enableval; do
+	IFS="$lt_save_ifs"
+	if test "X$pkg" = "X$p"; then
+	  enable_static=yes
+	fi
+      done
+      IFS="$lt_save_ifs"
+      ;;
+    esac],
+    [enable_static=]_LT_ENABLE_STATIC_DEFAULT)
+
+    _LT_DECL([build_old_libs], [enable_static], [0],
+	[Whether or not to build static libraries])
+])# _LT_ENABLE_STATIC
+
+LT_OPTION_DEFINE([LT_INIT], [static], [_LT_ENABLE_STATIC([yes])])
+LT_OPTION_DEFINE([LT_INIT], [disable-static], [_LT_ENABLE_STATIC([no])])
+
+# Old names:
+AC_DEFUN([AC_ENABLE_STATIC],
+[_LT_SET_OPTION([LT_INIT], m4_if([$1], [no], [disable-])[static])
+])
+
+AC_DEFUN([AC_DISABLE_STATIC],
+[_LT_SET_OPTION([LT_INIT], [disable-static])
+])
+
+AU_DEFUN([AM_ENABLE_STATIC], [AC_ENABLE_STATIC($@)])
+AU_DEFUN([AM_DISABLE_STATIC], [AC_DISABLE_STATIC($@)])
+
+dnl aclocal-1.4 backwards compatibility:
+dnl AC_DEFUN([AM_ENABLE_STATIC], [])
+dnl AC_DEFUN([AM_DISABLE_STATIC], [])
+
+
+
+# _LT_ENABLE_FAST_INSTALL([DEFAULT])
+# ----------------------------------
+# implement the --enable-fast-install flag, and support the `fast-install'
+# and `disable-fast-install' LT_INIT options.
+# DEFAULT is either `yes' or `no'.  If omitted, it defaults to `yes'.
+m4_define([_LT_ENABLE_FAST_INSTALL],
+[m4_define([_LT_ENABLE_FAST_INSTALL_DEFAULT], [m4_if($1, no, no, yes)])dnl
+AC_ARG_ENABLE([fast-install],
+    [AS_HELP_STRING([--enable-fast-install@<:@=PKGS@:>@],
+    [optimize for fast installation @<:@default=]_LT_ENABLE_FAST_INSTALL_DEFAULT[@:>@])],
+    [p=${PACKAGE-default}
+    case $enableval in
+    yes) enable_fast_install=yes ;;
+    no) enable_fast_install=no ;;
+    *)
+      enable_fast_install=no
+      # Look at the argument we got.  We use all the common list separators.
+      lt_save_ifs="$IFS"; IFS="${IFS}$PATH_SEPARATOR,"
+      for pkg in $enableval; do
+	IFS="$lt_save_ifs"
+	if test "X$pkg" = "X$p"; then
+	  enable_fast_install=yes
+	fi
+      done
+      IFS="$lt_save_ifs"
+      ;;
+    esac],
+    [enable_fast_install=]_LT_ENABLE_FAST_INSTALL_DEFAULT)
+
+_LT_DECL([fast_install], [enable_fast_install], [0],
+	 [Whether or not to optimize for fast installation])dnl
+])# _LT_ENABLE_FAST_INSTALL
+
+LT_OPTION_DEFINE([LT_INIT], [fast-install], [_LT_ENABLE_FAST_INSTALL([yes])])
+LT_OPTION_DEFINE([LT_INIT], [disable-fast-install], [_LT_ENABLE_FAST_INSTALL([no])])
+
+# Old names:
+AU_DEFUN([AC_ENABLE_FAST_INSTALL],
+[_LT_SET_OPTION([LT_INIT], m4_if([$1], [no], [disable-])[fast-install])
+AC_DIAGNOSE([obsolete],
+[$0: Remove this warning and the call to _LT_SET_OPTION when you put
+the `fast-install' option into LT_INIT's first parameter.])
+])
+
+AU_DEFUN([AC_DISABLE_FAST_INSTALL],
+[_LT_SET_OPTION([LT_INIT], [disable-fast-install])
+AC_DIAGNOSE([obsolete],
+[$0: Remove this warning and the call to _LT_SET_OPTION when you put
+the `disable-fast-install' option into LT_INIT's first parameter.])
+])
+
+dnl aclocal-1.4 backwards compatibility:
+dnl AC_DEFUN([AC_ENABLE_FAST_INSTALL], [])
+dnl AC_DEFUN([AM_DISABLE_FAST_INSTALL], [])
+
+
+# _LT_WITH_PIC([MODE])
+# --------------------
+# implement the --with-pic flag, and support the `pic-only' and `no-pic'
+# LT_INIT options.
+# MODE is either `yes' or `no'.  If omitted, it defaults to `both'.
+m4_define([_LT_WITH_PIC],
+[AC_ARG_WITH([pic],
+    [AS_HELP_STRING([--with-pic@<:@=PKGS@:>@],
+	[try to use only PIC/non-PIC objects @<:@default=use both@:>@])],
+    [lt_p=${PACKAGE-default}
+    case $withval in
+    yes|no) pic_mode=$withval ;;
+    *)
+      pic_mode=default
+      # Look at the argument we got.  We use all the common list separators.
+      lt_save_ifs="$IFS"; IFS="${IFS}$PATH_SEPARATOR,"
+      for lt_pkg in $withval; do
+	IFS="$lt_save_ifs"
+	if test "X$lt_pkg" = "X$lt_p"; then
+	  pic_mode=yes
+	fi
+      done
+      IFS="$lt_save_ifs"
+      ;;
+    esac],
+    [pic_mode=default])
+
+test -z "$pic_mode" && pic_mode=m4_default([$1], [default])
+
+_LT_DECL([], [pic_mode], [0], [What type of objects to build])dnl
+])# _LT_WITH_PIC
+
+LT_OPTION_DEFINE([LT_INIT], [pic-only], [_LT_WITH_PIC([yes])])
+LT_OPTION_DEFINE([LT_INIT], [no-pic], [_LT_WITH_PIC([no])])
+
+# Old name:
+AU_DEFUN([AC_LIBTOOL_PICMODE],
+[_LT_SET_OPTION([LT_INIT], [pic-only])
+AC_DIAGNOSE([obsolete],
+[$0: Remove this warning and the call to _LT_SET_OPTION when you
+put the `pic-only' option into LT_INIT's first parameter.])
+])
+
+dnl aclocal-1.4 backwards compatibility:
+dnl AC_DEFUN([AC_LIBTOOL_PICMODE], [])
+
+## ----------------- ##
+## LTDL_INIT Options ##
+## ----------------- ##
+
+m4_define([_LTDL_MODE], [])
+LT_OPTION_DEFINE([LTDL_INIT], [nonrecursive],
+		 [m4_define([_LTDL_MODE], [nonrecursive])])
+LT_OPTION_DEFINE([LTDL_INIT], [recursive],
+		 [m4_define([_LTDL_MODE], [recursive])])
+LT_OPTION_DEFINE([LTDL_INIT], [subproject],
+		 [m4_define([_LTDL_MODE], [subproject])])
+
+m4_define([_LTDL_TYPE], [])
+LT_OPTION_DEFINE([LTDL_INIT], [installable],
+		 [m4_define([_LTDL_TYPE], [installable])])
+LT_OPTION_DEFINE([LTDL_INIT], [convenience],
+		 [m4_define([_LTDL_TYPE], [convenience])])

m4/ltsugar.m4

@@ -0,0 +1,123 @@
+# ltsugar.m4 -- libtool m4 base layer.                         -*-Autoconf-*-
+#
+# Copyright (C) 2004, 2005, 2007, 2008 Free Software Foundation, Inc.
+# Written by Gary V. Vaughan, 2004
+#
+# This file is free software; the Free Software Foundation gives
+# unlimited permission to copy and/or distribute it, with or without
+# modifications, as long as this notice is preserved.
+
+# serial 6 ltsugar.m4
+
+# This is to help aclocal find these macros, as it can't see m4_define.
+AC_DEFUN([LTSUGAR_VERSION], [m4_if([0.1])])
+
+
+# lt_join(SEP, ARG1, [ARG2...])
+# -----------------------------
+# Produce ARG1SEPARG2...SEPARGn, omitting [] arguments and their
+# associated separator.
+# Needed until we can rely on m4_join from Autoconf 2.62, since all earlier
+# versions in m4sugar had bugs.
+m4_define([lt_join],
+[m4_if([$#], [1], [],
+       [$#], [2], [[$2]],
+       [m4_if([$2], [], [], [[$2]_])$0([$1], m4_shift(m4_shift($@)))])])
+m4_define([_lt_join],
+[m4_if([$#$2], [2], [],
+       [m4_if([$2], [], [], [[$1$2]])$0([$1], m4_shift(m4_shift($@)))])])
+
+
+# lt_car(LIST)
+# lt_cdr(LIST)
+# ------------
+# Manipulate m4 lists.
+# These macros are necessary as long as will still need to support
+# Autoconf-2.59 which quotes differently.
+m4_define([lt_car], [[$1]])
+m4_define([lt_cdr],
+[m4_if([$#], 0, [m4_fatal([$0: cannot be called without arguments])],
+       [$#], 1, [],
+       [m4_dquote(m4_shift($@))])])
+m4_define([lt_unquote], $1)
+
+
+# lt_append(MACRO-NAME, STRING, [SEPARATOR])
+# ------------------------------------------
+# Redefine MACRO-NAME to hold its former content plus `SEPARATOR'`STRING'.
+# Note that neither SEPARATOR nor STRING are expanded; they are appended
+# to MACRO-NAME as is (leaving the expansion for when MACRO-NAME is invoked).
+# No SEPARATOR is output if MACRO-NAME was previously undefined (different
+# than defined and empty).
+#
+# This macro is needed until we can rely on Autoconf 2.62, since earlier
+# versions of m4sugar mistakenly expanded SEPARATOR but not STRING.
+m4_define([lt_append],
+[m4_define([$1],
+	   m4_ifdef([$1], [m4_defn([$1])[$3]])[$2])])
+
+
+
+# lt_combine(SEP, PREFIX-LIST, INFIX, SUFFIX1, [SUFFIX2...])
+# ----------------------------------------------------------
+# Produce a SEP delimited list of all paired combinations of elements of
+# PREFIX-LIST with SUFFIX1 through SUFFIXn.  Each element of the list
+# has the form PREFIXmINFIXSUFFIXn.
+# Needed until we can rely on m4_combine added in Autoconf 2.62.
+m4_define([lt_combine],
+[m4_if(m4_eval([$# > 3]), [1],
+       [m4_pushdef([_Lt_sep], [m4_define([_Lt_sep], m4_defn([lt_car]))])]]dnl
+[[m4_foreach([_Lt_prefix], [$2],
+	     [m4_foreach([_Lt_suffix],
+		]m4_dquote(m4_dquote(m4_shift(m4_shift(m4_shift($@)))))[,
+	[_Lt_sep([$1])[]m4_defn([_Lt_prefix])[$3]m4_defn([_Lt_suffix])])])])])
+
+
+# lt_if_append_uniq(MACRO-NAME, VARNAME, [SEPARATOR], [UNIQ], [NOT-UNIQ])
+# -----------------------------------------------------------------------
+# Iff MACRO-NAME does not yet contain VARNAME, then append it (delimited
+# by SEPARATOR if supplied) and expand UNIQ, else NOT-UNIQ.
+m4_define([lt_if_append_uniq],
+[m4_ifdef([$1],
+	  [m4_if(m4_index([$3]m4_defn([$1])[$3], [$3$2$3]), [-1],
+		 [lt_append([$1], [$2], [$3])$4],
+		 [$5])],
+	  [lt_append([$1], [$2], [$3])$4])])
+
+
+# lt_dict_add(DICT, KEY, VALUE)
+# -----------------------------
+m4_define([lt_dict_add],
+[m4_define([$1($2)], [$3])])
+
+
+# lt_dict_add_subkey(DICT, KEY, SUBKEY, VALUE)
+# --------------------------------------------
+m4_define([lt_dict_add_subkey],
+[m4_define([$1($2:$3)], [$4])])
+
+
+# lt_dict_fetch(DICT, KEY, [SUBKEY])
+# ----------------------------------
+m4_define([lt_dict_fetch],
+[m4_ifval([$3],
+	m4_ifdef([$1($2:$3)], [m4_defn([$1($2:$3)])]),
+    m4_ifdef([$1($2)], [m4_defn([$1($2)])]))])
+
+
+# lt_if_dict_fetch(DICT, KEY, [SUBKEY], VALUE, IF-TRUE, [IF-FALSE])
+# -----------------------------------------------------------------
+m4_define([lt_if_dict_fetch],
+[m4_if(lt_dict_fetch([$1], [$2], [$3]), [$4],
+	[$5],
+    [$6])])
+
+
+# lt_dict_filter(DICT, [SUBKEY], VALUE, [SEPARATOR], KEY, [...])
+# --------------------------------------------------------------
+m4_define([lt_dict_filter],
+[m4_if([$5], [], [],
+  [lt_join(m4_quote(m4_default([$4], [[, ]])),
+           lt_unquote(m4_split(m4_normalize(m4_foreach(_Lt_key, lt_car([m4_shiftn(4, $@)]),
+		      [lt_if_dict_fetch([$1], _Lt_key, [$2], [$3], [_Lt_key ])])))))])[]dnl
+])

m4/ltversion.m4

@@ -0,0 +1,23 @@
+# ltversion.m4 -- version numbers			-*- Autoconf -*-
+#
+#   Copyright (C) 2004 Free Software Foundation, Inc.
+#   Written by Scott James Remnant, 2004
+#
+# This file is free software; the Free Software Foundation gives
+# unlimited permission to copy and/or distribute it, with or without
+# modifications, as long as this notice is preserved.
+
+# @configure_input@
+
+# serial 3337 ltversion.m4
+# This file is part of GNU Libtool
+
+m4_define([LT_PACKAGE_VERSION], [2.4.2])
+m4_define([LT_PACKAGE_REVISION], [1.3337])
+
+AC_DEFUN([LTVERSION_VERSION],
+[macro_version='2.4.2'
+macro_revision='1.3337'
+_LT_DECL(, macro_version, 0, [Which release of libtool.m4 was used?])
+_LT_DECL(, macro_revision, 0)
+])

m4/lt~obsolete.m4

@@ -0,0 +1,98 @@
+# lt~obsolete.m4 -- aclocal satisfying obsolete definitions.    -*-Autoconf-*-
+#
+#   Copyright (C) 2004, 2005, 2007, 2009 Free Software Foundation, Inc.
+#   Written by Scott James Remnant, 2004.
+#
+# This file is free software; the Free Software Foundation gives
+# unlimited permission to copy and/or distribute it, with or without
+# modifications, as long as this notice is preserved.
+
+# serial 5 lt~obsolete.m4
+
+# These exist entirely to fool aclocal when bootstrapping libtool.
+#
+# In the past libtool.m4 has provided macros via AC_DEFUN (or AU_DEFUN)
+# which have later been changed to m4_define as they aren't part of the
+# exported API, or moved to Autoconf or Automake where they belong.
+#
+# The trouble is, aclocal is a bit thick.  It'll see the old AC_DEFUN
+# in /usr/share/aclocal/libtool.m4 and remember it, then when it sees us
+# using a macro with the same name in our local m4/libtool.m4 it'll
+# pull the old libtool.m4 in (it doesn't see our shiny new m4_define
+# and doesn't know about Autoconf macros at all.)
+#
+# So we provide this file, which has a silly filename so it's always
+# included after everything else.  This provides aclocal with the
+# AC_DEFUNs it wants, but when m4 processes it, it doesn't do anything
+# because those macros already exist, or will be overwritten later.
+# We use AC_DEFUN over AU_DEFUN for compatibility with aclocal-1.6. 
+#
+# Anytime we withdraw an AC_DEFUN or AU_DEFUN, remember to add it here.
+# Yes, that means every name once taken will need to remain here until
+# we give up compatibility with versions before 1.7, at which point
+# we need to keep only those names which we still refer to.
+
+# This is to help aclocal find these macros, as it can't see m4_define.
+AC_DEFUN([LTOBSOLETE_VERSION], [m4_if([1])])
+
+m4_ifndef([AC_LIBTOOL_LINKER_OPTION],	[AC_DEFUN([AC_LIBTOOL_LINKER_OPTION])])
+m4_ifndef([AC_PROG_EGREP],		[AC_DEFUN([AC_PROG_EGREP])])
+m4_ifndef([_LT_AC_PROG_ECHO_BACKSLASH],	[AC_DEFUN([_LT_AC_PROG_ECHO_BACKSLASH])])
+m4_ifndef([_LT_AC_SHELL_INIT],		[AC_DEFUN([_LT_AC_SHELL_INIT])])
+m4_ifndef([_LT_AC_SYS_LIBPATH_AIX],	[AC_DEFUN([_LT_AC_SYS_LIBPATH_AIX])])
+m4_ifndef([_LT_PROG_LTMAIN],		[AC_DEFUN([_LT_PROG_LTMAIN])])
+m4_ifndef([_LT_AC_TAGVAR],		[AC_DEFUN([_LT_AC_TAGVAR])])
+m4_ifndef([AC_LTDL_ENABLE_INSTALL],	[AC_DEFUN([AC_LTDL_ENABLE_INSTALL])])
+m4_ifndef([AC_LTDL_PREOPEN],		[AC_DEFUN([AC_LTDL_PREOPEN])])
+m4_ifndef([_LT_AC_SYS_COMPILER],	[AC_DEFUN([_LT_AC_SYS_COMPILER])])
+m4_ifndef([_LT_AC_LOCK],		[AC_DEFUN([_LT_AC_LOCK])])
+m4_ifndef([AC_LIBTOOL_SYS_OLD_ARCHIVE],	[AC_DEFUN([AC_LIBTOOL_SYS_OLD_ARCHIVE])])
+m4_ifndef([_LT_AC_TRY_DLOPEN_SELF],	[AC_DEFUN([_LT_AC_TRY_DLOPEN_SELF])])
+m4_ifndef([AC_LIBTOOL_PROG_CC_C_O],	[AC_DEFUN([AC_LIBTOOL_PROG_CC_C_O])])
+m4_ifndef([AC_LIBTOOL_SYS_HARD_LINK_LOCKS], [AC_DEFUN([AC_LIBTOOL_SYS_HARD_LINK_LOCKS])])
+m4_ifndef([AC_LIBTOOL_OBJDIR],		[AC_DEFUN([AC_LIBTOOL_OBJDIR])])
+m4_ifndef([AC_LTDL_OBJDIR],		[AC_DEFUN([AC_LTDL_OBJDIR])])
+m4_ifndef([AC_LIBTOOL_PROG_LD_HARDCODE_LIBPATH], [AC_DEFUN([AC_LIBTOOL_PROG_LD_HARDCODE_LIBPATH])])
+m4_ifndef([AC_LIBTOOL_SYS_LIB_STRIP],	[AC_DEFUN([AC_LIBTOOL_SYS_LIB_STRIP])])
+m4_ifndef([AC_PATH_MAGIC],		[AC_DEFUN([AC_PATH_MAGIC])])
+m4_ifndef([AC_PROG_LD_GNU],		[AC_DEFUN([AC_PROG_LD_GNU])])
+m4_ifndef([AC_PROG_LD_RELOAD_FLAG],	[AC_DEFUN([AC_PROG_LD_RELOAD_FLAG])])
+m4_ifndef([AC_DEPLIBS_CHECK_METHOD],	[AC_DEFUN([AC_DEPLIBS_CHECK_METHOD])])
+m4_ifndef([AC_LIBTOOL_PROG_COMPILER_NO_RTTI], [AC_DEFUN([AC_LIBTOOL_PROG_COMPILER_NO_RTTI])])
+m4_ifndef([AC_LIBTOOL_SYS_GLOBAL_SYMBOL_PIPE], [AC_DEFUN([AC_LIBTOOL_SYS_GLOBAL_SYMBOL_PIPE])])
+m4_ifndef([AC_LIBTOOL_PROG_COMPILER_PIC], [AC_DEFUN([AC_LIBTOOL_PROG_COMPILER_PIC])])
+m4_ifndef([AC_LIBTOOL_PROG_LD_SHLIBS],	[AC_DEFUN([AC_LIBTOOL_PROG_LD_SHLIBS])])
+m4_ifndef([AC_LIBTOOL_POSTDEP_PREDEP],	[AC_DEFUN([AC_LIBTOOL_POSTDEP_PREDEP])])
+m4_ifndef([LT_AC_PROG_EGREP],		[AC_DEFUN([LT_AC_PROG_EGREP])])
+m4_ifndef([LT_AC_PROG_SED],		[AC_DEFUN([LT_AC_PROG_SED])])
+m4_ifndef([_LT_CC_BASENAME],		[AC_DEFUN([_LT_CC_BASENAME])])
+m4_ifndef([_LT_COMPILER_BOILERPLATE],	[AC_DEFUN([_LT_COMPILER_BOILERPLATE])])
+m4_ifndef([_LT_LINKER_BOILERPLATE],	[AC_DEFUN([_LT_LINKER_BOILERPLATE])])
+m4_ifndef([_AC_PROG_LIBTOOL],		[AC_DEFUN([_AC_PROG_LIBTOOL])])
+m4_ifndef([AC_LIBTOOL_SETUP],		[AC_DEFUN([AC_LIBTOOL_SETUP])])
+m4_ifndef([_LT_AC_CHECK_DLFCN],		[AC_DEFUN([_LT_AC_CHECK_DLFCN])])
+m4_ifndef([AC_LIBTOOL_SYS_DYNAMIC_LINKER],	[AC_DEFUN([AC_LIBTOOL_SYS_DYNAMIC_LINKER])])
+m4_ifndef([_LT_AC_TAGCONFIG],		[AC_DEFUN([_LT_AC_TAGCONFIG])])
+m4_ifndef([AC_DISABLE_FAST_INSTALL],	[AC_DEFUN([AC_DISABLE_FAST_INSTALL])])
+m4_ifndef([_LT_AC_LANG_CXX],		[AC_DEFUN([_LT_AC_LANG_CXX])])
+m4_ifndef([_LT_AC_LANG_F77],		[AC_DEFUN([_LT_AC_LANG_F77])])
+m4_ifndef([_LT_AC_LANG_GCJ],		[AC_DEFUN([_LT_AC_LANG_GCJ])])
+m4_ifndef([AC_LIBTOOL_LANG_C_CONFIG],	[AC_DEFUN([AC_LIBTOOL_LANG_C_CONFIG])])
+m4_ifndef([_LT_AC_LANG_C_CONFIG],	[AC_DEFUN([_LT_AC_LANG_C_CONFIG])])
+m4_ifndef([AC_LIBTOOL_LANG_CXX_CONFIG],	[AC_DEFUN([AC_LIBTOOL_LANG_CXX_CONFIG])])
+m4_ifndef([_LT_AC_LANG_CXX_CONFIG],	[AC_DEFUN([_LT_AC_LANG_CXX_CONFIG])])
+m4_ifndef([AC_LIBTOOL_LANG_F77_CONFIG],	[AC_DEFUN([AC_LIBTOOL_LANG_F77_CONFIG])])
+m4_ifndef([_LT_AC_LANG_F77_CONFIG],	[AC_DEFUN([_LT_AC_LANG_F77_CONFIG])])
+m4_ifndef([AC_LIBTOOL_LANG_GCJ_CONFIG],	[AC_DEFUN([AC_LIBTOOL_LANG_GCJ_CONFIG])])
+m4_ifndef([_LT_AC_LANG_GCJ_CONFIG],	[AC_DEFUN([_LT_AC_LANG_GCJ_CONFIG])])
+m4_ifndef([AC_LIBTOOL_LANG_RC_CONFIG],	[AC_DEFUN([AC_LIBTOOL_LANG_RC_CONFIG])])
+m4_ifndef([_LT_AC_LANG_RC_CONFIG],	[AC_DEFUN([_LT_AC_LANG_RC_CONFIG])])
+m4_ifndef([AC_LIBTOOL_CONFIG],		[AC_DEFUN([AC_LIBTOOL_CONFIG])])
+m4_ifndef([_LT_AC_FILE_LTDLL_C],	[AC_DEFUN([_LT_AC_FILE_LTDLL_C])])
+m4_ifndef([_LT_REQUIRED_DARWIN_CHECKS],	[AC_DEFUN([_LT_REQUIRED_DARWIN_CHECKS])])
+m4_ifndef([_LT_AC_PROG_CXXCPP],		[AC_DEFUN([_LT_AC_PROG_CXXCPP])])
+m4_ifndef([_LT_PREPARE_SED_QUOTE_VARS],	[AC_DEFUN([_LT_PREPARE_SED_QUOTE_VARS])])
+m4_ifndef([_LT_PROG_ECHO_BACKSLASH],	[AC_DEFUN([_LT_PROG_ECHO_BACKSLASH])])
+m4_ifndef([_LT_PROG_F77],		[AC_DEFUN([_LT_PROG_F77])])
+m4_ifndef([_LT_PROG_FC],		[AC_DEFUN([_LT_PROG_FC])])
+m4_ifndef([_LT_PROG_CXX],		[AC_DEFUN([_LT_PROG_CXX])])

