@leocuvee/wrkzcoin-multi-hashing 0.0.20

package/crypto/oaes_lib.h

@@ -0,0 +1,215 @@
+/* 
+ * ---------------------------------------------------------------------------
+ * OpenAES License
+ * ---------------------------------------------------------------------------
+ * Copyright (c) 2012, Nabil S. Al Ramli, www.nalramli.com
+ * All rights reserved.
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 
+ *   - Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *   - Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in the
+ *     documentation and/or other materials provided with the distribution.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ * ---------------------------------------------------------------------------
+ */
+
+#ifndef _OAES_LIB_H
+#define _OAES_LIB_H
+
+#include <stdint.h>
+#include <stdlib.h>
+
+#ifdef __cplusplus 
+extern "C" {
+#endif
+
+#ifdef _WIN32
+#	ifdef OAES_SHARED
+#		ifdef oaes_lib_EXPORTS
+#			define OAES_API __declspec(dllexport)
+#		else
+#			define OAES_API __declspec(dllimport)
+#		endif
+#	else
+#		define OAES_API
+#	endif
+#else
+#	define OAES_API
+#endif // WIN32
+
+#define OAES_VERSION "0.8.1"
+#define OAES_BLOCK_SIZE 16
+
+typedef void OAES_CTX;
+
+typedef enum
+{
+	OAES_RET_FIRST = 0,
+	OAES_RET_SUCCESS = 0,
+	OAES_RET_UNKNOWN,
+	OAES_RET_ARG1,
+	OAES_RET_ARG2,
+	OAES_RET_ARG3,
+	OAES_RET_ARG4,
+	OAES_RET_ARG5,
+	OAES_RET_NOKEY,
+	OAES_RET_MEM,
+	OAES_RET_BUF,
+	OAES_RET_HEADER,
+	OAES_RET_COUNT
+} OAES_RET;
+
+/*
+ * oaes_set_option() takes one of these values for its [option] parameter
+ * some options accept either an optional or a required [value] parameter
+ */
+// no option
+#define OAES_OPTION_NONE 0
+// enable ECB mode, disable CBC mode
+#define OAES_OPTION_ECB 1
+// enable CBC mode, disable ECB mode
+// value is optional, may pass uint8_t iv[OAES_BLOCK_SIZE] to specify
+// the value of the initialization vector, iv
+#define OAES_OPTION_CBC 2
+
+#ifdef OAES_DEBUG
+typedef int ( * oaes_step_cb ) (
+		const uint8_t state[OAES_BLOCK_SIZE],
+		const char * step_name,
+		int step_count,
+		void * user_data );
+// enable state stepping mode
+// value is required, must pass oaes_step_cb to receive the state at each step
+#define OAES_OPTION_STEP_ON 4
+// disable state stepping mode
+#define OAES_OPTION_STEP_OFF 8
+#endif // OAES_DEBUG
+
+typedef uint16_t OAES_OPTION;
+
+typedef struct _oaes_key
+{
+  size_t data_len;
+  uint8_t *data;
+  size_t exp_data_len;
+  uint8_t *exp_data;
+  size_t num_keys;
+  size_t key_base;
+} oaes_key;
+
+typedef struct _oaes_ctx
+{
+#ifdef OAES_HAVE_ISAAC
+  randctx * rctx;
+#endif // OAES_HAVE_ISAAC
+
+#ifdef OAES_DEBUG
+  oaes_step_cb step_cb;
+#endif // OAES_DEBUG
+
+  oaes_key * key;
+  OAES_OPTION options;
+  uint8_t iv[OAES_BLOCK_SIZE];
+} oaes_ctx;
+/*
+ * // usage:
+ * 
+ * OAES_CTX * ctx = oaes_alloc();
+ * .
+ * .
+ * .
+ * {
+ *   oaes_gen_key_xxx( ctx );
+ *   {
+ *     oaes_key_export( ctx, _buf, &_buf_len );
+ *     // or
+ *     oaes_key_export_data( ctx, _buf, &_buf_len );\
+ *   }
+ * }
+ * // or
+ * {
+ *   oaes_key_import( ctx, _buf, _buf_len );
+ *   // or
+ *   oaes_key_import_data( ctx, _buf, _buf_len );
+ * }
+ * .
+ * .
+ * .
+ * oaes_encrypt( ctx, m, m_len, c, &c_len );
+ * .
