Random123/ars.h
Go to the documentation of this file.
00001 /* 00002 Copyright 2010-2011, D. E. Shaw Research. 00003 All rights reserved. 00004 00005 Redistribution and use in source and binary forms, with or without 00006 modification, are permitted provided that the following conditions are 00007 met: 00008 00009 * Redistributions of source code must retain the above copyright 00010 notice, this list of conditions, and the following disclaimer. 00011 00012 * Redistributions in binary form must reproduce the above copyright 00013 notice, this list of conditions, and the following disclaimer in the 00014 documentation and/or other materials provided with the distribution. 00015 00016 * Neither the name of D. E. Shaw Research nor the names of its 00017 contributors may be used to endorse or promote products derived from 00018 this software without specific prior written permission. 00019 00020 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 00021 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 00022 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 00023 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 00024 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 00025 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 00026 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 00027 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 00028 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 00029 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 00030 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 00031 */ 00032 #ifndef __Random123_ars_dot_hpp__ 00033 #define __Random123_ars_dot_hpp__ 00034 00035 #include "features/compilerfeatures.h" 00036 #include "array.h" 00037 00038 #if R123_USE_AES_NI 00039 00040 #ifndef ARS1xm128i_DEFAULT_ROUNDS 00041 #define ARS1xm128i_DEFAULT_ROUNDS 7 00042 #endif 00043 00045 enum r123_enum_ars1xm128i {ars1xm128i_rounds = ARS1xm128i_DEFAULT_ROUNDS}; 00046 00047 /* ARS1xm128i with Weyl keys. Fast, and Crush-resistant, but NOT CRYPTO. */ 00049 typedef struct r123array1xm128i ars1xm128i_ctr_t; 00051 typedef struct r123array1xm128i ars1xm128i_key_t; 00053 typedef struct r123array1xm128i ars1xm128i_ukey_t; 00055 R123_STATIC_INLINE ars1xm128i_key_t ars1xm128ikeyinit(ars1xm128i_ukey_t uk) { return uk; } 00057 R123_STATIC_INLINE ars1xm128i_ctr_t ars1xm128i_R(unsigned int Nrounds, ars1xm128i_ctr_t in, ars1xm128i_key_t k){ 00058 __m128i kweyl = _mm_set_epi64x(R123_64BIT(0xBB67AE8584CAA73B), /* sqrt(3) - 1.0 */ 00059 R123_64BIT(0x9E3779B97F4A7C15)); /* golden ratio */ 00060 /* N.B. the aesenc instructions do the xor *after* 00061 // so if we want to follow the AES pattern, we 00062 // have to do the initial xor explicitly */ 00063 __m128i kk = k.v[0].m; 00064 __m128i v = _mm_xor_si128(in.v[0].m, kk); 00065 ars1xm128i_ctr_t ret; 00066 R123_ASSERT(Nrounds<=10); 00067 if( Nrounds>1 ){ 00068 kk = _mm_add_epi64(kk, kweyl); 00069 v = _mm_aesenc_si128(v, kk); 00070 } 00071 if( Nrounds>2 ){ 00072 kk = _mm_add_epi64(kk, kweyl); 00073 v = _mm_aesenc_si128(v, kk); 00074 } 00075 if( Nrounds>3 ){ 00076 kk = _mm_add_epi64(kk, kweyl); 00077 v = _mm_aesenc_si128(v, kk); 00078 } 00079 if( Nrounds>4 ){ 00080 kk = _mm_add_epi64(kk, kweyl); 00081 v = _mm_aesenc_si128(v, kk); 00082 } 00083 if( Nrounds>5 ){ 00084 kk = _mm_add_epi64(kk, kweyl); 00085 v = _mm_aesenc_si128(v, kk); 00086 } 00087 if( Nrounds>6 ){ 00088 kk = _mm_add_epi64(kk, kweyl); 00089 v = _mm_aesenc_si128(v, kk); 00090 } 00091 if( Nrounds>7 ){ 00092 kk = _mm_add_epi64(kk, kweyl); 00093 v = _mm_aesenc_si128(v, kk); 00094 } 00095 if( Nrounds>8 ){ 00096 kk = _mm_add_epi64(kk, kweyl); 00097 v = _mm_aesenc_si128(v, kk); 00098 } 00099 if( Nrounds>9 ){ 00100 kk = _mm_add_epi64(kk, kweyl); 00101 v = _mm_aesenc_si128(v, kk); 00102 } 00103 kk = _mm_add_epi64(kk, kweyl); 00104 v = _mm_aesenclast_si128(v, kk); 00105 ret.v[0].m = v; 00106 return ret; 00107 } 00108 00112 #define ars1xm128i(c,k) ars1xm128i_R(ars1xm128i_rounds, c, k) 00113 00115 typedef struct r123array4x32 ars4x32_ctr_t; 00117 typedef struct r123array4x32 ars4x32_key_t; 00119 typedef struct r123array4x32 ars4x32_ukey_t; 00121 enum r123_enum_ars4x32 {ars4x32_rounds = ARS1xm128i_DEFAULT_ROUNDS}; 00123 R123_STATIC_INLINE ars4x32_key_t ars4x32keyinit(ars4x32_ukey_t uk) { return uk; } 00125 R123_STATIC_INLINE ars4x32_ctr_t ars4x32_R(unsigned int Nrounds, ars4x32_ctr_t c, ars4x32_key_t k){ 00126 ars1xm128i_ctr_t c128; 00127 ars1xm128i_key_t k128; 00128 c128.v[0].m = _mm_set_epi32(c.v[3], c.v[2], c.v[1], c.v[0]); 00129 k128.v[0].m = _mm_set_epi32(k.v[3], k.v[2], k.v[1], k.v[0]); 00130 c128 = ars1xm128i_R(Nrounds, c128, k128); 00131 _mm_storeu_si128((__m128i*)&c.v[0], c128.v[0].m); 00132 return c; 00133 } 00134 00138 #define ars4x32(c,k) ars4x32_R(ars4x32_rounds, c, k) 00139 00140 #ifdef __cplusplus 00141 namespace r123{ 00163 template<unsigned int ROUNDS> 00164 struct ARS1xm128i_R{ 00165 typedef ars1xm128i_ctr_t ctr_type; 00166 typedef ars1xm128i_key_t key_type; 00167 typedef ars1xm128i_key_t ukey_type; 00168 static const unsigned int rounds=ROUNDS; 00169 R123_FORCE_INLINE(ctr_type operator()(ctr_type ctr, key_type key) const){ 00170 return ars1xm128i_R(ROUNDS, ctr, key); 00171 } 00172 }; 00173 00178 template<unsigned int ROUNDS> 00179 struct ARS4x32_R{ 00180 typedef ars4x32_ctr_t ctr_type; 00181 typedef ars4x32_key_t key_type; 00182 typedef ars4x32_key_t ukey_type; 00183 static const unsigned int rounds=ROUNDS; 00184 R123_FORCE_INLINE(ctr_type operator()(ctr_type ctr, key_type key) const){ 00185 return ars4x32_R(ROUNDS, ctr, key); 00186 } 00187 }; 00196 typedef ARS1xm128i_R<ars1xm128i_rounds> ARS1xm128i; 00197 typedef ARS4x32_R<ars4x32_rounds> ARS4x32; 00198 } // namespace r123 00199 00200 #endif /* __cplusplus */ 00201 00202 #endif /* R123_USE_AES_NI */ 00203 00204 #endif