src/Makefile.am

@@ -0,0 +1,152 @@
+#   BLAKE2 reference source code package - optimized C implementations
+#
+#   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+#
+#   To the extent possible under law, the author(s) have dedicated all copyright
+#   and related and neighboring rights to this software to the public domain
+#   worldwide. This software is distributed without any warranty.
+#
+#   You should have received a copy of the CC0 Public Domain Dedication along with
+#   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+
+EXTRA_DIST = 
+
+CPPFLAGS += $(LTDLINCL) $(OPENMP_CFLAGS)
+CFLAGS += $(OPENMP_CFLAGS)
+LDFLAGS += -version-info $(B2_LIBRARY_VERSION)
+
+lib_LTLIBRARIES = libb2.la
+libb2_la_LIBADD = # -lgomp -lpthread
+libb2_la_CPPFLAGS =  -DSUFFIX=  \
+                     $(LTDLINCL) \
+                     ${top_builddir}/src/
+
+include_HEADERS = blake2.h
+
+if USE_FAT
+noinst_LTLIBRARIES = libblake2b_ref.la \
+                     libblake2b_sse2.la \
+                     libblake2b_ssse3.la \
+                     libblake2b_sse41.la \
+                     libblake2b_avx.la  \
+                     libblake2b_xop.la \
+                     libblake2s_ref.la \
+                     libblake2s_sse2.la \
+                     libblake2s_ssse3.la \
+                     libblake2s_sse41.la \
+                     libblake2s_avx.la  \
+                     libblake2s_xop.la 
+
+libb2_la_SOURCES = blake2-dispatch.c blake2sp.c blake2bp.c
+libb2_la_LIBADD += libblake2b_ref.la \
+                  libblake2b_sse2.la \
+				          libblake2b_ssse3.la \
+				          libblake2b_sse41.la \
+				          libblake2b_avx.la  \
+				          libblake2b_xop.la \
+				          libblake2s_ref.la \
+                  libblake2s_sse2.la \
+                  libblake2s_ssse3.la \
+                  libblake2s_sse41.la \
+                  libblake2s_avx.la  \
+                  libblake2s_xop.la 
+
+
+libblake2b_ref_la_SOURCES = blake2b-ref.c
+libblake2b_ref_la_CPPFLAGS = -DSUFFIX=_ref
+libblake2b_ref_la_CFLAGS = 
+
+libblake2b_sse2_la_SOURCES = blake2b.c
+libblake2b_sse2_la_CPPFLAGS = -DSUFFIX=_sse2 
+libblake2b_sse2_la_CFLAGS = -msse2
+
+libblake2b_ssse3_la_SOURCES = blake2b.c
+libblake2b_ssse3_la_CPPFLAGS = -DSUFFIX=_ssse3
+libblake2b_ssse3_la_CFLAGS = -msse2 -mssse3
+
+libblake2b_sse41_la_SOURCES = blake2b.c
+libblake2b_sse41_la_CPPFLAGS = -DSUFFIX=_sse41 
+libblake2b_sse41_la_CFLAGS = -msse2 -mssse3 -msse4.1
+
+libblake2b_avx_la_SOURCES = blake2b.c
+libblake2b_avx_la_CPPFLAGS = -DSUFFIX=_avx
+libblake2b_avx_la_CFLAGS = -msse2 -mssse3 -msse4.1 -mavx
+
+libblake2b_xop_la_SOURCES = blake2b.c
+libblake2b_xop_la_CPPFLAGS = -DSUFFIX=_xop 
+libblake2b_xop_la_CFLAGS = -msse2 -mssse3 -msse4.1 -mavx -mxop
+
+
+libblake2s_ref_la_SOURCES = blake2s-ref.c
+libblake2s_ref_la_CPPFLAGS = -DSUFFIX=_ref
+libblake2s_ref_la_CFLAGS = 
+
+libblake2s_sse2_la_SOURCES = blake2s.c
+libblake2s_sse2_la_CPPFLAGS = -DSUFFIX=_sse2
+libblake2s_sse2_la_CFLAGS = -msse2
+
+libblake2s_ssse3_la_SOURCES = blake2s.c
+libblake2s_ssse3_la_CPPFLAGS = -DSUFFIX=_ssse3
+libblake2s_ssse3_la_CFLAGS = -msse2 -mssse3
+
+libblake2s_sse41_la_SOURCES = blake2s.c
+libblake2s_sse41_la_CPPFLAGS = -DSUFFIX=_sse41 
+libblake2s_sse41_la_CFLAGS = -msse2 -mssse3 -msse4.1
+
+libblake2s_avx_la_SOURCES = blake2s.c
+libblake2s_avx_la_CPPFLAGS = -DSUFFIX=_avx
+libblake2s_avx_la_CFLAGS = -msse2 -mssse3 -msse4.1 -mavx
+
+libblake2s_xop_la_SOURCES = blake2s.c
+libblake2s_xop_la_CPPFLAGS = -DSUFFIX=_xop 
+libblake2s_xop_la_CFLAGS = -msse2 -mssse3 -msse4.1 -mavx -mxop
+
+else
+
+if USE_SSE
+libb2_la_SOURCES = blake2sp.c \
+                   blake2bp.c \
+                   blake2s.c \
+                   blake2b.c \
+                   blake2-impl.h \
+                   blake2-config.h \
+                   blake2s-round.h \
+                   blake2b-round.h \
+                   blake2s-load-xop.h \
+                   blake2s-load-sse41.h \
+                   blake2s-load-sse2.h \
+                   blake2b-load-sse41.h \
+                   blake2b-load-sse2.h 
+else
+libb2_la_SOURCES = blake2s-ref.c \
+                   blake2b-ref.c \
+                   blake2.h \
+                   blake2-impl.h \
+                   blake2sp.c \
+                   blake2bp.c \
+                   blake2-kat.h 
+endif
+endif
+
+TESTS_TARGETS = blake2s-test \
+                blake2b-test \
+                blake2sp-test \
+                blake2bp-test
+
+check_PROGRAMS = $(TESTS_TARGETS)
+TESTS = $(TESTS_TARGETS)
+
+TESTS_LDADD = ${top_builddir}/src/libb2.la
+
+blake2s_test_SOURCE = blake2s-test.c blake2-kat.h 
+blake2s_test_LDADD = $(TESTS_LDADD)
+
+blake2b_test_SOURCE = blake2b-test.c blake2-kat.h 
+blake2b_test_LDADD = $(TESTS_LDADD)
+
+blake2sp_test_SOURCE = blake2sp-test.c blake2-kat.h
+blake2sp_test_LDADD = $(TESTS_LDADD)
+
+blake2bp_test_SOURCE = blake2bp-test.c blake2-kat.h
+blake2bp_test_LDADD = $(TESTS_LDADD)
+

src/blake2-config.h

@@ -0,0 +1,71 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2_CONFIG_H__
+#define __BLAKE2_CONFIG_H__
+
+#if defined(__SSE2__)
+#define HAVE_SSE2
+#endif
+
+#if defined(__SSSE3__)
+#define HAVE_SSSE3
+#endif
+
+#if defined(__SSE4_1__)
+#define HAVE_SSE4_1
+#endif
+
+#if defined(__AVX__)
+#define HAVE_AVX
+#endif
+
+#if defined(__XOP__)
+#define HAVE_XOP
+#endif
+
+
+#ifdef HAVE_AVX2
+#ifndef HAVE_AVX
+#define HAVE_AVX
+#endif
+#endif
+
+#ifdef HAVE_XOP
+#ifndef HAVE_AVX
+#define HAVE_AVX
+#endif
+#endif
+
+#ifdef HAVE_AVX
+#ifndef HAVE_SSE4_1
+#define HAVE_SSE4_1
+#endif
+#endif
+
+#ifdef HAVE_SSE41
+#ifndef HAVE_SSSE3
+#define HAVE_SSSE3
+#endif
+#endif
+
+#ifdef HAVE_SSSE3
+#define HAVE_SSE2
+#endif
+
+#if !defined(HAVE_SSE2)
+#error "This code requires at least SSE2."
+#endif
+
+#endif
+

src/blake2-dispatch.c

@@ -0,0 +1,560 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#include <stdio.h>
+#if defined(WIN32)
+#include <windows.h>
+#endif
+#include "blake2.h"
+
+#if defined(__x86_64__) || defined(__i386__) || defined(_M_IX86) || defined(_M_X64)
+#define HAVE_X86
+#endif
+
+typedef enum
+{
+  NONE  = 0,
+#if defined(HAVE_X86)
+  SSE2  = 1,
+  SSSE3 = 2,
+  SSE41 = 3,
+  AVX   = 4,
+  XOP   = 5,
+  /* AVX2  = 6, */
+#endif
+} cpu_feature_t;
+
+static const char feature_names[][8] =
+{
+  "none",
+#if defined(HAVE_X86)
+  "sse2",
+  "ssse3",
+  "sse41",
+  "avx",
+  "xop",
+  /* "avx2" */
+#endif
+};
+
+#if defined(HAVE_X86)
+
+#if defined(__GNUC__)
+static inline void cpuid( uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx )
+{
+  __asm__ __volatile__(
+#if defined(__i386__) /* This is needed for -fPIC to work on i386 */
+    "movl %%ebx, %%esi\n\t"
+#endif
+    "cpuid\n\t"
+#if defined(__i386__)
+    "xchgl %%ebx, %%esi\n\t"
+    : "=a"( *eax ), "=S"( *ebx ), "=c"( *ecx ), "=d"( *edx ) : "a"( *eax ) );
+#else
+    : "=a"( *eax ), "=b"( *ebx ), "=c"( *ecx ), "=d"( *edx ) : "a"( *eax ) );
+#endif
+}
+#elif defined(_MSC_VER)
+#include <intrin.h>
+static inline void cpuid( uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx )
+{
+  int regs[4];
+  __cpuid( regs, *eax );
+  *eax = regs[0];
+  *ebx = regs[1];
+  *ecx = regs[2];
+  *edx = regs[3];
+}
+#else
+#error "Don't know how to call cpuid on this compiler!"
+#endif
+
+#endif /* HAVE_X86 */
+
+static inline cpu_feature_t get_cpu_features( void )
+{
+#if defined(HAVE_X86)
+  static volatile int initialized = 0;
+  static cpu_feature_t feature = NONE; // Safe default
+  uint32_t eax, ecx, edx, ebx;
+
+  if( initialized )
+    return feature;
+
+  eax = 1;
+  cpuid( &eax, &ebx, &ecx, &edx );
+
+  if( 1 & ( edx >> 26 ) )
+    feature = SSE2;
+
+  if( 1 & ( ecx >> 9 ) )
+    feature = SSSE3;
+
+  if( 1 & ( ecx >> 19 ) )
+    feature = SSE41;
+
+#if defined(WIN32) /* Work around the fact that Windows <7 does NOT support AVX... */
+  if( IsProcessorFeaturePresent(17) ) /* Some environments don't know about PF_XSAVE_ENABLED */
+#endif
+  {
+    if( 1 & ( ecx >> 28 ) )
+      feature = AVX;
+
+
+    eax = 0x80000001;
+    cpuid( &eax, &ebx, &ecx, &edx );
+
+    if( 1 & ( ecx >> 11 ) )
+      feature = XOP;
+  }
+
+  /* For future architectures */
+  /* 
+      eax = 7; ecx = 0;
+      cpuid(&eax, &ebx, &ecx, &edx);
+
+      if(1&(ebx >> 5))
+        feature = AVX2;
+  */
+  /* fprintf( stderr, "Using %s engine\n", feature_names[feature] ); */
+  initialized = 1;
+  return feature;
+#else
+  return NONE;
+#endif
+}
+
+
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+  int blake2b_init_ref( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_ref( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_ref( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_ref( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_ref( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_ref( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+#if defined(HAVE_X86)
+
+  int blake2b_init_sse2( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_sse2( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_sse2( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_sse2( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_sse2( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_sse2( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2b_init_ssse3( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_ssse3( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_ssse3( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_ssse3( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_ssse3( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_ssse3( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2b_init_sse41( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_sse41( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_sse41( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_sse41( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_sse41( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_sse41( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2b_init_avx( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_avx( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_avx( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_avx( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_avx( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_avx( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2b_init_xop( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_xop( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_xop( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_xop( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_xop( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_xop( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+#endif /* HAVE_X86 */
+
+  int blake2s_init_ref( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_ref( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_ref( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_ref( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_ref( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_ref( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+#if defined(HAVE_X86)
+
+  int blake2s_init_sse2( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_sse2( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_sse2( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_sse2( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_sse2( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_sse2( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2s_init_ssse3( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_ssse3( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_ssse3( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_ssse3( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_ssse3( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_ssse3( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2s_init_sse41( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_sse41( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_sse41( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_sse41( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_sse41( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_sse41( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2s_init_avx( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_avx( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_avx( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_avx( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_avx( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_avx( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2s_init_xop( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_xop( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_xop( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_xop( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_xop( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_xop( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+#endif /* HAVE_X86 */
+
+#if defined(__cplusplus)
+}
+#endif
+
+typedef int ( *blake2b_init_fn )( blake2b_state *, size_t );
+typedef int ( *blake2b_init_key_fn )( blake2b_state *, size_t, const void *, size_t );
+typedef int ( *blake2b_init_param_fn )( blake2b_state *, const blake2b_param * );
+typedef int ( *blake2b_update_fn )( blake2b_state *, const uint8_t *, size_t );
+typedef int ( *blake2b_final_fn )( blake2b_state *, uint8_t *, size_t );
+typedef int ( *blake2b_fn )( uint8_t *, const void *, const void *, size_t, size_t, size_t );
+
+typedef int ( *blake2s_init_fn )( blake2s_state *, size_t );
+typedef int ( *blake2s_init_key_fn )( blake2s_state *, size_t, const void *, size_t );
+typedef int ( *blake2s_init_param_fn )( blake2s_state *, const blake2s_param * );
+typedef int ( *blake2s_update_fn )( blake2s_state *, const uint8_t *, size_t );
+typedef int ( *blake2s_final_fn )( blake2s_state *, uint8_t *, size_t );
+typedef int ( *blake2s_fn )( uint8_t *, const void *, const void *, size_t, size_t, size_t );
+
+static const blake2b_init_fn blake2b_init_table[] =
+{
+  blake2b_init_ref,
+#if defined(HAVE_X86)
+  blake2b_init_sse2,
+  blake2b_init_ssse3,
+  blake2b_init_sse41,
+  blake2b_init_avx,
+  blake2b_init_xop
+#endif
+};
+
+static const blake2b_init_key_fn blake2b_init_key_table[] =
+{
+  blake2b_init_key_ref,
+#if defined(HAVE_X86)
+  blake2b_init_key_sse2,
+  blake2b_init_key_ssse3,
+  blake2b_init_key_sse41,
+  blake2b_init_key_avx,
+  blake2b_init_key_xop
+#endif
+};
+
+static const blake2b_init_param_fn blake2b_init_param_table[] =
+{
+  blake2b_init_param_ref,
+#if defined(HAVE_X86)
+  blake2b_init_param_sse2,
+  blake2b_init_param_ssse3,
+  blake2b_init_param_sse41,
+  blake2b_init_param_avx,
+  blake2b_init_param_xop
+#endif
+};
+
+static const blake2b_update_fn blake2b_update_table[] =
+{
+  blake2b_update_ref,
+#if defined(HAVE_X86)
+  blake2b_update_sse2,
+  blake2b_update_ssse3,
+  blake2b_update_sse41,
+  blake2b_update_avx,
+  blake2b_update_xop
+#endif
+};
+
+static const blake2b_final_fn blake2b_final_table[] =
+{
+  blake2b_final_ref,
+#if defined(HAVE_X86)
+  blake2b_final_sse2,
+  blake2b_final_ssse3,
+  blake2b_final_sse41,
+  blake2b_final_avx,
+  blake2b_final_xop
+#endif
+};
+
+static const blake2b_fn blake2b_table[] =
+{
+  blake2b_ref,
+#if defined(HAVE_X86)
+  blake2b_sse2,
+  blake2b_ssse3,
+  blake2b_sse41,
+  blake2b_avx,
+  blake2b_xop
+#endif
+};
+
+static const blake2s_init_fn blake2s_init_table[] =
+{
+  blake2s_init_ref,
+#if defined(HAVE_X86)
+  blake2s_init_sse2,
+  blake2s_init_ssse3,
+  blake2s_init_sse41,
+  blake2s_init_avx,
+  blake2s_init_xop
+#endif
+};
+
+static const blake2s_init_key_fn blake2s_init_key_table[] =
+{
+  blake2s_init_key_ref,
+#if defined(HAVE_X86)
+  blake2s_init_key_sse2,
+  blake2s_init_key_ssse3,
+  blake2s_init_key_sse41,
+  blake2s_init_key_avx,
+  blake2s_init_key_xop
+#endif
+};
+
+static const blake2s_init_param_fn blake2s_init_param_table[] =
+{
+  blake2s_init_param_ref,
+#if defined(HAVE_X86)
+  blake2s_init_param_sse2,
+  blake2s_init_param_ssse3,
+  blake2s_init_param_sse41,
+  blake2s_init_param_avx,
+  blake2s_init_param_xop
+#endif
+};
+
+static const blake2s_update_fn blake2s_update_table[] =
+{
+  blake2s_update_ref,
+#if defined(HAVE_X86)
+  blake2s_update_sse2,
+  blake2s_update_ssse3,
+  blake2s_update_sse41,
+  blake2s_update_avx,
+  blake2s_update_xop
+#endif
+};
+
+static const blake2s_final_fn blake2s_final_table[] =
+{
+  blake2s_final_ref,
+#if defined(HAVE_X86)
+  blake2s_final_sse2,
+  blake2s_final_ssse3,
+  blake2s_final_sse41,
+  blake2s_final_avx,
+  blake2s_final_xop
+#endif
+};
+
+static const blake2s_fn blake2s_table[] =
+{
+  blake2s_ref,
+#if defined(HAVE_X86)
+  blake2s_sse2,
+  blake2s_ssse3,
+  blake2s_sse41,
+  blake2s_avx,
+  blake2s_xop
+#endif
+};
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+  int blake2b_init_dispatch( blake2b_state *S, size_t outlen );
+  int blake2b_init_key_dispatch( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_init_param_dispatch( blake2b_state *S, const blake2b_param *P );
+  int blake2b_update_dispatch( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final_dispatch( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b_dispatch( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  int blake2s_init_dispatch( blake2s_state *S, size_t outlen );
+  int blake2s_init_key_dispatch( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_init_param_dispatch( blake2s_state *S, const blake2s_param *P );
+  int blake2s_update_dispatch( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final_dispatch( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s_dispatch( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+#if defined(__cplusplus)
+}
+#endif
+
+static blake2b_init_fn blake2b_init_ptr = blake2b_init_dispatch;
+static blake2b_init_key_fn blake2b_init_key_ptr = blake2b_init_key_dispatch;
+static blake2b_init_param_fn blake2b_init_param_ptr = blake2b_init_param_dispatch;
+static blake2b_update_fn blake2b_update_ptr = blake2b_update_dispatch;
+static blake2b_final_fn blake2b_final_ptr = blake2b_final_dispatch;
+static blake2b_fn blake2b_ptr = blake2b_dispatch;
+
+static blake2s_init_fn blake2s_init_ptr = blake2s_init_dispatch;
+static blake2s_init_key_fn blake2s_init_key_ptr = blake2s_init_key_dispatch;
+static blake2s_init_param_fn blake2s_init_param_ptr = blake2s_init_param_dispatch;
+static blake2s_update_fn blake2s_update_ptr = blake2s_update_dispatch;
+static blake2s_final_fn blake2s_final_ptr = blake2s_final_dispatch;
+static blake2s_fn blake2s_ptr = blake2s_dispatch;
+
+int blake2b_init_dispatch( blake2b_state *S, size_t outlen )
+{
+  blake2b_init_ptr = blake2b_init_table[get_cpu_features()];
+  return blake2b_init_ptr( S, outlen );
+}
+
+int blake2b_init_key_dispatch( blake2b_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  blake2b_init_key_ptr = blake2b_init_key_table[get_cpu_features()];
+  return blake2b_init_key_ptr( S, outlen, key, keylen );
+}
+
+int blake2b_init_param_dispatch( blake2b_state *S, const blake2b_param *P )
+{
+  blake2b_init_param_ptr = blake2b_init_param_table[get_cpu_features()];
+  return blake2b_init_param_ptr( S, P );
+}
+
+int blake2b_update_dispatch( blake2b_state *S, const uint8_t *in, size_t inlen )
+{
+  blake2b_update_ptr = blake2b_update_table[get_cpu_features()];
+  return blake2b_update_ptr( S, in, inlen );
+}
+
+int blake2b_final_dispatch( blake2b_state *S, uint8_t *out, size_t outlen )
+{
+  blake2b_final_ptr = blake2b_final_table[get_cpu_features()];
+  return blake2b_final_ptr( S, out, outlen );
+}
+
+int blake2b_dispatch( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  blake2b_ptr = blake2b_table[get_cpu_features()];
+  return blake2b_ptr( out, in, key, outlen, inlen, keylen );
+}
+
+EXPORT_DLL int blake2b_init( blake2b_state *S, size_t outlen )
+{
+  return blake2b_init_ptr( S, outlen );
+}
+
+EXPORT_DLL int blake2b_init_key( blake2b_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  return blake2b_init_key_ptr( S, outlen, key, keylen );
+}
+
+EXPORT_DLL int blake2b_init_param( blake2b_state *S, const blake2b_param *P )
+{
+  return blake2b_init_param_ptr( S, P );
+}
+
+EXPORT_DLL int blake2b_update( blake2b_state *S, const uint8_t *in, size_t inlen )
+{
+  return blake2b_update_ptr( S, in, inlen );
+}
+
+EXPORT_DLL int blake2b_final( blake2b_state *S, uint8_t *out, size_t outlen )
+{
+  return blake2b_final_ptr( S, out, outlen );
+}
+
+EXPORT_DLL int blake2b( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  return blake2b_ptr( out, in, key, outlen, inlen, keylen );
+}
+
+int blake2s_init_dispatch( blake2s_state *S, size_t outlen )
+{
+  blake2s_init_ptr = blake2s_init_table[get_cpu_features()];
+  return blake2s_init_ptr( S, outlen );
+}
+
+int blake2s_init_key_dispatch( blake2s_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  blake2s_init_key_ptr = blake2s_init_key_table[get_cpu_features()];
+  return blake2s_init_key_ptr( S, outlen, key, keylen );
+}
+
+int blake2s_init_param_dispatch( blake2s_state *S, const blake2s_param *P )
+{
+  blake2s_init_param_ptr = blake2s_init_param_table[get_cpu_features()];
+  return blake2s_init_param_ptr( S, P );
+}
+
+int blake2s_update_dispatch( blake2s_state *S, const uint8_t *in, size_t inlen )
+{
+  blake2s_update_ptr = blake2s_update_table[get_cpu_features()];
+  return blake2s_update_ptr( S, in, inlen );
+}
+
+int blake2s_final_dispatch( blake2s_state *S, uint8_t *out, size_t outlen )
+{
+  blake2s_final_ptr = blake2s_final_table[get_cpu_features()];
+  return blake2s_final_ptr( S, out, outlen );
+}
+
+int blake2s_dispatch( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  blake2s_ptr = blake2s_table[get_cpu_features()];
+  return blake2s_ptr( out, in, key, outlen, inlen, keylen );
+}
+
+EXPORT_DLL int blake2s_init( blake2s_state *S, size_t outlen )
+{
+  return blake2s_init_ptr( S, outlen );
+}
+
+EXPORT_DLL int blake2s_init_key( blake2s_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  return blake2s_init_key_ptr( S, outlen, key, keylen );
+}
+
+EXPORT_DLL int blake2s_init_param( blake2s_state *S, const blake2s_param *P )
+{
+  return blake2s_init_param_ptr( S, P );
+}
+
+EXPORT_DLL int blake2s_update( blake2s_state *S, const uint8_t *in, size_t inlen )
+{
+  return blake2s_update_ptr( S, in, inlen );
+}
+
+EXPORT_DLL int blake2s_final( blake2s_state *S, uint8_t *out, size_t outlen )
+{
+  return blake2s_final_ptr( S, out, outlen );
+}
+
+EXPORT_DLL int blake2s( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  return blake2s_ptr( out, in, key, outlen, inlen, keylen );
+}
+