+ * .
+ * .
+ * oaes_decrypt( ctx, c, c_len, m, &m_len );
+ * .
+ * .
+ * .
+ * oaes_free( &ctx );
+ */
+
+OAES_API OAES_CTX * oaes_alloc(void);
+
+OAES_API OAES_RET oaes_free( OAES_CTX ** ctx );
+
+OAES_API OAES_RET oaes_set_option( OAES_CTX * ctx,
+		OAES_OPTION option, const void * value );
+
+OAES_API OAES_RET oaes_key_gen_128( OAES_CTX * ctx );
+
+OAES_API OAES_RET oaes_key_gen_192( OAES_CTX * ctx );
+
+OAES_API OAES_RET oaes_key_gen_256( OAES_CTX * ctx );
+
+// export key with header information
+// set data == NULL to get the required data_len
+OAES_API OAES_RET oaes_key_export( OAES_CTX * ctx,
+		uint8_t * data, size_t * data_len );
+
+// directly export the data from key
+// set data == NULL to get the required data_len
+OAES_API OAES_RET oaes_key_export_data( OAES_CTX * ctx,
+		uint8_t * data, size_t * data_len );
+
+// import key with header information
+OAES_API OAES_RET oaes_key_import( OAES_CTX * ctx,
+		const uint8_t * data, size_t data_len );
+
+// directly import data into key
+OAES_API OAES_RET oaes_key_import_data( OAES_CTX * ctx,
+		const uint8_t * data, size_t data_len );
+
+// set c == NULL to get the required c_len
+OAES_API OAES_RET oaes_encrypt( OAES_CTX * ctx,
+		const uint8_t * m, size_t m_len, uint8_t * c, size_t * c_len );
+
+// set m == NULL to get the required m_len
+OAES_API OAES_RET oaes_decrypt( OAES_CTX * ctx,
+		const uint8_t * c, size_t c_len, uint8_t * m, size_t * m_len );
+
+// set buf == NULL to get the required buf_len
+OAES_API OAES_RET oaes_sprintf(
+		char * buf, size_t * buf_len, const uint8_t * data, size_t data_len );
+
+OAES_API OAES_RET oaes_encryption_round( const uint8_t * key, uint8_t * c );
+
+OAES_API OAES_RET oaes_pseudo_encrypt_ecb( OAES_CTX * ctx, uint8_t * c );
+
+#ifdef __cplusplus 
+}
+#endif
+
+#endif // _OAES_LIB_H

package/crypto/skein_port.h

@@ -0,0 +1,190 @@
+#ifndef _SKEIN_PORT_H_
+#define _SKEIN_PORT_H_
+
+#include <limits.h>
+#include <stdint.h>
+
+#ifndef RETURN_VALUES
+#  define RETURN_VALUES
+#  if defined( DLL_EXPORT )
+#    if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+#      define VOID_RETURN    __declspec( dllexport ) void __stdcall
+#      define INT_RETURN     __declspec( dllexport ) int  __stdcall
+#    elif defined( __GNUC__ )
+#      define VOID_RETURN    __declspec( __dllexport__ ) void
+#      define INT_RETURN     __declspec( __dllexport__ ) int
+#    else
+#      error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+#    endif
+#  elif defined( DLL_IMPORT )
+#    if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+#      define VOID_RETURN    __declspec( dllimport ) void __stdcall
+#      define INT_RETURN     __declspec( dllimport ) int  __stdcall
+#    elif defined( __GNUC__ )
+#      define VOID_RETURN    __declspec( __dllimport__ ) void
+#      define INT_RETURN     __declspec( __dllimport__ ) int
+#    else
+#      error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+#    endif
+#  elif defined( __WATCOMC__ )
+#    define VOID_RETURN  void __cdecl
+#    define INT_RETURN   int  __cdecl
+#  else
+#    define VOID_RETURN  void
+#    define INT_RETURN   int
+#  endif
+#endif
+
+/*  These defines are used to declare buffers in a way that allows
+    faster operations on longer variables to be used.  In all these
+    defines 'size' must be a power of 2 and >= 8
+
