mirror of
https://github.com/JayDDee/cpuminer-opt.git
synced 2025-09-17 23:44:27 +00:00
Compare commits
4 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
1b76cee239 | ||
|
|
0681ca996d | ||
|
|
88f81fda0b | ||
|
|
103e6ad36c |
3
AUTHORS
3
AUTHORS
@@ -33,3 +33,6 @@ Jay D Dee
|
||||
xcouiz@gmail.com
|
||||
|
||||
Cryply
|
||||
|
||||
Colin Percival
|
||||
Alexander Peslyak
|
||||
|
||||
18
Makefile.am
18
Makefile.am
@@ -80,7 +80,6 @@ cpuminer_SOURCES = \
|
||||
algo/cryptonight/cryptonight-common.c\
|
||||
algo/cryptonight/cryptonight-aesni.c\
|
||||
algo/cryptonight/cryptonight.c\
|
||||
algo/cubehash/sph_cubehash.c \
|
||||
algo/cubehash/cubehash_sse2.c\
|
||||
algo/cubehash/cube-hash-2way.c \
|
||||
algo/echo/sph_echo.c \
|
||||
@@ -103,9 +102,6 @@ cpuminer_SOURCES = \
|
||||
algo/hamsi/hamsi-hash-4way.c \
|
||||
algo/haval/haval.c \
|
||||
algo/haval/haval-hash-4way.c \
|
||||
algo/heavy/sph_hefty1.c \
|
||||
algo/heavy/heavy.c \
|
||||
algo/heavy/bastion.c \
|
||||
algo/hodl/aes.c \
|
||||
algo/hodl/hodl-gate.c \
|
||||
algo/hodl/hodl-wolf.c \
|
||||
@@ -121,9 +117,9 @@ cpuminer_SOURCES = \
|
||||
algo/keccak/keccak-hash-4way.c \
|
||||
algo/keccak/keccak-4way.c\
|
||||
algo/keccak/keccak-gate.c \
|
||||
algo/keccak/sha3d-4way.c \
|
||||
algo/keccak/sha3d.c \
|
||||
algo/lanehash/lane.c \
|
||||
algo/luffa/sph_luffa.c \
|
||||
algo/luffa/luffa.c \
|
||||
algo/luffa/luffa_for_sse2.c \
|
||||
algo/luffa/luffa-hash-2way.c \
|
||||
algo/lyra2/lyra2.c \
|
||||
@@ -152,7 +148,6 @@ cpuminer_SOURCES = \
|
||||
algo/nist5/zr5.c \
|
||||
algo/panama/panama-hash-4way.c \
|
||||
algo/panama/sph_panama.c \
|
||||
algo/radiogatun/sph_radiogatun.c \
|
||||
algo/quark/quark-gate.c \
|
||||
algo/quark/quark.c \
|
||||
algo/quark/quark-4way.c \
|
||||
@@ -175,11 +170,11 @@ cpuminer_SOURCES = \
|
||||
algo/ripemd/lbry-4way.c \
|
||||
algo/scrypt/scrypt.c \
|
||||
algo/scrypt/neoscrypt.c \
|
||||
algo/scrypt/pluck.c \
|
||||
algo/sha/sph_sha2.c \
|
||||
algo/sha/sph_sha2big.c \
|
||||
algo/sha/sha256-hash-4way.c \
|
||||
algo/sha/sha512-hash-4way.c \
|
||||
algo/sha/hmac-sha256-hash.c \
|
||||
algo/sha/sha2.c \
|
||||
algo/sha/sha256t-gate.c \
|
||||
algo/sha/sha256t-4way.c \
|
||||
@@ -193,7 +188,6 @@ cpuminer_SOURCES = \
|
||||
algo/shavite/shavite-hash-2way.c \
|
||||
algo/shavite/shavite-hash-4way.c \
|
||||
algo/shavite/shavite.c \
|
||||
algo/simd/sph_simd.c \
|
||||
algo/simd/nist.c \
|
||||
algo/simd/vector.c \
|
||||
algo/simd/simd-hash-2way.c \
|
||||
@@ -231,7 +225,6 @@ cpuminer_SOURCES = \
|
||||
algo/x11/timetravel10-gate.c \
|
||||
algo/x11/timetravel10.c \
|
||||
algo/x11/timetravel10-4way.c \
|
||||
algo/x11/fresh.c \
|
||||
algo/x11/x11evo.c \
|
||||
algo/x11/x11evo-4way.c \
|
||||
algo/x11/x11evo-gate.c \
|
||||
@@ -250,7 +243,6 @@ cpuminer_SOURCES = \
|
||||
algo/x13/skunk-gate.c \
|
||||
algo/x13/skunk-4way.c \
|
||||
algo/x13/skunk.c \
|
||||
algo/x13/drop.c \
|
||||
algo/x13/x13bcd-4way.c \
|
||||
algo/x13/x13bcd.c \
|
||||
algo/x14/x14-gate.c \
|
||||
@@ -285,19 +277,17 @@ cpuminer_SOURCES = \
|
||||
algo/x17/sonoa-gate.c \
|
||||
algo/x17/sonoa-4way.c \
|
||||
algo/x17/sonoa.c \
|
||||
algo/x20/x20r.c \
|
||||
algo/x22/x22i-4way.c \
|
||||
algo/x22/x22i.c \
|
||||
algo/x22/x22i-gate.c \
|
||||
algo/x22/x25x.c \
|
||||
algo/x22/x25x-4way.c \
|
||||
algo/yescrypt/yescrypt.c \
|
||||
algo/yescrypt/sha256_Y.c \
|
||||
algo/yescrypt/yescrypt-best.c \
|
||||
algo/yespower/yespower-gate.c \
|
||||
algo/yespower/yespower-blake2b.c \
|
||||
algo/yespower/crypto/blake2b-yp.c \
|
||||
algo/yespower/sha256_p.c \
|
||||
algo/yespower/yescrypt-r8g.c \
|
||||
algo/yespower/yespower-opt.c
|
||||
|
||||
disable_flags =
|
||||
|
||||
@@ -53,7 +53,6 @@ Supported Algorithms
|
||||
argon2d500 argon2d-dyn, Dynamic (DYN)
|
||||
argon2d4096 argon2d-uis, Unitus, (UIS)
|
||||
axiom Shabal-256 MemoHash
|
||||
bastion
|
||||
blake Blake-256 (SFR)
|
||||
blake2b Blake2b 256
|
||||
blake2s Blake-2 S
|
||||
@@ -64,10 +63,7 @@ Supported Algorithms
|
||||
decred
|
||||
deep Deepcoin (DCN)
|
||||
dmd-gr Diamond-Groestl
|
||||
drop Dropcoin
|
||||
fresh Fresh
|
||||
groestl Groestl coin
|
||||
heavy Heavy
|
||||
hex x16r-hex
|
||||
hmq1725 Espers
|
||||
hodl Hodlcoin
|
||||
@@ -97,10 +93,10 @@ Supported Algorithms
|
||||
qubit Qubit
|
||||
scrypt scrypt(1024, 1, 1) (default)
|
||||
scrypt:N scrypt(N, 1, 1)
|
||||
scryptjane:nf
|
||||
sha256d Double SHA-256
|
||||
sha256q Quad SHA-256, Pyrite (PYE)
|
||||
sha256t Triple SHA-256, Onecoin (OC)
|
||||
sha3d Double keccak256 (BSHA3)
|
||||
shavite3 Shavite3
|
||||
skein Skein+Sha (Skeincoin)
|
||||
skein2 Double Skein (Woodcoin)
|
||||
@@ -134,6 +130,7 @@ Supported Algorithms
|
||||
xevan Bitsend (BSD)
|
||||
yescrypt Globalboost-Y (BSTY)
|
||||
yescryptr8 BitZeny (ZNY)
|
||||
yescryptr8g Koto (KOTO)
|
||||
yescryptr16 Eli
|
||||
yescryptr32 WAVI
|
||||
yespower Cryply
|
||||
|
||||
@@ -65,6 +65,50 @@ If not what makes it happen or not happen?
|
||||
Change Log
|
||||
----------
|
||||
|
||||
v3.11.9
|
||||
|
||||
Fixed x16r invalid shares when Luffa was first in hash order.
|
||||
|
||||
New startup message for status of stratum connection, API & extranonce.
|
||||
|
||||
New log report for CPU temperature, frequency of fastest and slowest cores.
|
||||
|
||||
Compile time is a little shorter and binary file size a little smaller
|
||||
using conditional compilation..
|
||||
|
||||
Removed code for Bastion, Drop, Heavy, Luffa an Pluck algos and other unused
|
||||
code.
|
||||
|
||||
v3.11.8
|
||||
|
||||
Fixed network hashrate showing incorrect data, should be close now.
|
||||
|
||||
Fixed compile errors when using GCC 10 with default flag -fno-common.
|
||||
|
||||
Faster x16r, x16rv2, x16rt, x16s, x21s, veil, hex with midstate prehash.
|
||||
|
||||
Decoupled sapling usage from block version 5 in yescryptr8g.
|
||||
|
||||
More detailed data reporting for low difficulty rejected shares.
|
||||
|
||||
v3.11.7
|
||||
|
||||
Added yescryptr8g algo for KOTO, including support for block version 5.
|
||||
|
||||
Added sha3d algo for BSHA3.
|
||||
|
||||
Removed memcmp and clean_job checks from get_new_work, now only check job_id.
|
||||
|
||||
Small improvement to sha512 and sha256 parallel implementations that don't
|
||||
use SHA.
|
||||
|
||||
v3.11.6
|
||||
|
||||
Fixed CPU temperature regression from v3.11.5.
|
||||
|
||||
More improvements to share log. More compact, highlight incremented counter,
|
||||
block height when solved, job id when stale.
|
||||
|
||||
v3.11.5
|
||||
|
||||
Fixed AVX512 detection that could cause compilation errors on CPUs
|
||||
|
||||
@@ -162,7 +162,6 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
case ALGO_ARGON2D500: register_argon2d_dyn_algo ( gate ); break;
|
||||
case ALGO_ARGON2D4096: register_argon2d4096_algo ( gate ); break;
|
||||
case ALGO_AXIOM: register_axiom_algo ( gate ); break;
|
||||
case ALGO_BASTION: register_bastion_algo ( gate ); break;
|
||||
case ALGO_BLAKE: register_blake_algo ( gate ); break;
|
||||
case ALGO_BLAKE2B: register_blake2b_algo ( gate ); break;
|
||||
case ALGO_BLAKE2S: register_blake2s_algo ( gate ); break;
|
||||
@@ -175,10 +174,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
case ALGO_DECRED: register_decred_algo ( gate ); break;
|
||||
case ALGO_DEEP: register_deep_algo ( gate ); break;
|
||||
case ALGO_DMD_GR: register_dmd_gr_algo ( gate ); break;
|
||||
case ALGO_DROP: register_drop_algo ( gate ); break;
|
||||
case ALGO_FRESH: register_fresh_algo ( gate ); break;
|
||||
case ALGO_GROESTL: register_groestl_algo ( gate ); break;
|
||||
case ALGO_HEAVY: register_heavy_algo ( gate ); break;
|
||||
case ALGO_HEX: register_hex_algo ( gate ); break;
|
||||
case ALGO_HMQ1725: register_hmq1725_algo ( gate ); break;
|
||||
case ALGO_HODL: register_hodl_algo ( gate ); break;
|
||||
@@ -186,7 +182,6 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
case ALGO_KECCAK: register_keccak_algo ( gate ); break;
|
||||
case ALGO_KECCAKC: register_keccakc_algo ( gate ); break;
|
||||
case ALGO_LBRY: register_lbry_algo ( gate ); break;
|
||||
case ALGO_LUFFA: register_luffa_algo ( gate ); break;
|
||||
case ALGO_LYRA2H: register_lyra2h_algo ( gate ); break;
|
||||
case ALGO_LYRA2RE: register_lyra2re_algo ( gate ); break;
|
||||
case ALGO_LYRA2REV2: register_lyra2rev2_algo ( gate ); break;
|
||||
@@ -200,7 +195,6 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
case ALGO_PENTABLAKE: register_pentablake_algo ( gate ); break;
|
||||
case ALGO_PHI1612: register_phi1612_algo ( gate ); break;
|
||||
case ALGO_PHI2: register_phi2_algo ( gate ); break;
|
||||
case ALGO_PLUCK: register_pluck_algo ( gate ); break;
|
||||
case ALGO_POLYTIMOS: register_polytimos_algo ( gate ); break;
|
||||
case ALGO_POWER2B: register_power2b_algo ( gate ); break;
|
||||
case ALGO_QUARK: register_quark_algo ( gate ); break;
|
||||
@@ -209,6 +203,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
case ALGO_SHA256D: register_sha256d_algo ( gate ); break;
|
||||
case ALGO_SHA256Q: register_sha256q_algo ( gate ); break;
|
||||
case ALGO_SHA256T: register_sha256t_algo ( gate ); break;
|
||||
case ALGO_SHA3D: register_sha3d_algo ( gate ); break;
|
||||
case ALGO_SHAVITE3: register_shavite_algo ( gate ); break;
|
||||
case ALGO_SKEIN: register_skein_algo ( gate ); break;
|
||||
case ALGO_SKEIN2: register_skein2_algo ( gate ); break;
|
||||
@@ -247,6 +242,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
*/
|
||||
case ALGO_YESCRYPT: register_yescrypt_algo ( gate ); break;
|
||||
case ALGO_YESCRYPTR8: register_yescryptr8_algo ( gate ); break;
|
||||
case ALGO_YESCRYPTR8G: register_yescryptr8g_algo ( gate ); break;
|
||||
case ALGO_YESCRYPTR16: register_yescryptr16_algo ( gate ); break;
|
||||
case ALGO_YESCRYPTR32: register_yescryptr32_algo ( gate ); break;
|
||||
case ALGO_YESPOWER: register_yespower_algo ( gate ); break;
|
||||
@@ -273,10 +269,6 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
|
||||
// override std defaults with jr2 defaults
|
||||
bool register_json_rpc2( algo_gate_t *gate )
|
||||
{
|
||||
applog(LOG_WARNING,"\nCryptonight algorithm and variants are no longer");
|
||||
applog(LOG_WARNING,"supported by cpuminer-opt. Shares submitted will");
|
||||
applog(LOG_WARNING,"likely be rejected. Proceed at your own risk.\n");
|
||||
|
||||
// gate->wait_for_diff = (void*)&do_nothing;
|
||||
gate->get_new_work = (void*)&jr2_get_new_work;
|
||||
gate->get_nonceptr = (void*)&jr2_get_nonceptr;
|
||||
@@ -358,7 +350,7 @@ void get_algo_alias( char** algo_or_alias )
|
||||
if ( !strcasecmp( *algo_or_alias, algo_alias_map[i][ ALIAS ] ) )
|
||||
{
|
||||
// found valid alias, return proper name
|
||||
*algo_or_alias = (char* const)( algo_alias_map[i][ PROPER ] );
|
||||
*algo_or_alias = (const char*)( algo_alias_map[i][ PROPER ] );
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -121,54 +121,55 @@ void ( *hash_suw ) ( void*, const void* );
|
||||
|
||||
// Allocate thread local buffers and other initialization specific to miner
|
||||
// threads.
|
||||
bool ( *miner_thread_init ) ( int );
|
||||
bool ( *miner_thread_init ) ( int );
|
||||
|
||||
// Generate global blockheader from stratum data.
|
||||
void ( *stratum_gen_work ) ( struct stratum_ctx*, struct work* );
|
||||
void ( *stratum_gen_work ) ( struct stratum_ctx*, struct work* );
|
||||
|
||||
// Get thread local copy of blockheader with unique nonce.
|
||||
void ( *get_new_work ) ( struct work*, struct work*, int, uint32_t*,
|
||||
bool );
|
||||
void ( *get_new_work ) ( struct work*, struct work*, int, uint32_t* );
|
||||
|
||||
// Return pointer to nonce in blockheader.
|
||||
uint32_t *( *get_nonceptr ) ( uint32_t* );
|
||||
uint32_t *( *get_nonceptr ) ( uint32_t* );
|
||||
|
||||
// Decode getwork blockheader
|
||||
bool ( *work_decode ) ( const json_t*, struct work* );
|
||||
bool ( *work_decode ) ( const json_t*, struct work* );
|
||||
|
||||
// Extra getwork data
|
||||
void ( *decode_extra_data ) ( struct work*, uint64_t* );
|
||||
void ( *decode_extra_data ) ( struct work*, uint64_t* );
|
||||
|
||||
bool ( *submit_getwork_result ) ( CURL*, struct work* );
|
||||
bool ( *submit_getwork_result ) ( CURL*, struct work* );
|
||||
|
||||
void ( *gen_merkle_root ) ( char*, struct stratum_ctx* );
|
||||
void ( *gen_merkle_root ) ( char*, struct stratum_ctx* );
|
||||
|
||||
// Increment extranonce
|
||||
void ( *build_extraheader ) ( struct work*, struct stratum_ctx* );
|
||||
void ( *build_extraheader ) ( struct work*, struct stratum_ctx* );
|
||||
|
||||
void ( *build_block_header ) ( struct work*, uint32_t, uint32_t*,
|
||||
uint32_t*, uint32_t, uint32_t,
|
||||
unsigned char* );
|
||||
|
||||
void ( *build_block_header ) ( struct work*, uint32_t, uint32_t*,
|
||||
uint32_t*, uint32_t, uint32_t );
|
||||
// Build mining.submit message
|
||||
void ( *build_stratum_request ) ( char*, struct work*, struct stratum_ctx* );
|
||||
void ( *build_stratum_request ) ( char*, struct work*, struct stratum_ctx* );
|
||||
|
||||
char* ( *malloc_txs_request ) ( struct work* );
|
||||
char* ( *malloc_txs_request ) ( struct work* );
|
||||
|
||||
// Big or little
|
||||
void ( *set_work_data_endian ) ( struct work* );
|
||||
void ( *set_work_data_endian ) ( struct work* );
|
||||
|
||||
double ( *calc_network_diff ) ( struct work* );
|
||||
double ( *calc_network_diff ) ( struct work* );
|
||||
|
||||
// Wait for first work
|
||||
bool ( *ready_to_mine ) ( struct work*, struct stratum_ctx*, int );
|
||||
bool ( *ready_to_mine ) ( struct work*, struct stratum_ctx*, int );
|
||||
|
||||
// Diverge mining threads
|
||||
bool ( *do_this_thread ) ( int );
|
||||
bool ( *do_this_thread ) ( int );
|
||||
|
||||
// After do_this_thread
|
||||
void ( *resync_threads ) ( struct work* );
|
||||
void ( *resync_threads ) ( struct work* );
|
||||
|
||||
json_t* (*longpoll_rpc_call) ( CURL*, int*, char* );
|
||||
bool ( *stratum_handle_response )( json_t* );
|
||||
json_t* (*longpoll_rpc_call) ( CURL*, int*, char* );
|
||||
bool ( *stratum_handle_response ) ( json_t* );
|
||||
set_t optimizations;
|
||||
int ( *get_work_data_size ) ();
|
||||
int ntime_index;
|
||||
@@ -225,7 +226,7 @@ uint32_t *std_get_nonceptr( uint32_t *work_data );
|
||||
uint32_t *jr2_get_nonceptr( uint32_t *work_data );
|
||||
|
||||
void std_get_new_work( struct work *work, struct work *g_work, int thr_id,
|
||||
uint32_t* end_nonce_ptr, bool clean_job );
|
||||
uint32_t* end_nonce_ptr );
|
||||
void jr2_get_new_work( struct work *work, struct work *g_work, int thr_id,
|
||||
uint32_t* end_nonce_ptr );
|
||||
|
||||
@@ -256,7 +257,8 @@ double std_calc_network_diff( struct work *work );
|
||||
|
||||
void std_build_block_header( struct work* g_work, uint32_t version,
|
||||
uint32_t *prevhash, uint32_t *merkle_root,
|
||||
uint32_t ntime, uint32_t nbits );
|
||||
uint32_t ntime, uint32_t nbits,
|
||||
unsigned char *final_sapling_hash );
|
||||
|
||||
void std_build_extraheader( struct work *work, struct stratum_ctx *sctx );
|
||||
|
||||
|
||||
@@ -1,4 +1,5 @@
|
||||
#include "argon2d-gate.h"
|
||||
#include "simd-utils.h"
|
||||
#include "argon2d/argon2.h"
|
||||
|
||||
static const size_t INPUT_BYTES = 80; // Lenth of a block header in bytes. Input Length = Salt Length (salt = input)
|
||||
@@ -36,7 +37,7 @@ void argon2d_crds_hash( void *output, const void *input )
|
||||
int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t _ALIGN(64) endiandata[20];
|
||||
uint32_t _ALIGN(64) edata[20];
|
||||
uint32_t _ALIGN(64) hash[8];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
@@ -45,11 +46,11 @@ int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
uint32_t nonce = first_nonce;
|
||||
|
||||
swab32_array( endiandata, pdata, 20 );
|
||||
swab32_array( edata, pdata, 20 );
|
||||
|
||||
do {
|
||||
be32enc(&endiandata[19], nonce);
|
||||
argon2d_crds_hash( hash, endiandata );
|
||||
be32enc(&edata[19], nonce);
|
||||
argon2d_crds_hash( hash, edata );
|
||||
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) && !opt_benchmark )
|
||||
{
|
||||
pdata[19] = nonce;
|
||||
@@ -103,31 +104,32 @@ void argon2d_dyn_hash( void *output, const void *input )
|
||||
int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t _ALIGN(64) endiandata[20];
|
||||
uint32_t _ALIGN(64) edata[20];
|
||||
uint32_t _ALIGN(64) hash[8];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const int thr_id = mythr->id;
|
||||
const uint32_t first_nonce = (const uint32_t)pdata[19];
|
||||
const uint32_t last_nonce = (const uint32_t)max_nonce;
|
||||
uint32_t nonce = first_nonce;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
swab32_array( endiandata, pdata, 20 );
|
||||
|
||||
mm128_bswap32_80( edata, pdata );
|
||||
do
|
||||
{
|
||||
be32enc(&endiandata[19], nonce);
|
||||
argon2d_dyn_hash( hash, endiandata );
|
||||
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) && !opt_benchmark )
|
||||
edata[19] = nonce;
|
||||
argon2d_dyn_hash( hash, edata );
|
||||
if ( unlikely( valid_hash( (uint64_t*)hash, (uint64_t*)ptarget )
|
||||
&& !bench ) )
|
||||
{
|
||||
pdata[19] = nonce;
|
||||
pdata[19] = bswap_32( nonce );;
|
||||
submit_solution( work, hash, mythr );
|
||||
}
|
||||
nonce++;
|
||||
} while (nonce < max_nonce && !work_restart[thr_id].restart);
|
||||
} while ( likely( nonce < last_nonce && !work_restart[thr_id].restart ) );
|
||||
|
||||
pdata[19] = nonce;
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
*hashes_done = pdata[19] - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
@@ -146,36 +148,34 @@ int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t _ALIGN(64) vhash[8];
|
||||
uint32_t _ALIGN(64) endiandata[20];
|
||||
uint32_t _ALIGN(64) edata[20];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = (const uint32_t)max_nonce;
|
||||
uint32_t n = first_nonce;
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
uint32_t t_cost = 1; // 1 iteration
|
||||
uint32_t m_cost = 4096; // use 4MB
|
||||
uint32_t parallelism = 1; // 1 thread, 2 lanes
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
for ( int i = 0; i < 19; i++ )
|
||||
be32enc( &endiandata[i], pdata[i] );
|
||||
mm128_bswap32_80( edata, pdata );
|
||||
|
||||
do {
|
||||
be32enc( &endiandata[19], n );
|
||||
argon2d_hash_raw( t_cost, m_cost, parallelism, (char*) endiandata, 80,
|
||||
(char*) endiandata, 80, (char*) vhash, 32, ARGON2_VERSION_13 );
|
||||
if ( vhash[7] < Htarg && fulltest( vhash, ptarget ) && !opt_benchmark )
|
||||
edata[19] = n;
|
||||
argon2d_hash_raw( t_cost, m_cost, parallelism, (char*) edata, 80,
|
||||
(char*) edata, 80, (char*) vhash, 32, ARGON2_VERSION_13 );
|
||||
if ( unlikely( valid_hash( vhash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = n;
|
||||
be32enc( &pdata[19], n );
|
||||
submit_solution( work, vhash, mythr );
|
||||
}
|
||||
n++;
|
||||
} while ( likely( n < last_nonce && !work_restart[thr_id].restart ) );
|
||||
|
||||
} while (n < max_nonce && !work_restart[thr_id].restart);
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
*hashes_done = n - first_nonce;
|
||||
pdata[19] = n;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
@@ -33,6 +33,8 @@
|
||||
|
||||
#include "blake2b-hash-4way.h"
|
||||
|
||||
#if defined(__AVX2__)
|
||||
|
||||
static const uint8_t sigma[12][16] =
|
||||
{
|
||||
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
|
||||
@@ -203,9 +205,9 @@ void blake2b_8way_final( blake2b_8way_ctx *ctx, void *out )
|
||||
casti_m512i( out, 3 ) = ctx->h[3];
|
||||
}
|
||||
|
||||
#endif
|
||||
#endif // AVX512
|
||||
|
||||
#if defined(__AVX2__)
|
||||
// AVX2
|
||||
|
||||
// G Mixing function.
|
||||
|
||||
@@ -369,4 +371,4 @@ void blake2b_4way_final( blake2b_4way_ctx *ctx, void *out )
|
||||
casti_m256i( out, 3 ) = ctx->h[3];
|
||||
}
|
||||
|
||||
#endif
|
||||
#endif // AVX2
|
||||
|
||||
@@ -4,6 +4,9 @@
|
||||
*/
|
||||
|
||||
#include "blake2b-gate.h"
|
||||
|
||||
#if !defined(BLAKE2B_8WAY) && !defined(BLAKE2B_4WAY)
|
||||
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
#include "algo/blake/sph_blake2b.h"
|
||||
@@ -58,3 +61,4 @@ int scanhash_blake2b( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,5 +1,7 @@
|
||||
#include "blake2s-gate.h"
|
||||
|
||||
#if !defined(BLAKE2S_16WAY) && !defined(BLAKE2S_8WAY) && !defined(BLAKE2S)
|
||||
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
|
||||
@@ -70,3 +72,4 @@ int scanhash_blake2s( struct work *work,
|
||||
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "blakecoin-gate.h"
|
||||
|
||||
#if !defined(BLAKECOIN_8WAY) && !defined(BLAKECOIN_4WAY)
|
||||
|
||||
#define BLAKE32_ROUNDS 8
|
||||
#include "sph_blake.h"
|
||||
|
||||
@@ -93,3 +96,4 @@ int scanhash_blakecoin( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "decred-gate.h"
|
||||
|
||||
#if !defined(DECRED_8WAY) && !defined(DECRED_4WAY)
|
||||
|
||||
#include "sph_blake.h"
|
||||
|
||||
#include <string.h>
|
||||
@@ -275,3 +278,5 @@ bool register_decred_algo( algo_gate_t* gate )
|
||||
return true;
|
||||
}
|
||||
*/
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "pentablake-gate.h"
|
||||
|
||||
#if !defined(PENTABLAKE_8WAY) && !defined(PENTABLAKE_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -111,3 +114,4 @@ int scanhash_pentablake( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,5 +1,7 @@
|
||||
#include "algo-gate-api.h"
|
||||
|
||||
#if !defined(BMW512_8WAY) && !defined(BMW512_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
@@ -50,4 +52,4 @@ int scanhash_bmw512( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -48,6 +48,8 @@ extern "C"{
|
||||
#pragma warning (disable: 4146)
|
||||
#endif
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
static const sph_u32 IV224[] = {
|
||||
SPH_C32(0x00010203), SPH_C32(0x04050607),
|
||||
SPH_C32(0x08090A0B), SPH_C32(0x0C0D0E0F),
|
||||
@@ -70,6 +72,8 @@ static const sph_u32 IV256[] = {
|
||||
SPH_C32(0x78797A7B), SPH_C32(0x7C7D7E7F)
|
||||
};
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
static const sph_u64 IV384[] = {
|
||||
@@ -135,6 +139,8 @@ static const sph_u64 IV512[] = {
|
||||
#define M16_30 14, 15, 1, 2, 5, 8, 9
|
||||
#define M16_31 15, 16, 2, 3, 6, 9, 10
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
#define ss0(x) (((x) >> 1) ^ SPH_T32((x) << 3) \
|
||||
^ SPH_ROTL32(x, 4) ^ SPH_ROTL32(x, 19))
|
||||
#define ss1(x) (((x) >> 1) ^ SPH_T32((x) << 2) \
|
||||
@@ -189,6 +195,8 @@ static const sph_u64 IV512[] = {
|
||||
#define expand2s_(qf, mf, hf, i16, ix, iy) \
|
||||
expand2s_inner LPAR qf, mf, hf, i16, ix, iy)
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
#define sb0(x) (((x) >> 1) ^ SPH_T64((x) << 3) \
|
||||
@@ -291,6 +299,8 @@ static const sph_u64 Kb_tab[] = {
|
||||
tt((M(i0) ^ H(i0)) op01 (M(i1) ^ H(i1)) op12 (M(i2) ^ H(i2)) \
|
||||
op23 (M(i3) ^ H(i3)) op34 (M(i4) ^ H(i4)))
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
#define Ws0 MAKE_W(SPH_T32, 5, -, 7, +, 10, +, 13, +, 14)
|
||||
#define Ws1 MAKE_W(SPH_T32, 6, -, 8, +, 11, +, 14, -, 15)
|
||||
#define Ws2 MAKE_W(SPH_T32, 0, +, 7, +, 9, -, 12, +, 15)
|
||||
@@ -407,6 +417,8 @@ static const sph_u64 Kb_tab[] = {
|
||||
|
||||
#define Qs(j) (qt[j])
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
#define Wb0 MAKE_W(SPH_T64, 5, -, 7, +, 10, +, 13, +, 14)
|
||||
@@ -557,7 +569,6 @@ static const sph_u64 Kb_tab[] = {
|
||||
+ ((xl >> 2) ^ qf(22) ^ qf(15))); \
|
||||
} while (0)
|
||||
|
||||
#define FOLDs FOLD(sph_u32, MAKE_Qs, SPH_T32, SPH_ROTL32, M, Qs, dH)
|
||||
|
||||
#if SPH_64
|
||||
|
||||
@@ -565,6 +576,10 @@ static const sph_u64 Kb_tab[] = {
|
||||
|
||||
#endif
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
#define FOLDs FOLD(sph_u32, MAKE_Qs, SPH_T32, SPH_ROTL32, M, Qs, dH)
|
||||
|
||||
static void
|
||||
compress_small(const unsigned char *data, const sph_u32 h[16], sph_u32 dh[16])
|
||||
{
|
||||
@@ -711,6 +726,8 @@ bmw32_close(sph_bmw_small_context *sc, unsigned ub, unsigned n,
|
||||
sph_enc32le(out + 4 * u, h1[v]);
|
||||
}
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
static void
|
||||
@@ -840,6 +857,8 @@ bmw64_close(sph_bmw_big_context *sc, unsigned ub, unsigned n,
|
||||
|
||||
#endif
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
/* see sph_bmw.h */
|
||||
void
|
||||
sph_bmw224_init(void *cc)
|
||||
@@ -898,6 +917,8 @@ sph_bmw256_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
|
||||
// sph_bmw256_init(cc);
|
||||
}
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
/* see sph_bmw.h */
|
||||
|
||||
@@ -77,6 +77,9 @@ extern "C"{
|
||||
* computation can be cloned by copying the context (e.g. with a simple
|
||||
* <code>memcpy()</code>).