src/blake2-impl.h

@@ -0,0 +1,139 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2_IMPL_H__
+#define __BLAKE2_IMPL_H__
+
+#include <stddef.h>
+#include <stdint.h>
+#include "config.h"
+
+#define BLAKE2_IMPL_CAT(x,y) x ## y
+#define BLAKE2_IMPL_EVAL(x,y)  BLAKE2_IMPL_CAT(x,y)
+#define BLAKE2_IMPL_NAME(fun)  BLAKE2_IMPL_EVAL(fun, SUFFIX)
+
+static inline uint32_t load32( const void *src )
+{
+#if defined(NATIVE_LITTLE_ENDIAN) && !defined(HAVE_ALIGNED_ACCESS_REQUIRED)
+  return *( uint32_t * )( src );
+#else
+  const uint8_t *p = ( uint8_t * )src;
+  uint32_t w = *p++;
+  w |= ( uint32_t )( *p++ ) <<  8;
+  w |= ( uint32_t )( *p++ ) << 16;
+  w |= ( uint32_t )( *p++ ) << 24;
+  return w;
+#endif
+}
+
+static inline uint64_t load64( const void *src )
+{
+#if defined(NATIVE_LITTLE_ENDIAN) && !defined(HAVE_ALIGNED_ACCESS_REQUIRED)
+  return *( uint64_t * )( src );
+#else
+  const uint8_t *p = ( uint8_t * )src;
+  uint64_t w = *p++;
+  w |= ( uint64_t )( *p++ ) <<  8;
+  w |= ( uint64_t )( *p++ ) << 16;
+  w |= ( uint64_t )( *p++ ) << 24;
+  w |= ( uint64_t )( *p++ ) << 32;
+  w |= ( uint64_t )( *p++ ) << 40;
+  w |= ( uint64_t )( *p++ ) << 48;
+  w |= ( uint64_t )( *p++ ) << 56;
+  return w;
+#endif
+}
+
+static inline void store32( void *dst, uint32_t w )
+{
+#if defined(NATIVE_LITTLE_ENDIAN) && !defined(HAVE_ALIGNED_ACCESS_REQUIRED)
+  *( uint32_t * )( dst ) = w;
+#else
+  uint8_t *p = ( uint8_t * )dst;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w;
+#endif
+}
+
+static inline void store64( void *dst, uint64_t w )
+{
+#if defined(NATIVE_LITTLE_ENDIAN) && !defined(HAVE_ALIGNED_ACCESS_REQUIRED)
+  *( uint64_t * )( dst ) = w;
+#else
+  uint8_t *p = ( uint8_t * )dst;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w;
+#endif
+}
+
+static inline uint64_t load48( const void *src )
+{
+  const uint8_t *p = ( const uint8_t * )src;
+  uint64_t w = *p++;
+  w |= ( uint64_t )( *p++ ) <<  8;
+  w |= ( uint64_t )( *p++ ) << 16;
+  w |= ( uint64_t )( *p++ ) << 24;
+  w |= ( uint64_t )( *p++ ) << 32;
+  w |= ( uint64_t )( *p++ ) << 40;
+  return w;
+}
+
+static inline void store48( void *dst, uint64_t w )
+{
+  uint8_t *p = ( uint8_t * )dst;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w; w >>= 8;
+  *p++ = ( uint8_t )w;
+}
+
+static inline uint32_t rotl32( const uint32_t w, const unsigned c )
+{
+  return ( w << c ) | ( w >> ( 32 - c ) );
+}
+
+static inline uint64_t rotl64( const uint64_t w, const unsigned c )
+{
+  return ( w << c ) | ( w >> ( 64 - c ) );
+}
+
+static inline uint32_t rotr32( const uint32_t w, const unsigned c )
+{
+  return ( w >> c ) | ( w << ( 32 - c ) );
+}
+
+static inline uint64_t rotr64( const uint64_t w, const unsigned c )
+{
+  return ( w >> c ) | ( w << ( 64 - c ) );
+}
+
+/* prevents compiler optimizing out memset() */
+static inline void secure_zero_memory( void *v, size_t n )
+{
+  volatile uint8_t *p = ( volatile uint8_t * )v;
+
+  while( n-- ) *p++ = 0;
+}
+
+#endif
+

src/blake2.h

@@ -0,0 +1,168 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2_H__
+#define __BLAKE2_H__
+
+#include <stddef.h>
+#include <stdint.h>
+
+#if defined(_MSC_VER)
+  #if defined(DYNAMIC_LIB)
+    #if defined(DYNAMIC_LIB_EXPORT)
+      #define EXPORT_DLL __declspec(dllexport)
+    #else
+      #define EXPORT_DLL __declspec(dllimport)
+    #endif
+  #else
+    #define EXPORT_DLL
+  #endif
+#else
+  #define EXPORT_DLL
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+  enum blake2s_constant
+  {
+    BLAKE2S_BLOCKBYTES = 64,
+    BLAKE2S_OUTBYTES   = 32,
+    BLAKE2S_KEYBYTES   = 32,
+    BLAKE2S_SALTBYTES  = 8,
+    BLAKE2S_PERSONALBYTES = 8
+  };
+
+  enum blake2b_constant
+  {
+    BLAKE2B_BLOCKBYTES = 128,
+    BLAKE2B_OUTBYTES   = 64,
+    BLAKE2B_KEYBYTES   = 64,
+    BLAKE2B_SALTBYTES  = 16,
+    BLAKE2B_PERSONALBYTES = 16
+  };
+
+#pragma pack(push, 1)
+  typedef struct __blake2s_param
+  {
+    uint8_t  digest_length; // 1
+    uint8_t  key_length;    // 2
+    uint8_t  fanout;        // 3
+    uint8_t  depth;         // 4
+    uint32_t leaf_length;   // 8
+    uint8_t  node_offset[6];// 14
+    uint8_t  node_depth;    // 15
+    uint8_t  inner_length;  // 16
+    // uint8_t  reserved[0];
+    uint8_t  salt[BLAKE2S_SALTBYTES]; // 24
+    uint8_t  personal[BLAKE2S_PERSONALBYTES];  // 32
+  } blake2s_param;
+
+  typedef struct __blake2s_state
+  {
+    uint32_t h[8];
+    uint32_t t[2];
+    uint32_t f[2];
+    uint8_t  buf[2 * BLAKE2S_BLOCKBYTES];
+    uint32_t buflen;
+    uint8_t  outlen;
+    uint8_t  last_node;
+  } blake2s_state;
+
+  typedef struct __blake2b_param
+  {
+    uint8_t  digest_length; // 1
+    uint8_t  key_length;    // 2
+    uint8_t  fanout;        // 3
+    uint8_t  depth;         // 4
+    uint32_t leaf_length;   // 8
+    uint64_t node_offset;   // 16
+    uint8_t  node_depth;    // 17
+    uint8_t  inner_length;  // 18
+    uint8_t  reserved[14];  // 32
+    uint8_t  salt[BLAKE2B_SALTBYTES]; // 48
+    uint8_t  personal[BLAKE2B_PERSONALBYTES];  // 64
+  } blake2b_param;
+
+  typedef struct __blake2b_state
+  {
+    uint64_t h[8];
+    uint64_t t[2];
+    uint64_t f[2];
+    uint8_t  buf[2 * BLAKE2B_BLOCKBYTES];
+    uint32_t buflen;
+    uint8_t  outlen;
+    uint8_t  last_node;
+  } blake2b_state;
+
+  typedef struct __blake2sp_state
+  {
+    blake2s_state S[8][1];
+    blake2s_state R[1];
+    uint8_t  buf[8 * BLAKE2S_BLOCKBYTES];
+    uint32_t buflen;
+    uint8_t  outlen;
+  } blake2sp_state;
+
+  typedef struct __blake2bp_state
+  {
+    blake2b_state S[4][1];
+    blake2b_state R[1];
+    uint8_t  buf[4 * BLAKE2B_BLOCKBYTES];
+    uint32_t buflen;
+    uint8_t  outlen;
+  } blake2bp_state;
+#pragma pack(pop)
+
+  // Streaming API
+  EXPORT_DLL int blake2s_init( blake2s_state *S, size_t outlen );
+  EXPORT_DLL int blake2s_init_key( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  EXPORT_DLL int blake2s_init_param( blake2s_state *S, const blake2s_param *P );
+  EXPORT_DLL int blake2s_update( blake2s_state *S, const uint8_t *in, size_t inlen );
+  EXPORT_DLL int blake2s_final( blake2s_state *S, uint8_t *out, size_t outlen );
+
+  EXPORT_DLL int blake2b_init( blake2b_state *S, size_t outlen );
+  EXPORT_DLL int blake2b_init_key( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  EXPORT_DLL int blake2b_init_param( blake2b_state *S, const blake2b_param *P );
+  EXPORT_DLL int blake2b_update( blake2b_state *S, const uint8_t *in, size_t inlen );
+  EXPORT_DLL int blake2b_final( blake2b_state *S, uint8_t *out, size_t outlen );
+
+  EXPORT_DLL int blake2sp_init( blake2sp_state *S, size_t outlen );
+  EXPORT_DLL int blake2sp_init_key( blake2sp_state *S, size_t outlen, const void *key, size_t keylen );
+  EXPORT_DLL int blake2sp_update( blake2sp_state *S, const uint8_t *in, size_t inlen );
+  EXPORT_DLL int blake2sp_final( blake2sp_state *S, uint8_t *out, size_t outlen );
+
+  EXPORT_DLL int blake2bp_init( blake2bp_state *S, size_t outlen );
+  EXPORT_DLL int blake2bp_init_key( blake2bp_state *S, size_t outlen, const void *key, size_t keylen );
+  EXPORT_DLL int blake2bp_update( blake2bp_state *S, const uint8_t *in, size_t inlen );
+  EXPORT_DLL int blake2bp_final( blake2bp_state *S, uint8_t *out, size_t outlen );
+
+  // Simple API
+  EXPORT_DLL int blake2s( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+  EXPORT_DLL int blake2b( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  EXPORT_DLL int blake2sp( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+  EXPORT_DLL int blake2bp( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+
+  static inline int blake2( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+  {
+    return blake2b( out, in, key, outlen, inlen, keylen );
+  }
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif
+

src/blake2b-load-sse2.h

@@ -0,0 +1,68 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2B_LOAD_SSE2_H__
+#define __BLAKE2B_LOAD_SSE2_H__
+
+#define LOAD_MSG_0_1(b0, b1) b0 = _mm_set_epi64x(m2, m0); b1 = _mm_set_epi64x(m6, m4)
+#define LOAD_MSG_0_2(b0, b1) b0 = _mm_set_epi64x(m3, m1); b1 = _mm_set_epi64x(m7, m5)
+#define LOAD_MSG_0_3(b0, b1) b0 = _mm_set_epi64x(m10, m8); b1 = _mm_set_epi64x(m14, m12)
+#define LOAD_MSG_0_4(b0, b1) b0 = _mm_set_epi64x(m11, m9); b1 = _mm_set_epi64x(m15, m13)
+#define LOAD_MSG_1_1(b0, b1) b0 = _mm_set_epi64x(m4, m14); b1 = _mm_set_epi64x(m13, m9)
+#define LOAD_MSG_1_2(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 = _mm_set_epi64x(m6, m15)
+#define LOAD_MSG_1_3(b0, b1) b0 = _mm_set_epi64x(m0, m1); b1 = _mm_set_epi64x(m5, m11)
+#define LOAD_MSG_1_4(b0, b1) b0 = _mm_set_epi64x(m2, m12); b1 = _mm_set_epi64x(m3, m7)
+#define LOAD_MSG_2_1(b0, b1) b0 = _mm_set_epi64x(m12, m11); b1 = _mm_set_epi64x(m15, m5)
+#define LOAD_MSG_2_2(b0, b1) b0 = _mm_set_epi64x(m0, m8); b1 = _mm_set_epi64x(m13, m2)
+#define LOAD_MSG_2_3(b0, b1) b0 = _mm_set_epi64x(m3, m10); b1 = _mm_set_epi64x(m9, m7)
+#define LOAD_MSG_2_4(b0, b1) b0 = _mm_set_epi64x(m6, m14); b1 = _mm_set_epi64x(m4, m1)
+#define LOAD_MSG_3_1(b0, b1) b0 = _mm_set_epi64x(m3, m7); b1 = _mm_set_epi64x(m11, m13)
+#define LOAD_MSG_3_2(b0, b1) b0 = _mm_set_epi64x(m1, m9); b1 = _mm_set_epi64x(m14, m12)
+#define LOAD_MSG_3_3(b0, b1) b0 = _mm_set_epi64x(m5, m2); b1 = _mm_set_epi64x(m15, m4)
+#define LOAD_MSG_3_4(b0, b1) b0 = _mm_set_epi64x(m10, m6); b1 = _mm_set_epi64x(m8, m0)
+#define LOAD_MSG_4_1(b0, b1) b0 = _mm_set_epi64x(m5, m9); b1 = _mm_set_epi64x(m10, m2)
+#define LOAD_MSG_4_2(b0, b1) b0 = _mm_set_epi64x(m7, m0); b1 = _mm_set_epi64x(m15, m4)
+#define LOAD_MSG_4_3(b0, b1) b0 = _mm_set_epi64x(m11, m14); b1 = _mm_set_epi64x(m3, m6)
+#define LOAD_MSG_4_4(b0, b1) b0 = _mm_set_epi64x(m12, m1); b1 = _mm_set_epi64x(m13, m8)
+#define LOAD_MSG_5_1(b0, b1) b0 = _mm_set_epi64x(m6, m2); b1 = _mm_set_epi64x(m8, m0)
+#define LOAD_MSG_5_2(b0, b1) b0 = _mm_set_epi64x(m10, m12); b1 = _mm_set_epi64x(m3, m11)
+#define LOAD_MSG_5_3(b0, b1) b0 = _mm_set_epi64x(m7, m4); b1 = _mm_set_epi64x(m1, m15)
+#define LOAD_MSG_5_4(b0, b1) b0 = _mm_set_epi64x(m5, m13); b1 = _mm_set_epi64x(m9, m14)
+#define LOAD_MSG_6_1(b0, b1) b0 = _mm_set_epi64x(m1, m12); b1 = _mm_set_epi64x(m4, m14)
+#define LOAD_MSG_6_2(b0, b1) b0 = _mm_set_epi64x(m15, m5); b1 = _mm_set_epi64x(m10, m13)
+#define LOAD_MSG_6_3(b0, b1) b0 = _mm_set_epi64x(m6, m0); b1 = _mm_set_epi64x(m8, m9)
+#define LOAD_MSG_6_4(b0, b1) b0 = _mm_set_epi64x(m3, m7); b1 = _mm_set_epi64x(m11, m2)
+#define LOAD_MSG_7_1(b0, b1) b0 = _mm_set_epi64x(m7, m13); b1 = _mm_set_epi64x(m3, m12)
+#define LOAD_MSG_7_2(b0, b1) b0 = _mm_set_epi64x(m14, m11); b1 = _mm_set_epi64x(m9, m1)
+#define LOAD_MSG_7_3(b0, b1) b0 = _mm_set_epi64x(m15, m5); b1 = _mm_set_epi64x(m2, m8)
+#define LOAD_MSG_7_4(b0, b1) b0 = _mm_set_epi64x(m4, m0); b1 = _mm_set_epi64x(m10, m6)
+#define LOAD_MSG_8_1(b0, b1) b0 = _mm_set_epi64x(m14, m6); b1 = _mm_set_epi64x(m0, m11)
+#define LOAD_MSG_8_2(b0, b1) b0 = _mm_set_epi64x(m9, m15); b1 = _mm_set_epi64x(m8, m3)
+#define LOAD_MSG_8_3(b0, b1) b0 = _mm_set_epi64x(m13, m12); b1 = _mm_set_epi64x(m10, m1)
+#define LOAD_MSG_8_4(b0, b1) b0 = _mm_set_epi64x(m7, m2); b1 = _mm_set_epi64x(m5, m4)
+#define LOAD_MSG_9_1(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 = _mm_set_epi64x(m1, m7)
+#define LOAD_MSG_9_2(b0, b1) b0 = _mm_set_epi64x(m4, m2); b1 = _mm_set_epi64x(m5, m6)
+#define LOAD_MSG_9_3(b0, b1) b0 = _mm_set_epi64x(m9, m15); b1 = _mm_set_epi64x(m13, m3)
+#define LOAD_MSG_9_4(b0, b1) b0 = _mm_set_epi64x(m14, m11); b1 = _mm_set_epi64x(m0, m12)
+#define LOAD_MSG_10_1(b0, b1) b0 = _mm_set_epi64x(m2, m0); b1 = _mm_set_epi64x(m6, m4)
+#define LOAD_MSG_10_2(b0, b1) b0 = _mm_set_epi64x(m3, m1); b1 = _mm_set_epi64x(m7, m5)
+#define LOAD_MSG_10_3(b0, b1) b0 = _mm_set_epi64x(m10, m8); b1 = _mm_set_epi64x(m14, m12)
+#define LOAD_MSG_10_4(b0, b1) b0 = _mm_set_epi64x(m11, m9); b1 = _mm_set_epi64x(m15, m13)
+#define LOAD_MSG_11_1(b0, b1) b0 = _mm_set_epi64x(m4, m14); b1 = _mm_set_epi64x(m13, m9)
+#define LOAD_MSG_11_2(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 = _mm_set_epi64x(m6, m15)
+#define LOAD_MSG_11_3(b0, b1) b0 = _mm_set_epi64x(m0, m1); b1 = _mm_set_epi64x(m5, m11)
+#define LOAD_MSG_11_4(b0, b1) b0 = _mm_set_epi64x(m2, m12); b1 = _mm_set_epi64x(m3, m7)
+
+
+#endif
+