+    dec_unit_type(size,x)       declares a variable 'x' of length 
+                                'size' bits
+
+    dec_bufr_type(size,bsize,x) declares a buffer 'x' of length 'bsize' 
+                                bytes defined as an array of variables
+                                each of 'size' bits (bsize must be a 
+                                multiple of size / 8)
+
+    ptr_cast(x,size)            casts a pointer to a pointer to a 
+                                varaiable of length 'size' bits
+*/
+
+#define ui_type(size)               uint##size##_t
+#define dec_unit_type(size,x)       typedef ui_type(size) x
+#define dec_bufr_type(size,bsize,x) typedef ui_type(size) x[bsize / (size >> 3)]
+#define ptr_cast(x,size)            ((ui_type(size)*)(x))
+
+typedef unsigned int    uint_t;             /* native unsigned integer */
+typedef uint8_t         u08b_t;             /*  8-bit unsigned integer */
+typedef uint64_t        u64b_t;             /* 64-bit unsigned integer */
+
+#ifndef RotL_64
+#define RotL_64(x,N)    (((x) << (N)) | ((x) >> (64-(N))))
+#endif
+
+/*
+ * Skein is "natively" little-endian (unlike SHA-xxx), for optimal
+ * performance on x86 CPUs.  The Skein code requires the following
+ * definitions for dealing with endianness:
+ *
+ *    SKEIN_NEED_SWAP:  0 for little-endian, 1 for big-endian
+ *    Skein_Put64_LSB_First
+ *    Skein_Get64_LSB_First
+ *    Skein_Swap64
+ *
+ * If SKEIN_NEED_SWAP is defined at compile time, it is used here
+ * along with the portable versions of Put64/Get64/Swap64, which 
+ * are slow in general.
+ *
+ * Otherwise, an "auto-detect" of endianness is attempted below.
+ * If the default handling doesn't work well, the user may insert
+ * platform-specific code instead (e.g., for big-endian CPUs).
+ *
+ */
+#ifndef SKEIN_NEED_SWAP /* compile-time "override" for endianness? */
+
+
+#include "int-util.h"
+
+#define IS_BIG_ENDIAN      4321 /* byte 0 is most significant (mc68k) */
+#define IS_LITTLE_ENDIAN   1234 /* byte 0 is least significant (i386) */
+
+#if BYTE_ORDER == LITTLE_ENDIAN
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if BYTE_ORDER == BIG_ENDIAN
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#endif
+
+/* special handler for IA64, which may be either endianness (?)  */
+/* here we assume little-endian, but this may need to be changed */
+#if defined(__ia64) || defined(__ia64__) || defined(_M_IA64)
+#  define PLATFORM_MUST_ALIGN (1)
+#ifndef PLATFORM_BYTE_ORDER
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+#endif
+
+#ifndef   PLATFORM_MUST_ALIGN
+#  define PLATFORM_MUST_ALIGN (0)
+#endif
+
+
+#if   PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN
+    /* here for big-endian CPUs */
+#define SKEIN_NEED_SWAP   (1)
+#elif PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN
+    /* here for x86 and x86-64 CPUs (and other detected little-endian CPUs) */
+#define SKEIN_NEED_SWAP   (0)
+#if   PLATFORM_MUST_ALIGN == 0              /* ok to use "fast" versions? */
+#define Skein_Put64_LSB_First(dst08,src64,bCnt) memcpy(dst08,src64,bCnt)
+#define Skein_Get64_LSB_First(dst64,src08,wCnt) memcpy(dst64,src08,8*(wCnt))
+#endif
+#else
+#error "Skein needs endianness setting!"
+#endif
+
+#endif /* ifndef SKEIN_NEED_SWAP */
+
+/*
+ ******************************************************************
+ *      Provide any definitions still needed.