|
||||
*/
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
typedef struct {
|
||||
#ifndef DOXYGEN_IGNORE
|
||||
unsigned char buf[64]; /* first field, for alignment */
|
||||
@@ -102,6 +105,8 @@ typedef sph_bmw_small_context sph_bmw224_context;
|
||||
*/
|
||||
typedef sph_bmw_small_context sph_bmw256_context;
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
/**
|
||||
@@ -137,6 +142,8 @@ typedef sph_bmw_big_context sph_bmw512_context;
|
||||
|
||||
#endif
|
||||
|
||||
#if !defined(__AVX2__)
|
||||
|
||||
/**
|
||||
* Initialize a BMW-224 context. This process performs no memory allocation.
|
||||
*
|
||||
@@ -227,6 +234,8 @@ void sph_bmw256_close(void *cc, void *dst);
|
||||
void sph_bmw256_addbits_and_close(
|
||||
void *cc, unsigned ub, unsigned n, void *dst);
|
||||
|
||||
#endif // !AVX2
|
||||
|
||||
#if SPH_64
|
||||
|
||||
/**
|
||||
|
||||
@@ -358,6 +358,9 @@ int scanhash_cryptolight( struct work *work,
|
||||
|
||||
bool register_cryptolight_algo( algo_gate_t* gate )
|
||||
{
|
||||
applog(LOG_WARNING,"Cryptonight algorithm and variants are no longer");
|
||||
applog(LOG_WARNING,"supported by cpuminer-opt. Shares submitted will");
|
||||
applog(LOG_WARNING,"likely be rejected. Proceed at your own risk.\n");
|
||||
register_json_rpc2( gate );
|
||||
gate->optimizations = SSE2_OPT | AES_OPT;
|
||||
gate->scanhash = (void*)&scanhash_cryptolight;
|
||||
|
||||
@@ -105,6 +105,9 @@ int scanhash_cryptonight( struct work *work, uint32_t max_nonce,
|
||||
|
||||
bool register_cryptonight_algo( algo_gate_t* gate )
|
||||
{
|
||||
applog(LOG_WARNING,"Cryptonight algorithm and variants are no longer");
|
||||
applog(LOG_WARNING,"supported by cpuminer-opt. Shares submitted will");
|
||||
applog(LOG_WARNING,"likely be rejected. Proceed at your own risk.\n");
|
||||
cryptonightV7 = false;
|
||||
register_json_rpc2( gate );
|
||||
gate->optimizations = SSE2_OPT | AES_OPT;
|
||||
@@ -116,6 +119,9 @@ bool register_cryptonight_algo( algo_gate_t* gate )
|
||||
|
||||
bool register_cryptonightv7_algo( algo_gate_t* gate )
|
||||
{
|
||||
applog(LOG_WARNING,"Cryptonight algorithm and variants are no longer");
|
||||
applog(LOG_WARNING,"supported by cpuminer-opt. Shares submitted will");
|
||||
applog(LOG_WARNING,"likely be rejected. Proceed at your own risk.\n");
|
||||
cryptonightV7 = true;
|
||||
register_json_rpc2( gate );
|
||||
gate->optimizations = SSE2_OPT | AES_OPT;
|
||||
|
||||
@@ -230,11 +230,10 @@ int cubehashDigest( cubehashParam *sp, byte *digest )
|
||||
|
||||
// pos is zero for 64 byte data, 1 for 80 byte data.
|
||||
sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ],
|
||||
_mm_set_epi8( 0,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 ) );
|
||||
m128_const_64( 0, 0x80 ) );
|
||||
transform( sp );
|
||||
|
||||
sp->x[7] = _mm_xor_si128( sp->x[7], _mm_set_epi32( 1,0,0,0 ) );
|
||||
sp->x[7] = _mm_xor_si128( sp->x[7], m128_const_64( 0x100000000, 0 ) );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
@@ -276,11 +275,89 @@ int cubehashUpdateDigest( cubehashParam *sp, byte *digest,
|
||||
|
||||
// pos is zero for 64 byte data, 1 for 80 byte data.
|
||||
sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ],
|
||||
_mm_set_epi8( 0,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 ) );
|
||||
m128_const_64( 0, 0x80 ) );
|
||||
transform( sp );
|
||||
|
||||
sp->x[7] = _mm_xor_si128( sp->x[7], _mm_set_epi32( 1,0,0,0 ) );
|
||||
sp->x[7] = _mm_xor_si128( sp->x[7], m128_const_64( 0x100000000, 0 ) );
|
||||
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
|
||||
for ( i = 0; i < sp->hashlen; i++ )
|
||||
hash[i] = sp->x[i];
|
||||
|
||||
return SUCCESS;
|
||||
}
|
||||
|
||||
int cubehash_full( cubehashParam *sp, byte *digest, int hashbitlen,
|
||||
const byte *data, size_t size )
|
||||
{
|
||||
__m128i *x = (__m128i*)sp->x;
|
||||
sp->hashlen = hashbitlen/128;
|
||||
sp->blocksize = 32/16;
|
||||
sp->rounds = 16;
|
||||
sp->pos = 0;
|
||||
|
||||
if ( hashbitlen == 512 )
|
||||
{
|
||||
|
||||
x[0] = m128_const_64( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
|
||||
x[1] = m128_const_64( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
|
||||
x[2] = m128_const_64( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
|
||||
x[3] = m128_const_64( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
|
||||
x[4] = m128_const_64( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
|
||||
x[5] = m128_const_64( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
|
||||
x[6] = m128_const_64( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
|
||||
x[7] = m128_const_64( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
|
||||
}
|
||||
else
|
||||
{
|
||||
x[0] = m128_const_64( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
|
||||
x[1] = m128_const_64( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
|
||||
x[2] = m128_const_64( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
|
||||
x[3] = m128_const_64( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
|
||||
x[4] = m128_const_64( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
|
||||
x[5] = m128_const_64( 0x93CB628565C892FD, 0x5FA2560309392549 );
|
||||
x[6] = m128_const_64( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
|
||||
x[7] = m128_const_64( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
|
||||
}
|
||||
|
||||
|
||||
|
||||
|
||||
const int len = size / 16;
|
||||
const __m128i* in = (__m128i*)data;
|
||||
__m128i* hash = (__m128i*)digest;
|
||||
int i;
|
||||
|
||||
// It is assumed data is aligned to 256 bits and is a multiple of 128 bits.
|
||||
// Current usage sata is either 64 or 80 bytes.
|
||||
|
||||
for ( i = 0; i < len; i++ )
|
||||
{
|
||||
sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ], in[i] );
|
||||
sp->pos++;
|
||||
if ( sp->pos == sp->blocksize )
|
||||
{
|
||||
transform( sp );
|
||||
sp->pos = 0;
|
||||
}
|
||||
}
|
||||
|
||||
// pos is zero for 64 byte data, 1 for 80 byte data.
|
||||
sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ],
|
||||
m128_const_64( 0, 0x80 ) );
|
||||
transform( sp );
|
||||
|
||||
sp->x[7] = _mm_xor_si128( sp->x[7], m128_const_64( 0x100000000, 0 ) );
|
||||
|
||||
transform( sp );
|
||||
transform( sp );
|
||||
|
||||
@@ -19,7 +19,7 @@ struct _cubehashParam
|
||||
int rounds;
|
||||
int blocksize; // __m128i
|
||||
int pos; // number of __m128i read into x from current block
|
||||
__m128i _ALIGN(256) x[8]; // aligned for __m256i
|
||||
__m128i _ALIGN(64) x[8]; // aligned for __m256i
|
||||
};
|
||||
|
||||
typedef struct _cubehashParam cubehashParam;
|
||||
@@ -39,6 +39,9 @@ int cubehashDigest(cubehashParam* sp, byte *digest);
|
||||
int cubehashUpdateDigest( cubehashParam *sp, byte *digest, const byte *data,
|
||||
size_t size );
|
||||
|
||||
int cubehash_full( cubehashParam* sp, byte *digest, int hashbitlen,
|
||||
const byte *data, size_t size );
|
||||
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -36,6 +36,8 @@
|
||||
|
||||
#include "sph_echo.h"
|
||||
|
||||
#if !defined(__AES__)
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C"{
|
||||
#endif
|
||||
@@ -1028,4 +1030,5 @@ sph_echo512_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
|
||||
}
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
#endif
|
||||
#endif
|
||||
#endif // !AES
|
||||
|
||||
@@ -36,6 +36,8 @@
|
||||
#ifndef SPH_ECHO_H__
|
||||
#define SPH_ECHO_H__
|
||||
|
||||
#if !defined(__AES__)
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C"{
|
||||
#endif
|
||||
@@ -316,5 +318,5 @@ void sph_echo512_addbits_and_close(
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
#endif
|
||||
|
||||
#endif // !AES
|
||||
#endif
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@@ -1,3 +1,6 @@
|
||||
#if !defined GROESTL_INTR_AES_H__
|
||||
#define GROESTL_INTR_AES_H__
|
||||
|
||||
/* groestl-intr-aes.h Aug 2011
|
||||
*
|
||||
* Groestl implementation with intrinsics using ssse3, sse4.1, and aes
|
||||
@@ -11,16 +14,51 @@
|
||||
#include <wmmintrin.h>
|
||||
#include "hash-groestl.h"
|
||||
|
||||
/* global constants */
|
||||
__m128i ROUND_CONST_Lx;
|
||||
//__m128i ROUND_CONST_L0[ROUNDS512];
|
||||
//__m128i ROUND_CONST_L7[ROUNDS512];
|
||||
__m128i ROUND_CONST_P[ROUNDS1024];
|
||||
__m128i ROUND_CONST_Q[ROUNDS1024];
|
||||
__m128i TRANSP_MASK;
|
||||
__m128i SUBSH_MASK[8];
|
||||
__m128i ALL_1B;
|
||||
__m128i ALL_FF;
|
||||
static const __m128i round_const_p[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x7060504030201000, 0xf0e0d0c0b0a09080 },
|
||||
{ 0x7161514131211101, 0xf1e1d1c1b1a19181 },
|
||||
{ 0x7262524232221202, 0xf2e2d2c2b2a29282 },
|
||||
{ 0x7363534333231303, 0xf3e3d3c3b3a39383 },
|
||||
{ 0x7464544434241404, 0xf4e4d4c4b4a49484 },
|
||||
{ 0x7565554535251505, 0xf5e5d5c5b5a59585 },
|
||||
{ 0x7666564636261606, 0xf6e6d6c6b6a69686 },
|
||||
{ 0x7767574737271707, 0xf7e7d7c7b7a79787 },
|
||||
{ 0x7868584838281808, 0xf8e8d8c8b8a89888 },
|
||||
{ 0x7969594939291909, 0xf9e9d9c9b9a99989 },
|
||||
{ 0x7a6a5a4a3a2a1a0a, 0xfaeadacabaaa9a8a },
|
||||
{ 0x7b6b5b4b3b2b1b0b, 0xfbebdbcbbbab9b8b },
|
||||
{ 0x7c6c5c4c3c2c1c0c, 0xfcecdcccbcac9c8c },
|
||||
{ 0x7d6d5d4d3d2d1d0d, 0xfdedddcdbdad9d8d }
|
||||
};
|
||||
|
||||
static const __m128i round_const_q[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x8f9fafbfcfdfefff, 0x0f1f2f3f4f5f6f7f },
|
||||
{ 0x8e9eaebecedeeefe, 0x0e1e2e3e4e5e6e7e },
|
||||
{ 0x8d9dadbdcdddedfd, 0x0d1d2d3d4d5d6d7d },
|
||||
{ 0x8c9cacbcccdcecfc, 0x0c1c2c3c4c5c6c7c },
|
||||
{ 0x8b9babbbcbdbebfb, 0x0b1b2b3b4b5b6b7b },
|
||||
{ 0x8a9aaabacadaeafa, 0x0a1a2a3a4a5a6a7a },
|
||||
{ 0x8999a9b9c9d9e9f9, 0x0919293949596979 },
|
||||
{ 0x8898a8b8c8d8e8f8, 0x0818283848586878 },
|
||||
{ 0x8797a7b7c7d7e7f7, 0x0717273747576777 },
|
||||
{ 0x8696a6b6c6d6e6f6, 0x0616263646566676 },
|
||||
{ 0x8595a5b5c5d5e5f5, 0x0515253545556575 },
|
||||
{ 0x8494a4b4c4d4e4f4, 0x0414243444546474 },
|
||||
{ 0x8393a3b3c3d3e3f3, 0x0313233343536373 },
|
||||
{ 0x8292a2b2c2d2e2f2, 0x0212223242526272 }
|
||||
};
|
||||
|
||||
static const __m128i TRANSP_MASK = { 0x0d0509010c040800, 0x0f070b030e060a02 };
|
||||
static const __m128i SUBSH_MASK0 = { 0x0b0e0104070a0d00, 0x0306090c0f020508 };
|
||||
static const __m128i SUBSH_MASK1 = { 0x0c0f0205080b0e01, 0x04070a0d00030609 };
|
||||
static const __m128i SUBSH_MASK2 = { 0x0d000306090c0f02, 0x05080b0e0104070a };
|
||||
static const __m128i SUBSH_MASK3 = { 0x0e0104070a0d0003, 0x06090c0f0205080b };
|
||||
static const __m128i SUBSH_MASK4 = { 0x0f0205080b0e0104, 0x070a0d000306090c };
|
||||
static const __m128i SUBSH_MASK5 = { 0x000306090c0f0205, 0x080b0e0104070a0d };
|
||||
static const __m128i SUBSH_MASK6 = { 0x0104070a0d000306, 0x090c0f0205080b0e };
|
||||
static const __m128i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003 };
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
@@ -111,7 +149,7 @@ __m128i ALL_FF;
|
||||
\
|
||||
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
|
||||
/* compute w_i : add y_{i+4} */\
|
||||
b1 = ALL_1B;\
|
||||
b1 = m128_const1_64( 0x1b1b1b1b1b1b1b1b );\
|
||||
MUL2(a0, b0, b1);\
|
||||
a0 = _mm_xor_si128(a0, TEMP0);\
|
||||
MUL2(a1, b0, b1);\
|
||||
@@ -152,25 +190,6 @@ __m128i ALL_FF;
|
||||
}/*MixBytes*/
|
||||
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
ALL_FF = _mm_set_epi32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff);\
|
||||
ALL_1B = _mm_set_epi32(0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b);\
|
||||
TRANSP_MASK = _mm_set_epi32(0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800);\
|
||||
SUBSH_MASK[0] = _mm_set_epi32(0x0306090c, 0x0f020508, 0x0b0e0104, 0x070a0d00);\
|
||||
SUBSH_MASK[1] = _mm_set_epi32(0x04070a0d, 0x00030609, 0x0c0f0205, 0x080b0e01);\
|
||||
SUBSH_MASK[2] = _mm_set_epi32(0x05080b0e, 0x0104070a, 0x0d000306, 0x090c0f02);\
|
||||
SUBSH_MASK[3] = _mm_set_epi32(0x06090c0f, 0x0205080b, 0x0e010407, 0x0a0d0003);\
|
||||
SUBSH_MASK[4] = _mm_set_epi32(0x070a0d00, 0x0306090c, 0x0f020508, 0x0b0e0104);\
|
||||
SUBSH_MASK[5] = _mm_set_epi32(0x080b0e01, 0x04070a0d, 0x00030609, 0x0c0f0205);\
|
||||
SUBSH_MASK[6] = _mm_set_epi32(0x090c0f02, 0x05080b0e, 0x0104070a, 0x0d000306);\
|
||||
SUBSH_MASK[7] = _mm_set_epi32(0x0e010407, 0x0a0d0003, 0x06090c0f, 0x0205080b);\
|
||||
for(i = 0; i < ROUNDS1024; i++)\
|
||||
{\
|
||||
ROUND_CONST_P[i] = _mm_set_epi32(0xf0e0d0c0 ^ (i * 0x01010101), 0xb0a09080 ^ (i * 0x01010101), 0x70605040 ^ (i * 0x01010101), 0x30201000 ^ (i * 0x01010101));\
|
||||
ROUND_CONST_Q[i] = _mm_set_epi32(0x0f1f2f3f ^ (i * 0x01010101), 0x4f5f6f7f ^ (i * 0x01010101), 0x8f9fafbf ^ (i * 0x01010101), 0xcfdfefff ^ (i * 0x01010101));\
|
||||
}\
|
||||
}while(0);\
|
||||
|
||||
/* one round
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
@@ -194,30 +213,34 @@ __m128i ALL_FF;
|
||||
u8 round_counter = 0;\
|
||||
for(round_counter = 0; round_counter < 14; round_counter+=2) {\
|
||||
/* AddRoundConstant P1024 */\
|
||||
xmm8 = _mm_xor_si128(xmm8, (ROUND_CONST_P[round_counter]));\
|
||||
xmm8 = _mm_xor_si128( xmm8, \
|
||||
casti_m128i( round_const_p, round_counter ) ); \
|
||||
/* ShiftBytes P1024 + pre-AESENCLAST */\
|
||||
xmm8 = _mm_shuffle_epi8(xmm8, (SUBSH_MASK[0]));\
|
||||
xmm9 = _mm_shuffle_epi8(xmm9, (SUBSH_MASK[1]));\
|
||||
xmm10 = _mm_shuffle_epi8(xmm10, (SUBSH_MASK[2]));\
|
||||
xmm11 = _mm_shuffle_epi8(xmm11, (SUBSH_MASK[3]));\
|
||||
xmm12 = _mm_shuffle_epi8(xmm12, (SUBSH_MASK[4]));\
|
||||
xmm13 = _mm_shuffle_epi8(xmm13, (SUBSH_MASK[5]));\
|
||||
xmm14 = _mm_shuffle_epi8(xmm14, (SUBSH_MASK[6]));\
|
||||
xmm15 = _mm_shuffle_epi8(xmm15, (SUBSH_MASK[7]));\
|
||||
xmm8 = _mm_shuffle_epi8( xmm8, SUBSH_MASK0 ); \
|
||||
xmm9 = _mm_shuffle_epi8( xmm9, SUBSH_MASK1 ); \
|
||||
xmm10 = _mm_shuffle_epi8( xmm10, SUBSH_MASK2 ); \
|
||||
xmm11 = _mm_shuffle_epi8( xmm11, SUBSH_MASK3 ); \
|
||||
xmm12 = _mm_shuffle_epi8( xmm12, SUBSH_MASK4 ); \
|
||||
xmm13 = _mm_shuffle_epi8( xmm13, SUBSH_MASK5 ); \
|
||||
xmm14 = _mm_shuffle_epi8( xmm14, SUBSH_MASK6 ); \
|
||||
xmm15 = _mm_shuffle_epi8( xmm15, SUBSH_MASK7 ); \
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
SUBMIX( xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, \
|
||||
xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7 ); \
|
||||
\
|
||||
/* AddRoundConstant P1024 */\
|
||||
xmm0 = _mm_xor_si128(xmm0, (ROUND_CONST_P[round_counter+1]));\
|
||||
xmm0 = _mm_shuffle_epi8(xmm0, (SUBSH_MASK[0]));\
|
||||
xmm1 = _mm_shuffle_epi8(xmm1, (SUBSH_MASK[1]));\
|
||||
xmm2 = _mm_shuffle_epi8(xmm2, (SUBSH_MASK[2]));\
|
||||
xmm3 = _mm_shuffle_epi8(xmm3, (SUBSH_MASK[3]));\
|
||||
xmm4 = _mm_shuffle_epi8(xmm4, (SUBSH_MASK[4]));\
|
||||
xmm5 = _mm_shuffle_epi8(xmm5, (SUBSH_MASK[5]));\
|
||||
xmm6 = _mm_shuffle_epi8(xmm6, (SUBSH_MASK[6]));\
|
||||
xmm7 = _mm_shuffle_epi8(xmm7, (SUBSH_MASK[7]));\
|
||||
SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
xmm0 = _mm_xor_si128( xmm0, \
|
||||
casti_m128i( round_const_p, round_counter+1 ) ); \
|
||||
xmm0 = _mm_shuffle_epi8( xmm0, SUBSH_MASK0 ); \
|
||||
xmm1 = _mm_shuffle_epi8( xmm1, SUBSH_MASK1 ); \
|
||||
xmm2 = _mm_shuffle_epi8( xmm2, SUBSH_MASK2 ); \
|
||||
xmm3 = _mm_shuffle_epi8( xmm3, SUBSH_MASK3 ); \
|
||||
xmm4 = _mm_shuffle_epi8( xmm4, SUBSH_MASK4 ); \
|
||||
xmm5 = _mm_shuffle_epi8( xmm5, SUBSH_MASK5 ); \
|
||||
xmm6 = _mm_shuffle_epi8( xmm6, SUBSH_MASK6 ); \
|
||||
xmm7 = _mm_shuffle_epi8( xmm7, SUBSH_MASK7 ); \
|
||||
SUBMIX( xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, \
|
||||
xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15 ); \
|
||||
}\
|
||||
}
|
||||
|
||||
@@ -225,48 +248,52 @@ __m128i ALL_FF;
|
||||
u8 round_counter = 0;\
|
||||
for(round_counter = 0; round_counter < 14; round_counter+=2) {\
|
||||
/* AddRoundConstant Q1024 */\
|
||||
xmm1 = ALL_FF;\
|
||||
xmm8 = _mm_xor_si128(xmm8, xmm1);\
|
||||
xmm9 = _mm_xor_si128(xmm9, xmm1);\
|
||||
xmm10 = _mm_xor_si128(xmm10, xmm1);\
|
||||
xmm11 = _mm_xor_si128(xmm11, xmm1);\
|
||||
xmm12 = _mm_xor_si128(xmm12, xmm1);\
|
||||
xmm13 = _mm_xor_si128(xmm13, xmm1);\
|
||||
xmm14 = _mm_xor_si128(xmm14, xmm1);\
|
||||
xmm15 = _mm_xor_si128(xmm15, (ROUND_CONST_Q[round_counter]));\
|
||||
xmm1 = m128_neg1;\
|
||||
xmm8 = _mm_xor_si128( xmm8, xmm1 ); \
|
||||
xmm9 = _mm_xor_si128( xmm9, xmm1 ); \
|
||||
xmm10 = _mm_xor_si128( xmm10, xmm1 ); \
|
||||
xmm11 = _mm_xor_si128( xmm11, xmm1 ); \
|
||||
xmm12 = _mm_xor_si128( xmm12, xmm1 ); \
|
||||
xmm13 = _mm_xor_si128( xmm13, xmm1 ); \
|
||||
xmm14 = _mm_xor_si128( xmm14, xmm1 ); \
|
||||
xmm15 = _mm_xor_si128( xmm15, \
|
||||
casti_m128i( round_const_q, round_counter ) ); \
|
||||
/* ShiftBytes Q1024 + pre-AESENCLAST */\
|
||||
xmm8 = _mm_shuffle_epi8(xmm8, (SUBSH_MASK[1]));\
|
||||
xmm9 = _mm_shuffle_epi8(xmm9, (SUBSH_MASK[3]));\
|
||||
xmm10 = _mm_shuffle_epi8(xmm10, (SUBSH_MASK[5]));\
|
||||
xmm11 = _mm_shuffle_epi8(xmm11, (SUBSH_MASK[7]));\
|
||||
xmm12 = _mm_shuffle_epi8(xmm12, (SUBSH_MASK[0]));\
|
||||
xmm13 = _mm_shuffle_epi8(xmm13, (SUBSH_MASK[2]));\
|
||||
xmm14 = _mm_shuffle_epi8(xmm14, (SUBSH_MASK[4]));\
|
||||
xmm15 = _mm_shuffle_epi8(xmm15, (SUBSH_MASK[6]));\
|
||||
xmm8 = _mm_shuffle_epi8( xmm8, SUBSH_MASK1 ); \
|
||||
xmm9 = _mm_shuffle_epi8( xmm9, SUBSH_MASK3 ); \
|
||||
xmm10 = _mm_shuffle_epi8( xmm10, SUBSH_MASK5 ); \
|
||||
xmm11 = _mm_shuffle_epi8( xmm11, SUBSH_MASK7 ); \
|
||||
xmm12 = _mm_shuffle_epi8( xmm12, SUBSH_MASK0 ); \
|
||||
xmm13 = _mm_shuffle_epi8( xmm13, SUBSH_MASK2 ); \
|
||||
xmm14 = _mm_shuffle_epi8( xmm14, SUBSH_MASK4 ); \
|
||||
xmm15 = _mm_shuffle_epi8( xmm15, SUBSH_MASK6 ); \
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
SUBMIX( xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, \
|
||||
xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6 , xmm7 ); \
|
||||
\
|
||||
/* AddRoundConstant Q1024 */\
|
||||
xmm9 = ALL_FF;\
|
||||
xmm0 = _mm_xor_si128(xmm0, xmm9);\
|
||||
xmm1 = _mm_xor_si128(xmm1, xmm9);\
|
||||
xmm2 = _mm_xor_si128(xmm2, xmm9);\
|
||||
xmm3 = _mm_xor_si128(xmm3, xmm9);\
|
||||
xmm4 = _mm_xor_si128(xmm4, xmm9);\
|
||||
xmm5 = _mm_xor_si128(xmm5, xmm9);\
|
||||
xmm6 = _mm_xor_si128(xmm6, xmm9);\
|
||||
xmm7 = _mm_xor_si128(xmm7, (ROUND_CONST_Q[round_counter+1]));\
|
||||
xmm9 = m128_neg1;\
|
||||
xmm0 = _mm_xor_si128( xmm0, xmm9 ); \
|
||||
xmm1 = _mm_xor_si128( xmm1, xmm9 ); \
|
||||
xmm2 = _mm_xor_si128( xmm2, xmm9 ); \
|
||||
xmm3 = _mm_xor_si128( xmm3, xmm9 ); \
|
||||
xmm4 = _mm_xor_si128( xmm4, xmm9 ); \
|
||||
xmm5 = _mm_xor_si128( xmm5, xmm9 ); \
|
||||
xmm6 = _mm_xor_si128( xmm6, xmm9 ); \
|
||||
xmm7 = _mm_xor_si128( xmm7, \
|
||||
casti_m128i( round_const_q, round_counter+1 ) ); \
|
||||
/* ShiftBytes Q1024 + pre-AESENCLAST */\
|
||||
xmm0 = _mm_shuffle_epi8(xmm0, (SUBSH_MASK[1]));\
|
||||
xmm1 = _mm_shuffle_epi8(xmm1, (SUBSH_MASK[3]));\
|
||||
xmm2 = _mm_shuffle_epi8(xmm2, (SUBSH_MASK[5]));\
|
||||
xmm3 = _mm_shuffle_epi8(xmm3, (SUBSH_MASK[7]));\
|
||||
xmm4 = _mm_shuffle_epi8(xmm4, (SUBSH_MASK[0]));\
|
||||
xmm5 = _mm_shuffle_epi8(xmm5, (SUBSH_MASK[2]));\
|
||||
xmm6 = _mm_shuffle_epi8(xmm6, (SUBSH_MASK[4]));\
|
||||
xmm7 = _mm_shuffle_epi8(xmm7, (SUBSH_MASK[6]));\
|
||||
xmm0 = _mm_shuffle_epi8( xmm0, SUBSH_MASK1 ); \
|
||||
xmm1 = _mm_shuffle_epi8( xmm1, SUBSH_MASK3 ); \
|
||||
xmm2 = _mm_shuffle_epi8( xmm2, SUBSH_MASK5 ); \
|
||||
xmm3 = _mm_shuffle_epi8( xmm3, SUBSH_MASK7 ); \
|
||||
xmm4 = _mm_shuffle_epi8( xmm4, SUBSH_MASK0 ); \
|
||||
xmm5 = _mm_shuffle_epi8( xmm5, SUBSH_MASK2 ); \
|
||||
xmm6 = _mm_shuffle_epi8( xmm6, SUBSH_MASK4 ); \
|
||||
xmm7 = _mm_shuffle_epi8( xmm7, SUBSH_MASK6 ); \
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
SUBMIX( xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, \
|
||||
xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15 ); \
|
||||
}\
|
||||
}
|
||||
|
||||
@@ -278,7 +305,7 @@ __m128i ALL_FF;
|
||||
* clobbers: t0-t7
|
||||
*/
|
||||
#define Matrix_Transpose(i0, i1, i2, i3, i4, i5, i6, i7, t0, t1, t2, t3, t4, t5, t6, t7){\
|
||||
t0 = TRANSP_MASK;\
|
||||
t0 = TRANSP_MASK; \
|
||||
\
|
||||
i6 = _mm_shuffle_epi8(i6, t0);\
|
||||
i0 = _mm_shuffle_epi8(i0, t0);\
|
||||
@@ -366,7 +393,7 @@ __m128i ALL_FF;
|
||||
i4 = _mm_unpacklo_epi64(i4, i5);\
|
||||
t1 = _mm_unpackhi_epi64(t1, i5);\
|
||||
t2 = i6;\
|
||||
o0 = TRANSP_MASK;\
|
||||
o0 = TRANSP_MASK; \
|
||||
i6 = _mm_unpacklo_epi64(i6, i7);\
|
||||
t2 = _mm_unpackhi_epi64(t2, i7);\
|
||||
/* load transpose mask into a register, because it will be used 8 times */\
|
||||
@@ -607,3 +634,4 @@ void OF1024( __m128i* chaining )
|
||||
return;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@@ -1,10 +0,0 @@
|
||||
// specify assembly or intrinsics implementation
|
||||
//#define TASM
|
||||
#define TINTR
|
||||
|
||||
// Not to be confused with AVX512VAES
|
||||
#define VAES
|
||||
// #define VAVX
|
||||
// #define VVPERM
|
||||
|
||||
//#endif
|
||||
@@ -1,529 +0,0 @@
|
||||
/* groestl-asm-aes.h Aug 2011
|
||||
*
|
||||
* Groestl implementation with inline assembly using ssse3, sse4.1, and aes
|
||||
* instructions.