src/blake2b-load-sse41.h

@@ -0,0 +1,402 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2B_LOAD_SSE41_H__
+#define __BLAKE2B_LOAD_SSE41_H__
+
+#define LOAD_MSG_0_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m1); \
+b1 = _mm_unpacklo_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_0_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m0, m1); \
+b1 = _mm_unpackhi_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_0_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m4, m5); \
+b1 = _mm_unpacklo_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_0_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m5); \
+b1 = _mm_unpackhi_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_1_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m7, m2); \
+b1 = _mm_unpackhi_epi64(m4, m6); \
+} while(0)
+
+
+#define LOAD_MSG_1_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m5, m4); \
+b1 = _mm_alignr_epi8(m3, m7, 8); \
+} while(0)
+
+
+#define LOAD_MSG_1_3(b0, b1) \
+do \
+{ \
+b0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
+b1 = _mm_unpackhi_epi64(m5, m2); \
+} while(0)
+
+
+#define LOAD_MSG_1_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m6, m1); \
+b1 = _mm_unpackhi_epi64(m3, m1); \
+} while(0)
+
+
+#define LOAD_MSG_2_1(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m6, m5, 8); \
+b1 = _mm_unpackhi_epi64(m2, m7); \
+} while(0)
+
+
+#define LOAD_MSG_2_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m4, m0); \
+b1 = _mm_blend_epi16(m1, m6, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_2_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m5, m1, 0xF0); \
+b1 = _mm_unpackhi_epi64(m3, m4); \
+} while(0)
+
+
+#define LOAD_MSG_2_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m7, m3); \
+b1 = _mm_alignr_epi8(m2, m0, 8); \
+} while(0)
+
+
+#define LOAD_MSG_3_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m3, m1); \
+b1 = _mm_unpackhi_epi64(m6, m5); \
+} while(0)
+
+
+#define LOAD_MSG_3_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m0); \
+b1 = _mm_unpacklo_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_3_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m1, m2, 0xF0); \
+b1 = _mm_blend_epi16(m2, m7, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_3_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m3, m5); \
+b1 = _mm_unpacklo_epi64(m0, m4); \
+} while(0)
+
+
+#define LOAD_MSG_4_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m2); \
+b1 = _mm_unpacklo_epi64(m1, m5); \
+} while(0)
+
+
+#define LOAD_MSG_4_2(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m0, m3, 0xF0); \
+b1 = _mm_blend_epi16(m2, m7, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_4_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m7, m5, 0xF0); \
+b1 = _mm_blend_epi16(m3, m1, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_4_4(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m6, m0, 8); \
+b1 = _mm_blend_epi16(m4, m6, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_5_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m1, m3); \
+b1 = _mm_unpacklo_epi64(m0, m4); \
+} while(0)
+
+
+#define LOAD_MSG_5_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m6, m5); \
+b1 = _mm_unpackhi_epi64(m5, m1); \
+} while(0)
+
+
+#define LOAD_MSG_5_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m2, m3, 0xF0); \
+b1 = _mm_unpackhi_epi64(m7, m0); \
+} while(0)
+
+
+#define LOAD_MSG_5_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m6, m2); \
+b1 = _mm_blend_epi16(m7, m4, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_6_1(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m6, m0, 0xF0); \
+b1 = _mm_unpacklo_epi64(m7, m2); \
+} while(0)
+
+
+#define LOAD_MSG_6_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m2, m7); \
+b1 = _mm_alignr_epi8(m5, m6, 8); \
+} while(0)
+
+
+#define LOAD_MSG_6_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m3); \
+b1 = _mm_shuffle_epi32(m4, _MM_SHUFFLE(1,0,3,2)); \
+} while(0)
+
+
+#define LOAD_MSG_6_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m3, m1); \
+b1 = _mm_blend_epi16(m1, m5, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_7_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m6, m3); \
+b1 = _mm_blend_epi16(m6, m1, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_7_2(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m7, m5, 8); \
+b1 = _mm_unpackhi_epi64(m0, m4); \
+} while(0)
+
+
+#define LOAD_MSG_7_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m2, m7); \
+b1 = _mm_unpacklo_epi64(m4, m1); \
+} while(0)
+
+
+#define LOAD_MSG_7_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m2); \
+b1 = _mm_unpacklo_epi64(m3, m5); \
+} while(0)
+
+
+#define LOAD_MSG_8_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m3, m7); \
+b1 = _mm_alignr_epi8(m0, m5, 8); \
+} while(0)
+
+
+#define LOAD_MSG_8_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m7, m4); \
+b1 = _mm_alignr_epi8(m4, m1, 8); \
+} while(0)
+
+
+#define LOAD_MSG_8_3(b0, b1) \
+do \
+{ \
+b0 = m6; \
+b1 = _mm_alignr_epi8(m5, m0, 8); \
+} while(0)
+
+
+#define LOAD_MSG_8_4(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m1, m3, 0xF0); \
+b1 = m2; \
+} while(0)
+
+
+#define LOAD_MSG_9_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m5, m4); \
+b1 = _mm_unpackhi_epi64(m3, m0); \
+} while(0)
+
+
+#define LOAD_MSG_9_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m1, m2); \
+b1 = _mm_blend_epi16(m3, m2, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_9_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m7, m4); \
+b1 = _mm_unpackhi_epi64(m1, m6); \
+} while(0)
+
+
+#define LOAD_MSG_9_4(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m7, m5, 8); \
+b1 = _mm_unpacklo_epi64(m6, m0); \
+} while(0)
+
+
+#define LOAD_MSG_10_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m1); \
+b1 = _mm_unpacklo_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_10_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m0, m1); \
+b1 = _mm_unpackhi_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_10_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m4, m5); \
+b1 = _mm_unpacklo_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_10_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m5); \
+b1 = _mm_unpackhi_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_11_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m7, m2); \
+b1 = _mm_unpackhi_epi64(m4, m6); \
+} while(0)
+
+
+#define LOAD_MSG_11_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m5, m4); \
+b1 = _mm_alignr_epi8(m3, m7, 8); \
+} while(0)
+
+
+#define LOAD_MSG_11_3(b0, b1) \
+do \
+{ \
+b0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
+b1 = _mm_unpackhi_epi64(m5, m2); \
+} while(0)
+
+
+#define LOAD_MSG_11_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m6, m1); \
+b1 = _mm_unpackhi_epi64(m3, m1); \
+} while(0)
+
+
+#endif
+

src/blake2b-ref.c

@@ -0,0 +1,381 @@
+/*
+   BLAKE2 reference source code package - reference C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+
+#include <stdint.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "blake2.h"
+#include "blake2-impl.h"
+
+static const uint64_t blake2b_IV[8] =
+{
+  0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
+  0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
+  0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
+  0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
+};
+
+static const uint8_t blake2b_sigma[12][16] =
+{
+  {  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 } ,
+  { 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 } ,
+  { 11,  8, 12,  0,  5,  2, 15, 13, 10, 14,  3,  6,  7,  1,  9,  4 } ,
+  {  7,  9,  3,  1, 13, 12, 11, 14,  2,  6,  5, 10,  4,  0, 15,  8 } ,
+  {  9,  0,  5,  7,  2,  4, 10, 15, 14,  1, 11, 12,  6,  8,  3, 13 } ,
+  {  2, 12,  6, 10,  0, 11,  8,  3,  4, 13,  7,  5, 15, 14,  1,  9 } ,
+  { 12,  5,  1, 15, 14, 13,  4, 10,  0,  7,  6,  3,  9,  2,  8, 11 } ,
+  { 13, 11,  7, 14, 12,  1,  3,  9,  5,  0, 15,  4,  8,  6,  2, 10 } ,
+  {  6, 15, 14,  9, 11,  3,  0,  8, 12,  2, 13,  7,  1,  4, 10,  5 } ,
+  { 10,  2,  8,  4,  7,  6,  1,  5, 15, 11,  9, 14,  3, 12, 13 , 0 } ,
+  {  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 } ,
+  { 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 }
+};
+
+
+static inline int blake2b_set_lastnode( blake2b_state *S )
+{
+  S->f[1] = ~0ULL;
+  return 0;
+}
+
+static inline int blake2b_clear_lastnode( blake2b_state *S )
+{
+  S->f[1] = 0ULL;
+  return 0;
+}
+
+/* Some helper functions, not necessarily useful */
+static inline int blake2b_set_lastblock( blake2b_state *S )
+{
+  if( S->last_node ) blake2b_set_lastnode( S );
+
+  S->f[0] = ~0ULL;
+  return 0;
+}
+
+static inline int blake2b_clear_lastblock( blake2b_state *S )
+{
+  if( S->last_node ) blake2b_clear_lastnode( S );
+
+  S->f[0] = 0ULL;
+  return 0;
+}
+
+static inline int blake2b_increment_counter( blake2b_state *S, const uint64_t inc )
+{
+  S->t[0] += inc;
+  S->t[1] += ( S->t[0] < inc );
+  return 0;
+}
+
+
+
+// Parameter-related functions
+static inline int blake2b_param_set_digest_length( blake2b_param *P, const uint8_t digest_length )
+{
+  P->digest_length = digest_length;
+  return 0;
+}
+
+static inline int blake2b_param_set_fanout( blake2b_param *P, const uint8_t fanout )
+{
+  P->fanout = fanout;
+  return 0;
+}
+
+static inline int blake2b_param_set_max_depth( blake2b_param *P, const uint8_t depth )
+{
+  P->depth = depth;
+  return 0;
+}
+
+static inline int blake2b_param_set_leaf_length( blake2b_param *P, const uint32_t leaf_length )
+{
+  store32( &P->leaf_length, leaf_length );
+  return 0;
+}
+
+static inline int blake2b_param_set_node_offset( blake2b_param *P, const uint64_t node_offset )
+{
+  store64( &P->node_offset, node_offset );
+  return 0;
+}
+
+static inline int blake2b_param_set_node_depth( blake2b_param *P, const uint8_t node_depth )
+{
+  P->node_depth = node_depth;
+  return 0;
+}
+
+static inline int blake2b_param_set_inner_length( blake2b_param *P, const uint8_t inner_length )
+{
+  P->inner_length = inner_length;
+  return 0;
+}
+
+static inline int blake2b_param_set_salt( blake2b_param *P, const uint8_t salt[BLAKE2B_SALTBYTES] )
+{
+  memcpy( P->salt, salt, BLAKE2B_SALTBYTES );
+  return 0;
+}
+
+static inline int blake2b_param_set_personal( blake2b_param *P, const uint8_t personal[BLAKE2B_PERSONALBYTES] )
+{
+  memcpy( P->personal, personal, BLAKE2B_PERSONALBYTES );
+  return 0;
+}
+
+static inline int blake2b_init0( blake2b_state *S )
+{
+  memset( S, 0, sizeof( blake2b_state ) );
+
+  for( int i = 0; i < 8; ++i ) S->h[i] = blake2b_IV[i];
+
+  return 0;
+}
+
+#define blake2b_init BLAKE2_IMPL_NAME(blake2b_init)
+#define blake2b_init_param BLAKE2_IMPL_NAME(blake2b_init_param)
+#define blake2b_init_key BLAKE2_IMPL_NAME(blake2b_init_key)
+#define blake2b_update BLAKE2_IMPL_NAME(blake2b_update)
+#define blake2b_final BLAKE2_IMPL_NAME(blake2b_final)
+#define blake2b BLAKE2_IMPL_NAME(blake2b)
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+  int blake2b_init( blake2b_state *S, size_t outlen );
+  int blake2b_init_param( blake2b_state *S, const blake2b_param *P );
+  int blake2b_init_key( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_update( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+#if defined(__cplusplus)
+}
+#endif
+
+/* init xors IV with input parameter block */
+int blake2b_init_param( blake2b_state *S, const blake2b_param *P )
+{
+  blake2b_init0( S );
+  uint8_t *p = ( uint8_t * )( P );
+
+  /* IV XOR ParamBlock */
+  for( size_t i = 0; i < 8; ++i )
+    S->h[i] ^= load64( p + sizeof( S->h[i] ) * i );
+
+  S->outlen = P->digest_length;
+  return 0;
+}
+
+
+
+int blake2b_init( blake2b_state *S, size_t outlen )
+{
+  blake2b_param P[1];
+
+  if ( ( !outlen ) || ( outlen > BLAKE2B_OUTBYTES ) ) return -1;
+
+  P->digest_length = outlen;
+  P->key_length    = 0;
+  P->fanout        = 1;
+  P->depth         = 1;
+  store32( &P->leaf_length, 0 );
+  store64( &P->node_offset, 0 );
+  P->node_depth    = 0;
+  P->inner_length  = 0;
+  memset( P->reserved, 0, sizeof( P->reserved ) );
+  memset( P->salt,     0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+  return blake2b_init_param( S, P );
+}
+
+
+int blake2b_init_key( blake2b_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  blake2b_param P[1];
+
+  if ( ( !outlen ) || ( outlen > BLAKE2B_OUTBYTES ) ) return -1;
+
+  if ( !key || !keylen || keylen > BLAKE2B_KEYBYTES ) return -1;
+
+  P->digest_length = outlen;
+  P->key_length    = keylen;
+  P->fanout        = 1;
+  P->depth         = 1;
+  store32( &P->leaf_length, 0 );
+  store64( &P->node_offset, 0 );
+  P->node_depth    = 0;
+  P->inner_length  = 0;
+  memset( P->reserved, 0, sizeof( P->reserved ) );
+  memset( P->salt,     0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+
+  if( blake2b_init_param( S, P ) < 0 ) return -1;
+
+  {
+    uint8_t block[BLAKE2B_BLOCKBYTES];
+    memset( block, 0, BLAKE2B_BLOCKBYTES );
+    memcpy( block, key, keylen );
+    blake2b_update( S, block, BLAKE2B_BLOCKBYTES );
+    secure_zero_memory( block, BLAKE2B_BLOCKBYTES ); /* Burn the key from stack */
+  }
+  return 0;
+}
+
+static int blake2b_compress( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] )
+{
+  uint64_t m[16];
+  uint64_t v[16];
+  int i;
+
+  for( i = 0; i < 16; ++i )
+    m[i] = load64( block + i * sizeof( m[i] ) );
+
+  for( i = 0; i < 8; ++i )
+    v[i] = S->h[i];
+
+  v[ 8] = blake2b_IV[0];
+  v[ 9] = blake2b_IV[1];
+  v[10] = blake2b_IV[2];
+  v[11] = blake2b_IV[3];
+  v[12] = S->t[0] ^ blake2b_IV[4];
+  v[13] = S->t[1] ^ blake2b_IV[5];
+  v[14] = S->f[0] ^ blake2b_IV[6];
+  v[15] = S->f[1] ^ blake2b_IV[7];
+#define G(r,i,a,b,c,d) \
+  do { \
+    a = a + b + m[blake2b_sigma[r][2*i+0]]; \
+    d = rotr64(d ^ a, 32); \
+    c = c + d; \
+    b = rotr64(b ^ c, 24); \
+    a = a + b + m[blake2b_sigma[r][2*i+1]]; \
+    d = rotr64(d ^ a, 16); \
+    c = c + d; \
+    b = rotr64(b ^ c, 63); \
+  } while(0)
+#define ROUND(r)  \
+  do { \
+    G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
+    G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
+    G(r,2,v[ 2],v[ 6],v[10],v[14]); \
+    G(r,3,v[ 3],v[ 7],v[11],v[15]); \
+    G(r,4,v[ 0],v[ 5],v[10],v[15]); \
+    G(r,5,v[ 1],v[ 6],v[11],v[12]); \
+    G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
+    G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
+  } while(0)
+  ROUND( 0 );
+  ROUND( 1 );
+  ROUND( 2 );
+  ROUND( 3 );
+  ROUND( 4 );
+  ROUND( 5 );
+  ROUND( 6 );
+  ROUND( 7 );
+  ROUND( 8 );
+  ROUND( 9 );
+  ROUND( 10 );
+  ROUND( 11 );
+
+  for( i = 0; i < 8; ++i )
+    S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
+
+#undef G
+#undef ROUND
+  return 0;
+}
+
+
+int blake2b_update( blake2b_state *S, const uint8_t *in, size_t inlen )
+{
+  while( inlen > 0 )
+  {
+    size_t left = S->buflen;
+    size_t fill = 2 * BLAKE2B_BLOCKBYTES - left;
+
+    if( inlen > fill )
+    {
+      memcpy( S->buf + left, in, fill ); // Fill buffer
+      S->buflen += fill;
+      blake2b_increment_counter( S, BLAKE2B_BLOCKBYTES );
+      blake2b_compress( S, S->buf ); // Compress
+      memcpy( S->buf, S->buf + BLAKE2B_BLOCKBYTES, BLAKE2B_BLOCKBYTES ); // Shift buffer left
+      S->buflen -= BLAKE2B_BLOCKBYTES;
+      in += fill;
+      inlen -= fill;
+    }
+    else // inlen <= fill
+    {
+      memcpy( S->buf + left, in, inlen );
+      S->buflen += inlen; // Be lazy, do not compress
+      in += inlen;
+      inlen -= inlen;
+    }
+  }
+
+  return 0;
+}
+
+int blake2b_final( blake2b_state *S, uint8_t *out, size_t outlen )
+{
+  uint8_t buffer[BLAKE2B_OUTBYTES];
+
+  if(S->outlen != outlen) return -1;
+
+  if( S->buflen > BLAKE2B_BLOCKBYTES )
+  {
+    blake2b_increment_counter( S, BLAKE2B_BLOCKBYTES );
+    blake2b_compress( S, S->buf );
+    S->buflen -= BLAKE2B_BLOCKBYTES;
+    memcpy( S->buf, S->buf + BLAKE2B_BLOCKBYTES, S->buflen );
+  }
+
+  blake2b_increment_counter( S, S->buflen );
+  blake2b_set_lastblock( S );
+  memset( S->buf + S->buflen, 0, 2 * BLAKE2B_BLOCKBYTES - S->buflen ); /* Padding */
+  blake2b_compress( S, S->buf );
+
+  for( int i = 0; i < 8; ++i ) /* Output full hash to temp buffer */
+    store64( buffer + sizeof( S->h[i] ) * i, S->h[i] );
+
+  memcpy( out, buffer, outlen );
+  return 0;
+}
+
+int blake2b( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  blake2b_state S[1];
+
+  /* Verify parameters */
+  if ( NULL == in ) return -1;
+
+  if ( NULL == out ) return -1;
+
+  if( NULL == key && keylen > 0 ) return -1;
+
+  if( keylen > 0 )
+  {
+    if( blake2b_init_key( S, outlen, key, keylen ) < 0 ) return -1;
+  }
+  else
+  {
+    if( blake2b_init( S, outlen ) < 0 ) return -1;
+  }
+
+  if( blake2b_update( S, ( uint8_t * )in, inlen ) < 0 ) return -1;
+  return blake2b_final( S, out, outlen );
+}
+
+

src/blake2b-round.h

@@ -0,0 +1,160 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2B_ROUND_H__
+#define __BLAKE2B_ROUND_H__
+
+#define LOAD(p)  _mm_load_si128( (__m128i *)(p) )
+#define STORE(p,r) _mm_store_si128((__m128i *)(p), r)
+
+#define LOADU(p)  _mm_loadu_si128( (__m128i *)(p) )
+#define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r)
+
+#define TOF(reg) _mm_castsi128_ps((reg))
+#define TOI(reg) _mm_castps_si128((reg))
+
+#define LIKELY(x) __builtin_expect((x),1)
+
+
+/* Microarchitecture-specific macros */
+#ifndef HAVE_XOP
+#ifdef HAVE_SSSE3
+#define _mm_roti_epi64(x, c) \
+    (-(c) == 32) ? _mm_shuffle_epi32((x), _MM_SHUFFLE(2,3,0,1))  \
+    : (-(c) == 24) ? _mm_shuffle_epi8((x), r24) \
+    : (-(c) == 16) ? _mm_shuffle_epi8((x), r16) \
+    : (-(c) == 63) ? _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_add_epi64((x), (x)))  \
+    : _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_slli_epi64((x), 64-(-(c))))
+#else
+#define _mm_roti_epi64(r, c) _mm_xor_si128(_mm_srli_epi64( (r), -(c) ),_mm_slli_epi64( (r), 64-(-(c)) ))
+#endif
+#else
+/* ... */
+#endif
+
+
+
+#define G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
+  row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
+  row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
+  \
+  row4l = _mm_xor_si128(row4l, row1l); \
+  row4h = _mm_xor_si128(row4h, row1h); \
+  \
+  row4l = _mm_roti_epi64(row4l, -32); \
+  row4h = _mm_roti_epi64(row4h, -32); \
+  \
+  row3l = _mm_add_epi64(row3l, row4l); \
+  row3h = _mm_add_epi64(row3h, row4h); \
+  \
+  row2l = _mm_xor_si128(row2l, row3l); \
+  row2h = _mm_xor_si128(row2h, row3h); \
+  \
+  row2l = _mm_roti_epi64(row2l, -24); \
+  row2h = _mm_roti_epi64(row2h, -24); \
+ 
+#define G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
+  row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
+  row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
+  \
+  row4l = _mm_xor_si128(row4l, row1l); \
+  row4h = _mm_xor_si128(row4h, row1h); \
+  \
+  row4l = _mm_roti_epi64(row4l, -16); \
+  row4h = _mm_roti_epi64(row4h, -16); \
+  \
+  row3l = _mm_add_epi64(row3l, row4l); \
+  row3h = _mm_add_epi64(row3h, row4h); \
+  \
+  row2l = _mm_xor_si128(row2l, row3l); \
+  row2h = _mm_xor_si128(row2h, row3h); \
+  \
+  row2l = _mm_roti_epi64(row2l, -63); \
+  row2h = _mm_roti_epi64(row2h, -63); \
+ 
+#if defined(HAVE_SSSE3)
+#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+  t0 = _mm_alignr_epi8(row2h, row2l, 8); \
+  t1 = _mm_alignr_epi8(row2l, row2h, 8); \
+  row2l = t0; \
+  row2h = t1; \
+  \
+  t0 = row3l; \
+  row3l = row3h; \
+  row3h = t0;    \
+  \
+  t0 = _mm_alignr_epi8(row4h, row4l, 8); \
+  t1 = _mm_alignr_epi8(row4l, row4h, 8); \
+  row4l = t1; \
+  row4h = t0;
+
+#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+  t0 = _mm_alignr_epi8(row2l, row2h, 8); \
+  t1 = _mm_alignr_epi8(row2h, row2l, 8); \
+  row2l = t0; \
+  row2h = t1; \
+  \
+  t0 = row3l; \
+  row3l = row3h; \
+  row3h = t0; \
+  \
+  t0 = _mm_alignr_epi8(row4l, row4h, 8); \
+  t1 = _mm_alignr_epi8(row4h, row4l, 8); \
+  row4l = t1; \
+  row4h = t0;
+#else
+
+#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+  t0 = row4l;\
+  t1 = row2l;\
+  row4l = row3l;\
+  row3l = row3h;\
+  row3h = row4l;\
+  row4l = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t0, t0)); \
+  row4h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row4h, row4h)); \
+  row2l = _mm_unpackhi_epi64(row2l, _mm_unpacklo_epi64(row2h, row2h)); \
+  row2h = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(t1, t1))
+
+#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+  t0 = row3l;\
+  row3l = row3h;\
+  row3h = t0;\
+  t0 = row2l;\
+  t1 = row4l;\
+  row2l = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(row2l, row2l)); \
+  row2h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row2h, row2h)); \
+  row4l = _mm_unpackhi_epi64(row4l, _mm_unpacklo_epi64(row4h, row4h)); \
+  row4h = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t1, t1))
+
+#endif
+
+#if defined(HAVE_SSE4_1)
+#include "blake2b-load-sse41.h"
+#else
+#include "blake2b-load-sse2.h"
+#endif
+
+#define ROUND(r) \
+  LOAD_MSG_ ##r ##_1(b0, b1); \
+  G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+  LOAD_MSG_ ##r ##_2(b0, b1); \
+  G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+  DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
+  LOAD_MSG_ ##r ##_3(b0, b1); \
+  G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+  LOAD_MSG_ ##r ##_4(b0, b1); \
+  G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+  UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
+
+#endif
+