+ ******************************************************************
+ */
+#ifndef Skein_Swap64  /* swap for big-endian, nop for little-endian */
+#if     SKEIN_NEED_SWAP
+#define Skein_Swap64(w64)                       \
+  ( (( ((u64b_t)(w64))       & 0xFF) << 56) |   \
+    (((((u64b_t)(w64)) >> 8) & 0xFF) << 48) |   \
+    (((((u64b_t)(w64)) >>16) & 0xFF) << 40) |   \
+    (((((u64b_t)(w64)) >>24) & 0xFF) << 32) |   \
+    (((((u64b_t)(w64)) >>32) & 0xFF) << 24) |   \
+    (((((u64b_t)(w64)) >>40) & 0xFF) << 16) |   \
+    (((((u64b_t)(w64)) >>48) & 0xFF) <<  8) |   \
+    (((((u64b_t)(w64)) >>56) & 0xFF)      ) )
+#else
+#define Skein_Swap64(w64)  (w64)
+#endif
+#endif  /* ifndef Skein_Swap64 */
+
+
+#ifndef Skein_Put64_LSB_First
+void    Skein_Put64_LSB_First(u08b_t *dst,const u64b_t *src,size_t bCnt)
+#ifdef  SKEIN_PORT_CODE /* instantiate the function code here? */
+    { /* this version is fully portable (big-endian or little-endian), but slow */
+    size_t n;
+
+    for (n=0;n<bCnt;n++)
+        dst[n] = (u08b_t) (src[n>>3] >> (8*(n&7)));
+    }
+#else
+    ;    /* output only the function prototype */
+#endif
+#endif   /* ifndef Skein_Put64_LSB_First */
+
+
+#ifndef Skein_Get64_LSB_First
+void    Skein_Get64_LSB_First(u64b_t *dst,const u08b_t *src,size_t wCnt)
+#ifdef  SKEIN_PORT_CODE /* instantiate the function code here? */
+    { /* this version is fully portable (big-endian or little-endian), but slow */
+    size_t n;
+
+    for (n=0;n<8*wCnt;n+=8)
+        dst[n/8] = (((u64b_t) src[n  ])      ) +
+                   (((u64b_t) src[n+1]) <<  8) +
+                   (((u64b_t) src[n+2]) << 16) +
+                   (((u64b_t) src[n+3]) << 24) +
+                   (((u64b_t) src[n+4]) << 32) +
+                   (((u64b_t) src[n+5]) << 40) +
+                   (((u64b_t) src[n+6]) << 48) +
+                   (((u64b_t) src[n+7]) << 56) ;
+    }
+#else
+    ;    /* output only the function prototype */
+#endif
+#endif   /* ifndef Skein_Get64_LSB_First */
+
+#endif   /* ifndef _SKEIN_PORT_H_ */

package/crypto/variant2_int_sqrt.h

@@ -0,0 +1,168 @@
+#ifndef VARIANT2_INT_SQRT_H
+#define VARIANT2_INT_SQRT_H
+
+#include <math.h>
+#include <float.h>
+
+#define VARIANT2_INTEGER_MATH_SQRT_STEP_SSE2() \
+  do { \
+    const __m128i exp_double_bias = _mm_set_epi64x(0, 1023ULL << 52); \
+    __m128d x = _mm_castsi128_pd(_mm_add_epi64(_mm_cvtsi64_si128(sqrt_input >> 12), exp_double_bias)); \
+    x = _mm_sqrt_sd(_mm_setzero_pd(), x); \
+    sqrt_result = (uint64_t)(_mm_cvtsi128_si64(_mm_sub_epi64(_mm_castpd_si128(x), exp_double_bias))) >> 19; \
+  } while(0)
+
+#define VARIANT2_INTEGER_MATH_SQRT_STEP_FP64() \
+  do { \
+    sqrt_result = sqrt(sqrt_input + 18446744073709551616.0) * 2.0 - 8589934592.0; \
+  } while(0)
+
+//#define VARIANT2_INTEGER_MATH_SQRT_STEP_REF() \
+//  sqrt_result = integer_square_root_v2(sqrt_input)
+
+// Reference implementation of the integer square root for Cryptonight variant 2
+// Computes integer part of "sqrt(2^64 + n) * 2 - 2^33"
+//
+// In other words, given 64-bit unsigned integer n:
+// 1) Write it as x = 1.NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN000... in binary (1 <= x < 2, all 64 bits of n are used)
+// 2) Calculate sqrt(x) = 1.0RRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR... (1 <= sqrt(x) < sqrt(2), so it will always start with "1.0" in binary)
+// 3) Take 32 bits that come after "1.0" and return them as a 32-bit unsigned integer, discard all remaining bits
+//
+// Some sample inputs and outputs:
+//
+// Input            | Output     | Exact value of "sqrt(2^64 + n) * 2 - 2^33"
+// -----------------|------------|-------------------------------------------
+// 0                | 0          | 0
+// 2^32             | 0          | 0.99999999994179233909330885695244...
+// 2^32 + 1         | 1          | 1.0000000001746229827200734316305...
+// 2^50             | 262140     | 262140.00012206565608606978175873...
+// 2^55 + 20963331  | 8384515    | 8384515.9999999997673963974959744...