|
||||
* Authors: Günther A. Roland, Martin Schläffer, Krystian Matusiewicz
|
||||
*
|
||||
* This code is placed in the public domain
|
||||
*/
|
||||
|
||||
#include "hash-groestl256.h"
|
||||
/* global constants */
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_Lx[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_L0[ROUNDS512*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_L7[ROUNDS512*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_P[ROUNDS1024*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_Q[ROUNDS1024*16];
|
||||
__attribute__ ((aligned (16))) unsigned char TRANSP_MASK[16];
|
||||
__attribute__ ((aligned (16))) unsigned char SUBSH_MASK[8*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_1B[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_FF[16];
|
||||
|
||||
/* temporary variables */
|
||||
__attribute__ ((aligned (16))) unsigned char QTEMP[8*16];
|
||||
__attribute__ ((aligned (16))) unsigned char TEMP[3*16];
|
||||
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
|
||||
|
||||
/* xmm[i] will be multiplied by 2
|
||||
* xmm[j] will be lost
|
||||
* xmm[k] has to be all 0x1b */
|
||||
#define MUL2(i, j, k){\
|
||||
asm("pxor xmm"tostr(j)", xmm"tostr(j)"");\
|
||||
asm("pcmpgtb xmm"tostr(j)", xmm"tostr(i)"");\
|
||||
asm("paddb xmm"tostr(i)", xmm"tostr(i)"");\
|
||||
asm("pand xmm"tostr(j)", xmm"tostr(k)"");\
|
||||
asm("pxor xmm"tostr(i)", xmm"tostr(j)"");\
|
||||
}/**/
|
||||
|
||||
/* Yet another implementation of MixBytes.
|
||||
This time we use the formulae (3) from the paper "Byte Slicing Groestl".
|
||||
Input: a0, ..., a7
|
||||
Output: b0, ..., b7 = MixBytes(a0,...,a7).
|
||||
but we use the relations:
|
||||
t_i = a_i + a_{i+3}
|
||||
x_i = t_i + t_{i+3}
|
||||
y_i = t_i + t+{i+2} + a_{i+6}
|
||||
z_i = 2*x_i
|
||||
w_i = z_i + y_{i+4}
|
||||
v_i = 2*w_i
|
||||
b_i = v_{i+3} + y_{i+4}
|
||||
We keep building b_i in registers xmm8..xmm15 by first building y_{i+4} there
|
||||
and then adding v_i computed in the meantime in registers xmm0..xmm7.
|
||||
We almost fit into 16 registers, need only 3 spills to memory.
|
||||
This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
|
||||
K. Matusiewicz, 2011/05/29 */
|
||||
#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* t_i = a_i + a_{i+1} */\
|
||||
asm("movdqa xmm"tostr(b6)", xmm"tostr(a0)"");\
|
||||
asm("movdqa xmm"tostr(b7)", xmm"tostr(a1)"");\
|
||||
asm("pxor xmm"tostr(a0)", xmm"tostr(a1)"");\
|
||||
asm("movdqa xmm"tostr(b0)", xmm"tostr(a2)"");\
|
||||
asm("pxor xmm"tostr(a1)", xmm"tostr(a2)"");\
|
||||
asm("movdqa xmm"tostr(b1)", xmm"tostr(a3)"");\
|
||||
asm("pxor xmm"tostr(a2)", xmm"tostr(a3)"");\
|
||||
asm("movdqa xmm"tostr(b2)", xmm"tostr(a4)"");\
|
||||
asm("pxor xmm"tostr(a3)", xmm"tostr(a4)"");\
|
||||
asm("movdqa xmm"tostr(b3)", xmm"tostr(a5)"");\
|
||||
asm("pxor xmm"tostr(a4)", xmm"tostr(a5)"");\
|
||||
asm("movdqa xmm"tostr(b4)", xmm"tostr(a6)"");\
|
||||
asm("pxor xmm"tostr(a5)", xmm"tostr(a6)"");\
|
||||
asm("movdqa xmm"tostr(b5)", xmm"tostr(a7)"");\
|
||||
asm("pxor xmm"tostr(a6)", xmm"tostr(a7)"");\
|
||||
asm("pxor xmm"tostr(a7)", xmm"tostr(b6)"");\
|
||||
\
|
||||
/* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
|
||||
asm("pxor xmm"tostr(b0)", xmm"tostr(a4)"");\
|
||||
asm("pxor xmm"tostr(b6)", xmm"tostr(a4)"");\
|
||||
asm("pxor xmm"tostr(b1)", xmm"tostr(a5)"");\
|
||||
asm("pxor xmm"tostr(b7)", xmm"tostr(a5)"");\
|
||||
asm("pxor xmm"tostr(b2)", xmm"tostr(a6)"");\
|
||||
asm("pxor xmm"tostr(b0)", xmm"tostr(a6)"");\
|
||||
/* spill values y_4, y_5 to memory */\
|
||||
asm("movaps [TEMP+0*16], xmm"tostr(b0)"");\
|
||||
asm("pxor xmm"tostr(b3)", xmm"tostr(a7)"");\
|
||||
asm("pxor xmm"tostr(b1)", xmm"tostr(a7)"");\
|
||||
asm("movaps [TEMP+1*16], xmm"tostr(b1)"");\
|
||||
asm("pxor xmm"tostr(b4)", xmm"tostr(a0)"");\
|
||||
asm("pxor xmm"tostr(b2)", xmm"tostr(a0)"");\
|
||||
/* save values t0, t1, t2 to xmm8, xmm9 and memory */\
|
||||
asm("movdqa xmm"tostr(b0)", xmm"tostr(a0)"");\
|
||||
asm("pxor xmm"tostr(b5)", xmm"tostr(a1)"");\
|
||||
asm("pxor xmm"tostr(b3)", xmm"tostr(a1)"");\
|
||||
asm("movdqa xmm"tostr(b1)", xmm"tostr(a1)"");\
|
||||
asm("pxor xmm"tostr(b6)", xmm"tostr(a2)"");\
|
||||
asm("pxor xmm"tostr(b4)", xmm"tostr(a2)"");\
|
||||
asm("movaps [TEMP+2*16], xmm"tostr(a2)"");\
|
||||
asm("pxor xmm"tostr(b7)", xmm"tostr(a3)"");\
|
||||
asm("pxor xmm"tostr(b5)", xmm"tostr(a3)"");\
|
||||
\
|
||||
/* compute x_i = t_i + t_{i+3} */\
|
||||
asm("pxor xmm"tostr(a0)", xmm"tostr(a3)"");\
|
||||
asm("pxor xmm"tostr(a1)", xmm"tostr(a4)"");\
|
||||
asm("pxor xmm"tostr(a2)", xmm"tostr(a5)"");\
|
||||
asm("pxor xmm"tostr(a3)", xmm"tostr(a6)"");\
|
||||
asm("pxor xmm"tostr(a4)", xmm"tostr(a7)"");\
|
||||
asm("pxor xmm"tostr(a5)", xmm"tostr(b0)"");\
|
||||
asm("pxor xmm"tostr(a6)", xmm"tostr(b1)"");\
|
||||
asm("pxor xmm"tostr(a7)", [TEMP+2*16]");\
|
||||
\
|
||||
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
|
||||
/* compute w_i : add y_{i+4} */\
|
||||
asm("movaps xmm"tostr(b1)", [ALL_1B]");\
|
||||
MUL2(a0, b0, b1);\
|
||||
asm("pxor xmm"tostr(a0)", [TEMP+0*16]");\
|
||||
MUL2(a1, b0, b1);\
|
||||
asm("pxor xmm"tostr(a1)", [TEMP+1*16]");\
|
||||
MUL2(a2, b0, b1);\
|
||||
asm("pxor xmm"tostr(a2)", xmm"tostr(b2)"");\
|
||||
MUL2(a3, b0, b1);\
|
||||
asm("pxor xmm"tostr(a3)", xmm"tostr(b3)"");\
|
||||
MUL2(a4, b0, b1);\
|
||||
asm("pxor xmm"tostr(a4)", xmm"tostr(b4)"");\
|
||||
MUL2(a5, b0, b1);\
|
||||
asm("pxor xmm"tostr(a5)", xmm"tostr(b5)"");\
|
||||
MUL2(a6, b0, b1);\
|
||||
asm("pxor xmm"tostr(a6)", xmm"tostr(b6)"");\
|
||||
MUL2(a7, b0, b1);\
|
||||
asm("pxor xmm"tostr(a7)", xmm"tostr(b7)"");\
|
||||
\
|
||||
/* compute v_i : double w_i */\
|
||||
/* add to y_4 y_5 .. v3, v4, ... */\
|
||||
MUL2(a0, b0, b1);\
|
||||
asm("pxor xmm"tostr(b5)", xmm"tostr(a0)"");\
|
||||
MUL2(a1, b0, b1);\
|
||||
asm("pxor xmm"tostr(b6)", xmm"tostr(a1)"");\
|
||||
MUL2(a2, b0, b1);\
|
||||
asm("pxor xmm"tostr(b7)", xmm"tostr(a2)"");\
|
||||
MUL2(a5, b0, b1);\
|
||||
asm("pxor xmm"tostr(b2)", xmm"tostr(a5)"");\
|
||||
MUL2(a6, b0, b1);\
|
||||
asm("pxor xmm"tostr(b3)", xmm"tostr(a6)"");\
|
||||
MUL2(a7, b0, b1);\
|
||||
asm("pxor xmm"tostr(b4)", xmm"tostr(a7)"");\
|
||||
MUL2(a3, b0, b1);\
|
||||
MUL2(a4, b0, b1);\
|
||||
asm("movaps xmm"tostr(b0)", [TEMP+0*16]");\
|
||||
asm("movaps xmm"tostr(b1)", [TEMP+1*16]");\
|
||||
asm("pxor xmm"tostr(b0)", xmm"tostr(a3)"");\
|
||||
asm("pxor xmm"tostr(b1)", xmm"tostr(a4)"");\
|
||||
}/*MixBytes*/
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
((u64*)ALL_1B)[0] = 0x1b1b1b1b1b1b1b1bULL;\
|
||||
((u64*)ALL_1B)[1] = 0x1b1b1b1b1b1b1b1bULL;\
|
||||
((u64*)TRANSP_MASK)[0] = 0x0d0509010c040800ULL;\
|
||||
((u64*)TRANSP_MASK)[1] = 0x0f070b030e060a02ULL;\
|
||||
((u64*)SUBSH_MASK)[ 0] = 0x0c0f0104070b0e00ULL;\
|
||||
((u64*)SUBSH_MASK)[ 1] = 0x03060a0d08020509ULL;\
|
||||
((u64*)SUBSH_MASK)[ 2] = 0x0e090205000d0801ULL;\
|
||||
((u64*)SUBSH_MASK)[ 3] = 0x04070c0f0a03060bULL;\
|
||||
((u64*)SUBSH_MASK)[ 4] = 0x080b0306010f0a02ULL;\
|
||||
((u64*)SUBSH_MASK)[ 5] = 0x05000e090c04070dULL;\
|
||||
((u64*)SUBSH_MASK)[ 6] = 0x0a0d040702090c03ULL;\
|
||||
((u64*)SUBSH_MASK)[ 7] = 0x0601080b0e05000fULL;\
|
||||
((u64*)SUBSH_MASK)[ 8] = 0x0b0e0500030a0d04ULL;\
|
||||
((u64*)SUBSH_MASK)[ 9] = 0x0702090c0f060108ULL;\
|
||||
((u64*)SUBSH_MASK)[10] = 0x0d080601040c0f05ULL;\
|
||||
((u64*)SUBSH_MASK)[11] = 0x00030b0e0907020aULL;\
|
||||
((u64*)SUBSH_MASK)[12] = 0x0f0a0702050e0906ULL;\
|
||||
((u64*)SUBSH_MASK)[13] = 0x01040d080b00030cULL;\
|
||||
((u64*)SUBSH_MASK)[14] = 0x090c000306080b07ULL;\
|
||||
((u64*)SUBSH_MASK)[15] = 0x02050f0a0d01040eULL;\
|
||||
for(i = 0; i < ROUNDS512; i++)\
|
||||
{\
|
||||
((u64*)ROUND_CONST_L0)[i*2+1] = 0xffffffffffffffffULL;\
|
||||
((u64*)ROUND_CONST_L0)[i*2+0] = (i * 0x0101010101010101ULL) ^ 0x7060504030201000ULL;\
|
||||
((u64*)ROUND_CONST_L7)[i*2+1] = (i * 0x0101010101010101ULL) ^ 0x8f9fafbfcfdfefffULL;\
|
||||
((u64*)ROUND_CONST_L7)[i*2+0] = 0x0000000000000000ULL;\
|
||||
}\
|
||||
((u64*)ROUND_CONST_Lx)[1] = 0xffffffffffffffffULL;\
|
||||
((u64*)ROUND_CONST_Lx)[0] = 0x0000000000000000ULL;\
|
||||
}while(0);
|
||||
|
||||
#define Push_All_Regs() do{\
|
||||
/* not using any...
|
||||
asm("push rax");\
|
||||
asm("push rbx");\
|
||||
asm("push rcx");*/\
|
||||
}while(0);
|
||||
|
||||
#define Pop_All_Regs() do{\
|
||||
/* not using any...
|
||||
asm("pop rcx");\
|
||||
asm("pop rbx");\
|
||||
asm("pop rax");*/\
|
||||
}while(0);
|
||||
|
||||
/* one round
|
||||
* i = round number
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
*/
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* AddRoundConstant */\
|
||||
asm ("movaps xmm"tostr(b1)", [ROUND_CONST_Lx]");\
|
||||
asm ("pxor xmm"tostr(a0)", [ROUND_CONST_L0+"tostr(i)"*16]");\
|
||||
asm ("pxor xmm"tostr(a1)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a2)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a3)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a4)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a5)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a6)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a7)", [ROUND_CONST_L7+"tostr(i)"*16]");\
|
||||
/* ShiftBytes + SubBytes (interleaved) */\
|
||||
asm ("pxor xmm"tostr(b0)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a0)", [SUBSH_MASK+0*16]");\
|
||||
asm ("aesenclast xmm"tostr(a0)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a1)", [SUBSH_MASK+1*16]");\
|
||||
asm ("aesenclast xmm"tostr(a1)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a2)", [SUBSH_MASK+2*16]");\
|
||||
asm ("aesenclast xmm"tostr(a2)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a3)", [SUBSH_MASK+3*16]");\
|
||||
asm ("aesenclast xmm"tostr(a3)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a4)", [SUBSH_MASK+4*16]");\
|
||||
asm ("aesenclast xmm"tostr(a4)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a5)", [SUBSH_MASK+5*16]");\
|
||||
asm ("aesenclast xmm"tostr(a5)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a6)", [SUBSH_MASK+6*16]");\
|
||||
asm ("aesenclast xmm"tostr(a6)", xmm"tostr(b0)"");\
|
||||
asm ("pshufb xmm"tostr(a7)", [SUBSH_MASK+7*16]");\
|
||||
asm ("aesenclast xmm"tostr(a7)", xmm"tostr(b0)"");\
|
||||
/* MixBytes */\
|
||||
MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
|
||||
}
|
||||
|
||||
/* 10 rounds, P and Q in parallel */
|
||||
#define ROUNDS_P_Q(){\
|
||||
ROUND(0, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(2, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(3, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(4, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(5, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(6, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(7, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(8, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(9, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
}
|
||||
|
||||
/* Matrix Transpose Step 1
|
||||
* input is a 512-bit state with two columns in one xmm
|
||||
* output is a 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i3
|
||||
* outputs: i0, o1-o3
|
||||
* clobbers: t0
|
||||
*/
|
||||
#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
|
||||
asm ("movaps xmm"tostr(t0)", [TRANSP_MASK]");\
|
||||
\
|
||||
asm ("pshufb xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(i1)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(i3)", xmm"tostr(t0)"");\
|
||||
\
|
||||
asm ("movdqa xmm"tostr(o1)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(t0)", xmm"tostr(i2)"");\
|
||||
\
|
||||
asm ("punpcklwd xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("punpckhwd xmm"tostr(o1)", xmm"tostr(i1)"");\
|
||||
asm ("punpcklwd xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("punpckhwd xmm"tostr(t0)", xmm"tostr(i3)"");\
|
||||
\
|
||||
asm ("pshufd xmm"tostr(i0)", xmm"tostr(i0)", 216");\
|
||||
asm ("pshufd xmm"tostr(o1)", xmm"tostr(o1)", 216");\
|
||||
asm ("pshufd xmm"tostr(i2)", xmm"tostr(i2)", 216");\
|
||||
asm ("pshufd xmm"tostr(t0)", xmm"tostr(t0)", 216");\
|
||||
\
|
||||
asm ("movdqa xmm"tostr(o2)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(o3)", xmm"tostr(o1)"");\
|
||||
\
|
||||
asm ("punpckldq xmm"tostr(i0)", xmm"tostr(i2)"");\
|
||||
asm ("punpckldq xmm"tostr(o1)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhdq xmm"tostr(o2)", xmm"tostr(i2)"");\
|
||||
asm ("punpckhdq xmm"tostr(o3)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Step 2
|
||||
* input are two 512-bit states with two rows in one xmm
|
||||
* output are two 512-bit states with one row of each state in one xmm
|
||||
* inputs: i0-i3 = P, i4-i7 = Q
|
||||
* outputs: (i0, o1-o7) = (P|Q)
|
||||
* possible reassignments: (output reg = input reg)
|
||||
* * i1 -> o3-7
|
||||
* * i2 -> o5-7
|
||||
* * i3 -> o7
|
||||
* * i4 -> o3-7
|
||||
* * i5 -> o6-7
|
||||
*/
|
||||
#define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
|
||||
asm ("movdqa xmm"tostr(o1)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(o2)", xmm"tostr(i1)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(i4)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o1)", xmm"tostr(i4)"");\
|
||||
asm ("movdqa xmm"tostr(o3)", xmm"tostr(i1)"");\
|
||||
asm ("movdqa xmm"tostr(o4)", xmm"tostr(i2)"");\
|
||||
asm ("punpcklqdq xmm"tostr(o2)", xmm"tostr(i5)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o3)", xmm"tostr(i5)"");\
|
||||
asm ("movdqa xmm"tostr(o5)", xmm"tostr(i2)"");\
|
||||
asm ("movdqa xmm"tostr(o6)", xmm"tostr(i3)"");\
|
||||
asm ("punpcklqdq xmm"tostr(o4)", xmm"tostr(i6)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o5)", xmm"tostr(i6)"");\
|
||||
asm ("movdqa xmm"tostr(o7)", xmm"tostr(i3)"");\
|
||||
asm ("punpcklqdq xmm"tostr(o6)", xmm"tostr(i7)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o7)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Inverse Step 2
|
||||
* input are two 512-bit states with one row of each state in one xmm
|
||||
* output are two 512-bit states with two rows in one xmm
|
||||
* inputs: i0-i7 = (P|Q)
|
||||
* outputs: (i0, i2, i4, i6) = P, (o0-o3) = Q
|
||||
*/
|
||||
#define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
|
||||
asm ("movdqa xmm"tostr(o0)", xmm"tostr(i0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o0)", xmm"tostr(i1)"");\
|
||||
asm ("movdqa xmm"tostr(o1)", xmm"tostr(i2)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o1)", xmm"tostr(i3)"");\
|
||||
asm ("movdqa xmm"tostr(o2)", xmm"tostr(i4)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o2)", xmm"tostr(i5)"");\
|
||||
asm ("movdqa xmm"tostr(o3)", xmm"tostr(i6)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o3)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Step 2
|
||||
* input is one 512-bit state with two rows in one xmm
|
||||
* output is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* inputs: i0,i2,i4,i6 = S
|
||||
* outputs: (i0-7) = (0|S)
|
||||
*/
|
||||
#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
|
||||
asm ("pxor xmm"tostr(t0)", xmm"tostr(t0)"");\
|
||||
asm ("movdqa xmm"tostr(i1)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(i3)", xmm"tostr(i2)"");\
|
||||
asm ("movdqa xmm"tostr(i5)", xmm"tostr(i4)"");\
|
||||
asm ("movdqa xmm"tostr(i7)", xmm"tostr(i6)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i1)", xmm"tostr(t0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i3)", xmm"tostr(t0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i4)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i5)", xmm"tostr(t0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i6)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i7)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Inverse Step 2
|
||||
* input is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* output is one 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i7 = (0|S)
|
||||
* outputs: (i0, i2, i4, i6) = S
|
||||
*/
|
||||
#define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
|
||||
void INIT256(u64* h)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
asm volatile ("emms");
|
||||
|
||||
/* load IV into registers xmm12 - xmm15 */
|
||||
asm ("movaps xmm12, [rdi+0*16]");
|
||||
asm ("movaps xmm13, [rdi+1*16]");
|
||||
asm ("movaps xmm14, [rdi+2*16]");
|
||||
asm ("movaps xmm15, [rdi+3*16]");
|
||||
|
||||
/* transform chaining value from column ordering into row ordering */
|
||||
/* we put two rows (64 bit) of the IV into one 128-bit XMM register */
|
||||
Matrix_Transpose_A(12, 13, 14, 15, 2, 6, 7, 0);
|
||||
|
||||
/* store transposed IV */
|
||||
asm ("movaps [rdi+0*16], xmm12");
|
||||
asm ("movaps [rdi+1*16], xmm2");
|
||||
asm ("movaps [rdi+2*16], xmm6");
|
||||
asm ("movaps [rdi+3*16], xmm7");
|
||||
|
||||
asm volatile ("emms");
|
||||
asm (".att_syntax noprefix");
|
||||
}
|
||||
|
||||
void TF512(u64* h, u64* m)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
/* message M in rsi */
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_START;
|
||||
#endif
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
Push_All_Regs();
|
||||
|
||||
/* load message into registers xmm12 - xmm15 (Q = message) */
|
||||
asm ("movaps xmm12, [rsi+0*16]");
|
||||
asm ("movaps xmm13, [rsi+1*16]");
|
||||
asm ("movaps xmm14, [rsi+2*16]");
|
||||
asm ("movaps xmm15, [rsi+3*16]");
|
||||
|
||||
/* transform message M from column ordering into row ordering */
|
||||
/* we first put two rows (2x64 bit) of the message into one 128-bit xmm register */
|
||||
Matrix_Transpose_A(12, 13, 14, 15, 2, 6, 7, 0);
|
||||
|
||||
/* load previous chaining value */
|
||||
/* we first put two rows (64 bit) of the CV into one 128-bit xmm register */
|
||||
asm ("movaps xmm8, [rdi+0*16]");
|
||||
asm ("movaps xmm0, [rdi+1*16]");
|
||||
asm ("movaps xmm4, [rdi+2*16]");
|
||||
asm ("movaps xmm5, [rdi+3*16]");
|
||||
|
||||
/* xor message to CV get input of P */
|
||||
/* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
|
||||
asm ("pxor xmm8, xmm12");
|
||||
asm ("pxor xmm0, xmm2");
|
||||
asm ("pxor xmm4, xmm6");
|
||||
asm ("pxor xmm5, xmm7");
|
||||
|
||||
/* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
|
||||
/* result: the 8 rows of P and Q in xmm8 - xmm12 */
|
||||
Matrix_Transpose_B(8, 0, 4, 5, 12, 2, 6, 7, 9, 10, 11, 12, 13, 14, 15);
|
||||
|
||||
/* compute the two permutations P and Q in parallel */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P or two rows of Q in one xmm register */
|
||||
Matrix_Transpose_B_INV(8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3);
|
||||
|
||||
/* xor output of P and Q */
|
||||
/* result: P(CV+M)+Q(M) in xmm0...xmm3 */
|
||||
asm ("pxor xmm0, xmm8");
|
||||
asm ("pxor xmm1, xmm10");
|
||||
asm ("pxor xmm2, xmm12");
|
||||
asm ("pxor xmm3, xmm14");
|
||||
|
||||
/* xor CV (feed-forward) */
|
||||
/* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
|
||||
asm ("pxor xmm0, [rdi+0*16]");
|
||||
asm ("pxor xmm1, [rdi+1*16]");
|
||||
asm ("pxor xmm2, [rdi+2*16]");
|
||||
asm ("pxor xmm3, [rdi+3*16]");
|
||||
|
||||
/* store CV */
|
||||
asm ("movaps [rdi+0*16], xmm0");
|
||||
asm ("movaps [rdi+1*16], xmm1");
|
||||
asm ("movaps [rdi+2*16], xmm2");
|
||||
asm ("movaps [rdi+3*16], xmm3");
|
||||
|
||||
Pop_All_Regs();
|
||||
asm (".att_syntax noprefix");
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_END;
|
||||
#endif
|
||||
return;
|
||||
}
|
||||
|
||||
void OF512(u64* h)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
Push_All_Regs();
|
||||
|
||||
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
|
||||
asm ("movaps xmm8, [rdi+0*16]");
|
||||
asm ("movaps xmm10, [rdi+1*16]");
|
||||
asm ("movaps xmm12, [rdi+2*16]");
|
||||
asm ("movaps xmm14, [rdi+3*16]");
|
||||
|
||||
/* there are now 2 rows of the CV in one xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) into one half of an xmm register */
|
||||
/* result: the 8 input rows of P in xmm8 - xmm15 */
|
||||
Matrix_Transpose_O_B(8, 9, 10, 11, 12, 13, 14, 15, 0);
|
||||
|
||||
/* compute the permutation P */
|
||||
/* result: the output of P(CV) in xmm8 - xmm15 */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P in one xmm register */
|
||||
/* result: P(CV) in xmm8, xmm10, xmm12, xmm14 */
|
||||
Matrix_Transpose_O_B_INV(8, 9, 10, 11, 12, 13, 14, 15);
|
||||
|
||||
/* xor CV to P output (feed-forward) */
|
||||
/* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
|
||||
asm ("pxor xmm8, [rdi+0*16]");
|
||||
asm ("pxor xmm10, [rdi+1*16]");
|
||||
asm ("pxor xmm12, [rdi+2*16]");
|
||||
asm ("pxor xmm14, [rdi+3*16]");
|
||||
|
||||
/* transform state back from row ordering into column ordering */
|
||||
/* result: final hash value in xmm9, xmm11 */
|
||||
Matrix_Transpose_A(8, 10, 12, 14, 4, 9, 11, 0);
|
||||
|
||||
/* we only need to return the truncated half of the state */
|
||||
asm ("movaps [rdi+2*16], xmm9");
|
||||
asm ("movaps [rdi+3*16], xmm11");
|
||||
|
||||
Pop_All_Regs();
|
||||
asm (".att_syntax noprefix");
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
@@ -1,519 +0,0 @@
|
||||
/* groestl-asm-avx.h Aug 2011
|
||||
*
|
||||
* Groestl implementation with inline assembly using ssse3, sse4.1, aes and avx
|
||||
* instructions.