src/blake2b-test.c

@@ -0,0 +1,43 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#include <stdio.h>
+#include <string.h>
+#include "blake2.h"
+#include "blake2-kat.h"
+int main( int argc, char **argv )
+{
+  uint8_t key[BLAKE2B_KEYBYTES];
+  uint8_t buf[KAT_LENGTH];
+
+  for( size_t i = 0; i < BLAKE2B_KEYBYTES; ++i )
+    key[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+    buf[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+  {
+    uint8_t hash[BLAKE2B_OUTBYTES];
+
+    if( blake2b( hash, buf, key, BLAKE2B_OUTBYTES, i, BLAKE2B_KEYBYTES ) < 0 || 
+        0 != memcmp( hash, blake2b_keyed_kat[i], BLAKE2B_OUTBYTES ) )
+    {
+      puts( "error" );
+      return -1;
+    }
+  }
+
+  puts( "ok" );
+  return 0;
+}
+

src/blake2b.c

@@ -0,0 +1,439 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+
+#include <stdint.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "blake2.h"
+#include "blake2-impl.h"
+
+#include "blake2-config.h"
+
+#if defined(_MSC_VER)
+#include <intrin.h>
+#endif
+
+#if defined(HAVE_SSE2)
+#include <emmintrin.h>
+// MSVC only defines  _mm_set_epi64x for x86_64...
+#if defined(_MSC_VER) && !defined(_M_X64)
+static inline __m128i _mm_set_epi64x( const uint64_t u1, const uint64_t u0 )
+{
+  return _mm_set_epi32( u1 >> 32, u1, u0 >> 32, u0 );
+}
+#endif
+#endif
+
+#if defined(HAVE_SSSE3)
+#include <tmmintrin.h>
+#endif
+#if defined(HAVE_SSE4_1)
+#include <smmintrin.h>
+#endif
+#if defined(HAVE_AVX)
+#include <immintrin.h>
+#endif
+#if defined(HAVE_XOP) && !defined(_MSC_VER)
+#include <x86intrin.h>
+#endif
+
+
+
+#include "blake2b-round.h"
+
+static const uint64_t blake2b_IV[8] =
+{
+  0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
+  0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
+  0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
+  0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
+};
+
+static const uint8_t blake2b_sigma[12][16] =
+{
+  {  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 } ,
+  { 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 } ,
+  { 11,  8, 12,  0,  5,  2, 15, 13, 10, 14,  3,  6,  7,  1,  9,  4 } ,
+  {  7,  9,  3,  1, 13, 12, 11, 14,  2,  6,  5, 10,  4,  0, 15,  8 } ,
+  {  9,  0,  5,  7,  2,  4, 10, 15, 14,  1, 11, 12,  6,  8,  3, 13 } ,
+  {  2, 12,  6, 10,  0, 11,  8,  3,  4, 13,  7,  5, 15, 14,  1,  9 } ,
+  { 12,  5,  1, 15, 14, 13,  4, 10,  0,  7,  6,  3,  9,  2,  8, 11 } ,
+  { 13, 11,  7, 14, 12,  1,  3,  9,  5,  0, 15,  4,  8,  6,  2, 10 } ,
+  {  6, 15, 14,  9, 11,  3,  0,  8, 12,  2, 13,  7,  1,  4, 10,  5 } ,
+  { 10,  2,  8,  4,  7,  6,  1,  5, 15, 11,  9, 14,  3, 12, 13 , 0 } ,
+  {  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 } ,
+  { 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 }
+};
+
+
+/* Some helper functions, not necessarily useful */
+static inline int blake2b_set_lastnode( blake2b_state *S )
+{
+  S->f[1] = ~0ULL;
+  return 0;
+}
+
+static inline int blake2b_clear_lastnode( blake2b_state *S )
+{
+  S->f[1] = 0ULL;
+  return 0;
+}
+
+static inline int blake2b_set_lastblock( blake2b_state *S )
+{
+  if( S->last_node ) blake2b_set_lastnode( S );
+
+  S->f[0] = ~0ULL;
+  return 0;
+}
+
+static inline int blake2b_clear_lastblock( blake2b_state *S )
+{
+  if( S->last_node ) blake2b_clear_lastnode( S );
+
+  S->f[0] = 0ULL;
+  return 0;
+}
+
+
+static inline int blake2b_increment_counter( blake2b_state *S, const uint64_t inc )
+{
+#if defined(__x86_64__) && (defined(__GNUC__) || defined(__clang__))
+  // ADD/ADC chain
+  __uint128_t t = ( ( __uint128_t )S->t[1] << 64 ) | S->t[0];
+  t += inc;
+  S->t[0] = ( uint64_t )( t >>  0 );
+  S->t[1] = ( uint64_t )( t >> 64 );
+#else
+  S->t[0] += inc;
+  S->t[1] += ( S->t[0] < inc );
+#endif
+  return 0;
+}
+
+
+// Parameter-related functions
+static inline int blake2b_param_set_digest_length( blake2b_param *P, const uint8_t digest_length )
+{
+  P->digest_length = digest_length;
+  return 0;
+}
+
+static inline int blake2b_param_set_fanout( blake2b_param *P, const uint8_t fanout )
+{
+  P->fanout = fanout;
+  return 0;
+}
+
+static inline int blake2b_param_set_max_depth( blake2b_param *P, const uint8_t depth )
+{
+  P->depth = depth;
+  return 0;
+}
+
+static inline int blake2b_param_set_leaf_length( blake2b_param *P, const uint32_t leaf_length )
+{
+  P->leaf_length = leaf_length;
+  return 0;
+}
+
+static inline int blake2b_param_set_node_offset( blake2b_param *P, const uint64_t node_offset )
+{
+  P->node_offset = node_offset;
+  return 0;
+}
+
+static inline int blake2b_param_set_node_depth( blake2b_param *P, const uint8_t node_depth )
+{
+  P->node_depth = node_depth;
+  return 0;
+}
+
+static inline int blake2b_param_set_inner_length( blake2b_param *P, const uint8_t inner_length )
+{
+  P->inner_length = inner_length;
+  return 0;
+}
+
+static inline int blake2b_param_set_salt( blake2b_param *P, const uint8_t salt[BLAKE2B_SALTBYTES] )
+{
+  memcpy( P->salt, salt, BLAKE2B_SALTBYTES );
+  return 0;
+}
+
+static inline int blake2b_param_set_personal( blake2b_param *P, const uint8_t personal[BLAKE2B_PERSONALBYTES] )
+{
+  memcpy( P->personal, personal, BLAKE2B_PERSONALBYTES );
+  return 0;
+}
+
+static inline int blake2b_init0( blake2b_state *S )
+{
+  memset( S, 0, sizeof( blake2b_state ) );
+
+  for( int i = 0; i < 8; ++i ) S->h[i] = blake2b_IV[i];
+
+  return 0;
+}
+
+
+
+#define blake2b_init BLAKE2_IMPL_NAME(blake2b_init)
+#define blake2b_init_param BLAKE2_IMPL_NAME(blake2b_init_param)
+#define blake2b_init_key BLAKE2_IMPL_NAME(blake2b_init_key)
+#define blake2b_update BLAKE2_IMPL_NAME(blake2b_update)
+#define blake2b_final BLAKE2_IMPL_NAME(blake2b_final)
+#define blake2b BLAKE2_IMPL_NAME(blake2b)
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+  int blake2b_init( blake2b_state *S, size_t outlen );
+  int blake2b_init_param( blake2b_state *S, const blake2b_param *P );
+  int blake2b_init_key( blake2b_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2b_update( blake2b_state *S, const uint8_t *in, size_t inlen );
+  int blake2b_final( blake2b_state *S, uint8_t *out, size_t outlen );
+  int blake2b( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+#if defined(__cplusplus)
+}
+#endif
+
+/* init xors IV with input parameter block */
+int blake2b_init_param( blake2b_state *S, const blake2b_param *P )
+{
+  uint8_t *p, *h, *v;
+  //blake2b_init0( S );
+  v = ( uint8_t * )( blake2b_IV );
+  h = ( uint8_t * )( S->h );
+  p = ( uint8_t * )( P );
+  /* IV XOR ParamBlock */
+  memset( S, 0, sizeof( blake2b_state ) );
+
+  for( int i = 0; i < BLAKE2B_OUTBYTES; ++i ) h[i] = v[i] ^ p[i];
+
+  S->outlen = P->digest_length;
+  return 0;
+}
+
+
+/* Some sort of default parameter block initialization, for sequential blake2b */
+
+int blake2b_init( blake2b_state *S, size_t outlen )
+{
+  if ( ( !outlen ) || ( outlen > BLAKE2B_OUTBYTES ) ) return -1;
+
+  const blake2b_param P =
+  {
+    outlen,
+    0,
+    1,
+    1,
+    0,
+    0,
+    0,
+    0,
+    {0},
+    {0},
+    {0}
+  };
+  return blake2b_init_param( S, &P );
+}
+
+int blake2b_init_key( blake2b_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  if ( ( !outlen ) || ( outlen > BLAKE2B_OUTBYTES ) ) return -1;
+
+  if ( ( !keylen ) || keylen > BLAKE2B_KEYBYTES ) return -1;
+
+  const blake2b_param P =
+  {
+    outlen,
+    keylen,
+    1,
+    1,
+    0,
+    0,
+    0,
+    0,
+    {0},
+    {0},
+    {0}
+  };
+
+  if( blake2b_init_param( S, &P ) < 0 )
+    return 0;
+
+  {
+    uint8_t block[BLAKE2B_BLOCKBYTES];
+    memset( block, 0, BLAKE2B_BLOCKBYTES );
+    memcpy( block, key, keylen );
+    blake2b_update( S, block, BLAKE2B_BLOCKBYTES );
+    secure_zero_memory( block, BLAKE2B_BLOCKBYTES ); /* Burn the key from stack */
+  }
+  return 0;
+}
+
+static inline int blake2b_compress( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] )
+{
+  __m128i row1l, row1h;
+  __m128i row2l, row2h;
+  __m128i row3l, row3h;
+  __m128i row4l, row4h;
+  __m128i b0, b1;
+  __m128i t0, t1;
+#if defined(HAVE_SSSE3) && !defined(HAVE_XOP)
+  const __m128i r16 = _mm_setr_epi8( 2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9 );
+  const __m128i r24 = _mm_setr_epi8( 3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10 );
+#endif
+#if defined(HAVE_SSE4_1)
+  const __m128i m0 = LOADU( block + 00 );
+  const __m128i m1 = LOADU( block + 16 );
+  const __m128i m2 = LOADU( block + 32 );
+  const __m128i m3 = LOADU( block + 48 );
+  const __m128i m4 = LOADU( block + 64 );
+  const __m128i m5 = LOADU( block + 80 );
+  const __m128i m6 = LOADU( block + 96 );
+  const __m128i m7 = LOADU( block + 112 );
+#else
+  const uint64_t  m0 = ( ( uint64_t * )block )[ 0];
+  const uint64_t  m1 = ( ( uint64_t * )block )[ 1];
+  const uint64_t  m2 = ( ( uint64_t * )block )[ 2];
+  const uint64_t  m3 = ( ( uint64_t * )block )[ 3];
+  const uint64_t  m4 = ( ( uint64_t * )block )[ 4];
+  const uint64_t  m5 = ( ( uint64_t * )block )[ 5];
+  const uint64_t  m6 = ( ( uint64_t * )block )[ 6];
+  const uint64_t  m7 = ( ( uint64_t * )block )[ 7];
+  const uint64_t  m8 = ( ( uint64_t * )block )[ 8];
+  const uint64_t  m9 = ( ( uint64_t * )block )[ 9];
+  const uint64_t m10 = ( ( uint64_t * )block )[10];
+  const uint64_t m11 = ( ( uint64_t * )block )[11];
+  const uint64_t m12 = ( ( uint64_t * )block )[12];
+  const uint64_t m13 = ( ( uint64_t * )block )[13];
+  const uint64_t m14 = ( ( uint64_t * )block )[14];
+  const uint64_t m15 = ( ( uint64_t * )block )[15];
+#endif
+  row1l = LOADU( &S->h[0] );
+  row1h = LOADU( &S->h[2] );
+  row2l = LOADU( &S->h[4] );
+  row2h = LOADU( &S->h[6] );
+  row3l = LOADU( &blake2b_IV[0] );
+  row3h = LOADU( &blake2b_IV[2] );
+  row4l = _mm_xor_si128( LOADU( &blake2b_IV[4] ), LOADU( &S->t[0] ) );
+  row4h = _mm_xor_si128( LOADU( &blake2b_IV[6] ), LOADU( &S->f[0] ) );
+  ROUND( 0 );
+  ROUND( 1 );
+  ROUND( 2 );
+  ROUND( 3 );
+  ROUND( 4 );
+  ROUND( 5 );
+  ROUND( 6 );
+  ROUND( 7 );
+  ROUND( 8 );
+  ROUND( 9 );
+  ROUND( 10 );
+  ROUND( 11 );
+  row1l = _mm_xor_si128( row3l, row1l );
+  row1h = _mm_xor_si128( row3h, row1h );
+  STOREU( &S->h[0], _mm_xor_si128( LOADU( &S->h[0] ), row1l ) );
+  STOREU( &S->h[2], _mm_xor_si128( LOADU( &S->h[2] ), row1h ) );
+  row2l = _mm_xor_si128( row4l, row2l );
+  row2h = _mm_xor_si128( row4h, row2h );
+  STOREU( &S->h[4], _mm_xor_si128( LOADU( &S->h[4] ), row2l ) );
+  STOREU( &S->h[6], _mm_xor_si128( LOADU( &S->h[6] ), row2h ) );
+  return 0;
+}
+
+
+int blake2b_update( blake2b_state *S, const uint8_t *in, size_t inlen )
+{
+  while( inlen > 0 )
+  {
+    size_t left = S->buflen;
+    size_t fill = 2 * BLAKE2B_BLOCKBYTES - left;
+
+    if( inlen > fill )
+    {
+      memcpy( S->buf + left, in, fill ); // Fill buffer
+      S->buflen += fill;
+      blake2b_increment_counter( S, BLAKE2B_BLOCKBYTES );
+      blake2b_compress( S, S->buf ); // Compress
+      memcpy( S->buf, S->buf + BLAKE2B_BLOCKBYTES, BLAKE2B_BLOCKBYTES ); // Shift buffer left
+      S->buflen -= BLAKE2B_BLOCKBYTES;
+      in += fill;
+      inlen -= fill;
+    }
+    else // inlen <= fill
+    {
+      memcpy( S->buf + left, in, inlen );
+      S->buflen += inlen; // Be lazy, do not compress
+      in += inlen;
+      inlen -= inlen;
+    }
+  }
+
+  return 0;
+}
+
+
+int blake2b_final( blake2b_state *S, uint8_t *out, size_t outlen )
+{
+  if(S->outlen != outlen) return -1;
+
+  if( S->buflen > BLAKE2B_BLOCKBYTES )
+  {
+    blake2b_increment_counter( S, BLAKE2B_BLOCKBYTES );
+    blake2b_compress( S, S->buf );
+    S->buflen -= BLAKE2B_BLOCKBYTES;
+    memcpy( S->buf, S->buf + BLAKE2B_BLOCKBYTES, S->buflen );
+  }
+
+  blake2b_increment_counter( S, S->buflen );
+  blake2b_set_lastblock( S );
+  memset( S->buf + S->buflen, 0, 2 * BLAKE2B_BLOCKBYTES - S->buflen ); /* Padding */
+  blake2b_compress( S, S->buf );
+  memcpy( out, &S->h[0], outlen );
+  return 0;
+}
+
+
+int blake2b( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  blake2b_state S[1];
+
+  /* Verify parameters */
+  if ( NULL == in ) return -1;
+
+  if ( NULL == out ) return -1;
+
+  if( NULL == key && keylen > 0 ) return -1;
+
+  if( keylen )
+  {
+    if( blake2b_init_key( S, outlen, key, keylen ) < 0 ) return -1;
+  }
+  else
+  {
+    if( blake2b_init( S, outlen ) < 0 ) return -1;
+  }
+
+  if( blake2b_update( S, ( uint8_t * )in, inlen ) < 0) return -1;
+  return blake2b_final( S, out, outlen );
+}
+
+#if defined(SUPERCOP)
+int crypto_hash( unsigned char *out, unsigned char *in, unsigned long long inlen )
+{
+  return blake2b( out, in, NULL, BLAKE2B_OUTBYTES, inlen, 0 );
+}
+#endif

src/blake2bp-test.c

@@ -0,0 +1,44 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#include <stdio.h>
+#include <string.h>
+#include "blake2.h"
+#include "blake2-kat.h"
+
+int main( int argc, char **argv )
+{
+  uint8_t key[BLAKE2B_KEYBYTES];
+  uint8_t buf[KAT_LENGTH];
+
+  for( size_t i = 0; i < BLAKE2B_KEYBYTES; ++i )
+    key[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+    buf[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+  {
+    uint8_t hash[BLAKE2B_OUTBYTES];
+
+    if( blake2bp( hash, buf, key, BLAKE2B_OUTBYTES, i, BLAKE2B_KEYBYTES ) < 0 || 
+        0 != memcmp( hash, blake2bp_keyed_kat[i], BLAKE2B_OUTBYTES ) )
+    {
+      puts( "error" );
+      return -1;
+    }
+  }
+
+  puts( "ok" );
+  return 0;
+}
+