+// 2^55 + 20963332  | 8384516    | 8384516
+// 2^62 + 26599786  | 1013904242 | 1013904242.9999999999479374853545...
+// 2^62 + 26599787  | 1013904243 | 1013904243.0000000001561875439364...
+// 2^64 - 1         | 3558067407 | 3558067407.9041987696409179931096...
+
+// The reference implementation as it is now uses only unsigned int64 arithmetic, so it can't have undefined behavior
+// It was tested once for all edge cases and confirmed correct
+//
+// !!! Note: if you're modifying this code, uncomment the test in monero/tests/hash/main.cpp !!!
+//
+/*
+static inline uint64_t integer_square_root_v2(uint64_t n)
+{
+  uint64_t r = 1ULL << 63;
+
+  for (uint64_t bit = 1ULL << 60; bit; bit >>= 2)
+  {
+    const bool b = (n < r + bit);
+    const uint64_t n_next = n - (r + bit);
+    const uint64_t r_next = r + bit * 2;
+    n = b ? n : n_next;
+    r = b ? r : r_next;
+    r >>= 1;
+  }
+
+  return r * 2 + ((n > r) ? 1 : 0);
+}
+*/
+
+/*
+VARIANT2_INTEGER_MATH_SQRT_FIXUP checks that "r" is an integer part of "sqrt(2^64 + sqrt_input) * 2 - 2^33" and adds or subtracts 1 if needed
+It's hard to understand how it works, so here is a full calculation of formulas used in VARIANT2_INTEGER_MATH_SQRT_FIXUP
+
+The following inequalities must hold for r if it's an integer part of "sqrt(2^64 + sqrt_input) * 2 - 2^33":
+1) r <= sqrt(2^64 + sqrt_input) * 2 - 2^33
+2) r + 1 > sqrt(2^64 + sqrt_input) * 2 - 2^33
+
+We need to check them using only unsigned integer arithmetic to avoid rounding errors and undefined behavior
+
+First inequality: r <= sqrt(2^64 + sqrt_input) * 2 - 2^33
+-----------------------------------------------------------------------------------
+r <= sqrt(2^64 + sqrt_input) * 2 - 2^33
+r + 2^33 <= sqrt(2^64 + sqrt_input) * 2
+r/2 + 2^32 <= sqrt(2^64 + sqrt_input)
+(r/2 + 2^32)^2 <= 2^64 + sqrt_input
+
+Rewrite r as r = s * 2 + b (s = trunc(r/2), b is 0 or 1)
+
+((s*2+b)/2 + 2^32)^2 <= 2^64 + sqrt_input
+(s*2+b)^2/4 + 2*2^32*(s*2+b)/2 + 2^64 <= 2^64 + sqrt_input
+(s*2+b)^2/4 + 2*2^32*(s*2+b)/2 <= sqrt_input
+(s*2+b)^2/4 + 2^32*r <= sqrt_input
+(s^2*4+2*s*2*b+b^2)/4 + 2^32*r <= sqrt_input
+s^2+s*b+b^2/4 + 2^32*r <= sqrt_input
+s*(s+b) + b^2/4 + 2^32*r <= sqrt_input
+
+Let r2 = s*(s+b) + r*2^32
+r2 + b^2/4 <= sqrt_input
+
+If this inequality doesn't hold, then we must decrement r: IF "r2 + b^2/4 > sqrt_input" THEN r = r - 1
+
+b can be 0 or 1
+If b is 0 then we need to compare "r2 > sqrt_input"
+If b is 1 then b^2/4 = 0.25, so we need to compare "r2 + 0.25 > sqrt_input"
+Since both r2 and sqrt_input are integers, we can safely replace it with "r2 + 1 > sqrt_input"
+-----------------------------------------------------------------------------------
+Both cases can be merged to a single expression "r2 + b > sqrt_input"
+-----------------------------------------------------------------------------------
+There will be no overflow when calculating "r2 + b", so it's safe to compare with sqrt_input:
+r2 + b = s*(s+b) + r*2^32 + b
+The largest value s, b and r can have is s = 1779033703, b = 1, r = 3558067407 when sqrt_input = 2^64 - 1
+r2 + b <= 1779033703*1779033704 + 3558067407*2^32 + 1 = 18446744068217447385 < 2^64