|
||||
* Author: Günther A. Roland, Martin Schläffer, Krystian Matusiewicz
|
||||
*
|
||||
* This code is placed in the public domain
|
||||
*/
|
||||
|
||||
#include "hash-groestl256.h"
|
||||
|
||||
/* global variables */
|
||||
__attribute__ ((aligned (32))) unsigned char ROUND_CONST_Lx[16];
|
||||
__attribute__ ((aligned (32))) unsigned char ROUND_CONST_L0[ROUNDS512*16];
|
||||
__attribute__ ((aligned (32))) unsigned char ROUND_CONST_L7[ROUNDS512*16];
|
||||
__attribute__ ((aligned (32))) unsigned char ROUND_CONST_P[ROUNDS1024*16];
|
||||
__attribute__ ((aligned (32))) unsigned char ROUND_CONST_Q[ROUNDS1024*16];
|
||||
__attribute__ ((aligned (32))) unsigned char TRANSP_MASK[16];
|
||||
__attribute__ ((aligned (32))) unsigned char SUBSH_MASK[8*16];
|
||||
__attribute__ ((aligned (32))) unsigned char ALL_1B[32];
|
||||
__attribute__ ((aligned (32))) unsigned char ALL_FF[32];
|
||||
|
||||
/* temporary variables */
|
||||
__attribute__ ((aligned (32))) unsigned char TEMP[6*32];
|
||||
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
((u64*)TRANSP_MASK)[0] = 0x0d0509010c040800ULL;\
|
||||
((u64*)TRANSP_MASK)[1] = 0x0f070b030e060a02ULL;\
|
||||
((u64*)ALL_1B)[0] = 0x1b1b1b1b1b1b1b1bULL;\
|
||||
((u64*)ALL_1B)[1] = 0x1b1b1b1b1b1b1b1bULL;\
|
||||
((u64*)SUBSH_MASK)[ 0] = 0x0c0f0104070b0e00ULL;\
|
||||
((u64*)SUBSH_MASK)[ 1] = 0x03060a0d08020509ULL;\
|
||||
((u64*)SUBSH_MASK)[ 2] = 0x0e090205000d0801ULL;\
|
||||
((u64*)SUBSH_MASK)[ 3] = 0x04070c0f0a03060bULL;\
|
||||
((u64*)SUBSH_MASK)[ 4] = 0x080b0306010f0a02ULL;\
|
||||
((u64*)SUBSH_MASK)[ 5] = 0x05000e090c04070dULL;\
|
||||
((u64*)SUBSH_MASK)[ 6] = 0x0a0d040702090c03ULL;\
|
||||
((u64*)SUBSH_MASK)[ 7] = 0x0601080b0e05000fULL;\
|
||||
((u64*)SUBSH_MASK)[ 8] = 0x0b0e0500030a0d04ULL;\
|
||||
((u64*)SUBSH_MASK)[ 9] = 0x0702090c0f060108ULL;\
|
||||
((u64*)SUBSH_MASK)[10] = 0x0d080601040c0f05ULL;\
|
||||
((u64*)SUBSH_MASK)[11] = 0x00030b0e0907020aULL;\
|
||||
((u64*)SUBSH_MASK)[12] = 0x0f0a0702050e0906ULL;\
|
||||
((u64*)SUBSH_MASK)[13] = 0x01040d080b00030cULL;\
|
||||
((u64*)SUBSH_MASK)[14] = 0x090c000306080b07ULL;\
|
||||
((u64*)SUBSH_MASK)[15] = 0x02050f0a0d01040eULL;\
|
||||
for(i = 0; i < ROUNDS512; i++)\
|
||||
{\
|
||||
((u64*)ROUND_CONST_L0)[i*2+1] = 0xffffffffffffffffULL;\
|
||||
((u64*)ROUND_CONST_L0)[i*2+0] = (i * 0x0101010101010101ULL) ^ 0x7060504030201000ULL;\
|
||||
((u64*)ROUND_CONST_L7)[i*2+1] = (i * 0x0101010101010101ULL) ^ 0x8f9fafbfcfdfefffULL;\
|
||||
((u64*)ROUND_CONST_L7)[i*2+0] = 0x0000000000000000ULL;\
|
||||
}\
|
||||
((u64*)ROUND_CONST_Lx)[1] = 0xffffffffffffffffULL;\
|
||||
((u64*)ROUND_CONST_Lx)[0] = 0x0000000000000000ULL;\
|
||||
}while(0);
|
||||
|
||||
#define Push_All_Regs() do{\
|
||||
/* not using any...
|
||||
asm("push rax");\
|
||||
asm("push rbx");\
|
||||
asm("push rcx");*/\
|
||||
}while(0);
|
||||
|
||||
#define Pop_All_Regs() do{\
|
||||
/* not using any...
|
||||
asm("pop rcx");\
|
||||
asm("pop rbx");\
|
||||
asm("pop rax");*/\
|
||||
}while(0);
|
||||
|
||||
/* xmm[i] will be multiplied by 2
|
||||
* xmm[j] will be lost
|
||||
* xmm[k] has to be all 0x1b
|
||||
* xmm[z] has to be zero */
|
||||
#define VMUL2(i, j, k, z){\
|
||||
asm("vpcmpgtb xmm"tostr(j)", xmm"tostr(z)", xmm"tostr(i)"");\
|
||||
asm("vpaddb xmm"tostr(i)", xmm"tostr(i)", xmm"tostr(i)"");\
|
||||
asm("vpand xmm"tostr(j)", xmm"tostr(j)", xmm"tostr(k)"");\
|
||||
asm("vpxor xmm"tostr(i)", xmm"tostr(i)", xmm"tostr(j)"");\
|
||||
}/**/
|
||||
|
||||
/* xmm[i] will be multiplied by 2
|
||||
* xmm[j] will be lost
|
||||
* xmm[k] has to be all 0x1b
|
||||
* xmm[z] has to be zero */
|
||||
#define VMUL2v2(i, j, k, z){\
|
||||
asm("vpblendvb xmm"tostr(j)", xmm"tostr(z)", xmm"tostr(k)", xmm"tostr(i)"");\
|
||||
asm("vpaddb xmm"tostr(i)", xmm"tostr(i)", xmm"tostr(i)"");\
|
||||
asm("vpxor xmm"tostr(i)", xmm"tostr(i)", xmm"tostr(j)"");\
|
||||
}/**/
|
||||
|
||||
/* Yet another implementation of MixBytes.
|
||||
This time we use the formulae (3) from the paper "Byte Slicing Groestl".
|
||||
Input: a0, ..., a7
|
||||
Output: b0, ..., b7 = MixBytes(a0,...,a7).
|
||||
but we use the relations:
|
||||
t_i = a_i + a_{i+3}
|
||||
x_i = t_i + t_{i+3}
|
||||
y_i = t_i + t+{i+2} + a_{i+6}
|
||||
z_i = 2*x_i
|
||||
w_i = z_i + y_{i+4}
|
||||
v_i = 2*w_i
|
||||
b_i = v_{i+3} + y_{i+4}
|
||||
We keep building b_i in registers xmm8..xmm15 by first building y_{i+4} there
|
||||
and then adding v_i computed in the meantime in registers xmm0..xmm7.
|
||||
We almost fit into 16 registers, need only 3 spills to memory.
|
||||
This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
|
||||
K. Matusiewicz, 2011/05/29 */
|
||||
#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* xmm"tostr(8..xmm"tostr(15 = a2 a3... a0 a1 */\
|
||||
asm("vmovdqa xmm"tostr(b0)", xmm"tostr(a2)"");\
|
||||
asm("vmovdqa xmm"tostr(b1)", xmm"tostr(a3)"");\
|
||||
asm("vmovdqa xmm"tostr(b2)", xmm"tostr(a4)"");\
|
||||
asm("vmovdqa xmm"tostr(b3)", xmm"tostr(a5)"");\
|
||||
asm("vmovdqa xmm"tostr(b4)", xmm"tostr(a6)"");\
|
||||
asm("vmovdqa xmm"tostr(b5)", xmm"tostr(a7)"");\
|
||||
asm("vmovdqa xmm"tostr(b6)", xmm"tostr(a0)"");\
|
||||
asm("vmovdqa xmm"tostr(b7)", xmm"tostr(a1)"");\
|
||||
\
|
||||
/* t_i = a_i + a_{i+1} */\
|
||||
asm("vpxor xmm"tostr(a0)", xmm"tostr(a0)", xmm"tostr(a1)"");\
|
||||
asm("vpxor xmm"tostr(a1)", xmm"tostr(a1)", xmm"tostr(a2)"");\
|
||||
asm("vpxor xmm"tostr(a2)", xmm"tostr(a2)", xmm"tostr(a3)"");\
|
||||
asm("vpxor xmm"tostr(a3)", xmm"tostr(a3)", xmm"tostr(a4)"");\
|
||||
asm("vpxor xmm"tostr(a4)", xmm"tostr(a4)", xmm"tostr(a5)"");\
|
||||
asm("vpxor xmm"tostr(a5)", xmm"tostr(a5)", xmm"tostr(a6)"");\
|
||||
asm("vpxor xmm"tostr(a6)", xmm"tostr(a6)", xmm"tostr(a7)"");\
|
||||
asm("vpxor xmm"tostr(a7)", xmm"tostr(a7)", xmm"tostr(b6)"");\
|
||||
\
|
||||
/* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
|
||||
asm("vpxor xmm"tostr(b0)", xmm"tostr(b0)", xmm"tostr(a4)"");\
|
||||
asm("vpxor xmm"tostr(b1)", xmm"tostr(b1)", xmm"tostr(a5)"");\
|
||||
asm("vpxor xmm"tostr(b2)", xmm"tostr(b2)", xmm"tostr(a6)"");\
|
||||
asm("vpxor xmm"tostr(b3)", xmm"tostr(b3)", xmm"tostr(a7)"");\
|
||||
asm("vpxor xmm"tostr(b4)", xmm"tostr(b4)", xmm"tostr(a0)"");\
|
||||
asm("vpxor xmm"tostr(b5)", xmm"tostr(b5)", xmm"tostr(a1)"");\
|
||||
asm("vpxor xmm"tostr(b6)", xmm"tostr(b6)", xmm"tostr(a2)"");\
|
||||
asm("vpxor xmm"tostr(b7)", xmm"tostr(b7)", xmm"tostr(a3)"");\
|
||||
\
|
||||
asm("vpxor xmm"tostr(b0)", xmm"tostr(b0)", xmm"tostr(a6)"");\
|
||||
asm("vpxor xmm"tostr(b1)", xmm"tostr(b1)", xmm"tostr(a7)"");\
|
||||
asm("vpxor xmm"tostr(b2)", xmm"tostr(b2)", xmm"tostr(a0)"");\
|
||||
asm("vpxor xmm"tostr(b3)", xmm"tostr(b3)", xmm"tostr(a1)"");\
|
||||
asm("vpxor xmm"tostr(b4)", xmm"tostr(b4)", xmm"tostr(a2)"");\
|
||||
asm("vpxor xmm"tostr(b5)", xmm"tostr(b5)", xmm"tostr(a3)"");\
|
||||
asm("vpxor xmm"tostr(b6)", xmm"tostr(b6)", xmm"tostr(a4)"");\
|
||||
asm("vpxor xmm"tostr(b7)", xmm"tostr(b7)", xmm"tostr(a5)"");\
|
||||
\
|
||||
/* spill values y_4, y_5 to memory */\
|
||||
asm("vmovaps [TEMP+0*16], xmm"tostr(b0)"");\
|
||||
asm("vmovaps [TEMP+1*16], xmm"tostr(b1)"");\
|
||||
asm("vmovaps [TEMP+2*16], xmm"tostr(b2)"");\
|
||||
\
|
||||
/* save values t0, t1, t2 to xmm8, xmm9 and memory */\
|
||||
asm("vmovdqa xmm"tostr(b0)", xmm"tostr(a0)"");\
|
||||
asm("vmovdqa xmm"tostr(b1)", xmm"tostr(a1)"");\
|
||||
asm("vmovaps [TEMP+3*16], xmm"tostr(a2)"");\
|
||||
\
|
||||
/* compute x_i = t_i + t_{i+3} */\
|
||||
asm("vpxor xmm"tostr(a0)", xmm"tostr(a0)", xmm"tostr(a3)"");\
|
||||
asm("vpxor xmm"tostr(a1)", xmm"tostr(a1)", xmm"tostr(a4)"");\
|
||||
asm("vpxor xmm"tostr(a2)", xmm"tostr(a2)", xmm"tostr(a5)"");\
|
||||
asm("vpxor xmm"tostr(a3)", xmm"tostr(a3)", xmm"tostr(a6)"");\
|
||||
asm("vpxor xmm"tostr(a4)", xmm"tostr(a4)", xmm"tostr(a7)"");\
|
||||
asm("vpxor xmm"tostr(a5)", xmm"tostr(a5)", xmm"tostr(b0)"");\
|
||||
asm("vpxor xmm"tostr(a6)", xmm"tostr(a6)", xmm"tostr(b1)"");\
|
||||
asm("vpxor xmm"tostr(a7)", xmm"tostr(a7)", [TEMP+3*16]");\
|
||||
\
|
||||
/*compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
|
||||
asm("vmovaps xmm"tostr(b1)", [ALL_1B]");\
|
||||
asm("vpxor xmm"tostr(b2)", xmm"tostr(b2)", xmm"tostr(b2)"");\
|
||||
VMUL2(a7, b0, b1, b2);\
|
||||
VMUL2(a6, b0, b1, b2);\
|
||||
VMUL2(a5, b0, b1, b2);\
|
||||
VMUL2(a4, b0, b1, b2);\
|
||||
VMUL2(a3, b0, b1, b2);\
|
||||
VMUL2(a2, b0, b1, b2);\
|
||||
VMUL2(a1, b0, b1, b2);\
|
||||
VMUL2(a0, b0, b1, b2);\
|
||||
\
|
||||
/* compute w_i : add y_{i+4} */\
|
||||
asm("vpxor xmm"tostr(a0)", xmm"tostr(a0)", [TEMP+0*16]");\
|
||||
asm("vpxor xmm"tostr(a1)", xmm"tostr(a1)", [TEMP+1*16]");\
|
||||
asm("vpxor xmm"tostr(a2)", xmm"tostr(a2)", [TEMP+2*16]");\
|
||||
asm("vpxor xmm"tostr(a3)", xmm"tostr(a3)", xmm"tostr(b3)"");\
|
||||
asm("vpxor xmm"tostr(a4)", xmm"tostr(a4)", xmm"tostr(b4)"");\
|
||||
asm("vpxor xmm"tostr(a5)", xmm"tostr(a5)", xmm"tostr(b5)"");\
|
||||
asm("vpxor xmm"tostr(a6)", xmm"tostr(a6)", xmm"tostr(b6)"");\
|
||||
asm("vpxor xmm"tostr(a7)", xmm"tostr(a7)", xmm"tostr(b7)"");\
|
||||
\
|
||||
/*compute v_i: double w_i */\
|
||||
VMUL2(a0, b0, b1, b2);\
|
||||
VMUL2(a1, b0, b1, b2);\
|
||||
VMUL2(a2, b0, b1, b2);\
|
||||
VMUL2(a3, b0, b1, b2);\
|
||||
VMUL2(a4, b0, b1, b2);\
|
||||
VMUL2(a5, b0, b1, b2);\
|
||||
VMUL2(a6, b0, b1, b2);\
|
||||
VMUL2(a7, b0, b1, b2);\
|
||||
\
|
||||
/* add to y_4 y_5 .. v3, v4, ... */\
|
||||
asm("vpxor xmm"tostr(b0)", xmm"tostr(a3)", [TEMP+0*16]");\
|
||||
asm("vpxor xmm"tostr(b1)", xmm"tostr(a4)", [TEMP+1*16]");\
|
||||
asm("vpxor xmm"tostr(b2)", xmm"tostr(a5)", [TEMP+2*16]");\
|
||||
asm("vpxor xmm"tostr(b3)", xmm"tostr(b3)", xmm"tostr(a6)"");\
|
||||
asm("vpxor xmm"tostr(b4)", xmm"tostr(b4)", xmm"tostr(a7)"");\
|
||||
asm("vpxor xmm"tostr(b5)", xmm"tostr(b5)", xmm"tostr(a0)"");\
|
||||
asm("vpxor xmm"tostr(b6)", xmm"tostr(b6)", xmm"tostr(a1)"");\
|
||||
asm("vpxor xmm"tostr(b7)", xmm"tostr(b7)", xmm"tostr(a2)"");\
|
||||
}/*MixBytes*/
|
||||
|
||||
/* one round
|
||||
* i = round number
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
*/
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* AddRoundConstant */\
|
||||
asm ("vmovaps xmm"tostr(b1)", [ROUND_CONST_Lx]");\
|
||||
asm ("vpxor xmm"tostr(a0)", xmm"tostr(a0)", [ROUND_CONST_L0+"tostr(i)"*16]");\
|
||||
asm ("vpxor xmm"tostr(a1)", xmm"tostr(a1)", xmm"tostr(b1)"");\
|
||||
asm ("vpxor xmm"tostr(a2)", xmm"tostr(a2)", xmm"tostr(b1)"");\
|
||||
asm ("vpxor xmm"tostr(a3)", xmm"tostr(a3)", xmm"tostr(b1)"");\
|
||||
asm ("vpxor xmm"tostr(a4)", xmm"tostr(a4)", xmm"tostr(b1)"");\
|
||||
asm ("vpxor xmm"tostr(a5)", xmm"tostr(a5)", xmm"tostr(b1)"");\
|
||||
asm ("vpxor xmm"tostr(a6)", xmm"tostr(a6)", xmm"tostr(b1)"");\
|
||||
asm ("vpxor xmm"tostr(a7)", xmm"tostr(a7)", [ROUND_CONST_L7+"tostr(i)"*16]");\
|
||||
/* ShiftBytes + SubBytes (interleaved) */\
|
||||
asm ("vpxor xmm"tostr(b0)", xmm"tostr(b0)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a0)", xmm"tostr(a0)", [SUBSH_MASK+0*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a0)", xmm"tostr(a0)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a1)", xmm"tostr(a1)", [SUBSH_MASK+1*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a1)", xmm"tostr(a1)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a2)", xmm"tostr(a2)", [SUBSH_MASK+2*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a2)", xmm"tostr(a2)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a3)", xmm"tostr(a3)", [SUBSH_MASK+3*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a3)", xmm"tostr(a3)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a4)", xmm"tostr(a4)", [SUBSH_MASK+4*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a4)", xmm"tostr(a4)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a5)", xmm"tostr(a5)", [SUBSH_MASK+5*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a5)", xmm"tostr(a5)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a6)", xmm"tostr(a6)", [SUBSH_MASK+6*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a6)", xmm"tostr(a6)", xmm"tostr(b0)"");\
|
||||
asm ("vpshufb xmm"tostr(a7)", xmm"tostr(a7)", [SUBSH_MASK+7*16]");\
|
||||
asm ("vaesenclast xmm"tostr(a7)", xmm"tostr(a7)", xmm"tostr(b0)"");\
|
||||
/* MixBytes */\
|
||||
MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
|
||||
}
|
||||
|
||||
/* 10 rounds, P and Q in parallel */
|
||||
#define ROUNDS_P_Q(){\
|
||||
ROUND(0, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(2, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(3, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(4, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(5, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(6, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(7, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(8, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(9, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
}
|
||||
|
||||
/* Matrix Transpose Step 1
|
||||
* input is a 512-bit state with two columns in one xmm
|
||||
* output is a 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i3
|
||||
|
||||
* outputs: i0, o1-o3
|
||||
* clobbers: t0
|
||||
*/
|
||||
#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
|
||||
asm ("vmovaps xmm"tostr(t0)", [TRANSP_MASK]");\
|
||||
\
|
||||
asm ("vpshufb xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("vpshufb xmm"tostr(i1)", xmm"tostr(i1)", xmm"tostr(t0)"");\
|
||||
asm ("vpshufb xmm"tostr(i2)", xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("vpshufb xmm"tostr(i3)", xmm"tostr(i3)", xmm"tostr(t0)"");\
|
||||
\
|
||||
asm ("vpunpckhwd xmm"tostr(o1)", xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("vpunpcklwd xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("vpunpckhwd xmm"tostr(t0)", xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("vpunpcklwd xmm"tostr(i2)", xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
\
|
||||
asm ("vpshufd xmm"tostr(i0)", xmm"tostr(i0)", 216");\
|
||||
asm ("vpshufd xmm"tostr(o1)", xmm"tostr(o1)", 216");\
|
||||
asm ("vpshufd xmm"tostr(i2)", xmm"tostr(i2)", 216");\
|
||||
asm ("vpshufd xmm"tostr(t0)", xmm"tostr(t0)", 216");\
|
||||
\
|
||||
asm ("vpunpckhdq xmm"tostr(o2)", xmm"tostr(i0)", xmm"tostr(i2)"");\
|
||||
asm ("vpunpckhdq xmm"tostr(o3)", xmm"tostr(o1)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpckldq xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(i2)"");\
|
||||
asm ("vpunpckldq xmm"tostr(o1)", xmm"tostr(o1)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Step 2
|
||||
* input are two 512-bit states with two rows in one xmm
|
||||
* output are two 512-bit states with one row of each state in one xmm
|
||||
* inputs: i0-i3 = P, i4-i7 = Q
|
||||
* outputs: (i0, o1-o7) = (P|Q)
|
||||
* possible reassignments: (output reg = input reg)
|
||||
* * i1 -> o3-7
|
||||
* * i2 -> o5-7
|
||||
* * i3 -> o7
|
||||
* * i4 -> o3-7
|
||||
* * i5 -> o6-7
|
||||
*/
|
||||
#define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
|
||||
asm ("vpunpckhqdq xmm"tostr(o1)", xmm"tostr(i0)", xmm"tostr(i4)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(i4)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(o2)", xmm"tostr(i1)", xmm"tostr(i5)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(o3)", xmm"tostr(i1)", xmm"tostr(i5)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(o4)", xmm"tostr(i2)", xmm"tostr(i6)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(o5)", xmm"tostr(i2)", xmm"tostr(i6)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(o6)", xmm"tostr(i3)", xmm"tostr(i7)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(o7)", xmm"tostr(i3)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Inverse Step 2
|
||||
* input are two 512-bit states with one row of each state in one xmm
|
||||
* output are two 512-bit states with two rows in one xmm
|
||||
* inputs: i0-i7 = (P|Q)
|
||||
* outputs: (i0, i2, i4, i6) = P, (o0-o3) = Q
|
||||
*/
|
||||
#define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
|
||||
asm ("vpunpckhqdq xmm"tostr(o0)", xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(o1)", xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i2)", xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(o2)", xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i4)", xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(o3)", xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i6)", xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Step 2
|
||||
* input is one 512-bit state with two rows in one xmm
|
||||
* output is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* inputs: i0,i2,i4,i6 = S
|
||||
* outputs: (i0-7) = (0|S)
|
||||
*/
|
||||
#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
|
||||
asm ("vpxor xmm"tostr(t0)", xmm"tostr(t0)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(i1)", xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(i3)", xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i2)", xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(i5)", xmm"tostr(i4)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i4)", xmm"tostr(i4)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpckhqdq xmm"tostr(i7)", xmm"tostr(i6)", xmm"tostr(t0)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i6)", xmm"tostr(i6)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Inverse Step 2
|
||||
* input is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* output is one 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i7 = (0|S)
|
||||
* outputs: (i0, i2, i4, i6) = S
|
||||
*/
|
||||
#define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
|
||||
asm ("vpunpcklqdq xmm"tostr(i0)", xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i2)", xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i4)", xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("vpunpcklqdq xmm"tostr(i6)", xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
|
||||
void INIT256(u64* h)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
asm volatile ("emms");
|
||||
|
||||
/* load IV into registers xmm12 - xmm15 */
|
||||
asm ("vmovaps xmm12, [rdi+0*16]");
|
||||
asm ("vmovaps xmm13, [rdi+1*16]");
|
||||
asm ("vmovaps xmm14, [rdi+2*16]");
|
||||
asm ("vmovaps xmm15, [rdi+3*16]");
|
||||
|
||||
/* transform chaining value from column ordering into row ordering */
|
||||
/* we put two rows (64 bit) of the IV into one 128-bit XMM register */
|
||||
Matrix_Transpose_A(12, 13, 14, 15, 2, 6, 7, 0);
|
||||
|
||||
/* store transposed IV */
|
||||
asm ("vmovaps [rdi+0*16], xmm12");
|
||||
asm ("vmovaps [rdi+1*16], xmm2");
|
||||
asm ("vmovaps [rdi+2*16], xmm6");
|
||||
asm ("vmovaps [rdi+3*16], xmm7");
|
||||
|
||||
asm volatile ("emms");
|
||||
asm (".att_syntax noprefix");
|
||||
}
|
||||
|
||||
void TF512(u64* h, u64* m)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
/* message M in rsi */
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_START;
|
||||
#endif
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
Push_All_Regs();
|
||||
|
||||
/* load message into registers xmm12 - xmm15 (Q = message) */
|
||||
asm ("vmovaps xmm12, [rsi+0*16]");
|
||||
asm ("vmovaps xmm13, [rsi+1*16]");
|
||||
asm ("vmovaps xmm14, [rsi+2*16]");
|
||||
asm ("vmovaps xmm15, [rsi+3*16]");
|
||||
|
||||
/* transform message M from column ordering into row ordering */
|
||||
/* we first put two rows (64 bit) of the message into one 128-bit xmm register */
|
||||
Matrix_Transpose_A(12, 13, 14, 15, 2, 6, 7, 0);
|
||||
|
||||
/* load previous chaining value and xor message to CV to get input of P */
|
||||
/* we first put two rows (2x64 bit) of the CV into one 128-bit xmm register */
|
||||
/* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
|
||||
asm ("vpxor xmm8, xmm12, [rdi+0*16]");
|
||||
asm ("vpxor xmm0, xmm2, [rdi+1*16]");
|
||||
asm ("vpxor xmm4, xmm6, [rdi+2*16]");
|
||||
asm ("vpxor xmm5, xmm7, [rdi+3*16]");
|
||||
|
||||
/* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
|
||||
/* result: the 8 rows of P and Q in xmm8 - xmm12 */
|
||||
Matrix_Transpose_B(8, 0, 4, 5, 12, 2, 6, 7, 9, 10, 11, 12, 13, 14, 15);
|
||||
|
||||
/* compute the two permutations P and Q in parallel */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P or two rows of Q in one xmm register */
|
||||
Matrix_Transpose_B_INV(8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3);
|
||||
|
||||
/* xor output of P and Q */
|
||||
/* result: P(CV+M)+Q(M) in xmm0...xmm3 */
|
||||
asm ("vpxor xmm0, xmm0, xmm8");
|
||||
asm ("vpxor xmm1, xmm1, xmm10");
|
||||
asm ("vpxor xmm2, xmm2, xmm12");
|
||||
asm ("vpxor xmm3, xmm3, xmm14");
|
||||
|
||||
/* xor CV (feed-forward) */
|
||||
/* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
|
||||
asm ("vpxor xmm0, xmm0, [rdi+0*16]");
|
||||
asm ("vpxor xmm1, xmm1, [rdi+1*16]");
|
||||
asm ("vpxor xmm2, xmm2, [rdi+2*16]");
|
||||
asm ("vpxor xmm3, xmm3, [rdi+3*16]");
|
||||
|
||||
/* store CV */
|
||||
asm ("vmovaps [rdi+0*16], xmm0");
|
||||
asm ("vmovaps [rdi+1*16], xmm1");
|
||||
asm ("vmovaps [rdi+2*16], xmm2");
|
||||
asm ("vmovaps [rdi+3*16], xmm3");
|
||||
|
||||
Pop_All_Regs();
|
||||
asm (".att_syntax noprefix");
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_END;
|
||||
#endif
|
||||
return;
|
||||
}
|
||||
|
||||
void OF512(u64* h)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
Push_All_Regs();
|
||||
|
||||
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
|
||||
asm ("vmovaps xmm8, [rdi+0*16]");
|
||||
asm ("vmovaps xmm10, [rdi+1*16]");
|
||||
asm ("vmovaps xmm12, [rdi+2*16]");
|
||||
asm ("vmovaps xmm14, [rdi+3*16]");
|
||||
|
||||
/* there are now 2 rows of the CV in one xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) into one half of an xmm register */
|
||||
/* result: the 8 input rows of P in xmm8 - xmm15 */
|
||||
Matrix_Transpose_O_B(8, 9, 10, 11, 12, 13, 14, 15, 0);
|
||||
|
||||
/* compute the permutation P */
|
||||
/* result: the output of P(CV) in xmm8 - xmm15 */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P in one xmm register */
|
||||
/* result: P(CV) in xmm8, xmm10, xmm12, xmm14 */
|
||||
Matrix_Transpose_O_B_INV(8, 9, 10, 11, 12, 13, 14, 15);
|
||||
|
||||
/* xor CV to P output (feed-forward) */
|
||||
/* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
|
||||
asm ("vpxor xmm8, xmm8, [rdi+0*16]");
|
||||
asm ("vpxor xmm10, xmm10, [rdi+1*16]");
|
||||
asm ("vpxor xmm12, xmm12, [rdi+2*16]");
|
||||
asm ("vpxor xmm14, xmm14, [rdi+3*16]");
|
||||
|
||||
/* transform state back from row ordering into column ordering */
|
||||
/* result: final hash value in xmm9, xmm11 */
|
||||
Matrix_Transpose_A(8, 10, 12, 14, 4, 9, 11, 0);
|
||||
|
||||
/* we only need to return the truncated half of the state */
|
||||
asm ("vmovaps [rdi+2*16], xmm9");
|
||||
asm ("vmovaps [rdi+3*16], xmm11");
|
||||
|
||||
Pop_All_Regs();
|
||||
asm (".att_syntax noprefix");
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
@@ -1,856 +0,0 @@
|
||||
/* groestl-asm-vperm.h Aug 2011
|
||||
*
|
||||
* Groestl implementation with inline assembly using ssse3 instructions.