src/blake2bp.c

@@ -0,0 +1,272 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+
+#if defined(_OPENMP)
+#include <omp.h>
+#endif
+
+#include "blake2.h"
+#include "blake2-impl.h"
+
+#define PARALLELISM_DEGREE 4
+
+static int blake2bp_init_leaf( blake2b_state *S, uint8_t outlen, uint8_t keylen, uint64_t offset )
+{
+  blake2b_param P[1];
+  P->digest_length = outlen;
+  P->key_length = keylen;
+  P->fanout = PARALLELISM_DEGREE;
+  P->depth = 2;
+  P->leaf_length = 0;
+  P->node_offset = offset;
+  P->node_depth = 0;
+  P->inner_length = BLAKE2B_OUTBYTES;
+  memset( P->reserved, 0, sizeof( P->reserved ) );
+  memset( P->salt, 0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+  blake2b_init_param( S, P );
+  S->outlen = P->inner_length;
+  return 0;
+}
+
+static int blake2bp_init_root( blake2b_state *S, uint8_t outlen, uint8_t keylen )
+{
+  blake2b_param P[1];
+  P->digest_length = outlen;
+  P->key_length = keylen;
+  P->fanout = PARALLELISM_DEGREE;
+  P->depth = 2;
+  P->leaf_length = 0;
+  P->node_offset = 0;
+  P->node_depth = 1;
+  P->inner_length = BLAKE2B_OUTBYTES;
+  memset( P->reserved, 0, sizeof( P->reserved ) );
+  memset( P->salt, 0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+  blake2b_init_param( S, P );
+  S->outlen = P->digest_length;
+  return 0;
+}
+
+
+int blake2bp_init( blake2bp_state *S, size_t outlen )
+{
+  if( !outlen || outlen > BLAKE2B_OUTBYTES ) return -1;
+
+  memset( S->buf, 0, sizeof( S->buf ) );
+  S->buflen = 0;
+
+  if( blake2bp_init_root( S->R, outlen, 0 ) < 0 )
+    return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    if( blake2bp_init_leaf( S->S[i], outlen, 0, i ) < 0 ) return -1;
+
+  S->R->last_node = 1;
+  S->S[PARALLELISM_DEGREE - 1]->last_node = 1;
+  S->outlen = outlen;
+  return 0;
+}
+
+int blake2bp_init_key( blake2bp_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  if( !outlen || outlen > BLAKE2B_OUTBYTES ) return -1;
+
+  if( !key || !keylen || keylen > BLAKE2B_KEYBYTES ) return -1;
+
+  memset( S->buf, 0, sizeof( S->buf ) );
+  S->buflen = 0;
+
+  if( blake2bp_init_root( S->R, outlen, keylen ) < 0 )
+    return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    if( blake2bp_init_leaf( S->S[i], outlen, keylen, i ) < 0 ) return -1;
+
+  S->R->last_node = 1;
+  S->S[PARALLELISM_DEGREE - 1]->last_node = 1;
+  S->outlen = outlen;
+  {
+    uint8_t block[BLAKE2B_BLOCKBYTES];
+    memset( block, 0, BLAKE2B_BLOCKBYTES );
+    memcpy( block, key, keylen );
+
+    for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+      blake2b_update( S->S[i], block, BLAKE2B_BLOCKBYTES );
+
+    secure_zero_memory( block, BLAKE2B_BLOCKBYTES ); /* Burn the key from stack */
+  }
+  return 0;
+}
+
+
+int blake2bp_update( blake2bp_state *S, const uint8_t *in, size_t inlen )
+{
+  size_t left = S->buflen;
+  size_t fill = sizeof( S->buf ) - left;
+
+  if( left && inlen >= fill )
+  {
+    memcpy( S->buf + left, in, fill );
+
+    for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+      blake2b_update( S->S[i], S->buf + i * BLAKE2B_BLOCKBYTES, BLAKE2B_BLOCKBYTES );
+
+    in += fill;
+    inlen -= fill;
+    left = 0;
+  }
+
+#if defined(_OPENMP)
+  omp_set_num_threads(PARALLELISM_DEGREE);
+  #pragma omp parallel shared(S)
+#else
+  for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
+#endif
+  {
+#if defined(_OPENMP)
+    size_t      id__ = omp_get_thread_num();
+#endif
+    size_t inlen__ = inlen;
+    const uint8_t *in__ = ( const uint8_t * )in;
+    in__ += id__ * BLAKE2B_BLOCKBYTES;
+
+    while( inlen__ >= PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES )
+    {
+      blake2b_update( S->S[id__], in__, BLAKE2B_BLOCKBYTES );
+      in__ += PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES;
+      inlen__ -= PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES;
+    }
+  }
+
+  in += inlen - inlen % ( PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES );
+  inlen %= PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES;
+
+  if( inlen > 0 )
+    memcpy( S->buf + left, in, inlen );
+
+  S->buflen = left + inlen;
+  return 0;
+}
+
+
+
+int blake2bp_final( blake2bp_state *S, uint8_t *out, size_t outlen )
+{
+  uint8_t hash[PARALLELISM_DEGREE][BLAKE2B_OUTBYTES];
+
+  if(S->outlen != outlen) return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+  {
+    if( S->buflen > i * BLAKE2B_BLOCKBYTES )
+    {
+      size_t left = S->buflen - i * BLAKE2B_BLOCKBYTES;
+
+      if( left > BLAKE2B_BLOCKBYTES ) left = BLAKE2B_BLOCKBYTES;
+
+      blake2b_update( S->S[i], S->buf + i * BLAKE2B_BLOCKBYTES, left );
+    }
+
+    blake2b_final( S->S[i], hash[i], BLAKE2B_OUTBYTES );
+  }
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    blake2b_update( S->R, hash[i], BLAKE2B_OUTBYTES );
+
+  return blake2b_final( S->R, out, outlen );
+}
+
+int blake2bp( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  uint8_t hash[PARALLELISM_DEGREE][BLAKE2B_OUTBYTES];
+  blake2b_state S[PARALLELISM_DEGREE][1];
+  blake2b_state FS[1];
+
+  /* Verify parameters */
+  if ( NULL == in ) return -1;
+
+  if ( NULL == out ) return -1;
+
+  if ( NULL == key && keylen > 0) return -1;
+
+  if( !outlen || outlen > BLAKE2B_OUTBYTES ) return -1;
+
+  if( keylen > BLAKE2B_KEYBYTES ) return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    if( blake2bp_init_leaf( S[i], outlen, keylen, i ) < 0 ) return -1;
+
+  S[PARALLELISM_DEGREE - 1]->last_node = 1; // mark last node
+
+  if( keylen > 0 )
+  {
+    uint8_t block[BLAKE2B_BLOCKBYTES];
+    memset( block, 0, BLAKE2B_BLOCKBYTES );
+    memcpy( block, key, keylen );
+
+    for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+      blake2b_update( S[i], block, BLAKE2B_BLOCKBYTES );
+
+    secure_zero_memory( block, BLAKE2B_BLOCKBYTES ); /* Burn the key from stack */
+  }
+
+#if defined(_OPENMP)
+  omp_set_num_threads(PARALLELISM_DEGREE);
+  #pragma omp parallel shared(S,hash)
+#else
+  for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
+#endif
+  {
+#if defined(_OPENMP)
+    size_t      id__ = omp_get_thread_num();
+#endif
+    size_t inlen__ = inlen;
+    const uint8_t *in__ = ( const uint8_t * )in;
+    in__ += id__ * BLAKE2B_BLOCKBYTES;
+
+    while( inlen__ >= PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES )
+    {
+      blake2b_update( S[id__], in__, BLAKE2B_BLOCKBYTES );
+      in__ += PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES;
+      inlen__ -= PARALLELISM_DEGREE * BLAKE2B_BLOCKBYTES;
+    }
+
+    if( inlen__ > id__ * BLAKE2B_BLOCKBYTES )
+    {
+      const size_t left = inlen__ - id__ * BLAKE2B_BLOCKBYTES;
+      const size_t len = left <= BLAKE2B_BLOCKBYTES ? left : BLAKE2B_BLOCKBYTES;
+      blake2b_update( S[id__], in__, len );
+    }
+
+    blake2b_final( S[id__], hash[id__], BLAKE2B_OUTBYTES );
+  }
+
+  if( blake2bp_init_root( FS, outlen, keylen ) < 0 )
+    return -1;
+
+  FS->last_node = 1; // Mark as last node
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    blake2b_update( FS, hash[i], BLAKE2B_OUTBYTES );
+
+  return blake2b_final( FS, out, outlen );
+}
+
+
+

src/blake2s-load-sse2.h

@@ -0,0 +1,59 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2S_LOAD_SSE2_H__
+#define __BLAKE2S_LOAD_SSE2_H__
+
+#define LOAD_MSG_0_1(buf) buf = _mm_set_epi32(m6,m4,m2,m0)
+#define LOAD_MSG_0_2(buf) buf = _mm_set_epi32(m7,m5,m3,m1)
+#define LOAD_MSG_0_3(buf) buf = _mm_set_epi32(m14,m12,m10,m8)
+#define LOAD_MSG_0_4(buf) buf = _mm_set_epi32(m15,m13,m11,m9)
+#define LOAD_MSG_1_1(buf) buf = _mm_set_epi32(m13,m9,m4,m14)
+#define LOAD_MSG_1_2(buf) buf = _mm_set_epi32(m6,m15,m8,m10)
+#define LOAD_MSG_1_3(buf) buf = _mm_set_epi32(m5,m11,m0,m1)
+#define LOAD_MSG_1_4(buf) buf = _mm_set_epi32(m3,m7,m2,m12)
+#define LOAD_MSG_2_1(buf) buf = _mm_set_epi32(m15,m5,m12,m11)
+#define LOAD_MSG_2_2(buf) buf = _mm_set_epi32(m13,m2,m0,m8)
+#define LOAD_MSG_2_3(buf) buf = _mm_set_epi32(m9,m7,m3,m10)
+#define LOAD_MSG_2_4(buf) buf = _mm_set_epi32(m4,m1,m6,m14)
+#define LOAD_MSG_3_1(buf) buf = _mm_set_epi32(m11,m13,m3,m7)
+#define LOAD_MSG_3_2(buf) buf = _mm_set_epi32(m14,m12,m1,m9)
+#define LOAD_MSG_3_3(buf) buf = _mm_set_epi32(m15,m4,m5,m2)
+#define LOAD_MSG_3_4(buf) buf = _mm_set_epi32(m8,m0,m10,m6)
+#define LOAD_MSG_4_1(buf) buf = _mm_set_epi32(m10,m2,m5,m9)
+#define LOAD_MSG_4_2(buf) buf = _mm_set_epi32(m15,m4,m7,m0)
+#define LOAD_MSG_4_3(buf) buf = _mm_set_epi32(m3,m6,m11,m14)
+#define LOAD_MSG_4_4(buf) buf = _mm_set_epi32(m13,m8,m12,m1)
+#define LOAD_MSG_5_1(buf) buf = _mm_set_epi32(m8,m0,m6,m2)
+#define LOAD_MSG_5_2(buf) buf = _mm_set_epi32(m3,m11,m10,m12)
+#define LOAD_MSG_5_3(buf) buf = _mm_set_epi32(m1,m15,m7,m4)
+#define LOAD_MSG_5_4(buf) buf = _mm_set_epi32(m9,m14,m5,m13)
+#define LOAD_MSG_6_1(buf) buf = _mm_set_epi32(m4,m14,m1,m12)
+#define LOAD_MSG_6_2(buf) buf = _mm_set_epi32(m10,m13,m15,m5)
+#define LOAD_MSG_6_3(buf) buf = _mm_set_epi32(m8,m9,m6,m0)
+#define LOAD_MSG_6_4(buf) buf = _mm_set_epi32(m11,m2,m3,m7)
+#define LOAD_MSG_7_1(buf) buf = _mm_set_epi32(m3,m12,m7,m13)
+#define LOAD_MSG_7_2(buf) buf = _mm_set_epi32(m9,m1,m14,m11)
+#define LOAD_MSG_7_3(buf) buf = _mm_set_epi32(m2,m8,m15,m5)
+#define LOAD_MSG_7_4(buf) buf = _mm_set_epi32(m10,m6,m4,m0)
+#define LOAD_MSG_8_1(buf) buf = _mm_set_epi32(m0,m11,m14,m6)
+#define LOAD_MSG_8_2(buf) buf = _mm_set_epi32(m8,m3,m9,m15)
+#define LOAD_MSG_8_3(buf) buf = _mm_set_epi32(m10,m1,m13,m12)
+#define LOAD_MSG_8_4(buf) buf = _mm_set_epi32(m5,m4,m7,m2)
+#define LOAD_MSG_9_1(buf) buf = _mm_set_epi32(m1,m7,m8,m10)
+#define LOAD_MSG_9_2(buf) buf = _mm_set_epi32(m5,m6,m4,m2)
+#define LOAD_MSG_9_3(buf) buf = _mm_set_epi32(m13,m3,m9,m15)
+#define LOAD_MSG_9_4(buf) buf = _mm_set_epi32(m0,m12,m14,m11)
+
+
+#endif

src/blake2s-load-sse41.h

@@ -0,0 +1,229 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2S_LOAD_SSE41_H__
+#define __BLAKE2S_LOAD_SSE41_H__
+
+#define LOAD_MSG_0_1(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m0), TOF(m1), _MM_SHUFFLE(2,0,2,0)));
+
+#define LOAD_MSG_0_2(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m0), TOF(m1), _MM_SHUFFLE(3,1,3,1)));
+
+#define LOAD_MSG_0_3(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m2), TOF(m3), _MM_SHUFFLE(2,0,2,0)));
+
+#define LOAD_MSG_0_4(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m2), TOF(m3), _MM_SHUFFLE(3,1,3,1)));
+
+#define LOAD_MSG_1_1(buf) \
+t0 = _mm_blend_epi16(m1, m2, 0x0C); \
+t1 = _mm_slli_si128(m3, 4); \
+t2 = _mm_blend_epi16(t0, t1, 0xF0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,1,0,3));
+
+#define LOAD_MSG_1_2(buf) \
+t0 = _mm_shuffle_epi32(m2,_MM_SHUFFLE(0,0,2,0)); \
+t1 = _mm_blend_epi16(m1,m3,0xC0); \
+t2 = _mm_blend_epi16(t0, t1, 0xF0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
+
+#define LOAD_MSG_1_3(buf) \
+t0 = _mm_slli_si128(m1, 4); \
+t1 = _mm_blend_epi16(m2, t0, 0x30); \
+t2 = _mm_blend_epi16(m0, t1, 0xF0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
+
+#define LOAD_MSG_1_4(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_slli_si128(m3, 4); \
+t2 = _mm_blend_epi16(t0, t1, 0x0C); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
+
+#define LOAD_MSG_2_1(buf) \
+t0 = _mm_unpackhi_epi32(m2,m3); \
+t1 = _mm_blend_epi16(m3,m1,0x0C); \
+t2 = _mm_blend_epi16(t0, t1, 0x0F); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(3,1,0,2));
+
+#define LOAD_MSG_2_2(buf) \
+t0 = _mm_unpacklo_epi32(m2,m0); \
+t1 = _mm_blend_epi16(t0, m0, 0xF0); \
+t2 = _mm_slli_si128(m3, 8); \
+buf = _mm_blend_epi16(t1, t2, 0xC0);
+
+#define LOAD_MSG_2_3(buf) \
+t0 = _mm_blend_epi16(m0, m2, 0x3C); \
+t1 = _mm_srli_si128(m1, 12); \
+t2 = _mm_blend_epi16(t0,t1,0x03); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,0,3,2));
+
+#define LOAD_MSG_2_4(buf) \
+t0 = _mm_slli_si128(m3, 4); \
+t1 = _mm_blend_epi16(m0, m1, 0x33); \
+t2 = _mm_blend_epi16(t1, t0, 0xC0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(0,1,2,3));
+
+#define LOAD_MSG_3_1(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_unpackhi_epi32(t0, m2); \
+t2 = _mm_blend_epi16(t1, m3, 0x0C); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(3,1,0,2));
+
+#define LOAD_MSG_3_2(buf) \
+t0 = _mm_slli_si128(m2, 8); \
+t1 = _mm_blend_epi16(m3,m0,0x0C); \
+t2 = _mm_blend_epi16(t1, t0, 0xC0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,0,1,3));
+
+#define LOAD_MSG_3_3(buf) \
+t0 = _mm_blend_epi16(m0,m1,0x0F); \
+t1 = _mm_blend_epi16(t0, m3, 0xC0); \
+buf = _mm_shuffle_epi32(t1, _MM_SHUFFLE(3,0,1,2));
+
+#define LOAD_MSG_3_4(buf) \
+t0 = _mm_unpacklo_epi32(m0,m2); \
+t1 = _mm_unpackhi_epi32(m1,m2); \
+buf = _mm_unpacklo_epi64(t1,t0);
+
+#define LOAD_MSG_4_1(buf) \
+t0 = _mm_unpacklo_epi64(m1,m2); \
+t1 = _mm_unpackhi_epi64(m0,m2); \
+t2 = _mm_blend_epi16(t0,t1,0x33); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,0,1,3));
+
+#define LOAD_MSG_4_2(buf) \
+t0 = _mm_unpackhi_epi64(m1,m3); \
+t1 = _mm_unpacklo_epi64(m0,m1); \
+buf = _mm_blend_epi16(t0,t1,0x33);
+
+#define LOAD_MSG_4_3(buf) \
+t0 = _mm_unpackhi_epi64(m3,m1); \
+t1 = _mm_unpackhi_epi64(m2,m0); \
+buf = _mm_blend_epi16(t1,t0,0x33);
+
+#define LOAD_MSG_4_4(buf) \
+t0 = _mm_blend_epi16(m0,m2,0x03); \
+t1 = _mm_slli_si128(t0, 8); \
+t2 = _mm_blend_epi16(t1,m3,0x0F); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,2,0,3));
+
+#define LOAD_MSG_5_1(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_unpacklo_epi32(m0,m2); \
+buf = _mm_unpacklo_epi64(t0,t1);
+
+#define LOAD_MSG_5_2(buf) \
+t0 = _mm_srli_si128(m2, 4); \
+t1 = _mm_blend_epi16(m0,m3,0x03); \
+buf = _mm_blend_epi16(t1,t0,0x3C);
+
+#define LOAD_MSG_5_3(buf) \
+t0 = _mm_blend_epi16(m1,m0,0x0C); \
+t1 = _mm_srli_si128(m3, 4); \
+t2 = _mm_blend_epi16(t0,t1,0x30); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,2,3,0));
+
+#define LOAD_MSG_5_4(buf) \
+t0 = _mm_unpacklo_epi64(m1,m2); \
+t1= _mm_shuffle_epi32(m3, _MM_SHUFFLE(0,2,0,1)); \
+buf = _mm_blend_epi16(t0,t1,0x33);
+
+#define LOAD_MSG_6_1(buf) \
+t0 = _mm_slli_si128(m1, 12); \
+t1 = _mm_blend_epi16(m0,m3,0x33); \
+buf = _mm_blend_epi16(t1,t0,0xC0);
+
+#define LOAD_MSG_6_2(buf) \
+t0 = _mm_blend_epi16(m3,m2,0x30); \
+t1 = _mm_srli_si128(m1, 4); \
+t2 = _mm_blend_epi16(t0,t1,0x03); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,1,3,0));
+
+#define LOAD_MSG_6_3(buf) \
+t0 = _mm_unpacklo_epi64(m0,m2); \
+t1 = _mm_srli_si128(m1, 4); \
+buf = _mm_shuffle_epi32(_mm_blend_epi16(t0,t1,0x0C), _MM_SHUFFLE(2,3,1,0));
+
+#define LOAD_MSG_6_4(buf) \
+t0 = _mm_unpackhi_epi32(m1,m2); \
+t1 = _mm_unpackhi_epi64(m0,t0); \
+buf = _mm_shuffle_epi32(t1, _MM_SHUFFLE(3,0,1,2));
+
+#define LOAD_MSG_7_1(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_blend_epi16(t0,m3,0x0F); \
+buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(2,0,3,1));
+
+#define LOAD_MSG_7_2(buf) \
+t0 = _mm_blend_epi16(m2,m3,0x30); \
+t1 = _mm_srli_si128(m0,4); \
+t2 = _mm_blend_epi16(t0,t1,0x03); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,0,2,3));
+
+#define LOAD_MSG_7_3(buf) \
+t0 = _mm_unpackhi_epi64(m0,m3); \
+t1 = _mm_unpacklo_epi64(m1,m2); \
+t2 = _mm_blend_epi16(t0,t1,0x3C); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(0,2,3,1));
+
+#define LOAD_MSG_7_4(buf) \
+t0 = _mm_unpacklo_epi32(m0,m1); \
+t1 = _mm_unpackhi_epi32(m1,m2); \
+buf = _mm_unpacklo_epi64(t0,t1);
+
+#define LOAD_MSG_8_1(buf) \
+t0 = _mm_unpackhi_epi32(m1,m3); \
+t1 = _mm_unpacklo_epi64(t0,m0); \
+t2 = _mm_blend_epi16(t1,m2,0xC0); \
+buf = _mm_shufflehi_epi16(t2,_MM_SHUFFLE(1,0,3,2));
+
+#define LOAD_MSG_8_2(buf) \
+t0 = _mm_unpackhi_epi32(m0,m3); \
+t1 = _mm_blend_epi16(m2,t0,0xF0); \
+buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(0,2,1,3));
+
+#define LOAD_MSG_8_3(buf) \
+t0 = _mm_blend_epi16(m2,m0,0x0C); \
+t1 = _mm_slli_si128(t0,4); \
+buf = _mm_blend_epi16(t1,m3,0x0F);
+
+#define LOAD_MSG_8_4(buf) \
+t0 = _mm_blend_epi16(m1,m0,0x30); \
+buf = _mm_shuffle_epi32(t0,_MM_SHUFFLE(1,0,3,2));
+
+#define LOAD_MSG_9_1(buf) \
+t0 = _mm_blend_epi16(m0,m2,0x03); \
+t1 = _mm_blend_epi16(m1,m2,0x30); \
+t2 = _mm_blend_epi16(t1,t0,0x0F); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(1,3,0,2));
+
+#define LOAD_MSG_9_2(buf) \
+t0 = _mm_slli_si128(m0,4); \
+t1 = _mm_blend_epi16(m1,t0,0xC0); \
+buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(1,2,0,3));
+
+#define LOAD_MSG_9_3(buf) \
+t0 = _mm_unpackhi_epi32(m0,m3); \
+t1 = _mm_unpacklo_epi32(m2,m3); \
+t2 = _mm_unpackhi_epi64(t0,t1); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(3,0,2,1));
+
+#define LOAD_MSG_9_4(buf) \
+t0 = _mm_blend_epi16(m3,m2,0xC0); \
+t1 = _mm_unpacklo_epi32(m0,m3); \
+t2 = _mm_blend_epi16(t0,t1,0x0F); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(0,1,2,3));
+
+#endif
+