+
+Second inequality: r + 1 > sqrt(2^64 + sqrt_input) * 2 - 2^33
+-----------------------------------------------------------------------------------
+r + 1 > sqrt(2^64 + sqrt_input) * 2 - 2^33
+r + 1 + 2^33 > sqrt(2^64 + sqrt_input) * 2
+((r+1)/2 + 2^32)^2 > 2^64 + sqrt_input
+
+Rewrite r as r = s * 2 + b (s = trunc(r/2), b is 0 or 1)
+
+((s*2+b+1)/2 + 2^32)^2 > 2^64 + sqrt_input
+(s*2+b+1)^2/4 + 2*(s*2+b+1)/2*2^32 + 2^64 > 2^64 + sqrt_input
+(s*2+b+1)^2/4 + (s*2+b+1)*2^32 > sqrt_input
+(s*2+b+1)^2/4 + (r+1)*2^32 > sqrt_input
+(s*2+(b+1))^2/4 + r*2^32 + 2^32 > sqrt_input
+(s^2*4+2*s*2*(b+1)+(b+1)^2)/4 + r*2^32 + 2^32 > sqrt_input
+s^2+s*(b+1)+(b+1)^2/4 + r*2^32 + 2^32 > sqrt_input
+s*(s+b) + s + (b+1)^2/4 + r*2^32 + 2^32 > sqrt_input
+
+Let r2 = s*(s+b) + r*2^32
+
+r2 + s + (b+1)^2/4 + 2^32 > sqrt_input
+r2 + 2^32 + (b+1)^2/4 > sqrt_input - s
+
+If this inequality doesn't hold, then we must decrement r: IF "r2 + 2^32 + (b+1)^2/4 <= sqrt_input - s" THEN r = r - 1
+b can be 0 or 1
+If b is 0 then we need to compare "r2 + 2^32 + 1/4 <= sqrt_input - s" which is equal to "r2 + 2^32 < sqrt_input - s" because all numbers here are integers
+If b is 1 then (b+1)^2/4 = 1, so we need to compare "r2 + 2^32 + 1 <= sqrt_input - s" which is also equal to "r2 + 2^32 < sqrt_input - s"
+-----------------------------------------------------------------------------------
+Both cases can be merged to a single expression "r2 + 2^32 < sqrt_input - s"
+-----------------------------------------------------------------------------------
+There will be no overflow when calculating "r2 + 2^32":
+r2 + 2^32 = s*(s+b) + r*2^32 + 2^32 = s*(s+b) + (r+1)*2^32
+The largest value s, b and r can have is s = 1779033703, b = 1, r = 3558067407 when sqrt_input = 2^64 - 1
+r2 + b <= 1779033703*1779033704 + 3558067408*2^32 = 18446744072512414680 < 2^64
+
+There will be no integer overflow when calculating "sqrt_input - s", i.e. "sqrt_input >= s" at all times:
+s = trunc(r/2) = trunc(sqrt(2^64 + sqrt_input) - 2^32) < sqrt(2^64 + sqrt_input) - 2^32 + 1
+sqrt_input > sqrt(2^64 + sqrt_input) - 2^32 + 1
+sqrt_input + 2^32 - 1 > sqrt(2^64 + sqrt_input)
+(sqrt_input + 2^32 - 1)^2 > sqrt_input + 2^64
+sqrt_input^2 + 2*sqrt_input*(2^32 - 1) + (2^32-1)^2 > sqrt_input + 2^64
+sqrt_input^2 + sqrt_input*(2^33 - 2) + (2^32-1)^2 > sqrt_input + 2^64
+sqrt_input^2 + sqrt_input*(2^33 - 3) + (2^32-1)^2 > 2^64
+sqrt_input^2 + sqrt_input*(2^33 - 3) + 2^64-2^33+1 > 2^64
+sqrt_input^2 + sqrt_input*(2^33 - 3) - 2^33 + 1 > 0
+This inequality is true if sqrt_input > 1 and it's easy to check that s = 0 if sqrt_input is 0 or 1, so there will be no integer overflow
+*/
+
+#define VARIANT2_INTEGER_MATH_SQRT_FIXUP(r) \
+  do { \
+    const uint64_t s = r >> 1; \
+    const uint64_t b = r & 1; \
+    const uint64_t r2 = (uint64_t)(s) * (s + b) + (r << 32); \
+    r += ((r2 + b > sqrt_input) ? -1 : 0) + ((r2 + (1ULL << 32) < sqrt_input - s) ? 1 : 0); \
+  } while(0)
+
+#endif

package/crypto/wild_keccak.cpp

@@ -0,0 +1,119 @@
+// keccak.c
+// 19-Nov-11  Markku-Juhani O. Saarinen <mjos@iki.fi>
+// A baseline Keccak (3rd round) implementation.