|
||||
* Author: Günther A. Roland, Martin Schläffer, Krystian Matusiewicz
|
||||
*
|
||||
* Based on the vperm and aes_ni implementations of the hash function Groestl
|
||||
* by Cagdas Calik <ccalik@metu.edu.tr> http://www.metu.edu.tr/~ccalik/
|
||||
* Institute of Applied Mathematics, Middle East Technical University, Turkey
|
||||
*
|
||||
* This code is placed in the public domain
|
||||
*/
|
||||
|
||||
#include "hash-groestl256.h"
|
||||
|
||||
/* global constants */
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_Lx[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_L0[ROUNDS512*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_L7[ROUNDS512*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_P[ROUNDS1024*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ROUND_CONST_Q[ROUNDS1024*16];
|
||||
__attribute__ ((aligned (16))) unsigned char TRANSP_MASK[16];
|
||||
__attribute__ ((aligned (16))) unsigned char SUBSH_MASK[8*16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_0F[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_15[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_1B[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_63[16];
|
||||
__attribute__ ((aligned (16))) unsigned char ALL_FF[16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_IPT[2*16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_OPT[2*16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_INV[2*16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_SB1[2*16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_SB2[2*16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_SB4[2*16];
|
||||
__attribute__ ((aligned (16))) unsigned char VPERM_SBO[2*16];
|
||||
|
||||
/* temporary variables */
|
||||
__attribute__ ((aligned (16))) unsigned char TEMP_MUL1[8*16];
|
||||
__attribute__ ((aligned (16))) unsigned char TEMP_MUL2[8*16];
|
||||
__attribute__ ((aligned (16))) unsigned char TEMP_MUL4[1*16];
|
||||
__attribute__ ((aligned (16))) unsigned char QTEMP[8*16];
|
||||
__attribute__ ((aligned (16))) unsigned char TEMP[8*16];
|
||||
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
|
||||
#define SET_SHARED_CONSTANTS(){\
|
||||
((u64*)TRANSP_MASK)[0] = 0x0d0509010c040800ULL;\
|
||||
((u64*)TRANSP_MASK)[1] = 0x0f070b030e060a02ULL;\
|
||||
((u64*)ALL_1B)[0] = 0x1b1b1b1b1b1b1b1bULL;\
|
||||
((u64*)ALL_1B)[1] = 0x1b1b1b1b1b1b1b1bULL;\
|
||||
((u64*)ALL_63)[ 0] = 0x6363636363636363ULL;\
|
||||
((u64*)ALL_63)[ 1] = 0x6363636363636363ULL;\
|
||||
((u64*)ALL_0F)[ 0] = 0x0F0F0F0F0F0F0F0FULL;\
|
||||
((u64*)ALL_0F)[ 1] = 0x0F0F0F0F0F0F0F0FULL;\
|
||||
((u64*)VPERM_IPT)[ 0] = 0x4C01307D317C4D00ULL;\
|
||||
((u64*)VPERM_IPT)[ 1] = 0xCD80B1FCB0FDCC81ULL;\
|
||||
((u64*)VPERM_IPT)[ 2] = 0xC2B2E8985A2A7000ULL;\
|
||||
((u64*)VPERM_IPT)[ 3] = 0xCABAE09052227808ULL;\
|
||||
((u64*)VPERM_OPT)[ 0] = 0x01EDBD5150BCEC00ULL;\
|
||||
((u64*)VPERM_OPT)[ 1] = 0xE10D5DB1B05C0CE0ULL;\
|
||||
((u64*)VPERM_OPT)[ 2] = 0xFF9F4929D6B66000ULL;\
|
||||
((u64*)VPERM_OPT)[ 3] = 0xF7974121DEBE6808ULL;\
|
||||
((u64*)VPERM_INV)[ 0] = 0x01040A060F0B0780ULL;\
|
||||
((u64*)VPERM_INV)[ 1] = 0x030D0E0C02050809ULL;\
|
||||
((u64*)VPERM_INV)[ 2] = 0x0E05060F0D080180ULL;\
|
||||
((u64*)VPERM_INV)[ 3] = 0x040703090A0B0C02ULL;\
|
||||
((u64*)VPERM_SB1)[ 0] = 0x3618D415FAE22300ULL;\
|
||||
((u64*)VPERM_SB1)[ 1] = 0x3BF7CCC10D2ED9EFULL;\
|
||||
((u64*)VPERM_SB1)[ 2] = 0xB19BE18FCB503E00ULL;\
|
||||
((u64*)VPERM_SB1)[ 3] = 0xA5DF7A6E142AF544ULL;\
|
||||
((u64*)VPERM_SB2)[ 0] = 0x69EB88400AE12900ULL;\
|
||||
((u64*)VPERM_SB2)[ 1] = 0xC2A163C8AB82234AULL;\
|
||||
((u64*)VPERM_SB2)[ 2] = 0xE27A93C60B712400ULL;\
|
||||
((u64*)VPERM_SB2)[ 3] = 0x5EB7E955BC982FCDULL;\
|
||||
((u64*)VPERM_SB4)[ 0] = 0x3D50AED7C393EA00ULL;\
|
||||
((u64*)VPERM_SB4)[ 1] = 0xBA44FE79876D2914ULL;\
|
||||
((u64*)VPERM_SB4)[ 2] = 0xE1E937A03FD64100ULL;\
|
||||
((u64*)VPERM_SB4)[ 3] = 0xA876DE9749087E9FULL;\
|
||||
/*((u64*)VPERM_SBO)[ 0] = 0xCFE474A55FBB6A00ULL;\
|
||||
((u64*)VPERM_SBO)[ 1] = 0x8E1E90D1412B35FAULL;\
|
||||
((u64*)VPERM_SBO)[ 2] = 0xD0D26D176FBDC700ULL;\
|
||||
((u64*)VPERM_SBO)[ 3] = 0x15AABF7AC502A878ULL;*/\
|
||||
((u64*)ALL_15)[ 0] = 0x1515151515151515ULL;\
|
||||
((u64*)ALL_15)[ 1] = 0x1515151515151515ULL;\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Transform w/o settings c*
|
||||
* transforms 2 rows to/from "vperm mode"
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0, a1 = 2 rows
|
||||
* table = transformation table to use
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* a0, a1 = 2 rows transformed with table
|
||||
* */
|
||||
#define VPERM_Transform_No_Const(a0, a1, t0, t1, t2, t3, c0, c1, c2){\
|
||||
asm ("movdqa xmm"tostr(t0)", xmm"tostr(c0)"");\
|
||||
asm ("movdqa xmm"tostr(t1)", xmm"tostr(c0)"");\
|
||||
asm ("pandn xmm"tostr(t0)", xmm"tostr(a0)"");\
|
||||
asm ("pandn xmm"tostr(t1)", xmm"tostr(a1)"");\
|
||||
asm ("psrld xmm"tostr(t0)", 4");\
|
||||
asm ("psrld xmm"tostr(t1)", 4");\
|
||||
asm ("pand xmm"tostr(a0)", xmm"tostr(c0)"");\
|
||||
asm ("pand xmm"tostr(a1)", xmm"tostr(c0)"");\
|
||||
asm ("movdqa xmm"tostr(t2)", xmm"tostr(c2)"");\
|
||||
asm ("movdqa xmm"tostr(t3)", xmm"tostr(c2)"");\
|
||||
asm ("pshufb xmm"tostr(t2)", xmm"tostr(a0)"");\
|
||||
asm ("pshufb xmm"tostr(t3)", xmm"tostr(a1)"");\
|
||||
asm ("movdqa xmm"tostr(a0)", xmm"tostr(c1)"");\
|
||||
asm ("movdqa xmm"tostr(a1)", xmm"tostr(c1)"");\
|
||||
asm ("pshufb xmm"tostr(a0)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(a1)", xmm"tostr(t1)"");\
|
||||
asm ("pxor xmm"tostr(a0)", xmm"tostr(t2)"");\
|
||||
asm ("pxor xmm"tostr(a1)", xmm"tostr(t3)"");\
|
||||
}/**/
|
||||
|
||||
#define VPERM_Transform_Set_Const(table, c0, c1, c2){\
|
||||
asm ("movaps xmm"tostr(c0)", [ALL_0F]");\
|
||||
asm ("movaps xmm"tostr(c1)", ["tostr(table)"+0*16]");\
|
||||
asm ("movaps xmm"tostr(c2)", ["tostr(table)"+1*16]");\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Transform
|
||||
* transforms 2 rows to/from "vperm mode"
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0, a1 = 2 rows
|
||||
* table = transformation table to use
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* a0, a1 = 2 rows transformed with table
|
||||
* */
|
||||
#define VPERM_Transform(a0, a1, table, t0, t1, t2, t3, c0, c1, c2){\
|
||||
VPERM_Transform_Set_Const(table, c0, c1, c2);\
|
||||
VPERM_Transform_No_Const(a0, a1, t0, t1, t2, t3, c0, c1, c2);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Transform State
|
||||
* inputs:
|
||||
* a0-a3 = state
|
||||
* table = transformation table to use
|
||||
* t* = clobbers
|
||||
* outputs:
|
||||
* a0-a3 = transformed state
|
||||
* */
|
||||
#define VPERM_Transform_State(a0, a1, a2, a3, table, t0, t1, t2, t3, c0, c1, c2){\
|
||||
VPERM_Transform_Set_Const(table, c0, c1, c2);\
|
||||
VPERM_Transform_No_Const(a0, a1, t0, t1, t2, t3, c0, c1, c2);\
|
||||
VPERM_Transform_No_Const(a2, a3, t0, t1, t2, t3, c0, c1, c2);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Add Constant to State
|
||||
* inputs:
|
||||
* a0-a7 = state
|
||||
* constant = constant to add
|
||||
* t0 = clobber
|
||||
* outputs:
|
||||
* a0-a7 = state + constant
|
||||
* */
|
||||
#define VPERM_Add_Constant(a0, a1, a2, a3, a4, a5, a6, a7, constant, t0){\
|
||||
asm ("movaps xmm"tostr(t0)", ["tostr(constant)"]");\
|
||||
asm ("pxor xmm"tostr(a0)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a1)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a2)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a3)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a4)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a5)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a6)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(a7)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Set Substitute Core Constants
|
||||
* */
|
||||
#define VPERM_Substitute_Core_Set_Const(c0, c1, c2){\
|
||||
VPERM_Transform_Set_Const(VPERM_INV, c0, c1, c2);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Substitute Core
|
||||
* first part of sbox inverse computation
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0 = 1 row
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* b0a, b0b = inputs for lookup step
|
||||
* */
|
||||
#define VPERM_Substitute_Core(a0, b0a, b0b, t0, t1, c0, c1, c2){\
|
||||
asm ("movdqa xmm"tostr(t0)", xmm"tostr(c0)"");\
|
||||
asm ("pandn xmm"tostr(t0)", xmm"tostr(a0)"");\
|
||||
asm ("psrld xmm"tostr(t0)", 4");\
|
||||
asm ("pand xmm"tostr(a0)", xmm"tostr(c0)"");\
|
||||
asm ("movdqa xmm"tostr(b0a)", "tostr(c1)"");\
|
||||
asm ("pshufb xmm"tostr(b0a)", xmm"tostr(a0)"");\
|
||||
asm ("pxor xmm"tostr(a0)", xmm"tostr(t0)"");\
|
||||
asm ("movdqa xmm"tostr(b0b)", xmm"tostr(c2)"");\
|
||||
asm ("pshufb xmm"tostr(b0b)", xmm"tostr(t0)"");\
|
||||
asm ("pxor xmm"tostr(b0b)", xmm"tostr(b0a)"");\
|
||||
asm ("movdqa xmm"tostr(t1)", xmm"tostr(c2)"");\
|
||||
asm ("pshufb xmm"tostr(t1)", xmm"tostr(a0)"");\
|
||||
asm ("pxor xmm"tostr(t1)", xmm"tostr(b0a)"");\
|
||||
asm ("movdqa xmm"tostr(b0a)", xmm"tostr(c2)"");\
|
||||
asm ("pshufb xmm"tostr(b0a)", xmm"tostr(b0b)"");\
|
||||
asm ("pxor xmm"tostr(b0a)", xmm"tostr(a0)"");\
|
||||
asm ("movdqa xmm"tostr(b0b)", xmm"tostr(c2)"");\
|
||||
asm ("pshufb xmm"tostr(b0b)", xmm"tostr(t1)"");\
|
||||
asm ("pxor xmm"tostr(b0b)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Lookup
|
||||
* second part of sbox inverse computation
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0a, a0b = output of Substitution Core
|
||||
* table = lookup table to use (*1 / *2 / *4)
|
||||
* t0 = clobber
|
||||
* outputs:
|
||||
* b0 = output of sbox + multiplication
|
||||
* */
|
||||
#define VPERM_Lookup(a0a, a0b, table, b0, t0){\
|
||||
asm ("movaps xmm"tostr(b0)", ["tostr(table)"+0*16]");\
|
||||
asm ("movaps xmm"tostr(t0)", ["tostr(table)"+1*16]");\
|
||||
asm ("pshufb xmm"tostr(b0)", xmm"tostr(a0b)"");\
|
||||
asm ("pshufb xmm"tostr(t0)", xmm"tostr(a0a)"");\
|
||||
asm ("pxor xmm"tostr(b0)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* SubBytes and *2 / *4
|
||||
* this function is derived from:
|
||||
* Constant-time SSSE3 AES core implementation
|
||||
* by Mike Hamburg
|
||||
* and
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0-a7 = state
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* a0-a7 = state * 4
|
||||
* c2 = row0 * 2 -> b0
|
||||
* c1 = row7 * 2 -> b3
|
||||
* c0 = row7 * 1 -> b4
|
||||
* t2 = row4 * 1 -> b7
|
||||
* TEMP_MUL1 = row(i) * 1
|
||||
* TEMP_MUL2 = row(i) * 2
|
||||
*
|
||||
* call:VPERM_SUB_MULTIPLY(a0, a1, a2, a3, a4, a5, a6, a7, b1, b2, b5, b6, b0, b3, b4, b7) */
|
||||
#define VPERM_SUB_MULTIPLY(a0, a1, a2, a3, a4, a5, a6, a7, t0, t1, t3, t4, c2, c1, c0, t2){\
|
||||
/* set Constants */\
|
||||
VPERM_Substitute_Core_Set_Const(c0, c1, c2);\
|
||||
/* row 1 */\
|
||||
VPERM_Substitute_Core(a1, t0, t1, t3, t4, c0, xmm##c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
asm ("movaps [TEMP_MUL1+1*16], xmm"tostr(t2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
asm ("movaps [TEMP_MUL2+1*16], xmm"tostr(t3)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a1, t4);\
|
||||
/* --- */\
|
||||
/* row 2 */\
|
||||
VPERM_Substitute_Core(a2, t0, t1, t3, t4, c0, xmm##c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
asm ("movaps [TEMP_MUL1+2*16], xmm"tostr(t2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
asm ("movaps [TEMP_MUL2+2*16], xmm"tostr(t3)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a2, t4);\
|
||||
/* --- */\
|
||||
/* row 3 */\
|
||||
VPERM_Substitute_Core(a3, t0, t1, t3, t4, c0, xmm##c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
asm ("movaps [TEMP_MUL1+3*16], xmm"tostr(t2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
asm ("movaps [TEMP_MUL2+3*16], xmm"tostr(t3)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a3, t4);\
|
||||
/* --- */\
|
||||
/* row 5 */\
|
||||
VPERM_Substitute_Core(a5, t0, t1, t3, t4, c0, xmm##c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
asm ("movaps [TEMP_MUL1+5*16], xmm"tostr(t2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
asm ("movaps [TEMP_MUL2+5*16], xmm"tostr(t3)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a5, t4);\
|
||||
/* --- */\
|
||||
/* row 6 */\
|
||||
VPERM_Substitute_Core(a6, t0, t1, t3, t4, c0, xmm##c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
asm ("movaps [TEMP_MUL1+6*16], xmm"tostr(t2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
asm ("movaps [TEMP_MUL2+6*16], xmm"tostr(t3)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a6, t4);\
|
||||
/* --- */\
|
||||
/* row 7 */\
|
||||
VPERM_Substitute_Core(a7, t0, t1, t3, t4, c0, xmm##c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
asm ("movaps [TEMP_MUL1+7*16], xmm"tostr(t2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, c1, t4); /*c1 -> b3*/\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a7, t4);\
|
||||
/* --- */\
|
||||
/* row 4 */\
|
||||
VPERM_Substitute_Core(a4, t0, t1, t3, t4, c0, [VPERM_INV+0*16], c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4); /*t2 -> b7*/\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
asm ("movaps [TEMP_MUL2+4*16], xmm"tostr(t3)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a4, t4);\
|
||||
/* --- */\
|
||||
/* row 0 */\
|
||||
VPERM_Substitute_Core(a0, t0, t1, t3, t4, c0, [VPERM_INV+0*16], c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, c0, t4); /*c0 -> b4*/\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, c2, t4); /*c2 -> b0*/\
|
||||
asm ("movaps [TEMP_MUL2+0*16], xmm"tostr(c2)"");\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a0, t4);\
|
||||
/* --- */\
|
||||
}/**/
|
||||
|
||||
|
||||
/* Optimized MixBytes
|
||||
* inputs:
|
||||
* a0-a7 = (row0-row7) * 4
|
||||
* b0 = row0 * 2
|
||||
* b3 = row7 * 2
|
||||
* b4 = row7 * 1
|
||||
* b7 = row4 * 1
|
||||
* all *1 and *2 values must also be in TEMP_MUL1, TEMP_MUL2
|
||||
* output: b0-b7
|
||||
* */
|
||||
#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* save one value */\
|
||||
asm ("movaps [TEMP_MUL4], xmm"tostr(a3)"");\
|
||||
/* 1 */\
|
||||
asm ("movdqa xmm"tostr(b1)", xmm"tostr(a0)"");\
|
||||
asm ("pxor xmm"tostr(b1)", xmm"tostr(a5)"");\
|
||||
asm ("pxor xmm"tostr(b1)", xmm"tostr(b4)""); /* -> helper! */\
|
||||
asm ("pxor xmm"tostr(b1)", [TEMP_MUL2+3*16]");\
|
||||
asm ("movdqa xmm"tostr(b2)", xmm"tostr(b1)"");\
|
||||
\
|
||||
/* 2 */\
|
||||
asm ("movdqa xmm"tostr(b5)", xmm"tostr(a1)"");\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(a4)"");\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(b7)""); /* -> helper! */\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(b3)""); /* -> helper! */\
|
||||
asm ("movdqa xmm"tostr(b6)", xmm"tostr(b5)"");\
|
||||
\
|
||||
/* 4 */\
|
||||
asm ("pxor xmm"tostr(b7)", xmm"tostr(a6)"");\
|
||||
/*asm ("pxor xmm"tostr(b7)", [TEMP_MUL1+4*16]"); -> helper! */\
|
||||
asm ("pxor xmm"tostr(b7)", [TEMP_MUL1+6*16]");\
|
||||
asm ("pxor xmm"tostr(b7)", [TEMP_MUL2+1*16]");\
|
||||
asm ("pxor xmm"tostr(b7)", xmm"tostr(b3)""); /* -> helper! */\
|
||||
asm ("pxor xmm"tostr(b2)", xmm"tostr(b7)"");\
|
||||
\
|
||||
/* 3 */\
|
||||
asm ("pxor xmm"tostr(b0)", xmm"tostr(a7)"");\
|
||||
asm ("pxor xmm"tostr(b0)", [TEMP_MUL1+5*16]");\
|
||||
asm ("pxor xmm"tostr(b0)", [TEMP_MUL1+7*16]");\
|
||||
/*asm ("pxor xmm"tostr(b0)", [TEMP_MUL2+0*16]"); -> helper! */\
|
||||
asm ("pxor xmm"tostr(b0)", [TEMP_MUL2+2*16]");\
|
||||
asm ("movdqa xmm"tostr(b3)", xmm"tostr(b0)"");\
|
||||
asm ("pxor xmm"tostr(b1)", xmm"tostr(b0)"");\
|
||||
asm ("pxor xmm"tostr(b0)", xmm"tostr(b7)""); /* moved from 4 */\
|
||||
\
|
||||
/* 5 */\
|
||||
asm ("pxor xmm"tostr(b4)", xmm"tostr(a2)"");\
|
||||
/*asm ("pxor xmm"tostr(b4)", [TEMP_MUL1+0*16]"); -> helper! */\
|
||||
asm ("pxor xmm"tostr(b4)", [TEMP_MUL1+2*16]");\
|
||||
asm ("pxor xmm"tostr(b4)", [TEMP_MUL2+3*16]");\
|
||||
asm ("pxor xmm"tostr(b4)", [TEMP_MUL2+5*16]");\
|
||||
asm ("pxor xmm"tostr(b3)", xmm"tostr(b4)"");\
|
||||
asm ("pxor xmm"tostr(b6)", xmm"tostr(b4)"");\
|
||||
\
|
||||
/* 6 */\
|
||||
asm ("pxor xmm"tostr(a3)", [TEMP_MUL1+1*16]");\
|
||||
asm ("pxor xmm"tostr(a3)", [TEMP_MUL1+3*16]");\
|
||||
asm ("pxor xmm"tostr(a3)", [TEMP_MUL2+4*16]");\
|
||||
asm ("pxor xmm"tostr(a3)", [TEMP_MUL2+6*16]");\
|
||||
asm ("pxor xmm"tostr(b4)", xmm"tostr(a3)"");\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(a3)"");\
|
||||
asm ("pxor xmm"tostr(b7)", xmm"tostr(a3)"");\
|
||||
\
|
||||
/* 7 */\
|
||||
asm ("pxor xmm"tostr(a1)", [TEMP_MUL1+1*16]");\
|
||||
asm ("pxor xmm"tostr(a1)", [TEMP_MUL2+4*16]");\
|
||||
asm ("pxor xmm"tostr(b2)", xmm"tostr(a1)"");\
|
||||
asm ("pxor xmm"tostr(b3)", xmm"tostr(a1)"");\
|
||||
\
|
||||
/* 8 */\
|
||||
asm ("pxor xmm"tostr(a5)", [TEMP_MUL1+5*16]");\
|
||||
asm ("pxor xmm"tostr(a5)", [TEMP_MUL2+0*16]");\
|
||||
asm ("pxor xmm"tostr(b6)", xmm"tostr(a5)"");\
|
||||
asm ("pxor xmm"tostr(b7)", xmm"tostr(a5)"");\
|
||||
\
|
||||
/* 9 */\
|
||||
asm ("movaps xmm"tostr(a3)", [TEMP_MUL1+2*16]");\
|
||||
asm ("pxor xmm"tostr(a3)", [TEMP_MUL2+5*16]");\
|
||||
asm ("pxor xmm"tostr(b0)", xmm"tostr(a3)"");\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(a3)"");\
|
||||
\
|
||||
/* 10 */\
|
||||
asm ("movaps xmm"tostr(a1)", [TEMP_MUL1+6*16]");\
|
||||
asm ("pxor xmm"tostr(a1)", [TEMP_MUL2+1*16]");\
|
||||
asm ("pxor xmm"tostr(b1)", xmm"tostr(a1)"");\
|
||||
asm ("pxor xmm"tostr(b4)", xmm"tostr(a1)"");\
|
||||
\
|
||||
/* 11 */\
|
||||
asm ("movaps xmm"tostr(a5)", [TEMP_MUL1+3*16]");\
|
||||
asm ("pxor xmm"tostr(a5)", [TEMP_MUL2+6*16]");\
|
||||
asm ("pxor xmm"tostr(b1)", xmm"tostr(a5)"");\
|
||||
asm ("pxor xmm"tostr(b6)", xmm"tostr(a5)"");\
|
||||
\
|
||||
/* 12 */\
|
||||
asm ("movaps xmm"tostr(a3)", [TEMP_MUL1+7*16]");\
|
||||
asm ("pxor xmm"tostr(a3)", [TEMP_MUL2+2*16]");\
|
||||
asm ("pxor xmm"tostr(b2)", xmm"tostr(a3)"");\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(a3)"");\
|
||||
\
|
||||
/* 13 */\
|
||||
asm ("pxor xmm"tostr(b0)", [TEMP_MUL4]");\
|
||||
asm ("pxor xmm"tostr(b0)", xmm"tostr(a4)"");\
|
||||
asm ("pxor xmm"tostr(b1)", xmm"tostr(a4)"");\
|
||||
asm ("pxor xmm"tostr(b3)", xmm"tostr(a6)"");\
|
||||
asm ("pxor xmm"tostr(b4)", xmm"tostr(a0)"");\
|
||||
asm ("pxor xmm"tostr(b4)", xmm"tostr(a7)"");\
|
||||
asm ("pxor xmm"tostr(b5)", xmm"tostr(a0)"");\
|
||||
asm ("pxor xmm"tostr(b7)", xmm"tostr(a2)"");\
|
||||
}/**/
|
||||
|
||||
//#if (LENGTH <= 256)
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
SET_SHARED_CONSTANTS();\
|
||||
((u64*)SUBSH_MASK)[ 0] = 0x0706050403020100ULL;\
|
||||
((u64*)SUBSH_MASK)[ 1] = 0x080f0e0d0c0b0a09ULL;\
|
||||
((u64*)SUBSH_MASK)[ 2] = 0x0007060504030201ULL;\
|
||||
((u64*)SUBSH_MASK)[ 3] = 0x0a09080f0e0d0c0bULL;\
|
||||
((u64*)SUBSH_MASK)[ 4] = 0x0100070605040302ULL;\
|
||||
((u64*)SUBSH_MASK)[ 5] = 0x0c0b0a09080f0e0dULL;\
|
||||
((u64*)SUBSH_MASK)[ 6] = 0x0201000706050403ULL;\
|
||||
((u64*)SUBSH_MASK)[ 7] = 0x0e0d0c0b0a09080fULL;\
|
||||
((u64*)SUBSH_MASK)[ 8] = 0x0302010007060504ULL;\
|
||||
((u64*)SUBSH_MASK)[ 9] = 0x0f0e0d0c0b0a0908ULL;\
|
||||
((u64*)SUBSH_MASK)[10] = 0x0403020100070605ULL;\
|
||||
((u64*)SUBSH_MASK)[11] = 0x09080f0e0d0c0b0aULL;\
|
||||
((u64*)SUBSH_MASK)[12] = 0x0504030201000706ULL;\
|
||||
((u64*)SUBSH_MASK)[13] = 0x0b0a09080f0e0d0cULL;\
|
||||
((u64*)SUBSH_MASK)[14] = 0x0605040302010007ULL;\
|
||||
((u64*)SUBSH_MASK)[15] = 0x0d0c0b0a09080f0eULL;\
|
||||
for(i = 0; i < ROUNDS512; i++)\
|
||||
{\
|
||||
((u64*)ROUND_CONST_L0)[i*2+1] = 0xffffffffffffffffULL;\
|
||||
((u64*)ROUND_CONST_L0)[i*2+0] = (i * 0x0101010101010101ULL) ^ 0x7060504030201000ULL;\
|
||||
((u64*)ROUND_CONST_L7)[i*2+1] = (i * 0x0101010101010101ULL) ^ 0x8f9fafbfcfdfefffULL;\
|
||||
((u64*)ROUND_CONST_L7)[i*2+0] = 0x0000000000000000ULL;\
|
||||
}\
|
||||
((u64*)ROUND_CONST_Lx)[1] = 0xffffffffffffffffULL;\
|
||||
((u64*)ROUND_CONST_Lx)[0] = 0x0000000000000000ULL;\
|
||||
}/**/
|
||||
|
||||
#define Push_All_Regs(){\
|
||||
/* not using any...
|
||||
asm("push rax");\
|
||||
asm("push rbx");\
|
||||
asm("push rcx");*/\
|
||||
}/**/
|
||||
|
||||
#define Pop_All_Regs(){\
|
||||
/* not using any...