src/blake2s-load-xop.h

@@ -0,0 +1,189 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2S_LOAD_XOP_H__
+#define __BLAKE2S_LOAD_XOP_H__
+
+#define TOB(x) ((x)*4*0x01010101 + 0x03020100) // ..or not TOB
+
+/* Basic VPPERM emulation, for testing purposes */
+/*static __m128i _mm_perm_epi8(const __m128i src1, const __m128i src2, const __m128i sel)
+{
+   const __m128i sixteen = _mm_set1_epi8(16);
+   const __m128i t0 = _mm_shuffle_epi8(src1, sel);
+   const __m128i s1 = _mm_shuffle_epi8(src2, _mm_sub_epi8(sel, sixteen));
+   const __m128i mask = _mm_or_si128(_mm_cmpeq_epi8(sel, sixteen),
+                                     _mm_cmpgt_epi8(sel, sixteen)); // (>=16) = 0xff : 00
+   return _mm_blendv_epi8(t0, s1, mask);
+}*/
+
+#define LOAD_MSG_0_1(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(6),TOB(4),TOB(2),TOB(0)) );
+
+#define LOAD_MSG_0_2(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(7),TOB(5),TOB(3),TOB(1)) );
+
+#define LOAD_MSG_0_3(buf) \
+buf = _mm_perm_epi8(m2, m3, _mm_set_epi32(TOB(6),TOB(4),TOB(2),TOB(0)) );
+
+#define LOAD_MSG_0_4(buf) \
+buf = _mm_perm_epi8(m2, m3, _mm_set_epi32(TOB(7),TOB(5),TOB(3),TOB(1)) );
+
+#define LOAD_MSG_1_1(buf) \
+t0 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(0),TOB(5),TOB(0),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_1_2(buf) \
+t1 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(2),TOB(0),TOB(4),TOB(6)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_1_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(5),TOB(0),TOB(0),TOB(1)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_1_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(3),TOB(7),TOB(2),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(4)) );
+
+#define LOAD_MSG_2_1(buf) \
+t0 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(0),TOB(1),TOB(0),TOB(7)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(7),TOB(2),TOB(4),TOB(0)) );
+
+#define LOAD_MSG_2_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(2),TOB(0),TOB(4)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_2_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(7),TOB(3),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_2_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(4),TOB(1),TOB(6),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_3_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(3),TOB(7)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(5),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_3_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(0),TOB(1),TOB(5)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(6),TOB(4),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_3_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(4),TOB(5),TOB(2)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_3_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(0),TOB(6)) ); \
+buf = _mm_perm_epi8(t1, m2, _mm_set_epi32(TOB(4),TOB(2),TOB(6),TOB(0)) );
+
+#define LOAD_MSG_4_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(2),TOB(5),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(6),TOB(2),TOB(1),TOB(5)) );
+
+#define LOAD_MSG_4_2(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(4),TOB(7),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_4_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(3),TOB(6),TOB(0),TOB(0)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(2),TOB(7),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_4_4(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(4),TOB(0),TOB(1)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(4),TOB(0)) );
+
+#define LOAD_MSG_5_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(6),TOB(2)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(4),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_5_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(3),TOB(7),TOB(6),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(4)) );
+
+#define LOAD_MSG_5_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(1),TOB(0),TOB(7),TOB(4)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_5_4(buf) \
+t1 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(5),TOB(0),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(6),TOB(1),TOB(5)) );
+
+#define LOAD_MSG_6_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(4),TOB(0),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(6),TOB(1),TOB(4)) );
+
+#define LOAD_MSG_6_2(buf) \
+t1 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(6),TOB(0),TOB(0),TOB(1)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(5),TOB(7),TOB(0)) );
+
+#define LOAD_MSG_6_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(6),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(4),TOB(5),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_6_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(2),TOB(3),TOB(7)) ); \
+buf = _mm_perm_epi8(t1, m2, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_7_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(3),TOB(0),TOB(7),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(4),TOB(1),TOB(5)) );
+
+#define LOAD_MSG_7_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(5),TOB(1),TOB(0),TOB(7)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(6),TOB(0)) );
+
+#define LOAD_MSG_7_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(2),TOB(0),TOB(0),TOB(5)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(4),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(7),TOB(0)) );
+
+#define LOAD_MSG_7_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(6),TOB(4),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m2, _mm_set_epi32(TOB(6),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_8_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(0),TOB(6)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(6),TOB(0)) );
+
+#define LOAD_MSG_8_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(4),TOB(3),TOB(5),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(7)) );
+
+#define LOAD_MSG_8_3(buf) \
+t0 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(6),TOB(1),TOB(0),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(5),TOB(4)) ); \
+ 
+#define LOAD_MSG_8_4(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(5),TOB(4),TOB(7),TOB(2)) );
+
+#define LOAD_MSG_9_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(1),TOB(7),TOB(0),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(2),TOB(4),TOB(6)) );
+
+#define LOAD_MSG_9_2(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(5),TOB(6),TOB(4),TOB(2)) );
+
+#define LOAD_MSG_9_3(buf) \
+t0 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(3),TOB(5),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(7)) );
+
+#define LOAD_MSG_9_4(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(0),TOB(0),TOB(7)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(4),TOB(6),TOB(0)) );
+
+#endif
+

src/blake2s-ref.c

@@ -0,0 +1,370 @@
+/*
+   BLAKE2 reference source code package - reference C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+
+#include <stdint.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "blake2.h"
+#include "blake2-impl.h"
+
+static const uint32_t blake2s_IV[8] =
+{
+  0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
+  0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
+};
+
+static const uint8_t blake2s_sigma[10][16] =
+{
+  {  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 } ,
+  { 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 } ,
+  { 11,  8, 12,  0,  5,  2, 15, 13, 10, 14,  3,  6,  7,  1,  9,  4 } ,
+  {  7,  9,  3,  1, 13, 12, 11, 14,  2,  6,  5, 10,  4,  0, 15,  8 } ,
+  {  9,  0,  5,  7,  2,  4, 10, 15, 14,  1, 11, 12,  6,  8,  3, 13 } ,
+  {  2, 12,  6, 10,  0, 11,  8,  3,  4, 13,  7,  5, 15, 14,  1,  9 } ,
+  { 12,  5,  1, 15, 14, 13,  4, 10,  0,  7,  6,  3,  9,  2,  8, 11 } ,
+  { 13, 11,  7, 14, 12,  1,  3,  9,  5,  0, 15,  4,  8,  6,  2, 10 } ,
+  {  6, 15, 14,  9, 11,  3,  0,  8, 12,  2, 13,  7,  1,  4, 10,  5 } ,
+  { 10,  2,  8,  4,  7,  6,  1,  5, 15, 11,  9, 14,  3, 12, 13 , 0 } ,
+};
+
+static inline int blake2s_set_lastnode( blake2s_state *S )
+{
+  S->f[1] = ~0U;
+  return 0;
+}
+
+static inline int blake2s_clear_lastnode( blake2s_state *S )
+{
+  S->f[1] = 0U;
+  return 0;
+}
+
+/* Some helper functions, not necessarily useful */
+static inline int blake2s_set_lastblock( blake2s_state *S )
+{
+  if( S->last_node ) blake2s_set_lastnode( S );
+
+  S->f[0] = ~0U;
+  return 0;
+}
+
+static inline int blake2s_clear_lastblock( blake2s_state *S )
+{
+  if( S->last_node ) blake2s_clear_lastnode( S );
+
+  S->f[0] = 0U;
+  return 0;
+}
+
+static inline int blake2s_increment_counter( blake2s_state *S, const uint32_t inc )
+{
+  S->t[0] += inc;
+  S->t[1] += ( S->t[0] < inc );
+  return 0;
+}
+
+// Parameter-related functions
+static inline int blake2s_param_set_digest_length( blake2s_param *P, const uint8_t digest_length )
+{
+  P->digest_length = digest_length;
+  return 0;
+}
+
+static inline int blake2s_param_set_fanout( blake2s_param *P, const uint8_t fanout )
+{
+  P->fanout = fanout;
+  return 0;
+}
+
+static inline int blake2s_param_set_max_depth( blake2s_param *P, const uint8_t depth )
+{
+  P->depth = depth;
+  return 0;
+}
+
+static inline int blake2s_param_set_leaf_length( blake2s_param *P, const uint32_t leaf_length )
+{
+  store32( &P->leaf_length, leaf_length );
+  return 0;
+}
+
+static inline int blake2s_param_set_node_offset( blake2s_param *P, const uint64_t node_offset )
+{
+  store48( P->node_offset, node_offset );
+  return 0;
+}
+
+static inline int blake2s_param_set_node_depth( blake2s_param *P, const uint8_t node_depth )
+{
+  P->node_depth = node_depth;
+  return 0;
+}
+
+static inline int blake2s_param_set_inner_length( blake2s_param *P, const uint8_t inner_length )
+{
+  P->inner_length = inner_length;
+  return 0;
+}
+
+static inline int blake2s_param_set_salt( blake2s_param *P, const uint8_t salt[BLAKE2S_SALTBYTES] )
+{
+  memcpy( P->salt, salt, BLAKE2S_SALTBYTES );
+  return 0;
+}
+
+static inline int blake2s_param_set_personal( blake2s_param *P, const uint8_t personal[BLAKE2S_PERSONALBYTES] )
+{
+  memcpy( P->personal, personal, BLAKE2S_PERSONALBYTES );
+  return 0;
+}
+
+static inline int blake2s_init0( blake2s_state *S )
+{
+  memset( S, 0, sizeof( blake2s_state ) );
+
+  for( int i = 0; i < 8; ++i ) S->h[i] = blake2s_IV[i];
+
+  return 0;
+}
+
+#define blake2s_init BLAKE2_IMPL_NAME(blake2s_init)
+#define blake2s_init_param BLAKE2_IMPL_NAME(blake2s_init_param)
+#define blake2s_init_key BLAKE2_IMPL_NAME(blake2s_init_key)
+#define blake2s_update BLAKE2_IMPL_NAME(blake2s_update)
+#define blake2s_final BLAKE2_IMPL_NAME(blake2s_final)
+#define blake2s BLAKE2_IMPL_NAME(blake2s)
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+  int blake2s_init( blake2s_state *S, size_t outlen );
+  int blake2s_init_param( blake2s_state *S, const blake2s_param *P );
+  int blake2s_init_key( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_update( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+#if defined(__cplusplus)
+}
+#endif
+
+/* init2 xors IV with input parameter block */
+int blake2s_init_param( blake2s_state *S, const blake2s_param *P )
+{
+  blake2s_init0( S );
+  uint32_t *p = ( uint32_t * )( P );
+
+  /* IV XOR ParamBlock */
+  for( size_t i = 0; i < 8; ++i )
+    S->h[i] ^= load32( &p[i] );
+
+  S->outlen = P->digest_length;
+  return 0;
+}
+
+
+// Sequential blake2s initialization
+int blake2s_init( blake2s_state *S, size_t outlen )
+{
+  blake2s_param P[1];
+
+  /* Move interval verification here? */
+  if ( ( !outlen ) || ( outlen > BLAKE2S_OUTBYTES ) ) return -1;
+
+  P->digest_length = outlen;
+  P->key_length    = 0;
+  P->fanout        = 1;
+  P->depth         = 1;
+  store32( &P->leaf_length, 0 );
+  store48( &P->node_offset, 0 );
+  P->node_depth    = 0;
+  P->inner_length  = 0;
+  // memset(P->reserved, 0, sizeof(P->reserved) );
+  memset( P->salt,     0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+  return blake2s_init_param( S, P );
+}
+
+int blake2s_init_key( blake2s_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  blake2s_param P[1];
+
+  if ( ( !outlen ) || ( outlen > BLAKE2S_OUTBYTES ) ) return -1;
+
+  if ( !key || !keylen || keylen > BLAKE2S_KEYBYTES ) return -1;
+
+  P->digest_length = outlen;
+  P->key_length    = keylen;
+  P->fanout        = 1;
+  P->depth         = 1;
+  store32( &P->leaf_length, 0 );
+  store48( &P->node_offset, 0 );
+  P->node_depth    = 0;
+  P->inner_length  = 0;
+  // memset(P->reserved, 0, sizeof(P->reserved) );
+  memset( P->salt,     0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+
+  if( blake2s_init_param( S, P ) < 0 ) return -1;
+
+  {
+    uint8_t block[BLAKE2S_BLOCKBYTES];
+    memset( block, 0, BLAKE2S_BLOCKBYTES );
+    memcpy( block, key, keylen );
+    blake2s_update( S, block, BLAKE2S_BLOCKBYTES );
+    secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
+  }
+  return 0;
+}
+
+static int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES] )
+{
+  uint32_t m[16];
+  uint32_t v[16];
+
+  for( size_t i = 0; i < 16; ++i )
+    m[i] = load32( block + i * sizeof( m[i] ) );
+
+  for( size_t i = 0; i < 8; ++i )
+    v[i] = S->h[i];
+
+  v[ 8] = blake2s_IV[0];
+  v[ 9] = blake2s_IV[1];
+  v[10] = blake2s_IV[2];
+  v[11] = blake2s_IV[3];
+  v[12] = S->t[0] ^ blake2s_IV[4];
+  v[13] = S->t[1] ^ blake2s_IV[5];
+  v[14] = S->f[0] ^ blake2s_IV[6];
+  v[15] = S->f[1] ^ blake2s_IV[7];
+#define G(r,i,a,b,c,d) \
+  do { \
+    a = a + b + m[blake2s_sigma[r][2*i+0]]; \
+    d = rotr32(d ^ a, 16); \
+    c = c + d; \
+    b = rotr32(b ^ c, 12); \
+    a = a + b + m[blake2s_sigma[r][2*i+1]]; \
+    d = rotr32(d ^ a, 8); \
+    c = c + d; \
+    b = rotr32(b ^ c, 7); \
+  } while(0)
+#define ROUND(r)  \
+  do { \
+    G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
+    G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
+    G(r,2,v[ 2],v[ 6],v[10],v[14]); \
+    G(r,3,v[ 3],v[ 7],v[11],v[15]); \
+    G(r,4,v[ 0],v[ 5],v[10],v[15]); \
+    G(r,5,v[ 1],v[ 6],v[11],v[12]); \
+    G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
+    G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
+  } while(0)
+  ROUND( 0 );
+  ROUND( 1 );
+  ROUND( 2 );
+  ROUND( 3 );
+  ROUND( 4 );
+  ROUND( 5 );
+  ROUND( 6 );
+  ROUND( 7 );
+  ROUND( 8 );
+  ROUND( 9 );
+
+  for( size_t i = 0; i < 8; ++i )
+    S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
+
+#undef G
+#undef ROUND
+  return 0;
+}
+
+
+int blake2s_update( blake2s_state *S, const uint8_t *in, size_t inlen )
+{
+  while( inlen > 0 )
+  {
+    size_t left = S->buflen;
+    size_t fill = 2 * BLAKE2S_BLOCKBYTES - left;
+
+    if( inlen > fill )
+    {
+      memcpy( S->buf + left, in, fill ); // Fill buffer
+      S->buflen += fill;
+      blake2s_increment_counter( S, BLAKE2S_BLOCKBYTES );
+      blake2s_compress( S, S->buf ); // Compress
+      memcpy( S->buf, S->buf + BLAKE2S_BLOCKBYTES, BLAKE2S_BLOCKBYTES ); // Shift buffer left
+      S->buflen -= BLAKE2S_BLOCKBYTES;
+      in += fill;
+      inlen -= fill;
+    }
+    else // inlen <= fill
+    {
+      memcpy( S->buf + left, in, inlen );
+      S->buflen += inlen; // Be lazy, do not compress
+      in += inlen;
+      inlen -= inlen;
+    }
+  }
+
+  return 0;
+}
+
+int blake2s_final( blake2s_state *S, uint8_t *out, size_t outlen )
+{
+  uint8_t buffer[BLAKE2S_OUTBYTES];
+
+  if(S->outlen != outlen) return -1;
+
+  if( S->buflen > BLAKE2S_BLOCKBYTES )
+  {
+    blake2s_increment_counter( S, BLAKE2S_BLOCKBYTES );
+    blake2s_compress( S, S->buf );
+    S->buflen -= BLAKE2S_BLOCKBYTES;
+    memcpy( S->buf, S->buf + BLAKE2S_BLOCKBYTES, S->buflen );
+  }
+
+  blake2s_increment_counter( S, ( uint32_t )S->buflen );
+  blake2s_set_lastblock( S );
+  memset( S->buf + S->buflen, 0, 2 * BLAKE2S_BLOCKBYTES - S->buflen ); /* Padding */
+  blake2s_compress( S, S->buf );
+
+  for( int i = 0; i < 8; ++i ) /* Output full hash to temp buffer */
+    store32( buffer + sizeof( S->h[i] ) * i, S->h[i] );
+
+  memcpy( out, buffer, outlen );
+  return 0;
+}
+
+int blake2s( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  blake2s_state S[1];
+
+  /* Verify parameters */
+  if ( NULL == in ) return -1;
+
+  if ( NULL == out ) return -1;
+
+  if ( NULL == key && keylen > 0 ) return -1;
+
+  if( keylen > 0 )
+  {
+    if( blake2s_init_key( S, outlen, key, keylen ) < 0 ) return -1;
+  }
+  else
+  {
+    if( blake2s_init( S, outlen ) < 0 ) return -1;
+  }
+
+  if( blake2s_update( S, ( uint8_t * )in, inlen ) < 0) return -1;
+  return blake2s_final( S, out, outlen );
+}
+

src/blake2s-round.h

@@ -0,0 +1,91 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#pragma once
+#ifndef __BLAKE2S_ROUND_H__
+#define __BLAKE2S_ROUND_H__
+
+#define LOAD(p)  _mm_load_si128( (__m128i *)(p) )
+#define STORE(p,r) _mm_store_si128((__m128i *)(p), r)
+
+#define LOADU(p)  _mm_loadu_si128( (__m128i *)(p) )
+#define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r)
+
+#define TOF(reg) _mm_castsi128_ps((reg))
+#define TOI(reg) _mm_castps_si128((reg))
+
+#define LIKELY(x) __builtin_expect((x),1)
+
+
+/* Microarchitecture-specific macros */
+#ifndef HAVE_XOP
+#ifdef HAVE_SSSE3
+#define _mm_roti_epi32(r, c) ( \
+                (8==-(c)) ? _mm_shuffle_epi8(r,r8) \
+              : (16==-(c)) ? _mm_shuffle_epi8(r,r16) \
+              : _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32( (r), 32-(-(c)) )) )
+#else
+#define _mm_roti_epi32(r, c) _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32( (r), 32-(-(c)) ))
+#endif
+#else
+/* ... */
+#endif
+
+
+#define G1(row1,row2,row3,row4,buf) \
+  row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \
+  row4 = _mm_xor_si128( row4, row1 ); \
+  row4 = _mm_roti_epi32(row4, -16); \
+  row3 = _mm_add_epi32( row3, row4 );   \
+  row2 = _mm_xor_si128( row2, row3 ); \
+  row2 = _mm_roti_epi32(row2, -12);
+
+#define G2(row1,row2,row3,row4,buf) \
+  row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \
+  row4 = _mm_xor_si128( row4, row1 ); \
+  row4 = _mm_roti_epi32(row4, -8); \
+  row3 = _mm_add_epi32( row3, row4 );   \
+  row2 = _mm_xor_si128( row2, row3 ); \
+  row2 = _mm_roti_epi32(row2, -7);
+
+#define DIAGONALIZE(row1,row2,row3,row4) \
+  row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(2,1,0,3) ); \
+  row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \
+  row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(0,3,2,1) );
+
+#define UNDIAGONALIZE(row1,row2,row3,row4) \
+  row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(0,3,2,1) ); \
+  row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \
+  row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(2,1,0,3) );
+
+#if defined(HAVE_XOP)
+#include "blake2s-load-xop.h"
+#elif defined(HAVE_SSE4_1)
+#include "blake2s-load-sse41.h"
+#else
+#include "blake2s-load-sse2.h"
+#endif
+
+#define ROUND(r)  \
+  LOAD_MSG_ ##r ##_1(buf1); \
+  G1(row1,row2,row3,row4,buf1); \
+  LOAD_MSG_ ##r ##_2(buf2); \
+  G2(row1,row2,row3,row4,buf2); \
+  DIAGONALIZE(row1,row2,row3,row4); \
+  LOAD_MSG_ ##r ##_3(buf3); \
+  G1(row1,row2,row3,row4,buf3); \
+  LOAD_MSG_ ##r ##_4(buf4); \
+  G2(row1,row2,row3,row4,buf4); \
+  UNDIAGONALIZE(row1,row2,row3,row4); \
+ 
+#endif
+

src/blake2s-test.c

@@ -0,0 +1,43 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#include <stdio.h>
+#include <string.h>
+#include "blake2.h"
+#include "blake2-kat.h"
+int main( int argc, char **argv )
+{
+  uint8_t key[BLAKE2S_KEYBYTES];
+  uint8_t buf[KAT_LENGTH];
+
+  for( size_t i = 0; i < BLAKE2S_KEYBYTES; ++i )
+    key[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+    buf[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+  {
+    uint8_t hash[BLAKE2S_OUTBYTES];
+
+    if( blake2s( hash, buf, key, BLAKE2S_OUTBYTES, i, BLAKE2S_KEYBYTES ) < 0 ||
+        0 != memcmp( hash, blake2s_keyed_kat[i], BLAKE2S_OUTBYTES ) )
+    {
+      puts( "error" );
+      return -1;
+    }
+  }
+
+  puts( "ok" );
+  return 0;
+}
+