+
+// Memory-hard extension of keccak for PoW 
+// Copyright (c) 2014 The Boolberry developers
+// Distributed under the MIT/X11 software license, see the accompanying
+// file COPYING or http://www.opensource.org/licenses/mit-license.php.
+
+
+#include "wild_keccak.h"
+namespace crypto
+{
+
+  const uint64_t keccakf_rndc[24] = 
+  {
+    0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
+    0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
+    0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
+    0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
+    0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
+    0x8000000000008003, 0x8000000000008002, 0x8000000000000080, 
+    0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
+    0x8000000000008080, 0x0000000080000001, 0x8000000080008008
+  };
+
+  const int keccakf_rotc[24] = 
+  {
+    1,  3,  6,  10, 15, 21, 28, 36, 45, 55, 2,  14, 
+    27, 41, 56, 8,  25, 43, 62, 18, 39, 61, 20, 44
+  };
+
+  const int keccakf_piln[24] = 
+  {
+    10, 7,  11, 17, 18, 3, 5,  16, 8,  21, 24, 4, 
+    15, 23, 19, 13, 12, 2, 20, 14, 22, 9,  6,  1 
+  };
+
+  // update the state with given number of rounds
+  void regular_f::keccakf(uint64_t st[25], int rounds)
+  {
+    int i, j, round;
+    uint64_t t, bc[5];
+
+    for (round = 0; round < rounds; round++) {
+
+      // Theta
+      for (i = 0; i < 5; i++)     
+        bc[i] = st[i] ^ st[i + 5] ^ st[i + 10] ^ st[i + 15] ^ st[i + 20];
+
+      for (i = 0; i < 5; i++) {
+        t = bc[(i + 4) % 5] ^ ROTL64(bc[(i + 1) % 5], 1);
+        for (j = 0; j < 25; j += 5)
+          st[j + i] ^= t;
+      }
+
+      // Rho Pi
+      t = st[1];
+      for (i = 0; i < 24; i++) {
+        j = keccakf_piln[i];
+        bc[0] = st[j];
+        st[j] = ROTL64(t, keccakf_rotc[i]);
+        t = bc[0];
+      }
+
+      //  Chi
+      for (j = 0; j < 25; j += 5) {
+        for (i = 0; i < 5; i++)
+          bc[i] = st[j + i];
+        for (i = 0; i < 5; i++)
+          st[j + i] ^= (~bc[(i + 1) % 5]) & bc[(i + 2) % 5];
+      }
+
+      //  Iota
+      st[0] ^= keccakf_rndc[round];
+    }
+  }
+
+  void mul_f::keccakf(uint64_t st[25], int rounds)
+  {
+    int i, j, round;
+    uint64_t t, bc[5];
+
+    for (round = 0; round < rounds; round++) {
+
+      // Theta
+      for (i = 0; i < 5; i++)     
+      {
+        bc[i] = st[i] ^ st[i + 5] ^ st[i + 10] * st[i + 15] * st[i + 20];//surprise
+      }
+
+      for (i = 0; i < 5; i++) {
+        t = bc[(i + 4) % 5] ^ ROTL64(bc[(i + 1) % 5], 1);
+        for (j = 0; j < 25; j += 5)
+          st[j + i] ^= t;
+      }
+
+      // Rho Pi
+      t = st[1];
+      for (i = 0; i < 24; i++) {
+        j = keccakf_piln[i];
+        bc[0] = st[j];
+        st[j] = ROTL64(t, keccakf_rotc[i]);
+        t = bc[0];
+      }
+
+      //  Chi
+      for (j = 0; j < 25; j += 5) {
+        for (i = 0; i < 5; i++)
+          bc[i] = st[j + i];
+        for (i = 0; i < 5; i++)
+          st[j + i] ^= (~bc[(i + 1) % 5]) & bc[(i + 2) % 5];
+      }
+
+      //  Iota
+      st[0] ^= keccakf_rndc[round];
+    }
+  }
+}