|
||||
asm("pop rcx");\
|
||||
asm("pop rbx");\
|
||||
asm("pop rax");*/\
|
||||
}/**/
|
||||
|
||||
|
||||
/* vperm:
|
||||
* transformation before rounds with ipt
|
||||
* first round add transformed constant
|
||||
* middle rounds: add constant XOR 0x15...15
|
||||
* last round: additionally add 0x15...15 after MB
|
||||
* transformation after rounds with opt
|
||||
*/
|
||||
/* one round
|
||||
* i = round number
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
*/
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* AddRoundConstant + ShiftBytes (interleaved) */\
|
||||
asm ("movaps xmm"tostr(b1)", [ROUND_CONST_Lx]");\
|
||||
asm ("pxor xmm"tostr(a0)", [ROUND_CONST_L0+"tostr(i)"*16]");\
|
||||
asm ("pxor xmm"tostr(a1)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a2)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a3)", xmm"tostr(b1)"");\
|
||||
asm ("pshufb xmm"tostr(a0)", [SUBSH_MASK+0*16]");\
|
||||
asm ("pshufb xmm"tostr(a1)", [SUBSH_MASK+1*16]");\
|
||||
asm ("pxor xmm"tostr(a4)", xmm"tostr(b1)"");\
|
||||
asm ("pshufb xmm"tostr(a2)", [SUBSH_MASK+2*16]");\
|
||||
asm ("pshufb xmm"tostr(a3)", [SUBSH_MASK+3*16]");\
|
||||
asm ("pxor xmm"tostr(a5)", xmm"tostr(b1)"");\
|
||||
asm ("pxor xmm"tostr(a6)", xmm"tostr(b1)"");\
|
||||
asm ("pshufb xmm"tostr(a4)", [SUBSH_MASK+4*16]");\
|
||||
asm ("pshufb xmm"tostr(a5)", [SUBSH_MASK+5*16]");\
|
||||
asm ("pxor xmm"tostr(a7)", [ROUND_CONST_L7+"tostr(i)"*16]");\
|
||||
asm ("pshufb xmm"tostr(a6)", [SUBSH_MASK+6*16]");\
|
||||
asm ("pshufb xmm"tostr(a7)", [SUBSH_MASK+7*16]");\
|
||||
/* SubBytes + Multiplication by 2 and 4 */\
|
||||
VPERM_SUB_MULTIPLY(a0, a1, a2, a3, a4, a5, a6, a7, b1, b2, b5, b6, b0, b3, b4, b7);\
|
||||
/* MixBytes */\
|
||||
MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
|
||||
}/**/
|
||||
|
||||
/* 10 rounds, P and Q in parallel */
|
||||
#define ROUNDS_P_Q(){\
|
||||
VPERM_Add_Constant(8, 9, 10, 11, 12, 13, 14, 15, ALL_15, 0);\
|
||||
ROUND(0, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(2, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(3, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(4, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(5, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(6, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(7, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
ROUND(8, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);\
|
||||
ROUND(9, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);\
|
||||
VPERM_Add_Constant(8, 9, 10, 11, 12, 13, 14, 15, ALL_15, 0);\
|
||||
}
|
||||
|
||||
|
||||
/* Matrix Transpose Step 1
|
||||
* input is a 512-bit state with two columns in one xmm
|
||||
* output is a 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i3
|
||||
* outputs: i0, o1-o3
|
||||
* clobbers: t0
|
||||
*/
|
||||
#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
|
||||
asm ("movaps xmm"tostr(t0)", [TRANSP_MASK]");\
|
||||
\
|
||||
asm ("pshufb xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(i1)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("pshufb xmm"tostr(i3)", xmm"tostr(t0)"");\
|
||||
\
|
||||
asm ("movdqa xmm"tostr(o1)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(t0)", xmm"tostr(i2)"");\
|
||||
\
|
||||
asm ("punpcklwd xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("punpckhwd xmm"tostr(o1)", xmm"tostr(i1)"");\
|
||||
asm ("punpcklwd xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("punpckhwd xmm"tostr(t0)", xmm"tostr(i3)"");\
|
||||
\
|
||||
asm ("pshufd xmm"tostr(i0)", xmm"tostr(i0)", 216");\
|
||||
asm ("pshufd xmm"tostr(o1)", xmm"tostr(o1)", 216");\
|
||||
asm ("pshufd xmm"tostr(i2)", xmm"tostr(i2)", 216");\
|
||||
asm ("pshufd xmm"tostr(t0)", xmm"tostr(t0)", 216");\
|
||||
\
|
||||
asm ("movdqa xmm"tostr(o2)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(o3)", xmm"tostr(o1)"");\
|
||||
\
|
||||
asm ("punpckldq xmm"tostr(i0)", xmm"tostr(i2)"");\
|
||||
asm ("punpckldq xmm"tostr(o1)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhdq xmm"tostr(o2)", xmm"tostr(i2)"");\
|
||||
asm ("punpckhdq xmm"tostr(o3)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Step 2
|
||||
* input are two 512-bit states with two rows in one xmm
|
||||
* output are two 512-bit states with one row of each state in one xmm
|
||||
* inputs: i0-i3 = P, i4-i7 = Q
|
||||
* outputs: (i0, o1-o7) = (P|Q)
|
||||
* possible reassignments: (output reg = input reg)
|
||||
* * i1 -> o3-7
|
||||
* * i2 -> o5-7
|
||||
* * i3 -> o7
|
||||
* * i4 -> o3-7
|
||||
* * i5 -> o6-7
|
||||
*/
|
||||
#define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
|
||||
asm ("movdqa xmm"tostr(o1)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(o2)", xmm"tostr(i1)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(i4)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o1)", xmm"tostr(i4)"");\
|
||||
asm ("movdqa xmm"tostr(o3)", xmm"tostr(i1)"");\
|
||||
asm ("movdqa xmm"tostr(o4)", xmm"tostr(i2)"");\
|
||||
asm ("punpcklqdq xmm"tostr(o2)", xmm"tostr(i5)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o3)", xmm"tostr(i5)"");\
|
||||
asm ("movdqa xmm"tostr(o5)", xmm"tostr(i2)"");\
|
||||
asm ("movdqa xmm"tostr(o6)", xmm"tostr(i3)"");\
|
||||
asm ("punpcklqdq xmm"tostr(o4)", xmm"tostr(i6)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o5)", xmm"tostr(i6)"");\
|
||||
asm ("movdqa xmm"tostr(o7)", xmm"tostr(i3)"");\
|
||||
asm ("punpcklqdq xmm"tostr(o6)", xmm"tostr(i7)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o7)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Inverse Step 2
|
||||
* input are two 512-bit states with one row of each state in one xmm
|
||||
* output are two 512-bit states with two rows in one xmm
|
||||
* inputs: i0-i7 = (P|Q)
|
||||
* outputs: (i0, i2, i4, i6) = P, (o0-o3) = Q
|
||||
*/
|
||||
#define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
|
||||
asm ("movdqa xmm"tostr(o0)", xmm"tostr(i0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o0)", xmm"tostr(i1)"");\
|
||||
asm ("movdqa xmm"tostr(o1)", xmm"tostr(i2)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o1)", xmm"tostr(i3)"");\
|
||||
asm ("movdqa xmm"tostr(o2)", xmm"tostr(i4)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o2)", xmm"tostr(i5)"");\
|
||||
asm ("movdqa xmm"tostr(o3)", xmm"tostr(i6)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
asm ("punpckhqdq xmm"tostr(o3)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Step 2
|
||||
* input is one 512-bit state with two rows in one xmm
|
||||
* output is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* inputs: i0,i2,i4,i6 = S
|
||||
* outputs: (i0-7) = (0|S)
|
||||
*/
|
||||
#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
|
||||
asm ("pxor xmm"tostr(t0)", xmm"tostr(t0)"");\
|
||||
asm ("movdqa xmm"tostr(i1)", xmm"tostr(i0)"");\
|
||||
asm ("movdqa xmm"tostr(i3)", xmm"tostr(i2)"");\
|
||||
asm ("movdqa xmm"tostr(i5)", xmm"tostr(i4)"");\
|
||||
asm ("movdqa xmm"tostr(i7)", xmm"tostr(i6)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i1)", xmm"tostr(t0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i2)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i3)", xmm"tostr(t0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i4)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i5)", xmm"tostr(t0)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i6)", xmm"tostr(t0)"");\
|
||||
asm ("punpckhqdq xmm"tostr(i7)", xmm"tostr(t0)"");\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Inverse Step 2
|
||||
* input is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* output is one 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i7 = (0|S)
|
||||
* outputs: (i0, i2, i4, i6) = S
|
||||
*/
|
||||
#define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
|
||||
asm ("punpcklqdq xmm"tostr(i0)", xmm"tostr(i1)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i2)", xmm"tostr(i3)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i4)", xmm"tostr(i5)"");\
|
||||
asm ("punpcklqdq xmm"tostr(i6)", xmm"tostr(i7)"");\
|
||||
}/**/
|
||||
|
||||
|
||||
/* transform round constants into VPERM mode */
|
||||
#define VPERM_Transform_RoundConst_CNT2(i, j){\
|
||||
asm ("movaps xmm0, [ROUND_CONST_L0+"tostr(i)"*16]");\
|
||||
asm ("movaps xmm1, [ROUND_CONST_L7+"tostr(i)"*16]");\
|
||||
asm ("movaps xmm2, [ROUND_CONST_L0+"tostr(j)"*16]");\
|
||||
asm ("movaps xmm3, [ROUND_CONST_L7+"tostr(j)"*16]");\
|
||||
VPERM_Transform_State(0, 1, 2, 3, VPERM_IPT, 4, 5, 6, 7, 8, 9, 10);\
|
||||
asm ("pxor xmm0, [ALL_15]");\
|
||||
asm ("pxor xmm1, [ALL_15]");\
|
||||
asm ("pxor xmm2, [ALL_15]");\
|
||||
asm ("pxor xmm3, [ALL_15]");\
|
||||
asm ("movaps [ROUND_CONST_L0+"tostr(i)"*16], xmm0");\
|
||||
asm ("movaps [ROUND_CONST_L7+"tostr(i)"*16], xmm1");\
|
||||
asm ("movaps [ROUND_CONST_L0+"tostr(j)"*16], xmm2");\
|
||||
asm ("movaps [ROUND_CONST_L7+"tostr(j)"*16], xmm3");\
|
||||
}/**/
|
||||
|
||||
/* transform round constants into VPERM mode */
|
||||
#define VPERM_Transform_RoundConst(){\
|
||||
asm ("movaps xmm0, [ROUND_CONST_Lx]");\
|
||||
VPERM_Transform(0, 1, VPERM_IPT, 4, 5, 6, 7, 8, 9, 10);\
|
||||
asm ("pxor xmm0, [ALL_15]");\
|
||||
asm ("movaps [ROUND_CONST_Lx], xmm0");\
|
||||
VPERM_Transform_RoundConst_CNT2(0, 1);\
|
||||
VPERM_Transform_RoundConst_CNT2(2, 3);\
|
||||
VPERM_Transform_RoundConst_CNT2(4, 5);\
|
||||
VPERM_Transform_RoundConst_CNT2(6, 7);\
|
||||
VPERM_Transform_RoundConst_CNT2(8, 9);\
|
||||
}/**/
|
||||
|
||||
void INIT256(u64* h)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
asm volatile ("emms");
|
||||
|
||||
/* transform round constants into VPERM mode */
|
||||
VPERM_Transform_RoundConst();
|
||||
|
||||
/* load IV into registers xmm12 - xmm15 */
|
||||
asm ("movaps xmm12, [rdi+0*16]");
|
||||
asm ("movaps xmm13, [rdi+1*16]");
|
||||
asm ("movaps xmm14, [rdi+2*16]");
|
||||
asm ("movaps xmm15, [rdi+3*16]");
|
||||
|
||||
/* transform chaining value from column ordering into row ordering */
|
||||
/* we put two rows (64 bit) of the IV into one 128-bit XMM register */
|
||||
VPERM_Transform_State(12, 13, 14, 15, VPERM_IPT, 1, 2, 3, 4, 5, 6, 7);
|
||||
Matrix_Transpose_A(12, 13, 14, 15, 2, 6, 7, 0);
|
||||
|
||||
/* store transposed IV */
|
||||
asm ("movaps [rdi+0*16], xmm12");
|
||||
asm ("movaps [rdi+1*16], xmm2");
|
||||
asm ("movaps [rdi+2*16], xmm6");
|
||||
asm ("movaps [rdi+3*16], xmm7");
|
||||
|
||||
asm volatile ("emms");
|
||||
asm (".att_syntax noprefix");
|
||||
}
|
||||
|
||||
void TF512(u64* h, u64* m)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
/* message M in rsi */
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_START;
|
||||
#endif
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
Push_All_Regs();
|
||||
|
||||
/* load message into registers xmm12 - xmm15 (Q = message) */
|
||||
asm ("movaps xmm12, [rsi+0*16]");
|
||||
asm ("movaps xmm13, [rsi+1*16]");
|
||||
asm ("movaps xmm14, [rsi+2*16]");
|
||||
asm ("movaps xmm15, [rsi+3*16]");
|
||||
|
||||
/* transform message M from column ordering into row ordering */
|
||||
/* we first put two rows (64 bit) of the message into one 128-bit xmm register */
|
||||
VPERM_Transform_State(12, 13, 14, 15, VPERM_IPT, 1, 2, 3, 4, 5, 6, 7);
|
||||
Matrix_Transpose_A(12, 13, 14, 15, 2, 6, 7, 0);
|
||||
|
||||
/* load previous chaining value */
|
||||
/* we first put two rows (64 bit) of the CV into one 128-bit xmm register */
|
||||
asm ("movaps xmm8, [rdi+0*16]");
|
||||
asm ("movaps xmm0, [rdi+1*16]");
|
||||
asm ("movaps xmm4, [rdi+2*16]");
|
||||
asm ("movaps xmm5, [rdi+3*16]");
|
||||
|
||||
/* xor message to CV get input of P */
|
||||
/* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
|
||||
asm ("pxor xmm8, xmm12");
|
||||
asm ("pxor xmm0, xmm2");
|
||||
asm ("pxor xmm4, xmm6");
|
||||
asm ("pxor xmm5, xmm7");
|
||||
|
||||
/* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
|
||||
/* result: the 8 rows of P and Q in xmm8 - xmm12 */
|
||||
Matrix_Transpose_B(8, 0, 4, 5, 12, 2, 6, 7, 9, 10, 11, 12, 13, 14, 15);
|
||||
|
||||
/* compute the two permutations P and Q in parallel */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P or two rows of Q in one xmm register */
|
||||
Matrix_Transpose_B_INV(8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3);
|
||||
|
||||
/* xor output of P and Q */
|
||||
/* result: P(CV+M)+Q(M) in xmm0...xmm3 */
|
||||
asm ("pxor xmm0, xmm8");
|
||||
asm ("pxor xmm1, xmm10");
|
||||
asm ("pxor xmm2, xmm12");
|
||||
asm ("pxor xmm3, xmm14");
|
||||
|
||||
/* xor CV (feed-forward) */
|
||||
/* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
|
||||
asm ("pxor xmm0, [rdi+0*16]");
|
||||
asm ("pxor xmm1, [rdi+1*16]");
|
||||
asm ("pxor xmm2, [rdi+2*16]");
|
||||
asm ("pxor xmm3, [rdi+3*16]");
|
||||
|
||||
/* store CV */
|
||||
asm ("movaps [rdi+0*16], xmm0");
|
||||
asm ("movaps [rdi+1*16], xmm1");
|
||||
asm ("movaps [rdi+2*16], xmm2");
|
||||
asm ("movaps [rdi+3*16], xmm3");
|
||||
|
||||
Pop_All_Regs();
|
||||
asm (".att_syntax noprefix");
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_END;
|
||||
#endif
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
void OF512(u64* h)
|
||||
{
|
||||
/* __cdecl calling convention: */
|
||||
/* chaining value CV in rdi */
|
||||
|
||||
asm (".intel_syntax noprefix");
|
||||
Push_All_Regs();
|
||||
|
||||
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
|
||||
asm ("movaps xmm8, [rdi+0*16]");
|
||||
asm ("movaps xmm10, [rdi+1*16]");
|
||||
asm ("movaps xmm12, [rdi+2*16]");
|
||||
asm ("movaps xmm14, [rdi+3*16]");
|
||||
|
||||
/* there are now 2 rows of the CV in one xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) into one half of an xmm register */
|
||||
/* result: the 8 input rows of P in xmm8 - xmm15 */
|
||||
Matrix_Transpose_O_B(8, 9, 10, 11, 12, 13, 14, 15, 0);
|
||||
|
||||
/* compute the permutation P */
|
||||
/* result: the output of P(CV) in xmm8 - xmm15 */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P in one xmm register */
|
||||
/* result: P(CV) in xmm8, xmm10, xmm12, xmm14 */
|
||||
Matrix_Transpose_O_B_INV(8, 9, 10, 11, 12, 13, 14, 15);
|
||||
|
||||
/* xor CV to P output (feed-forward) */
|
||||
/* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
|
||||
asm ("pxor xmm8, [rdi+0*16]");
|
||||
asm ("pxor xmm10, [rdi+1*16]");
|
||||
asm ("pxor xmm12, [rdi+2*16]");
|
||||
asm ("pxor xmm14, [rdi+3*16]");
|
||||
|
||||
/* transform state back from row ordering into column ordering */
|
||||
/* result: final hash value in xmm9, xmm11 */
|
||||
Matrix_Transpose_A(8, 10, 12, 14, 4, 9, 11, 0);
|
||||
VPERM_Transform(9, 11, VPERM_OPT, 0, 1, 2, 3, 5, 6, 7);
|
||||
|
||||
/* we only need to return the truncated half of the state */
|
||||
asm ("movaps [rdi+2*16], xmm9");
|
||||
asm ("movaps [rdi+3*16], xmm11");
|
||||
|
||||
Pop_All_Regs();
|
||||
asm (".att_syntax noprefix");
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
|
||||
@@ -11,17 +11,44 @@
|
||||
#include <wmmintrin.h>
|
||||
#include "hash-groestl256.h"
|
||||
|
||||
/* global constants */
|
||||
__m128i ROUND_CONST_Lx;
|
||||
__m128i ROUND_CONST_L0[ROUNDS512];
|
||||
__m128i ROUND_CONST_L7[ROUNDS512];
|
||||
//__m128i ROUND_CONST_P[ROUNDS1024];
|
||||
//__m128i ROUND_CONST_Q[ROUNDS1024];
|
||||
__m128i TRANSP_MASK;
|
||||
__m128i SUBSH_MASK[8];
|
||||
__m128i ALL_1B;
|
||||
__m128i ALL_FF;
|
||||
static const __m128i round_const_l0[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x7060504030201000, 0xffffffffffffffff },
|
||||
{ 0x7161514131211101, 0xffffffffffffffff },
|
||||
{ 0x7262524232221202, 0xffffffffffffffff },
|
||||
{ 0x7363534333231303, 0xffffffffffffffff },
|
||||
{ 0x7464544434241404, 0xffffffffffffffff },
|
||||
{ 0x7565554535251505, 0xffffffffffffffff },
|
||||
{ 0x7666564636261606, 0xffffffffffffffff },
|
||||
{ 0x7767574737271707, 0xffffffffffffffff },
|
||||
{ 0x7868584838281808, 0xffffffffffffffff },
|
||||
{ 0x7969594939291909, 0xffffffffffffffff }
|
||||
};
|
||||
|
||||
static const __m128i round_const_l7[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x0000000000000000, 0x8f9fafbfcfdfefff },
|
||||
{ 0x0000000000000000, 0x8e9eaebecedeeefe },
|
||||
{ 0x0000000000000000, 0x8d9dadbdcdddedfd },
|
||||
{ 0x0000000000000000, 0x8c9cacbcccdcecfc },
|
||||
{ 0x0000000000000000, 0x8b9babbbcbdbebfb },
|
||||
{ 0x0000000000000000, 0x8a9aaabacadaeafa },
|
||||
{ 0x0000000000000000, 0x8999a9b9c9d9e9f9 },
|
||||
{ 0x0000000000000000, 0x8898a8b8c8d8e8f8 },
|
||||
{ 0x0000000000000000, 0x8797a7b7c7d7e7f7 },
|
||||
{ 0x0000000000000000, 0x8696a6b6c6d6e6f6 }
|
||||
};
|
||||
|
||||
static const __m128i TRANSP_MASK = { 0x0d0509010c040800, 0x0f070b030e060a02 };
|
||||
|
||||
static const __m128i SUBSH_MASK0 = { 0x0c0f0104070b0e00, 0x03060a0d08020509 };
|
||||
static const __m128i SUBSH_MASK1 = { 0x0e090205000d0801, 0x04070c0f0a03060b };
|
||||
static const __m128i SUBSH_MASK2 = { 0x080b0306010f0a02, 0x05000e090c04070d };
|
||||
static const __m128i SUBSH_MASK3 = { 0x0a0d040702090c03, 0x0601080b0e05000f };
|
||||
static const __m128i SUBSH_MASK4 = { 0x0b0e0500030a0d04, 0x0702090c0f060108 };
|
||||
static const __m128i SUBSH_MASK5 = { 0x0d080601040c0f05, 0x00030b0e0907020a };
|
||||
static const __m128i SUBSH_MASK6 = { 0x0f0a0702050e0906, 0x01040d080b00030c };
|
||||
static const __m128i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e };
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
@@ -38,8 +65,6 @@ __m128i ALL_FF;
|
||||
i = _mm_xor_si128(i, j);\
|
||||
}
|
||||
|
||||
/**/
|
||||
|
||||
/* Yet another implementation of MixBytes.
|
||||
This time we use the formulae (3) from the paper "Byte Slicing Groestl".
|
||||
Input: a0, ..., a7
|
||||
@@ -113,7 +138,7 @@ __m128i ALL_FF;
|
||||
\
|
||||
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
|
||||
/* compute w_i : add y_{i+4} */\
|
||||
b1 = ALL_1B;\
|
||||
b1 = m128_const1_64( 0x1b1b1b1b1b1b1b1b );\
|
||||
MUL2(a0, b0, b1);\
|
||||
a0 = _mm_xor_si128(a0, TEMP0);\
|
||||
MUL2(a1, b0, b1);\
|
||||
@@ -153,25 +178,6 @@ __m128i ALL_FF;
|
||||
b1 = _mm_xor_si128(b1, a4);\
|
||||
}/*MixBytes*/
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
ALL_1B = _mm_set_epi32(0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b);\
|
||||
TRANSP_MASK = _mm_set_epi32(0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800);\
|
||||
SUBSH_MASK[0] = _mm_set_epi32(0x03060a0d, 0x08020509, 0x0c0f0104, 0x070b0e00);\
|
||||
SUBSH_MASK[1] = _mm_set_epi32(0x04070c0f, 0x0a03060b, 0x0e090205, 0x000d0801);\
|
||||
SUBSH_MASK[2] = _mm_set_epi32(0x05000e09, 0x0c04070d, 0x080b0306, 0x010f0a02);\
|
||||
SUBSH_MASK[3] = _mm_set_epi32(0x0601080b, 0x0e05000f, 0x0a0d0407, 0x02090c03);\
|
||||
SUBSH_MASK[4] = _mm_set_epi32(0x0702090c, 0x0f060108, 0x0b0e0500, 0x030a0d04);\
|
||||
SUBSH_MASK[5] = _mm_set_epi32(0x00030b0e, 0x0907020a, 0x0d080601, 0x040c0f05);\
|
||||
SUBSH_MASK[6] = _mm_set_epi32(0x01040d08, 0x0b00030c, 0x0f0a0702, 0x050e0906);\
|
||||
SUBSH_MASK[7] = _mm_set_epi32(0x02050f0a, 0x0d01040e, 0x090c0003, 0x06080b07);\
|
||||
for(i = 0; i < ROUNDS512; i++)\
|
||||
{\
|
||||
ROUND_CONST_L0[i] = _mm_set_epi32(0xffffffff, 0xffffffff, 0x70605040 ^ (i * 0x01010101), 0x30201000 ^ (i * 0x01010101));\
|
||||
ROUND_CONST_L7[i] = _mm_set_epi32(0x8f9fafbf ^ (i * 0x01010101), 0xcfdfefff ^ (i * 0x01010101), 0x00000000, 0x00000000);\
|
||||
}\
|
||||
ROUND_CONST_Lx = _mm_set_epi32(0xffffffff, 0xffffffff, 0x00000000, 0x00000000);\
|
||||
}while(0); \
|
||||
|
||||
/* one round
|
||||
* i = round number
|
||||
* a0-a7 = input rows
|
||||
@@ -179,34 +185,34 @@ __m128i ALL_FF;
|
||||
*/
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* AddRoundConstant */\
|
||||
b1 = ROUND_CONST_Lx;\
|
||||
a0 = _mm_xor_si128(a0, (ROUND_CONST_L0[i]));\
|
||||
a1 = _mm_xor_si128(a1, b1);\
|
||||
a2 = _mm_xor_si128(a2, b1);\
|
||||
a3 = _mm_xor_si128(a3, b1);\
|
||||
a4 = _mm_xor_si128(a4, b1);\
|
||||
a5 = _mm_xor_si128(a5, b1);\
|
||||
a6 = _mm_xor_si128(a6, b1);\
|
||||
a7 = _mm_xor_si128(a7, (ROUND_CONST_L7[i]));\
|
||||
b1 = m128_const_64( 0xffffffffffffffff, 0 ); \
|
||||
a0 = _mm_xor_si128( a0, casti_m128i( round_const_l0, i ) ); \
|
||||
a1 = _mm_xor_si128( a1, b1 ); \
|
||||
a2 = _mm_xor_si128( a2, b1 ); \
|
||||
a3 = _mm_xor_si128( a3, b1 ); \
|
||||
a4 = _mm_xor_si128( a4, b1 ); \
|
||||
a5 = _mm_xor_si128( a5, b1 ); \
|
||||
a6 = _mm_xor_si128( a6, b1 ); \
|
||||
a7 = _mm_xor_si128( a7, casti_m128i( round_const_l7, i ) ); \
|
||||
\
|
||||
/* ShiftBytes + SubBytes (interleaved) */\
|
||||
b0 = _mm_xor_si128(b0, b0);\
|
||||
a0 = _mm_shuffle_epi8(a0, (SUBSH_MASK[0]));\
|
||||
a0 = _mm_aesenclast_si128(a0, b0);\
|
||||
a1 = _mm_shuffle_epi8(a1, (SUBSH_MASK[1]));\
|
||||
a1 = _mm_aesenclast_si128(a1, b0);\
|
||||
a2 = _mm_shuffle_epi8(a2, (SUBSH_MASK[2]));\
|
||||
a2 = _mm_aesenclast_si128(a2, b0);\
|
||||
a3 = _mm_shuffle_epi8(a3, (SUBSH_MASK[3]));\
|
||||
a3 = _mm_aesenclast_si128(a3, b0);\
|
||||
a4 = _mm_shuffle_epi8(a4, (SUBSH_MASK[4]));\
|
||||
a4 = _mm_aesenclast_si128(a4, b0);\
|
||||
a5 = _mm_shuffle_epi8(a5, (SUBSH_MASK[5]));\
|
||||
a5 = _mm_aesenclast_si128(a5, b0);\
|
||||
a6 = _mm_shuffle_epi8(a6, (SUBSH_MASK[6]));\
|
||||
a6 = _mm_aesenclast_si128(a6, b0);\
|
||||
a7 = _mm_shuffle_epi8(a7, (SUBSH_MASK[7]));\
|
||||
a7 = _mm_aesenclast_si128(a7, b0);\
|
||||
a0 = _mm_shuffle_epi8( a0, SUBSH_MASK0 ); \
|
||||
a0 = _mm_aesenclast_si128( a0, b0 );\
|
||||
a1 = _mm_shuffle_epi8( a1, SUBSH_MASK1 ); \
|
||||
a1 = _mm_aesenclast_si128( a1, b0 );\
|
||||
a2 = _mm_shuffle_epi8( a2, SUBSH_MASK2 ); \
|
||||
a2 = _mm_aesenclast_si128( a2, b0 );\
|
||||
a3 = _mm_shuffle_epi8( a3, SUBSH_MASK3 ); \
|
||||
a3 = _mm_aesenclast_si128( a3, b0 );\
|
||||
a4 = _mm_shuffle_epi8( a4, SUBSH_MASK4 ); \
|
||||
a4 = _mm_aesenclast_si128( a4, b0 );\
|
||||
a5 = _mm_shuffle_epi8( a5, SUBSH_MASK5 ); \
|
||||
a5 = _mm_aesenclast_si128( a5, b0 );\
|
||||
a6 = _mm_shuffle_epi8( a6, SUBSH_MASK6 ); \
|
||||
a6 = _mm_aesenclast_si128( a6, b0 );\
|
||||
a7 = _mm_shuffle_epi8( a7, SUBSH_MASK7 ); \
|
||||
a7 = _mm_aesenclast_si128( a7, b0 );\
|
||||
\
|
||||
/* MixBytes */\
|
||||
MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
|
||||
@@ -234,8 +240,9 @@ __m128i ALL_FF;
|
||||
* outputs: i0, o1-o3
|
||||
* clobbers: t0
|
||||
*/
|
||||
|
||||
#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
|
||||
t0 = TRANSP_MASK;\
|
||||
t0 = TRANSP_MASK; \
|
||||
\
|
||||
i0 = _mm_shuffle_epi8(i0, t0);\
|
||||
i1 = _mm_shuffle_epi8(i1, t0);\
|
||||
|
||||
@@ -1,482 +0,0 @@
|
||||
/* groestl-intr-avx.h Aug 2011
|
||||
*
|
||||
* Groestl implementation with intrinsics using ssse3, sse4.1, aes and avx
|
||||
* instructions.
|
||||
* Author: Günther A. Roland, Martin Schläffer, Krystian Matusiewicz
|
||||
*
|
||||
* This code is placed in the public domain
|
||||
*/
|
||||
|
||||
#include <smmintrin.h>
|
||||
#include <wmmintrin.h>
|
||||
#include <immintrin.h>
|
||||
#include "hash-groestl256.h"
|
||||
|
||||
/* global constants */
|
||||
__m128i ROUND_CONST_Lx;
|
||||
__m128i ROUND_CONST_L0[ROUNDS512];
|
||||
__m128i ROUND_CONST_L7[ROUNDS512];
|
||||
__m128i ROUND_CONST_P[ROUNDS1024];
|
||||
__m128i ROUND_CONST_Q[ROUNDS1024];
|
||||
__m128i TRANSP_MASK;
|
||||
__m128i SUBSH_MASK[8];
|
||||
__m128i ALL_FF;
|
||||
//#if LENGTH <= 256
|
||||
__m128i ALL_1B;
|
||||
//#else
|
||||
//__m256d ALL_1B;
|
||||
//#endif
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
|
||||
#define insert_m128i_in_m256d(ymm, xmm, pos) (_mm256_castsi256_pd(_mm256_insertf128_si256(_mm256_castpd_si256(ymm), xmm, pos)))
|
||||
#define extract_m128i_from_m256d(ymm, pos) (_mm256_extractf128_si256(_mm256_castpd_si256(ymm), pos))
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
ALL_1B = _mm_set_epi32(0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b);\
|
||||
ALL_FF = _mm_set_epi32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff);\
|
||||
TRANSP_MASK = _mm_set_epi32(0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800);\
|
||||
SUBSH_MASK[0] = _mm_set_epi32(0x03060a0d, 0x08020509, 0x0c0f0104, 0x070b0e00);\
|
||||
SUBSH_MASK[1] = _mm_set_epi32(0x04070c0f, 0x0a03060b, 0x0e090205, 0x000d0801);\
|
||||
SUBSH_MASK[2] = _mm_set_epi32(0x05000e09, 0x0c04070d, 0x080b0306, 0x010f0a02);\
|
||||
SUBSH_MASK[3] = _mm_set_epi32(0x0601080b, 0x0e05000f, 0x0a0d0407, 0x02090c03);\
|
||||
SUBSH_MASK[4] = _mm_set_epi32(0x0702090c, 0x0f060108, 0x0b0e0500, 0x030a0d04);\
|
||||
SUBSH_MASK[5] = _mm_set_epi32(0x00030b0e, 0x0907020a, 0x0d080601, 0x040c0f05);\
|
||||
SUBSH_MASK[6] = _mm_set_epi32(0x01040d08, 0x0b00030c, 0x0f0a0702, 0x050e0906);\
|
||||
SUBSH_MASK[7] = _mm_set_epi32(0x02050f0a, 0x0d01040e, 0x090c0003, 0x06080b07);\
|
||||
for(i = 0; i < ROUNDS512; i++)\
|
||||
{\
|
||||
ROUND_CONST_L0[i] = _mm_set_epi32(0xffffffff, 0xffffffff, 0x70605040 ^ (i * 0x01010101), 0x30201000 ^ (i * 0x01010101));\
|
||||
ROUND_CONST_L7[i] = _mm_set_epi32(0x8f9fafbf ^ (i * 0x01010101), 0xcfdfefff ^ (i * 0x01010101), 0x00000000, 0x00000000);\
|
||||
}\
|
||||
ROUND_CONST_Lx = _mm_set_epi32(0xffffffff, 0xffffffff, 0x00000000, 0x00000000);\
|
||||
}while(0);
|
||||
|
||||
/* xmm[i] will be multiplied by 2
|
||||
* xmm[j] will be lost
|
||||
* xmm[k] has to be all 0x1b
|
||||
* xmm[z] has to be zero */
|
||||
#define VMUL2(i, j, k, z){\
|
||||
j = _mm_cmpgt_epi8(z, i);\
|
||||
i = _mm_add_epi8(i, i);\
|
||||
j = _mm_and_si128(j, k);\
|
||||
i = _mm_xor_si128(i, j);\
|
||||
}/**/
|
||||
|
||||
/* Yet another implementation of MixBytes.