src/blake2s.c

@@ -0,0 +1,418 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+
+#include <stdint.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "blake2.h"
+#include "blake2-impl.h"
+
+#include "blake2-config.h"
+
+#if defined(_MSC_VER)
+#include <intrin.h>
+#endif
+
+#if defined(HAVE_SSE2)
+#include <emmintrin.h>
+// MSVC only defines  _mm_set_epi64x for x86_64...
+#if defined(_MSC_VER) && !defined(_M_X64)
+static inline __m128i _mm_set_epi64x( const uint64_t u1, const uint64_t u0 )
+{
+  return _mm_set_epi32( u1 >> 32, u1, u0 >> 32, u0 );
+}
+#endif
+#endif
+
+
+#if defined(HAVE_SSSE3)
+#include <tmmintrin.h>
+#endif
+#if defined(HAVE_SSE4_1)
+#include <smmintrin.h>
+#endif
+#if defined(HAVE_AVX)
+#include <immintrin.h>
+#endif
+#if defined(HAVE_XOP) && !defined(_MSC_VER)
+#include <x86intrin.h>
+#endif
+
+#include "blake2s-round.h"
+
+static const uint32_t blake2s_IV[8] =
+{
+  0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
+  0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
+};
+
+static const uint8_t blake2s_sigma[10][16] =
+{
+  {  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 } ,
+  { 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 } ,
+  { 11,  8, 12,  0,  5,  2, 15, 13, 10, 14,  3,  6,  7,  1,  9,  4 } ,
+  {  7,  9,  3,  1, 13, 12, 11, 14,  2,  6,  5, 10,  4,  0, 15,  8 } ,
+  {  9,  0,  5,  7,  2,  4, 10, 15, 14,  1, 11, 12,  6,  8,  3, 13 } ,
+  {  2, 12,  6, 10,  0, 11,  8,  3,  4, 13,  7,  5, 15, 14,  1,  9 } ,
+  { 12,  5,  1, 15, 14, 13,  4, 10,  0,  7,  6,  3,  9,  2,  8, 11 } ,
+  { 13, 11,  7, 14, 12,  1,  3,  9,  5,  0, 15,  4,  8,  6,  2, 10 } ,
+  {  6, 15, 14,  9, 11,  3,  0,  8, 12,  2, 13,  7,  1,  4, 10,  5 } ,
+  { 10,  2,  8,  4,  7,  6,  1,  5, 15, 11,  9, 14,  3, 12, 13 , 0 } ,
+};
+
+
+/* Some helper functions, not necessarily useful */
+static inline int blake2s_set_lastnode( blake2s_state *S )
+{
+  S->f[1] = ~0U;
+  return 0;
+}
+
+static inline int blake2s_clear_lastnode( blake2s_state *S )
+{
+  S->f[1] = 0U;
+  return 0;
+}
+
+static inline int blake2s_set_lastblock( blake2s_state *S )
+{
+  if( S->last_node ) blake2s_set_lastnode( S );
+
+  S->f[0] = ~0U;
+  return 0;
+}
+
+static inline int blake2s_clear_lastblock( blake2s_state *S )
+{
+  if( S->last_node ) blake2s_clear_lastnode( S );
+
+  S->f[0] = 0U;
+  return 0;
+}
+
+static inline int blake2s_increment_counter( blake2s_state *S, const uint32_t inc )
+{
+  uint64_t t = ( ( uint64_t )S->t[1] << 32 ) | S->t[0];
+  t += inc;
+  S->t[0] = ( uint32_t )( t >>  0 );
+  S->t[1] = ( uint32_t )( t >> 32 );
+  return 0;
+}
+
+
+// Parameter-related functions
+static inline int blake2s_param_set_digest_length( blake2s_param *P, const uint8_t digest_length )
+{
+  P->digest_length = digest_length;
+  return 0;
+}
+
+static inline int blake2s_param_set_fanout( blake2s_param *P, const uint8_t fanout )
+{
+  P->fanout = fanout;
+  return 0;
+}
+
+static inline int blake2s_param_set_max_depth( blake2s_param *P, const uint8_t depth )
+{
+  P->depth = depth;
+  return 0;
+}
+
+static inline int blake2s_param_set_leaf_length( blake2s_param *P, const uint32_t leaf_length )
+{
+  P->leaf_length = leaf_length;
+  return 0;
+}
+
+static inline int blake2s_param_set_node_offset( blake2s_param *P, const uint64_t node_offset )
+{
+  store48( P->node_offset, node_offset );
+  return 0;
+}
+
+static inline int blake2s_param_set_node_depth( blake2s_param *P, const uint8_t node_depth )
+{
+  P->node_depth = node_depth;
+  return 0;
+}
+
+static inline int blake2s_param_set_inner_length( blake2s_param *P, const uint8_t inner_length )
+{
+  P->inner_length = inner_length;
+  return 0;
+}
+
+static inline int blake2s_param_set_salt( blake2s_param *P, const uint8_t salt[BLAKE2S_SALTBYTES] )
+{
+  memcpy( P->salt, salt, BLAKE2S_SALTBYTES );
+  return 0;
+}
+
+static inline int blake2s_param_set_personal( blake2s_param *P, const uint8_t personal[BLAKE2S_PERSONALBYTES] )
+{
+  memcpy( P->personal, personal, BLAKE2S_PERSONALBYTES );
+  return 0;
+}
+
+static inline int blake2s_init0( blake2s_state *S )
+{
+  memset( S, 0, sizeof( blake2s_state ) );
+
+  for( int i = 0; i < 8; ++i ) S->h[i] = blake2s_IV[i];
+
+  return 0;
+}
+
+#define blake2s_init BLAKE2_IMPL_NAME(blake2s_init)
+#define blake2s_init_param BLAKE2_IMPL_NAME(blake2s_init_param)
+#define blake2s_init_key BLAKE2_IMPL_NAME(blake2s_init_key)
+#define blake2s_update BLAKE2_IMPL_NAME(blake2s_update)
+#define blake2s_final BLAKE2_IMPL_NAME(blake2s_final)
+#define blake2s BLAKE2_IMPL_NAME(blake2s)
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+  int blake2s_init( blake2s_state *S, size_t outlen );
+  int blake2s_init_param( blake2s_state *S, const blake2s_param *P );
+  int blake2s_init_key( blake2s_state *S, size_t outlen, const void *key, size_t keylen );
+  int blake2s_update( blake2s_state *S, const uint8_t *in, size_t inlen );
+  int blake2s_final( blake2s_state *S, uint8_t *out, size_t outlen );
+  int blake2s( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen );
+#if defined(__cplusplus)
+}
+#endif
+
+
+/* init2 xors IV with input parameter block */
+int blake2s_init_param( blake2s_state *S, const blake2s_param *P )
+{
+  uint8_t *p, *h, *v;
+  //blake2s_init0( S );
+  v = ( uint8_t * )( blake2s_IV );
+  h = ( uint8_t * )( S->h );
+  p = ( uint8_t * )( P );
+  /* IV XOR ParamBlock */
+  memset( S, 0, sizeof( blake2s_state ) );
+
+  for( int i = 0; i < BLAKE2S_OUTBYTES; ++i ) h[i] = v[i] ^ p[i];
+
+  S->outlen = P->digest_length;
+  return 0;
+}
+
+
+/* Some sort of default parameter block initialization, for sequential blake2s */
+int blake2s_init( blake2s_state *S, size_t outlen )
+{
+  if ( ( !outlen ) || ( outlen > BLAKE2S_OUTBYTES ) ) return -1;
+
+  const blake2s_param P =
+  {
+    outlen,
+    0,
+    1,
+    1,
+    0,
+    {0},
+    0,
+    0,
+    {0},
+    {0}
+  };
+  return blake2s_init_param( S, &P );
+}
+
+
+int blake2s_init_key( blake2s_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  if ( ( !outlen ) || ( outlen > BLAKE2S_OUTBYTES ) ) return -1;
+
+  if ( ( !key ) || ( !keylen ) || keylen > BLAKE2S_KEYBYTES ) return -1;
+
+  const blake2s_param P =
+  {
+    outlen,
+    keylen,
+    1,
+    1,
+    0,
+    {0},
+    0,
+    0,
+    {0},
+    {0}
+  };
+
+  if( blake2s_init_param( S, &P ) < 0 )
+    return -1;
+
+  {
+    uint8_t block[BLAKE2S_BLOCKBYTES];
+    memset( block, 0, BLAKE2S_BLOCKBYTES );
+    memcpy( block, key, keylen );
+    blake2s_update( S, block, BLAKE2S_BLOCKBYTES );
+    secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
+  }
+  return 0;
+}
+
+
+static inline int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES] )
+{
+  __m128i row1, row2, row3, row4;
+  __m128i buf1, buf2, buf3, buf4;
+#if defined(HAVE_SSE4_1)
+  __m128i t0, t1;
+#if !defined(HAVE_XOP)
+  __m128i t2;
+#endif
+#endif
+  __m128i ff0, ff1;
+#if defined(HAVE_SSSE3) && !defined(HAVE_XOP)
+  const __m128i r8 = _mm_set_epi8( 12, 15, 14, 13, 8, 11, 10, 9, 4, 7, 6, 5, 0, 3, 2, 1 );
+  const __m128i r16 = _mm_set_epi8( 13, 12, 15, 14, 9, 8, 11, 10, 5, 4, 7, 6, 1, 0, 3, 2 );
+#endif
+#if defined(HAVE_SSE4_1)
+  const __m128i m0 = LOADU( block +  00 );
+  const __m128i m1 = LOADU( block +  16 );
+  const __m128i m2 = LOADU( block +  32 );
+  const __m128i m3 = LOADU( block +  48 );
+#else
+  const uint32_t  m0 = ( ( uint32_t * )block )[ 0];
+  const uint32_t  m1 = ( ( uint32_t * )block )[ 1];
+  const uint32_t  m2 = ( ( uint32_t * )block )[ 2];
+  const uint32_t  m3 = ( ( uint32_t * )block )[ 3];
+  const uint32_t  m4 = ( ( uint32_t * )block )[ 4];
+  const uint32_t  m5 = ( ( uint32_t * )block )[ 5];
+  const uint32_t  m6 = ( ( uint32_t * )block )[ 6];
+  const uint32_t  m7 = ( ( uint32_t * )block )[ 7];
+  const uint32_t  m8 = ( ( uint32_t * )block )[ 8];
+  const uint32_t  m9 = ( ( uint32_t * )block )[ 9];
+  const uint32_t m10 = ( ( uint32_t * )block )[10];
+  const uint32_t m11 = ( ( uint32_t * )block )[11];
+  const uint32_t m12 = ( ( uint32_t * )block )[12];
+  const uint32_t m13 = ( ( uint32_t * )block )[13];
+  const uint32_t m14 = ( ( uint32_t * )block )[14];
+  const uint32_t m15 = ( ( uint32_t * )block )[15];
+#endif
+  row1 = ff0 = LOADU( &S->h[0] );
+  row2 = ff1 = LOADU( &S->h[4] );
+  row3 = _mm_setr_epi32( 0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A );
+  row4 = _mm_xor_si128( _mm_setr_epi32( 0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19 ), LOADU( &S->t[0] ) );
+  ROUND( 0 );
+  ROUND( 1 );
+  ROUND( 2 );
+  ROUND( 3 );
+  ROUND( 4 );
+  ROUND( 5 );
+  ROUND( 6 );
+  ROUND( 7 );
+  ROUND( 8 );
+  ROUND( 9 );
+  STOREU( &S->h[0], _mm_xor_si128( ff0, _mm_xor_si128( row1, row3 ) ) );
+  STOREU( &S->h[4], _mm_xor_si128( ff1, _mm_xor_si128( row2, row4 ) ) );
+  return 0;
+}
+
+
+int blake2s_update( blake2s_state *S, const uint8_t *in, size_t inlen )
+{
+  while( inlen > 0 )
+  {
+    size_t left = S->buflen;
+    size_t fill = 2 * BLAKE2S_BLOCKBYTES - left;
+
+    if( inlen > fill )
+    {
+      memcpy( S->buf + left, in, fill ); // Fill buffer
+      S->buflen += fill;
+      blake2s_increment_counter( S, BLAKE2S_BLOCKBYTES );
+      blake2s_compress( S, S->buf ); // Compress
+      memcpy( S->buf, S->buf + BLAKE2S_BLOCKBYTES, BLAKE2S_BLOCKBYTES ); // Shift buffer left
+      S->buflen -= BLAKE2S_BLOCKBYTES;
+      in += fill;
+      inlen -= fill;
+    }
+    else /* inlen <= fill */
+    {
+      memcpy( S->buf + left, in, inlen );
+      S->buflen += inlen; // Be lazy, do not compress
+      in += inlen;
+      inlen -= inlen;
+    }
+  }
+
+  return 0;
+}
+
+
+int blake2s_final( blake2s_state *S, uint8_t *out, size_t outlen )
+{
+  uint8_t buffer[BLAKE2S_OUTBYTES];
+
+  if(outlen != S->outlen ) return -1;
+
+  if( S->buflen > BLAKE2S_BLOCKBYTES )
+  {
+    blake2s_increment_counter( S, BLAKE2S_BLOCKBYTES );
+    blake2s_compress( S, S->buf );
+    S->buflen -= BLAKE2S_BLOCKBYTES;
+    memcpy( S->buf, S->buf + BLAKE2S_BLOCKBYTES, S->buflen );
+  }
+
+  blake2s_increment_counter( S, ( uint32_t )S->buflen );
+  blake2s_set_lastblock( S );
+  memset( S->buf + S->buflen, 0, 2 * BLAKE2S_BLOCKBYTES - S->buflen ); /* Padding */
+  blake2s_compress( S, S->buf );
+
+  for( int i = 0; i < 8; ++i ) /* Output full hash to temp buffer */
+    store32( buffer + sizeof( S->h[i] ) * i, S->h[i] );
+
+  memcpy( out, buffer, outlen );
+  return 0;
+}
+
+int blake2s( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  blake2s_state S[1];
+
+  /* Verify parameters */
+  if ( NULL == in ) return -1;
+
+  if ( NULL == out ) return -1;
+
+  if ( NULL == key && keylen > 0) return -1;
+
+  if( keylen > 0 )
+  {
+    if( blake2s_init_key( S, outlen, key, keylen ) < 0 ) return -1;
+  }
+  else
+  {
+    if( blake2s_init( S, outlen ) < 0 ) return -1;
+  }
+
+  if( blake2s_update( S, ( uint8_t * )in, inlen ) < 0) return -1;
+  return blake2s_final( S, out, outlen );
+}
+
+#if defined(SUPERCOP)
+int crypto_hash( unsigned char *out, unsigned char *in, unsigned long long inlen )
+{
+  return blake2s( out, in, NULL, BLAKE2S_OUTBYTES, (size_t)inlen, 0 );
+}
+#endif
+

src/blake2sp-test.c

@@ -0,0 +1,43 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+#include <stdio.h>
+#include <string.h>
+#include "blake2.h"
+#include "blake2-kat.h"
+
+int main( int argc, char **argv )
+{
+  uint8_t key[BLAKE2S_KEYBYTES];
+  uint8_t buf[KAT_LENGTH];
+
+  for( size_t i = 0; i < BLAKE2S_KEYBYTES; ++i )
+    key[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+    buf[i] = ( uint8_t )i;
+
+  for( size_t i = 0; i < KAT_LENGTH; ++i )
+  {
+    uint8_t hash[BLAKE2S_OUTBYTES];
+    if( blake2sp( hash, buf, key, BLAKE2S_OUTBYTES, i, BLAKE2S_KEYBYTES ) < 0 ||
+        0 != memcmp( hash, blake2sp_keyed_kat[i], BLAKE2S_OUTBYTES ) )
+    {
+      puts( "error" );
+      return -1;
+    }
+  }
+
+  puts( "ok" );
+  return 0;
+}
+

src/blake2sp.c

@@ -0,0 +1,272 @@
+/*
+   BLAKE2 reference source code package - optimized C implementations
+
+   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
+
+   To the extent possible under law, the author(s) have dedicated all copyright
+   and related and neighboring rights to this software to the public domain
+   worldwide. This software is distributed without any warranty.
+
+   You should have received a copy of the CC0 Public Domain Dedication along with
+   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+*/
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+
+#if defined(_OPENMP)
+#include <omp.h>
+#endif
+
+#include "blake2.h"
+#include "blake2-impl.h"
+
+#define PARALLELISM_DEGREE 8
+
+static int blake2sp_init_leaf( blake2s_state *S, uint8_t outlen, uint8_t keylen, uint64_t offset )
+{
+  blake2s_param P[1];
+  P->digest_length = outlen;
+  P->key_length = keylen;
+  P->fanout = PARALLELISM_DEGREE;
+  P->depth = 2;
+  P->leaf_length = 0;
+  store48( P->node_offset, offset );
+  P->node_depth = 0;
+  P->inner_length = BLAKE2S_OUTBYTES;
+  memset( P->salt, 0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+  blake2s_init_param( S, P );
+  S->outlen = P->inner_length;
+  return 0;
+}
+
+static int blake2sp_init_root( blake2s_state *S, uint8_t outlen, uint8_t keylen )
+{
+  blake2s_param P[1];
+  P->digest_length = outlen;
+  P->key_length = keylen;
+  P->fanout = PARALLELISM_DEGREE;
+  P->depth = 2;
+  P->leaf_length = 0;
+  store48( P->node_offset, 0ULL );
+  P->node_depth = 1;
+  P->inner_length = BLAKE2S_OUTBYTES;
+  memset( P->salt, 0, sizeof( P->salt ) );
+  memset( P->personal, 0, sizeof( P->personal ) );
+  blake2s_init_param( S, P );
+  S->outlen = P->digest_length;
+  return 0;
+}
+
+
+int blake2sp_init( blake2sp_state *S, size_t outlen )
+{
+  if( !outlen || outlen > BLAKE2S_OUTBYTES ) return -1;
+
+  memset( S->buf, 0, sizeof( S->buf ) );
+  S->buflen = 0;
+
+  if( blake2sp_init_root( S->R, outlen, 0 ) < 0 )
+    return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    if( blake2sp_init_leaf( S->S[i], outlen, 0, i ) < 0 ) return -1;
+
+  S->R->last_node = 1;
+  S->S[PARALLELISM_DEGREE - 1]->last_node = 1;
+  S->outlen = outlen;
+  return 0;
+}
+
+int blake2sp_init_key( blake2sp_state *S, size_t outlen, const void *key, size_t keylen )
+{
+  if( !outlen || outlen > BLAKE2S_OUTBYTES ) return -1;
+
+  if( !key || !keylen || keylen > BLAKE2S_KEYBYTES ) return -1;
+
+  memset( S->buf, 0, sizeof( S->buf ) );
+  S->buflen = 0;
+
+  if( blake2sp_init_root( S->R, outlen, keylen ) < 0 )
+    return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    if( blake2sp_init_leaf( S->S[i], outlen, keylen, i ) < 0 ) return -1;
+
+  S->R->last_node = 1;
+  S->S[PARALLELISM_DEGREE - 1]->last_node = 1;
+  S->outlen = outlen;
+  {
+    uint8_t block[BLAKE2S_BLOCKBYTES];
+    memset( block, 0, BLAKE2S_BLOCKBYTES );
+    memcpy( block, key, keylen );
+
+    for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+      blake2s_update( S->S[i], block, BLAKE2S_BLOCKBYTES );
+
+    secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
+  }
+  return 0;
+}
+
+
+int blake2sp_update( blake2sp_state *S, const uint8_t *in, size_t inlen )
+{
+  size_t left = S->buflen;
+  size_t fill = sizeof( S->buf ) - left;
+
+  if( left && inlen >= fill )
+  {
+    memcpy( S->buf + left, in, fill );
+
+    for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+      blake2s_update( S->S[i], S->buf + i * BLAKE2S_BLOCKBYTES, BLAKE2S_BLOCKBYTES );
+
+    in += fill;
+    inlen -= fill;
+    left = 0;
+  }
+
+#if defined(_OPENMP)
+  omp_set_num_threads(PARALLELISM_DEGREE);
+  #pragma omp parallel shared(S)
+#else
+  for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
+#endif
+  {
+#if defined(_OPENMP)
+    size_t      id__ = omp_get_thread_num();
+#endif
+    size_t inlen__ = inlen;
+    const uint8_t *in__ = ( const uint8_t * )in;
+    in__ += id__ * BLAKE2S_BLOCKBYTES;
+
+    while( inlen__ >= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES )
+    {
+      blake2s_update( S->S[id__], in__, BLAKE2S_BLOCKBYTES );
+      in__ += PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
+      inlen__ -= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
+    }
+  }
+
+  in += inlen - inlen % ( PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES );
+  inlen %= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
+
+  if( inlen > 0 )
+    memcpy( S->buf + left, in, inlen );
+
+  S->buflen = left + inlen;
+  return 0;
+}
+
+
+int blake2sp_final( blake2sp_state *S, uint8_t *out, size_t outlen )
+{
+  uint8_t hash[PARALLELISM_DEGREE][BLAKE2S_OUTBYTES];
+
+  if(S->outlen != outlen) return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+  {
+    if( S->buflen > i * BLAKE2S_BLOCKBYTES )
+    {
+      size_t left = S->buflen - i * BLAKE2S_BLOCKBYTES;
+
+      if( left > BLAKE2S_BLOCKBYTES ) left = BLAKE2S_BLOCKBYTES;
+
+      blake2s_update( S->S[i], S->buf + i * BLAKE2S_BLOCKBYTES, left );
+    }
+
+    blake2s_final( S->S[i], hash[i], BLAKE2S_OUTBYTES );
+  }
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    blake2s_update( S->R, hash[i], BLAKE2S_OUTBYTES );
+
+  blake2s_final( S->R, out, outlen );
+  return 0;
+}
+
+
+int blake2sp( uint8_t *out, const void *in, const void *key, size_t outlen, size_t inlen, size_t keylen )
+{
+  uint8_t hash[PARALLELISM_DEGREE][BLAKE2S_OUTBYTES];
+  blake2s_state S[PARALLELISM_DEGREE][1];
+  blake2s_state FS[1];
+
+  /* Verify parameters */
+  if ( NULL == in ) return -1;
+
+  if ( NULL == out ) return -1;
+
+  if ( NULL == key && keylen > 0 ) return -1;
+
+  if( !outlen || outlen > BLAKE2S_OUTBYTES ) return -1;
+
+  if( keylen > BLAKE2S_KEYBYTES ) return -1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    if( blake2sp_init_leaf( S[i], outlen, keylen, i ) < 0 ) return -1;
+
+  S[PARALLELISM_DEGREE - 1]->last_node = 1; // mark last node
+
+  if( keylen > 0 )
+  {
+    uint8_t block[BLAKE2S_BLOCKBYTES];
+    memset( block, 0, BLAKE2S_BLOCKBYTES );
+    memcpy( block, key, keylen );
+
+    for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+      blake2s_update( S[i], block, BLAKE2S_BLOCKBYTES );
+
+    secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
+  }
+
+#if defined(_OPENMP)
+  omp_set_num_threads(PARALLELISM_DEGREE);
+  #pragma omp parallel shared(S,hash)
+#else
+
+  for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
+#endif
+  {
+#if defined(_OPENMP)
+    size_t      id__ = omp_get_thread_num();
+#endif
+    size_t inlen__ = inlen;
+    const uint8_t *in__ = ( const uint8_t * )in;
+    in__ += id__ * BLAKE2S_BLOCKBYTES;
+
+    while( inlen__ >= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES )
+    {
+      blake2s_update( S[id__], in__, BLAKE2S_BLOCKBYTES );
+      in__ += PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
+      inlen__ -= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
+    }
+
+    if( inlen__ > id__ * BLAKE2S_BLOCKBYTES )
+    {
+      const size_t left = inlen__ - id__ * BLAKE2S_BLOCKBYTES;
+      const size_t len = left <= BLAKE2S_BLOCKBYTES ? left : BLAKE2S_BLOCKBYTES;
+      blake2s_update( S[id__], in__, len );
+    }
+
+    blake2s_final( S[id__], hash[id__], BLAKE2S_OUTBYTES );
+  }
+
+  if( blake2sp_init_root( FS, outlen, keylen ) < 0 )
+    return -1;
+
+  FS->last_node = 1;
+
+  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
+    blake2s_update( FS, hash[i], BLAKE2S_OUTBYTES );
+
+  return blake2s_final( FS, out, outlen );
+}
+
+
+
+