/**
* Executes a reduced version of G function with only one round
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's or BlaMka's G function
*/
inline static void reducedSpongeLyra(uint64_t *v) {
#if (SPONGE == 0)
ROUND_LYRA(0);
#endif
#if (SPONGE == 1)
uint64_t t0,t1,t2;
ROUND_LYRA_BLAMKA(0);
#endif
}
/**
* Executes a reduced version of Blake2b's G function with only one round
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's G function
*/
inline static void reducedBlake2bLyra(uint64_t *v) {
ROUND_LYRA(0);
}
/**
* Execute Blake2b's G function, with all 12 rounds.
*
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's G function
*/
inline static void blake2bLyra(uint64_t *v) {
ROUND_LYRA(0);
ROUND_LYRA(1);
ROUND_LYRA(2);
ROUND_LYRA(3);
ROUND_LYRA(4);
ROUND_LYRA(5);
ROUND_LYRA(6);
ROUND_LYRA(7);
ROUND_LYRA(8);
ROUND_LYRA(9);
ROUND_LYRA(10);
ROUND_LYRA(11);
}
inline static void lyra_round( uint64_t *v )
{
#ifdef __AVX2__
__m256i a = _mm256_load_si256( (__m256i*)(&v[ 0]) );
__m256i b = _mm256_load_si256( (__m256i*)(&v[ 4]) );
__m256i c = _mm256_load_si256( (__m256i*)(&v[ 8]) );
__m256i d = _mm256_load_si256( (__m256i*)(&v[12]) );
G_4X64( a, b, c, d );
// swap words
b = mm256_rotl256_1x64( b );
c = mm256_swap128( c );
d = mm256_rotr256_1x64( d );
G_4X64( a, b, c, d );
// unswap
b = mm256_rotr256_1x64( b );
c = mm256_swap128( c );
d = mm256_rotl256_1x64( d );
_mm256_store_si256( (__m256i*)(&v[ 0]), a );
_mm256_store_si256( (__m256i*)(&v[ 4]), b );
_mm256_store_si256( (__m256i*)(&v[ 8]), c );
_mm256_store_si256( (__m256i*)(&v[12]), d );
#elif defined __AVX__
__m128i a0, a1, b0, b1, c0, c1, d0, d1;
a0 = _mm_load_si128( (__m128i*)(&v[ 0]) );
a1 = _mm_load_si128( (__m128i*)(&v[ 2]) );
b0 = _mm_load_si128( (__m128i*)(&v[ 4]) );
b1 = _mm_load_si128( (__m128i*)(&v[ 6]) );
c0 = _mm_load_si128( (__m128i*)(&v[ 8]) );
c1 = _mm_load_si128( (__m128i*)(&v[10]) );
d0 = _mm_load_si128( (__m128i*)(&v[12]) );
d1 = _mm_load_si128( (__m128i*)(&v[14]) );
G_2X64( a0, b0, c0, d0 );
G_2X64( a1, b1, c1, d1 );
// swap words
mm128_rotl256_1x64( b0, b1 );
mm128_swap128( c0, c1 );
mm128_rotr256_1x64( d0, d1 );
G_2X64( a0, b0, c0, d0 );
G_2X64( a1, b1, c1, d1 );
// unswap
mm128_rotr256_1x64( b0, b1 );
mm128_swap128( c0, c1 );
mm128_rotl256_1x64( d0, d1 );
_mm_store_si128( (__m128i*)(&v[ 0]), a0 );
_mm_store_si128( (__m128i*)(&v[ 2]), a1 );
_mm_store_si128( (__m128i*)(&v[ 4]), b0 );
_mm_store_si128( (__m128i*)(&v[ 6]), b1 );
_mm_store_si128( (__m128i*)(&v[ 8]), c0 );
_mm_store_si128( (__m128i*)(&v[10]), c1 );
_mm_store_si128( (__m128i*)(&v[12]), d0 );
_mm_store_si128( (__m128i*)(&v[14]), d1 );
#else
// macro assumes v is defined
ROUND_LYRA(0);
#endif
}
/**
* Execute G function, with all 12 rounds.
*
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's or BlaMka's G function
*/
inline static void spongeLyra(uint64_t *v) {
#if (SPONGE == 0)
ROUND_LYRA(0);
ROUND_LYRA(1);
ROUND_LYRA(2);
ROUND_LYRA(3);
ROUND_LYRA(4);
ROUND_LYRA(5);
ROUND_LYRA(6);
ROUND_LYRA(7);
ROUND_LYRA(8);
ROUND_LYRA(9);
ROUND_LYRA(10);
ROUND_LYRA(11);
#endif
#if (SPONGE == 1)
uint64_t t0,t1,t2;
ROUND_LYRA_BLAMKA(0);
ROUND_LYRA_BLAMKA(1);
ROUND_LYRA_BLAMKA(2);
ROUND_LYRA_BLAMKA(3);
ROUND_LYRA_BLAMKA(4);
ROUND_LYRA_BLAMKA(5);
ROUND_LYRA_BLAMKA(6);
ROUND_LYRA_BLAMKA(7);
ROUND_LYRA_BLAMKA(8);
ROUND_LYRA_BLAMKA(9);
ROUND_LYRA_BLAMKA(10);
ROUND_LYRA_BLAMKA(11);
ROUND_LYRA_BLAMKA(12);
ROUND_LYRA_BLAMKA(13);
ROUND_LYRA_BLAMKA(14);
ROUND_LYRA_BLAMKA(15);
ROUND_LYRA_BLAMKA(16);
ROUND_LYRA_BLAMKA(17);
ROUND_LYRA_BLAMKA(18);
ROUND_LYRA_BLAMKA(19);
ROUND_LYRA_BLAMKA(20);
ROUND_LYRA_BLAMKA(21);
ROUND_LYRA_BLAMKA(22);
ROUND_LYRA_BLAMKA(23);
#endif
}
/**
* Execute Blake2b's G function, with all 12 rounds.
*
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's G function
*/
inline static void blake2bLyra(uint64_t *v) {
#if defined __AVX2__
// may be still used by squeeze
LYRA_INIT_AVX2; // defines local a[4]
LYRA_12_ROUNDS_AVX2( a[0], a[1], a[2], a[3] );
LYRA_CLOSE_AVX2;
#elif defined __AVX__
LYRA_INIT_AVX; // defines locals a0[4], a1[4]
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_ROUND_AVX;
LYRA_CLOSE_AVX;
#else
ROUND_LYRA(0);
ROUND_LYRA(1);
ROUND_LYRA(2);
ROUND_LYRA(3);
ROUND_LYRA(4);
ROUND_LYRA(5);
ROUND_LYRA(6);
ROUND_LYRA(7);
ROUND_LYRA(8);
ROUND_LYRA(9);
ROUND_LYRA(10);
ROUND_LYRA(11);
#endif
}