|
||||
This time we use the formulae (3) from the paper "Byte Slicing Groestl".
|
||||
Input: a0, ..., a7
|
||||
Output: b0, ..., b7 = MixBytes(a0,...,a7).
|
||||
but we use the relations:
|
||||
t_i = a_i + a_{i+3}
|
||||
x_i = t_i + t_{i+3}
|
||||
y_i = t_i + t+{i+2} + a_{i+6}
|
||||
z_i = 2*x_i
|
||||
w_i = z_i + y_{i+4}
|
||||
v_i = 2*w_i
|
||||
b_i = v_{i+3} + y_{i+4}
|
||||
We keep building b_i in registers xmm8..xmm15 by first building y_{i+4} there
|
||||
and then adding v_i computed in the meantime in registers xmm0..xmm7.
|
||||
We almost fit into 16 registers, need only 3 spills to memory.
|
||||
This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
|
||||
K. Matusiewicz, 2011/05/29 */
|
||||
#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* xmm"tostr(8..xmm"tostr(15 = a2 a3... a0 a1 */\
|
||||
b0 = a2;\
|
||||
b1 = a3;\
|
||||
b2 = a4;\
|
||||
b3 = a5;\
|
||||
b4 = a6;\
|
||||
b5 = a7;\
|
||||
b6 = a0;\
|
||||
b7 = a1;\
|
||||
\
|
||||
/* t_i = a_i + a_{i+1} */\
|
||||
a0 = _mm_xor_si128(a0, a1);\
|
||||
a1 = _mm_xor_si128(a1, a2);\
|
||||
a2 = _mm_xor_si128(a2, a3);\
|
||||
a3 = _mm_xor_si128(a3, a4);\
|
||||
a4 = _mm_xor_si128(a4, a5);\
|
||||
a5 = _mm_xor_si128(a5, a6);\
|
||||
a6 = _mm_xor_si128(a6, a7);\
|
||||
a7 = _mm_xor_si128(a7, b6);\
|
||||
\
|
||||
/* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
|
||||
b0 = _mm_xor_si128(b0, a4);\
|
||||
b1 = _mm_xor_si128(b1, a5);\
|
||||
b2 = _mm_xor_si128(b2, a6);\
|
||||
b3 = _mm_xor_si128(b3, a7);\
|
||||
b4 = _mm_xor_si128(b4, a0);\
|
||||
b5 = _mm_xor_si128(b5, a1);\
|
||||
b6 = _mm_xor_si128(b6, a2);\
|
||||
b7 = _mm_xor_si128(b7, a3);\
|
||||
\
|
||||
b0 = _mm_xor_si128(b0, a6);\
|
||||
b1 = _mm_xor_si128(b1, a7);\
|
||||
b2 = _mm_xor_si128(b2, a0);\
|
||||
b3 = _mm_xor_si128(b3, a1);\
|
||||
b4 = _mm_xor_si128(b4, a2);\
|
||||
b5 = _mm_xor_si128(b5, a3);\
|
||||
b6 = _mm_xor_si128(b6, a4);\
|
||||
b7 = _mm_xor_si128(b7, a5);\
|
||||
\
|
||||
/* spill values y_4, y_5 to memory */\
|
||||
TEMP0 = b0;\
|
||||
TEMP1 = b1;\
|
||||
TEMP2 = b2;\
|
||||
\
|
||||
/* save values t0, t1, t2 to xmm8, xmm9 and memory */\
|
||||
b0 = a0;\
|
||||
b1 = a1;\
|
||||
TEMP3 = a2;\
|
||||
\
|
||||
/* compute x_i = t_i + t_{i+3} */\
|
||||
a0 = _mm_xor_si128(a0, a3);\
|
||||
a1 = _mm_xor_si128(a1, a4);\
|
||||
a2 = _mm_xor_si128(a2, a5);\
|
||||
a3 = _mm_xor_si128(a3, a6);\
|
||||
a4 = _mm_xor_si128(a4, a7);\
|
||||
a5 = _mm_xor_si128(a5, b0);\
|
||||
a6 = _mm_xor_si128(a6, b1);\
|
||||
a7 = _mm_xor_si128(a7, TEMP3);\
|
||||
\
|
||||
/*compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
|
||||
b1 = ALL_1B;\
|
||||
b2 = _mm_xor_si128(b2, b2);\
|
||||
VMUL2(a7, b0, b1, b2);\
|
||||
VMUL2(a6, b0, b1, b2);\
|
||||
VMUL2(a5, b0, b1, b2);\
|
||||
VMUL2(a4, b0, b1, b2);\
|
||||
VMUL2(a3, b0, b1, b2);\
|
||||
VMUL2(a2, b0, b1, b2);\
|
||||
VMUL2(a1, b0, b1, b2);\
|
||||
VMUL2(a0, b0, b1, b2);\
|
||||
\
|
||||
/* compute w_i : add y_{i+4} */\
|
||||
a0 = _mm_xor_si128(a0, TEMP0);\
|
||||
a1 = _mm_xor_si128(a1, TEMP1);\
|
||||
a2 = _mm_xor_si128(a2, TEMP2);\
|
||||
a3 = _mm_xor_si128(a3, b3);\
|
||||
a4 = _mm_xor_si128(a4, b4);\
|
||||
a5 = _mm_xor_si128(a5, b5);\
|
||||
a6 = _mm_xor_si128(a6, b6);\
|
||||
a7 = _mm_xor_si128(a7, b7);\
|
||||
\
|
||||
/*compute v_i: double w_i */\
|
||||
VMUL2(a0, b0, b1, b2);\
|
||||
VMUL2(a1, b0, b1, b2);\
|
||||
VMUL2(a2, b0, b1, b2);\
|
||||
VMUL2(a3, b0, b1, b2);\
|
||||
VMUL2(a4, b0, b1, b2);\
|
||||
VMUL2(a5, b0, b1, b2);\
|
||||
VMUL2(a6, b0, b1, b2);\
|
||||
VMUL2(a7, b0, b1, b2);\
|
||||
\
|
||||
/* add to y_4 y_5 .. v3, v4, ... */\
|
||||
b0 = _mm_xor_si128(a3, TEMP0);\
|
||||
b1 = _mm_xor_si128(a4, TEMP1);\
|
||||
b2 = _mm_xor_si128(a5, TEMP2);\
|
||||
b3 = _mm_xor_si128(b3, a6);\
|
||||
b4 = _mm_xor_si128(b4, a7);\
|
||||
b5 = _mm_xor_si128(b5, a0);\
|
||||
b6 = _mm_xor_si128(b6, a1);\
|
||||
b7 = _mm_xor_si128(b7, a2);\
|
||||
}/*MixBytes*/
|
||||
|
||||
/* one round
|
||||
* i = round number
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
*/
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* Add Round Constant */\
|
||||
b1 = ROUND_CONST_Lx;\
|
||||
a0 = _mm_xor_si128(a0, (ROUND_CONST_L0[i]));\
|
||||
a1 = _mm_xor_si128(a1, b1);\
|
||||
a2 = _mm_xor_si128(a2, b1);\
|
||||
a3 = _mm_xor_si128(a3, b1);\
|
||||
a4 = _mm_xor_si128(a4, b1);\
|
||||
a5 = _mm_xor_si128(a5, b1);\
|
||||
a6 = _mm_xor_si128(a6, b1);\
|
||||
a7 = _mm_xor_si128(a7, (ROUND_CONST_L7[i]));\
|
||||
\
|
||||
/* ShiftBytes + SubBytes (interleaved) */\
|
||||
b0 = _mm_xor_si128(b0, b0);\
|
||||
a0 = _mm_shuffle_epi8(a0, (SUBSH_MASK[0]));\
|
||||
a0 = _mm_aesenclast_si128(a0, b0);\
|
||||
a1 = _mm_shuffle_epi8(a1, (SUBSH_MASK[1]));\
|
||||
a1 = _mm_aesenclast_si128(a1, b0);\
|
||||
a2 = _mm_shuffle_epi8(a2, (SUBSH_MASK[2]));\
|
||||
a2 = _mm_aesenclast_si128(a2, b0);\
|
||||
a3 = _mm_shuffle_epi8(a3, (SUBSH_MASK[3]));\
|
||||
a3 = _mm_aesenclast_si128(a3, b0);\
|
||||
a4 = _mm_shuffle_epi8(a4, (SUBSH_MASK[4]));\
|
||||
a4 = _mm_aesenclast_si128(a4, b0);\
|
||||
a5 = _mm_shuffle_epi8(a5, (SUBSH_MASK[5]));\
|
||||
a5 = _mm_aesenclast_si128(a5, b0);\
|
||||
a6 = _mm_shuffle_epi8(a6, (SUBSH_MASK[6]));\
|
||||
a6 = _mm_aesenclast_si128(a6, b0);\
|
||||
a7 = _mm_shuffle_epi8(a7, (SUBSH_MASK[7]));\
|
||||
a7 = _mm_aesenclast_si128(a7, b0);\
|
||||
\
|
||||
/* MixBytes */\
|
||||
MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
|
||||
}
|
||||
|
||||
/* 10 rounds, P and Q in parallel */
|
||||
#define ROUNDS_P_Q(){\
|
||||
ROUND(0, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(1, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(2, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(3, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(4, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(5, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(6, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(7, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(8, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(9, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
}
|
||||
|
||||
/* Matrix Transpose Step 1
|
||||
* input is a 512-bit state with two columns in one xmm
|
||||
* output is a 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i3
|
||||
* outputs: i0, o1-o3
|
||||
* clobbers: t0
|
||||
*/
|
||||
#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
|
||||
t0 = TRANSP_MASK;\
|
||||
\
|
||||
i0 = _mm_shuffle_epi8(i0, t0);\
|
||||
i1 = _mm_shuffle_epi8(i1, t0);\
|
||||
i2 = _mm_shuffle_epi8(i2, t0);\
|
||||
i3 = _mm_shuffle_epi8(i3, t0);\
|
||||
\
|
||||
o1 = _mm_unpackhi_epi16(i0, i1);\
|
||||
i0 = _mm_unpacklo_epi16(i0, i1);\
|
||||
t0 = _mm_unpackhi_epi16(i2, i3);\
|
||||
i2 = _mm_unpacklo_epi16(i2, i3);\
|
||||
\
|
||||
i0 = _mm_shuffle_epi32(i0, 216);\
|
||||
o1 = _mm_shuffle_epi32(o1, 216);\
|
||||
i2 = _mm_shuffle_epi32(i2, 216);\
|
||||
t0 = _mm_shuffle_epi32(t0, 216);\
|
||||
\
|
||||
o2 = _mm_unpackhi_epi32(i0, i2);\
|
||||
o3 = _mm_unpackhi_epi32(o1, t0);\
|
||||
i0 = _mm_unpacklo_epi32(i0, i2);\
|
||||
o1 = _mm_unpacklo_epi32(o1, t0);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Step 2
|
||||
* input are two 512-bit states with two rows in one xmm
|
||||
* output are two 512-bit states with one row of each state in one xmm
|
||||
* inputs: i0-i3 = P, i4-i7 = Q
|
||||
* outputs: (i0, o1-o7) = (P|Q)
|
||||
* possible reassignments: (output reg = input reg)
|
||||
* * i1 -> o3-7
|
||||
* * i2 -> o5-7
|
||||
* * i3 -> o7
|
||||
* * i4 -> o3-7
|
||||
* * i5 -> o6-7
|
||||
*/
|
||||
#define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
|
||||
o1 = _mm_unpackhi_epi64(i0, i4);\
|
||||
i0 = _mm_unpacklo_epi64(i0, i4);\
|
||||
o2 = _mm_unpacklo_epi64(i1, i5);\
|
||||
o3 = _mm_unpackhi_epi64(i1, i5);\
|
||||
o4 = _mm_unpacklo_epi64(i2, i6);\
|
||||
o5 = _mm_unpackhi_epi64(i2, i6);\
|
||||
o6 = _mm_unpacklo_epi64(i3, i7);\
|
||||
o7 = _mm_unpackhi_epi64(i3, i7);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Inverse Step 2
|
||||
* input are two 512-bit states with one row of each state in one xmm
|
||||
* output are two 512-bit states with two rows in one xmm
|
||||
* inputs: i0-i7 = (P|Q)
|
||||
* outputs: (i0, i2, i4, i6) = P, (o0-o3) = Q
|
||||
*/
|
||||
#define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
|
||||
o0 = _mm_unpackhi_epi64(i0, i1);\
|
||||
i0 = _mm_unpacklo_epi64(i0, i1);\
|
||||
o1 = _mm_unpackhi_epi64(i2, i3);\
|
||||
i2 = _mm_unpacklo_epi64(i2, i3);\
|
||||
o2 = _mm_unpackhi_epi64(i4, i5);\
|
||||
i4 = _mm_unpacklo_epi64(i4, i5);\
|
||||
o3 = _mm_unpackhi_epi64(i6, i7);\
|
||||
i6 = _mm_unpacklo_epi64(i6, i7);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Step 2
|
||||
* input is one 512-bit state with two rows in one xmm
|
||||
* output is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* inputs: i0,i2,i4,i6 = S
|
||||
* outputs: (i0-7) = (0|S)
|
||||
*/
|
||||
#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
|
||||
t0 = _mm_xor_si128(t0, t0);\
|
||||
i1 = _mm_unpackhi_epi64(i0, t0);\
|
||||
i0 = _mm_unpacklo_epi64(i0, t0);\
|
||||
i3 = _mm_unpackhi_epi64(i2, t0);\
|
||||
i2 = _mm_unpacklo_epi64(i2, t0);\
|
||||
i5 = _mm_unpackhi_epi64(i4, t0);\
|
||||
i4 = _mm_unpacklo_epi64(i4, t0);\
|
||||
i7 = _mm_unpackhi_epi64(i6, t0);\
|
||||
i6 = _mm_unpacklo_epi64(i6, t0);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Inverse Step 2
|
||||
* input is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* output is one 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i7 = (0|S)
|
||||
* outputs: (i0, i2, i4, i6) = S
|
||||
*/
|
||||
#define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
|
||||
i0 = _mm_unpacklo_epi64(i0, i1);\
|
||||
i2 = _mm_unpacklo_epi64(i2, i3);\
|
||||
i4 = _mm_unpacklo_epi64(i4, i5);\
|
||||
i6 = _mm_unpacklo_epi64(i6, i7);\
|
||||
}/**/
|
||||
|
||||
|
||||
void INIT256(u64* h)
|
||||
{
|
||||
__m128i* const chaining = (__m128i*) h;
|
||||
static __m128i xmm0, /*xmm1,*/ xmm2, /*xmm3, xmm4, xmm5,*/ xmm6, xmm7;
|
||||
static __m128i /*xmm8, xmm9, xmm10, xmm11,*/ xmm12, xmm13, xmm14, xmm15;
|
||||
|
||||
/* load IV into registers xmm12 - xmm15 */
|
||||
xmm12 = chaining[0];
|
||||
xmm13 = chaining[1];
|
||||
xmm14 = chaining[2];
|
||||
xmm15 = chaining[3];
|
||||
|
||||
/* transform chaining value from column ordering into row ordering */
|
||||
/* we put two rows (64 bit) of the IV into one 128-bit XMM register */
|
||||
Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
|
||||
|
||||
/* store transposed IV */
|
||||
chaining[0] = xmm12;
|
||||
chaining[1] = xmm2;
|
||||
chaining[2] = xmm6;
|
||||
chaining[3] = xmm7;
|
||||
}
|
||||
|
||||
void TF512(u64* h, u64* m)
|
||||
{
|
||||
__m128i* const chaining = (__m128i*) h;
|
||||
__m128i* const message = (__m128i*) m;
|
||||
static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
|
||||
static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
|
||||
static __m128i TEMP0;
|
||||
static __m128i TEMP1;
|
||||
static __m128i TEMP2;
|
||||
static __m128i TEMP3;
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_START;
|
||||
#endif
|
||||
|
||||
/* load message into registers xmm12 - xmm15 */
|
||||
xmm12 = message[0];
|
||||
xmm13 = message[1];
|
||||
xmm14 = message[2];
|
||||
xmm15 = message[3];
|
||||
|
||||
/* transform message M from column ordering into row ordering */
|
||||
/* we first put two rows (64 bit) of the message into one 128-bit xmm register */
|
||||
Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
|
||||
|
||||
/* load previous chaining value and xor message to CV to get input of P */
|
||||
/* we first put two rows (2x64 bit) of the CV into one 128-bit xmm register */
|
||||
/* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
|
||||
xmm8 = _mm_xor_si128(xmm12, chaining[0]);
|
||||
xmm0 = _mm_xor_si128(xmm2, chaining[1]);
|
||||
xmm4 = _mm_xor_si128(xmm6, chaining[2]);
|
||||
xmm5 = _mm_xor_si128(xmm7, chaining[3]);
|
||||
|
||||
/* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
|
||||
/* result: the 8 rows of P and Q in xmm8 - xmm12 */
|
||||
Matrix_Transpose_B(xmm8, xmm0, xmm4, xmm5, xmm12, xmm2, xmm6, xmm7, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);
|
||||
|
||||
/* compute the two permutations P and Q in parallel */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P or two rows of Q in one xmm register */
|
||||
Matrix_Transpose_B_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3);
|
||||
|
||||
/* xor output of P and Q */
|
||||
/* result: P(CV+M)+Q(M) in xmm0...xmm3 */
|
||||
xmm0 = _mm_xor_si128(xmm0, xmm8);
|
||||
xmm1 = _mm_xor_si128(xmm1, xmm10);
|
||||
xmm2 = _mm_xor_si128(xmm2, xmm12);
|
||||
xmm3 = _mm_xor_si128(xmm3, xmm14);
|
||||
|
||||
/* xor CV (feed-forward) */
|
||||
/* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
|
||||
xmm0 = _mm_xor_si128(xmm0, chaining[0]);
|
||||
xmm1 = _mm_xor_si128(xmm1, chaining[1]);
|
||||
xmm2 = _mm_xor_si128(xmm2, chaining[2]);
|
||||
xmm3 = _mm_xor_si128(xmm3, chaining[3]);
|
||||
|
||||
/* store CV */
|
||||
chaining[0] = xmm0;
|
||||
chaining[1] = xmm1;
|
||||
chaining[2] = xmm2;
|
||||
chaining[3] = xmm3;
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_END;
|
||||
#endif
|
||||
return;
|
||||
}
|
||||
|
||||
void OF512(u64* h)
|
||||
{
|
||||
__m128i* const chaining = (__m128i*) h;
|
||||
static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
|
||||
static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
|
||||
static __m128i TEMP0;
|
||||
static __m128i TEMP1;
|
||||
static __m128i TEMP2;
|
||||
static __m128i TEMP3;
|
||||
|
||||
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
|
||||
xmm8 = chaining[0];
|
||||
xmm10 = chaining[1];
|
||||
xmm12 = chaining[2];
|
||||
xmm14 = chaining[3];
|
||||
|
||||
/* there are now 2 rows of the CV in one xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) into one half of an xmm register */
|
||||
/* result: the 8 input rows of P in xmm8 - xmm15 */
|
||||
Matrix_Transpose_O_B(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0);
|
||||
|
||||
/* compute the permutation P */
|
||||
/* result: the output of P(CV) in xmm8 - xmm15 */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P in one xmm register */
|
||||
/* result: P(CV) in xmm8, xmm10, xmm12, xmm14 */
|
||||
Matrix_Transpose_O_B_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);
|
||||
|
||||
/* xor CV to P output (feed-forward) */
|
||||
/* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
|
||||
xmm8 = _mm_xor_si128(xmm8, (chaining[0]));
|
||||
xmm10 = _mm_xor_si128(xmm10, (chaining[1]));
|
||||
xmm12 = _mm_xor_si128(xmm12, (chaining[2]));
|
||||
xmm14 = _mm_xor_si128(xmm14, (chaining[3]));
|
||||
|
||||
/* transform state back from row ordering into column ordering */
|
||||
/* result: final hash value in xmm9, xmm11 */
|
||||
Matrix_Transpose_A(xmm8, xmm10, xmm12, xmm14, xmm4, xmm9, xmm11, xmm0);
|
||||
|
||||
/* we only need to return the truncated half of the state */
|
||||
chaining[2] = xmm9;
|
||||
chaining[3] = xmm11;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,793 +0,0 @@
|
||||
/* groestl-intr-vperm.h Aug 2011
|
||||
*
|
||||
* Groestl implementation with intrinsics using ssse3 instructions.
|
||||
* Author: Günther A. Roland, Martin Schläffer
|
||||
*
|
||||
* Based on the vperm and aes_ni implementations of the hash function Groestl
|
||||
* by Cagdas Calik <ccalik@metu.edu.tr> http://www.metu.edu.tr/~ccalik/
|
||||
* Institute of Applied Mathematics, Middle East Technical University, Turkey
|
||||
*
|
||||
* This code is placed in the public domain
|
||||
*/
|
||||
|
||||
#include <tmmintrin.h>
|
||||
#include "hash-groestl256.h"
|
||||
|
||||
/* global constants */
|
||||
__m128i ROUND_CONST_Lx;
|
||||
__m128i ROUND_CONST_L0[ROUNDS512];
|
||||
__m128i ROUND_CONST_L7[ROUNDS512];
|
||||
__m128i ROUND_CONST_P[ROUNDS1024];
|
||||
__m128i ROUND_CONST_Q[ROUNDS1024];
|
||||
__m128i TRANSP_MASK;
|
||||
__m128i SUBSH_MASK[8];
|
||||
__m128i ALL_0F;
|
||||
__m128i ALL_15;
|
||||
__m128i ALL_1B;
|
||||
__m128i ALL_63;
|
||||
__m128i ALL_FF;
|
||||
__m128i VPERM_IPT[2];
|
||||
__m128i VPERM_OPT[2];
|
||||
__m128i VPERM_INV[2];
|
||||
__m128i VPERM_SB1[2];
|
||||
__m128i VPERM_SB2[2];
|
||||
__m128i VPERM_SB4[2];
|
||||
__m128i VPERM_SBO[2];
|
||||
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
|
||||
#define SET_SHARED_CONSTANTS(){\
|
||||
TRANSP_MASK = _mm_set_epi32(0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800);\
|
||||
ALL_1B = _mm_set_epi32(0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b);\
|
||||
ALL_63 = _mm_set_epi32(0x63636363, 0x63636363, 0x63636363, 0x63636363);\
|
||||
ALL_0F = _mm_set_epi32(0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f);\
|
||||
ALL_15 = _mm_set_epi32(0x15151515, 0x15151515, 0x15151515, 0x15151515);\
|
||||
VPERM_IPT[0] = _mm_set_epi32(0xCD80B1FC, 0xB0FDCC81, 0x4C01307D, 0x317C4D00);\
|
||||
VPERM_IPT[1] = _mm_set_epi32(0xCABAE090, 0x52227808, 0xC2B2E898, 0x5A2A7000);\
|
||||
VPERM_OPT[0] = _mm_set_epi32(0xE10D5DB1, 0xB05C0CE0, 0x01EDBD51, 0x50BCEC00);\
|
||||
VPERM_OPT[1] = _mm_set_epi32(0xF7974121, 0xDEBE6808, 0xFF9F4929, 0xD6B66000);\
|
||||
VPERM_INV[0] = _mm_set_epi32(0x030D0E0C, 0x02050809, 0x01040A06, 0x0F0B0780);\
|
||||
VPERM_INV[1] = _mm_set_epi32(0x04070309, 0x0A0B0C02, 0x0E05060F, 0x0D080180);\
|
||||
VPERM_SB1[0] = _mm_set_epi32(0x3BF7CCC1, 0x0D2ED9EF, 0x3618D415, 0xFAE22300);\
|
||||
VPERM_SB1[1] = _mm_set_epi32(0xA5DF7A6E, 0x142AF544, 0xB19BE18F, 0xCB503E00);\
|
||||
VPERM_SB2[0] = _mm_set_epi32(0xC2A163C8, 0xAB82234A, 0x69EB8840, 0x0AE12900);\
|
||||
VPERM_SB2[1] = _mm_set_epi32(0x5EB7E955, 0xBC982FCD, 0xE27A93C6, 0x0B712400);\
|
||||
VPERM_SB4[0] = _mm_set_epi32(0xBA44FE79, 0x876D2914, 0x3D50AED7, 0xC393EA00);\
|
||||
VPERM_SB4[1] = _mm_set_epi32(0xA876DE97, 0x49087E9F, 0xE1E937A0, 0x3FD64100);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Transform w/o settings c*
|
||||
* transforms 2 rows to/from "vperm mode"
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0, a1 = 2 rows
|
||||
* table = transformation table to use
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* a0, a1 = 2 rows transformed with table
|
||||
* */
|
||||
#define VPERM_Transform_No_Const(a0, a1, t0, t1, t2, t3, c0, c1, c2){\
|
||||
t0 = c0;\
|
||||
t1 = c0;\
|
||||
t0 = _mm_andnot_si128(t0, a0);\
|
||||
t1 = _mm_andnot_si128(t1, a1);\
|
||||
t0 = _mm_srli_epi32(t0, 4);\
|
||||
t1 = _mm_srli_epi32(t1, 4);\
|
||||
a0 = _mm_and_si128(a0, c0);\
|
||||
a1 = _mm_and_si128(a1, c0);\
|
||||
t2 = c2;\
|
||||
t3 = c2;\
|
||||
t2 = _mm_shuffle_epi8(t2, a0);\
|
||||
t3 = _mm_shuffle_epi8(t3, a1);\
|
||||
a0 = c1;\
|
||||
a1 = c1;\
|
||||
a0 = _mm_shuffle_epi8(a0, t0);\
|
||||
a1 = _mm_shuffle_epi8(a1, t1);\
|
||||
a0 = _mm_xor_si128(a0, t2);\
|
||||
a1 = _mm_xor_si128(a1, t3);\
|
||||
}/**/
|
||||
|
||||
#define VPERM_Transform_Set_Const(table, c0, c1, c2){\
|
||||
c0 = ALL_0F;\
|
||||
c1 = ((__m128i*) table )[0];\
|
||||
c2 = ((__m128i*) table )[1];\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Transform
|
||||
* transforms 2 rows to/from "vperm mode"
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0, a1 = 2 rows
|
||||
* table = transformation table to use
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* a0, a1 = 2 rows transformed with table
|
||||
* */
|
||||
#define VPERM_Transform(a0, a1, table, t0, t1, t2, t3, c0, c1, c2){\
|
||||
VPERM_Transform_Set_Const(table, c0, c1, c2);\
|
||||
VPERM_Transform_No_Const(a0, a1, t0, t1, t2, t3, c0, c1, c2);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Transform State
|
||||
* inputs:
|
||||
* a0-a3 = state
|
||||
* table = transformation table to use
|
||||
* t* = clobbers
|
||||
* outputs:
|
||||
* a0-a3 = transformed state
|
||||
* */
|
||||
#define VPERM_Transform_State(a0, a1, a2, a3, table, t0, t1, t2, t3, c0, c1, c2){\
|
||||
VPERM_Transform_Set_Const(table, c0, c1, c2);\
|
||||
VPERM_Transform_No_Const(a0, a1, t0, t1, t2, t3, c0, c1, c2);\
|
||||
VPERM_Transform_No_Const(a2, a3, t0, t1, t2, t3, c0, c1, c2);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Add Constant to State
|
||||
* inputs:
|
||||
* a0-a7 = state
|
||||
* constant = constant to add
|
||||
* t0 = clobber
|
||||
* outputs:
|
||||
* a0-a7 = state + constant
|
||||
* */
|
||||
#define VPERM_Add_Constant(a0, a1, a2, a3, a4, a5, a6, a7, constant, t0){\
|
||||
t0 = constant;\
|
||||
a0 = _mm_xor_si128(a0, t0);\
|
||||
a1 = _mm_xor_si128(a1, t0);\
|
||||
a2 = _mm_xor_si128(a2, t0);\
|
||||
a3 = _mm_xor_si128(a3, t0);\
|
||||
a4 = _mm_xor_si128(a4, t0);\
|
||||
a5 = _mm_xor_si128(a5, t0);\
|
||||
a6 = _mm_xor_si128(a6, t0);\
|
||||
a7 = _mm_xor_si128(a7, t0);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Set Substitute Core Constants
|
||||
* */
|
||||
#define VPERM_Substitute_Core_Set_Const(c0, c1, c2){\
|
||||
VPERM_Transform_Set_Const(VPERM_INV, c0, c1, c2);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Substitute Core
|
||||
* first part of sbox inverse computation
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0 = 1 row
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* b0a, b0b = inputs for lookup step
|
||||
* */
|
||||
#define VPERM_Substitute_Core(a0, b0a, b0b, t0, t1, c0, c1, c2){\
|
||||
t0 = c0;\
|
||||
t0 = _mm_andnot_si128(t0, a0);\
|
||||
t0 = _mm_srli_epi32(t0, 4);\
|
||||
a0 = _mm_and_si128(a0, c0);\
|
||||
b0a = c1;\
|
||||
b0a = _mm_shuffle_epi8(b0a, a0);\
|
||||
a0 = _mm_xor_si128(a0, t0);\
|
||||
b0b = c2;\
|
||||
b0b = _mm_shuffle_epi8(b0b, t0);\
|
||||
b0b = _mm_xor_si128(b0b, b0a);\
|
||||
t1 = c2;\
|
||||
t1 = _mm_shuffle_epi8(t1, a0);\
|
||||
t1 = _mm_xor_si128(t1, b0a);\
|
||||
b0a = c2;\
|
||||
b0a = _mm_shuffle_epi8(b0a, b0b);\
|
||||
b0a = _mm_xor_si128(b0a, a0);\
|
||||
b0b = c2;\
|
||||
b0b = _mm_shuffle_epi8(b0b, t1);\
|
||||
b0b = _mm_xor_si128(b0b, t0);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* Lookup
|
||||
* second part of sbox inverse computation
|
||||
* this function is derived from:
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0a, a0b = output of Substitution Core
|
||||
* table = lookup table to use (*1 / *2 / *4)
|
||||
* t0 = clobber
|
||||
* outputs:
|
||||
* b0 = output of sbox + multiplication
|
||||
* */
|
||||
#define VPERM_Lookup(a0a, a0b, table, b0, t0){\
|
||||
b0 = ((__m128i*) table )[0];\
|
||||
t0 = ((__m128i*) table )[1];\
|
||||
b0 = _mm_shuffle_epi8(b0, a0b);\
|
||||
t0 = _mm_shuffle_epi8(t0, a0a);\
|
||||
b0 = _mm_xor_si128(b0, t0);\
|
||||
}/**/
|
||||
|
||||
/* VPERM
|
||||
* SubBytes and *2 / *4
|
||||
* this function is derived from:
|
||||
* Constant-time SSSE3 AES core implementation
|
||||
* by Mike Hamburg
|
||||
* and
|
||||
* vperm and aes_ni implementations of hash function Grostl
|
||||
* by Cagdas CALIK
|
||||
* inputs:
|
||||
* a0-a7 = state
|
||||
* t*, c* = clobbers
|
||||
* outputs:
|
||||
* a0-a7 = state * 4
|
||||
* c2 = row0 * 2 -> b0
|
||||
* c1 = row7 * 2 -> b3
|
||||
* c0 = row7 * 1 -> b4
|
||||
* t2 = row4 * 1 -> b7
|
||||
* TEMP_MUL1 = row(i) * 1
|
||||
* TEMP_MUL2 = row(i) * 2
|
||||
*
|
||||
* call:VPERM_SUB_MULTIPLY(a0, a1, a2, a3, a4, a5, a6, a7, b1, b2, b5, b6, b0, b3, b4, b7) */
|
||||
#define VPERM_SUB_MULTIPLY(a0, a1, a2, a3, a4, a5, a6, a7, t0, t1, t3, t4, c2, c1, c0, t2){\
|
||||
/* set Constants */\
|
||||
VPERM_Substitute_Core_Set_Const(c0, c1, c2);\
|
||||
/* row 1 */\
|
||||
VPERM_Substitute_Core(a1, t0, t1, t3, t4, c0, c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
TEMP_MUL1[1] = t2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
TEMP_MUL2[1] = t3;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a1, t4);\
|
||||
/* --- */\
|
||||
/* row 2 */\
|
||||
VPERM_Substitute_Core(a2, t0, t1, t3, t4, c0, c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
TEMP_MUL1[2] = t2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
TEMP_MUL2[2] = t3;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a2, t4);\
|
||||
/* --- */\
|
||||
/* row 3 */\
|
||||
VPERM_Substitute_Core(a3, t0, t1, t3, t4, c0, c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
TEMP_MUL1[3] = t2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
TEMP_MUL2[3] = t3;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a3, t4);\
|
||||
/* --- */\
|
||||
/* row 5 */\
|
||||
VPERM_Substitute_Core(a5, t0, t1, t3, t4, c0, c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
TEMP_MUL1[5] = t2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
TEMP_MUL2[5] = t3;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a5, t4);\
|
||||
/* --- */\
|
||||
/* row 6 */\
|
||||
VPERM_Substitute_Core(a6, t0, t1, t3, t4, c0, c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
TEMP_MUL1[6] = t2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
TEMP_MUL2[6] = t3;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a6, t4);\
|
||||
/* --- */\
|
||||
/* row 7 */\
|
||||
VPERM_Substitute_Core(a7, t0, t1, t3, t4, c0, c1, c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4);\
|
||||
TEMP_MUL1[7] = t2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, c1, t4); /*c1 -> b3*/\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a7, t4);\
|
||||
/* --- */\
|
||||
/* row 4 */\
|
||||
VPERM_Substitute_Core(a4, t0, t1, t3, t4, c0, (VPERM_INV[0]), c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, t2, t4); /*t2 -> b7*/\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, t3, t4);\
|
||||
TEMP_MUL2[4] = t3;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a4, t4);\
|
||||
/* --- */\
|
||||
/* row 0 */\
|
||||
VPERM_Substitute_Core(a0, t0, t1, t3, t4, c0, (VPERM_INV[0]), c2);\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB1, c0, t4); /*c0 -> b4*/\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB2, c2, t4); /*c2 -> b0*/\
|
||||
TEMP_MUL2[0] = c2;\
|
||||
VPERM_Lookup(t0, t1, VPERM_SB4, a0, t4);\
|
||||
/* --- */\
|
||||
}/**/
|
||||
|
||||
|
||||
/* Optimized MixBytes
|
||||
* inputs:
|
||||
* a0-a7 = (row0-row7) * 4
|
||||
* b0 = row0 * 2
|
||||
* b3 = row7 * 2
|
||||
* b4 = row7 * 1
|
||||
* b7 = row4 * 1
|
||||
* all *1 and *2 values must also be in TEMP_MUL1, TEMP_MUL2
|
||||
* output: b0-b7
|
||||
* */
|
||||
#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* save one value */\
|
||||
TEMP_MUL4 = a3;\
|
||||
/* 1 */\
|
||||
b1 = a0;\
|
||||
b1 = _mm_xor_si128(b1, a5);\
|
||||
b1 = _mm_xor_si128(b1, b4); /* -> helper! */\
|
||||
b1 = _mm_xor_si128(b1, (TEMP_MUL2[3]));\
|
||||
b2 = b1;\
|
||||
\
|
||||
/* 2 */\
|
||||
b5 = a1;\
|
||||
b5 = _mm_xor_si128(b5, a4);\
|
||||
b5 = _mm_xor_si128(b5, b7); /* -> helper! */\
|
||||
b5 = _mm_xor_si128(b5, b3); /* -> helper! */\
|
||||
b6 = b5;\
|
||||
\
|
||||
/* 4 */\
|
||||
b7 = _mm_xor_si128(b7, a6);\
|
||||
/*b7 = _mm_xor_si128(b7, (TEMP_MUL1[4])); -> helper! */\
|
||||
b7 = _mm_xor_si128(b7, (TEMP_MUL1[6]));\
|
||||
b7 = _mm_xor_si128(b7, (TEMP_MUL2[1]));\
|
||||
b7 = _mm_xor_si128(b7, b3); /* -> helper! */\
|
||||
b2 = _mm_xor_si128(b2, b7);\
|
||||
\
|
||||
/* 3 */\
|
||||
b0 = _mm_xor_si128(b0, a7);\
|
||||
b0 = _mm_xor_si128(b0, (TEMP_MUL1[5]));\
|
||||
b0 = _mm_xor_si128(b0, (TEMP_MUL1[7]));\
|
||||
/*b0 = _mm_xor_si128(b0, (TEMP_MUL2[0])); -> helper! */\
|
||||
b0 = _mm_xor_si128(b0, (TEMP_MUL2[2]));\
|
||||
b3 = b0;\
|
||||
b1 = _mm_xor_si128(b1, b0);\
|
||||
b0 = _mm_xor_si128(b0, b7); /* moved from 4 */\
|
||||
\
|
||||
/* 5 */\
|
||||
b4 = _mm_xor_si128(b4, a2);\
|
||||
/*b4 = _mm_xor_si128(b4, (TEMP_MUL1[0])); -> helper! */\
|
||||
b4 = _mm_xor_si128(b4, (TEMP_MUL1[2]));\
|
||||
b4 = _mm_xor_si128(b4, (TEMP_MUL2[3]));\
|
||||
b4 = _mm_xor_si128(b4, (TEMP_MUL2[5]));\
|
||||
b3 = _mm_xor_si128(b3, b4);\
|
||||
b6 = _mm_xor_si128(b6, b4);\
|
||||
\
|
||||
/* 6 */\
|
||||
a3 = _mm_xor_si128(a3, (TEMP_MUL1[1]));\
|
||||
a3 = _mm_xor_si128(a3, (TEMP_MUL1[3]));\
|
||||
a3 = _mm_xor_si128(a3, (TEMP_MUL2[4]));\
|
||||
a3 = _mm_xor_si128(a3, (TEMP_MUL2[6]));\
|
||||
b4 = _mm_xor_si128(b4, a3);\
|
||||
b5 = _mm_xor_si128(b5, a3);\
|
||||
b7 = _mm_xor_si128(b7, a3);\
|
||||
\
|
||||
/* 7 */\
|
||||
a1 = _mm_xor_si128(a1, (TEMP_MUL1[1]));\
|
||||
a1 = _mm_xor_si128(a1, (TEMP_MUL2[4]));\
|
||||
b2 = _mm_xor_si128(b2, a1);\
|
||||
b3 = _mm_xor_si128(b3, a1);\
|
||||
\
|
||||
/* 8 */\
|
||||
a5 = _mm_xor_si128(a5, (TEMP_MUL1[5]));\
|
||||
a5 = _mm_xor_si128(a5, (TEMP_MUL2[0]));\
|
||||
b6 = _mm_xor_si128(b6, a5);\
|
||||
b7 = _mm_xor_si128(b7, a5);\
|
||||
\
|
||||
/* 9 */\
|
||||
a3 = TEMP_MUL1[2];\
|
||||
a3 = _mm_xor_si128(a3, (TEMP_MUL2[5]));\
|
||||
b0 = _mm_xor_si128(b0, a3);\
|
||||
b5 = _mm_xor_si128(b5, a3);\
|
||||
\
|
||||
/* 10 */\
|
||||
a1 = TEMP_MUL1[6];\
|
||||
a1 = _mm_xor_si128(a1, (TEMP_MUL2[1]));\
|
||||
b1 = _mm_xor_si128(b1, a1);\
|
||||
b4 = _mm_xor_si128(b4, a1);\
|
||||
\
|
||||
/* 11 */\
|
||||
a5 = TEMP_MUL1[3];\
|
||||
a5 = _mm_xor_si128(a5, (TEMP_MUL2[6]));\
|
||||
b1 = _mm_xor_si128(b1, a5);\
|
||||
b6 = _mm_xor_si128(b6, a5);\
|
||||
\
|
||||
/* 12 */\
|
||||
a3 = TEMP_MUL1[7];\
|
||||
a3 = _mm_xor_si128(a3, (TEMP_MUL2[2]));\
|
||||
b2 = _mm_xor_si128(b2, a3);\
|
||||
b5 = _mm_xor_si128(b5, a3);\
|
||||
\
|
||||
/* 13 */\
|
||||
b0 = _mm_xor_si128(b0, (TEMP_MUL4));\
|
||||
b0 = _mm_xor_si128(b0, a4);\
|
||||
b1 = _mm_xor_si128(b1, a4);\
|
||||
b3 = _mm_xor_si128(b3, a6);\
|
||||
b4 = _mm_xor_si128(b4, a0);\
|
||||
b4 = _mm_xor_si128(b4, a7);\
|
||||
b5 = _mm_xor_si128(b5, a0);\
|
||||
b7 = _mm_xor_si128(b7, a2);\
|
||||
}/**/
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
SET_SHARED_CONSTANTS();\
|
||||
SUBSH_MASK[0] = _mm_set_epi32(0x080f0e0d, 0x0c0b0a09, 0x07060504, 0x03020100);\
|
||||
SUBSH_MASK[1] = _mm_set_epi32(0x0a09080f, 0x0e0d0c0b, 0x00070605, 0x04030201);\
|
||||
SUBSH_MASK[2] = _mm_set_epi32(0x0c0b0a09, 0x080f0e0d, 0x01000706, 0x05040302);\
|
||||
SUBSH_MASK[3] = _mm_set_epi32(0x0e0d0c0b, 0x0a09080f, 0x02010007, 0x06050403);\
|
||||
SUBSH_MASK[4] = _mm_set_epi32(0x0f0e0d0c, 0x0b0a0908, 0x03020100, 0x07060504);\
|
||||
SUBSH_MASK[5] = _mm_set_epi32(0x09080f0e, 0x0d0c0b0a, 0x04030201, 0x00070605);\
|
||||
SUBSH_MASK[6] = _mm_set_epi32(0x0b0a0908, 0x0f0e0d0c, 0x05040302, 0x01000706);\
|
||||
SUBSH_MASK[7] = _mm_set_epi32(0x0d0c0b0a, 0x09080f0e, 0x06050403, 0x02010007);\
|
||||
for(i = 0; i < ROUNDS512; i++)\
|
||||
{\
|
||||
ROUND_CONST_L0[i] = _mm_set_epi32(0xffffffff, 0xffffffff, 0x70605040 ^ (i * 0x01010101), 0x30201000 ^ (i * 0x01010101));\
|
||||
ROUND_CONST_L7[i] = _mm_set_epi32(0x8f9fafbf ^ (i * 0x01010101), 0xcfdfefff ^ (i * 0x01010101), 0x00000000, 0x00000000);\
|
||||
}\
|
||||
ROUND_CONST_Lx = _mm_set_epi32(0xffffffff, 0xffffffff, 0x00000000, 0x00000000);\
|
||||
}/**/
|
||||
|
||||
/* vperm:
|
||||
* transformation before rounds with ipt
|
||||
* first round add transformed constant
|
||||
* middle rounds: add constant XOR 0x15...15
|
||||
* last round: additionally add 0x15...15 after MB
|
||||
* transformation after rounds with opt
|
||||
*/
|
||||
/* one round
|
||||
* i = round number
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
*/
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* AddRoundConstant + ShiftBytes (interleaved) */\
|
||||
b1 = ROUND_CONST_Lx;\
|
||||
a0 = _mm_xor_si128(a0, (ROUND_CONST_L0[i]));\
|
||||
a1 = _mm_xor_si128(a1, b1);\
|
||||
a2 = _mm_xor_si128(a2, b1);\
|
||||
a3 = _mm_xor_si128(a3, b1);\
|
||||
a0 = _mm_shuffle_epi8(a0, (SUBSH_MASK[0]));\
|
||||
a1 = _mm_shuffle_epi8(a1, (SUBSH_MASK[1]));\
|
||||
a4 = _mm_xor_si128(a4, b1);\
|
||||
a2 = _mm_shuffle_epi8(a2, (SUBSH_MASK[2]));\
|
||||
a3 = _mm_shuffle_epi8(a3, (SUBSH_MASK[3]));\
|
||||
a5 = _mm_xor_si128(a5, b1);\
|
||||
a6 = _mm_xor_si128(a6, b1);\
|
||||
a4 = _mm_shuffle_epi8(a4, (SUBSH_MASK[4]));\
|
||||
a5 = _mm_shuffle_epi8(a5, (SUBSH_MASK[5]));\
|
||||
a7 = _mm_xor_si128(a7, (ROUND_CONST_L7[i]));\
|
||||
a6 = _mm_shuffle_epi8(a6, (SUBSH_MASK[6]));\
|
||||
a7 = _mm_shuffle_epi8(a7, (SUBSH_MASK[7]));\
|
||||
/* SubBytes + Multiplication by 2 and 4 */\
|
||||
VPERM_SUB_MULTIPLY(a0, a1, a2, a3, a4, a5, a6, a7, b1, b2, b5, b6, b0, b3, b4, b7);\
|
||||
/* MixBytes */\
|
||||
MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
|
||||
}/**/
|
||||
|
||||
/* 10 rounds, P and Q in parallel */
|
||||
#define ROUNDS_P_Q(){\
|
||||
VPERM_Add_Constant(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, ALL_15, xmm0);\
|
||||
ROUND(0, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(1, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(2, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(3, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(4, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(5, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(6, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(7, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
ROUND(8, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
ROUND(9, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
VPERM_Add_Constant(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, ALL_15, xmm0);\
|
||||
}
|
||||
|
||||
|
||||
/* Matrix Transpose Step 1
|
||||
* input is a 512-bit state with two columns in one xmm
|
||||
* output is a 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i3
|
||||
* outputs: i0, o1-o3
|
||||
* clobbers: t0
|
||||
*/
|
||||
#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
|
||||
t0 = TRANSP_MASK;\
|
||||
\
|
||||
i0 = _mm_shuffle_epi8(i0, t0);\
|
||||
i1 = _mm_shuffle_epi8(i1, t0);\
|
||||
i2 = _mm_shuffle_epi8(i2, t0);\
|
||||
i3 = _mm_shuffle_epi8(i3, t0);\
|
||||
\
|
||||
o1 = i0;\
|
||||
t0 = i2;\
|
||||
\
|
||||
i0 = _mm_unpacklo_epi16(i0, i1);\
|
||||
o1 = _mm_unpackhi_epi16(o1, i1);\
|
||||
i2 = _mm_unpacklo_epi16(i2, i3);\
|
||||
t0 = _mm_unpackhi_epi16(t0, i3);\
|
||||
\
|
||||
i0 = _mm_shuffle_epi32(i0, 216);\
|
||||
o1 = _mm_shuffle_epi32(o1, 216);\
|
||||
i2 = _mm_shuffle_epi32(i2, 216);\
|
||||
t0 = _mm_shuffle_epi32(t0, 216);\
|
||||
\
|
||||
o2 = i0;\
|
||||
o3 = o1;\
|
||||
\
|
||||
i0 = _mm_unpacklo_epi32(i0, i2);\
|
||||
o1 = _mm_unpacklo_epi32(o1, t0);\
|
||||
o2 = _mm_unpackhi_epi32(o2, i2);\
|
||||
o3 = _mm_unpackhi_epi32(o3, t0);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Step 2
|
||||
* input are two 512-bit states with two rows in one xmm
|
||||
* output are two 512-bit states with one row of each state in one xmm
|
||||
* inputs: i0-i3 = P, i4-i7 = Q
|
||||
* outputs: (i0, o1-o7) = (P|Q)
|
||||
* possible reassignments: (output reg = input reg)
|
||||
* * i1 -> o3-7
|
||||
* * i2 -> o5-7
|
||||
* * i3 -> o7
|
||||
* * i4 -> o3-7
|
||||
* * i5 -> o6-7
|
||||
*/
|
||||
#define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
|
||||
o1 = i0;\
|
||||
o2 = i1;\
|
||||
i0 = _mm_unpacklo_epi64(i0, i4);\
|
||||
o1 = _mm_unpackhi_epi64(o1, i4);\
|
||||
o3 = i1;\
|
||||
o4 = i2;\
|
||||
o2 = _mm_unpacklo_epi64(o2, i5);\
|
||||
o3 = _mm_unpackhi_epi64(o3, i5);\
|
||||
o5 = i2;\
|
||||
o6 = i3;\
|
||||
o4 = _mm_unpacklo_epi64(o4, i6);\
|
||||
o5 = _mm_unpackhi_epi64(o5, i6);\
|
||||
o7 = i3;\
|
||||
o6 = _mm_unpacklo_epi64(o6, i7);\
|
||||
o7 = _mm_unpackhi_epi64(o7, i7);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Inverse Step 2
|
||||
* input are two 512-bit states with one row of each state in one xmm
|
||||
* output are two 512-bit states with two rows in one xmm
|
||||
* inputs: i0-i7 = (P|Q)
|
||||
* outputs: (i0, i2, i4, i6) = P, (o0-o3) = Q
|
||||
*/
|
||||
#define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
|
||||
o0 = i0;\
|
||||
i0 = _mm_unpacklo_epi64(i0, i1);\
|
||||
o0 = _mm_unpackhi_epi64(o0, i1);\
|
||||
o1 = i2;\
|
||||
i2 = _mm_unpacklo_epi64(i2, i3);\
|
||||
o1 = _mm_unpackhi_epi64(o1, i3);\
|
||||
o2 = i4;\
|
||||
i4 = _mm_unpacklo_epi64(i4, i5);\
|
||||
o2 = _mm_unpackhi_epi64(o2, i5);\
|
||||
o3 = i6;\
|
||||
i6 = _mm_unpacklo_epi64(i6, i7);\
|
||||
o3 = _mm_unpackhi_epi64(o3, i7);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Step 2
|
||||
* input is one 512-bit state with two rows in one xmm
|
||||
* output is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* inputs: i0,i2,i4,i6 = S
|
||||
* outputs: (i0-7) = (0|S)
|
||||
*/
|
||||
#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
|
||||
t0 = _mm_xor_si128(t0, t0);\
|
||||
i1 = i0;\
|
||||
i3 = i2;\
|
||||
i5 = i4;\
|
||||
i7 = i6;\
|
||||
i0 = _mm_unpacklo_epi64(i0, t0);\
|
||||
i1 = _mm_unpackhi_epi64(i1, t0);\
|
||||
i2 = _mm_unpacklo_epi64(i2, t0);\
|
||||
i3 = _mm_unpackhi_epi64(i3, t0);\
|
||||
i4 = _mm_unpacklo_epi64(i4, t0);\
|
||||
i5 = _mm_unpackhi_epi64(i5, t0);\
|
||||
i6 = _mm_unpacklo_epi64(i6, t0);\
|
||||
i7 = _mm_unpackhi_epi64(i7, t0);\
|
||||
}/**/
|
||||
|
||||
/* Matrix Transpose Output Inverse Step 2
|
||||
* input is one 512-bit state with one row in the low 64-bits of one xmm
|
||||
* output is one 512-bit state with two rows in one xmm
|
||||
* inputs: i0-i7 = (0|S)
|
||||
* outputs: (i0, i2, i4, i6) = S
|
||||
*/
|
||||
#define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
|
||||
i0 = _mm_unpacklo_epi64(i0, i1);\
|
||||
i2 = _mm_unpacklo_epi64(i2, i3);\
|
||||
i4 = _mm_unpacklo_epi64(i4, i5);\
|
||||
i6 = _mm_unpacklo_epi64(i6, i7);\
|
||||
}/**/
|
||||
|
||||
|
||||
/* transform round constants into VPERM mode */
|
||||
#define VPERM_Transform_RoundConst_CNT2(i, j){\
|
||||
xmm0 = ROUND_CONST_L0[i];\
|
||||
xmm1 = ROUND_CONST_L7[i];\
|
||||
xmm2 = ROUND_CONST_L0[j];\
|
||||
xmm3 = ROUND_CONST_L7[j];\
|
||||
VPERM_Transform_State(xmm0, xmm1, xmm2, xmm3, VPERM_IPT, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10);\
|
||||
xmm0 = _mm_xor_si128(xmm0, (ALL_15));\
|
||||
xmm1 = _mm_xor_si128(xmm1, (ALL_15));\
|
||||
xmm2 = _mm_xor_si128(xmm2, (ALL_15));\
|
||||
xmm3 = _mm_xor_si128(xmm3, (ALL_15));\
|
||||
ROUND_CONST_L0[i] = xmm0;\
|
||||
ROUND_CONST_L7[i] = xmm1;\
|
||||
ROUND_CONST_L0[j] = xmm2;\
|
||||
ROUND_CONST_L7[j] = xmm3;\
|
||||
}/**/
|
||||
|
||||
/* transform round constants into VPERM mode */
|
||||
#define VPERM_Transform_RoundConst(){\
|
||||
xmm0 = ROUND_CONST_Lx;\
|
||||
VPERM_Transform(xmm0, xmm1, VPERM_IPT, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10);\
|
||||
xmm0 = _mm_xor_si128(xmm0, (ALL_15));\
|
||||
ROUND_CONST_Lx = xmm0;\
|
||||
VPERM_Transform_RoundConst_CNT2(0, 1);\
|
||||
VPERM_Transform_RoundConst_CNT2(2, 3);\
|
||||
VPERM_Transform_RoundConst_CNT2(4, 5);\
|
||||
VPERM_Transform_RoundConst_CNT2(6, 7);\
|
||||
VPERM_Transform_RoundConst_CNT2(8, 9);\
|
||||
}/**/
|
||||
|
||||
void INIT256(u64* h)
|
||||
{
|
||||
__m128i* const chaining = (__m128i*) h;
|
||||
static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
|
||||
static __m128i xmm8, xmm9, xmm10, /*xmm11,*/ xmm12, xmm13, xmm14, xmm15;
|
||||
|
||||
/* transform round constants into VPERM mode */
|
||||
VPERM_Transform_RoundConst();
|
||||
|
||||
/* load IV into registers xmm12 - xmm15 */
|
||||
xmm12 = chaining[0];
|
||||
xmm13 = chaining[1];
|
||||
xmm14 = chaining[2];
|
||||
xmm15 = chaining[3];
|
||||
|
||||
/* transform chaining value from column ordering into row ordering */
|
||||
/* we put two rows (64 bit) of the IV into one 128-bit XMM register */
|
||||
VPERM_Transform_State(xmm12, xmm13, xmm14, xmm15, VPERM_IPT, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);
|
||||
Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
|
||||
|
||||
/* store transposed IV */
|
||||
chaining[0] = xmm12;
|
||||
chaining[1] = xmm2;
|
||||
chaining[2] = xmm6;
|
||||
chaining[3] = xmm7;
|
||||
}
|
||||
|
||||
void TF512(u64* h, u64* m)
|
||||
{
|
||||
__m128i* const chaining = (__m128i*) h;
|
||||
__m128i* const message = (__m128i*) m;
|
||||
static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
|
||||
static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
|
||||
static __m128i TEMP_MUL1[8];
|
||||
static __m128i TEMP_MUL2[8];
|
||||
static __m128i TEMP_MUL4;
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_START;
|
||||
#endif
|
||||
|
||||
/* load message into registers xmm12 - xmm15 */
|
||||
xmm12 = message[0];
|
||||
xmm13 = message[1];
|
||||
xmm14 = message[2];
|
||||
xmm15 = message[3];
|
||||
|
||||
/* transform message M from column ordering into row ordering */
|
||||
/* we first put two rows (64 bit) of the message into one 128-bit xmm register */
|
||||
VPERM_Transform_State(xmm12, xmm13, xmm14, xmm15, VPERM_IPT, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);
|
||||
Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
|
||||
|
||||
/* load previous chaining value */
|
||||
/* we first put two rows (64 bit) of the CV into one 128-bit xmm register */
|
||||
xmm8 = chaining[0];
|
||||
xmm0 = chaining[1];
|
||||
xmm4 = chaining[2];
|
||||
xmm5 = chaining[3];
|
||||
|
||||
/* xor message to CV get input of P */
|
||||
/* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
|
||||
xmm8 = _mm_xor_si128(xmm8, xmm12);
|
||||
xmm0 = _mm_xor_si128(xmm0, xmm2);
|
||||
xmm4 = _mm_xor_si128(xmm4, xmm6);
|
||||
xmm5 = _mm_xor_si128(xmm5, xmm7);
|
||||
|
||||
/* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
|
||||
/* result: the 8 rows of P and Q in xmm8 - xmm12 */
|
||||
Matrix_Transpose_B(xmm8, xmm0, xmm4, xmm5, xmm12, xmm2, xmm6, xmm7, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);
|
||||
|
||||
/* compute the two permutations P and Q in parallel */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P or two rows of Q in one xmm register */
|
||||
Matrix_Transpose_B_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3);
|
||||
|
||||
/* xor output of P and Q */
|
||||
/* result: P(CV+M)+Q(M) in xmm0...xmm3 */
|
||||
xmm0 = _mm_xor_si128(xmm0, xmm8);
|
||||
xmm1 = _mm_xor_si128(xmm1, xmm10);
|
||||
xmm2 = _mm_xor_si128(xmm2, xmm12);
|
||||
xmm3 = _mm_xor_si128(xmm3, xmm14);
|
||||
|
||||
/* xor CV (feed-forward) */
|
||||
/* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
|
||||
xmm0 = _mm_xor_si128(xmm0, (chaining[0]));
|
||||
xmm1 = _mm_xor_si128(xmm1, (chaining[1]));
|
||||
xmm2 = _mm_xor_si128(xmm2, (chaining[2]));
|
||||
xmm3 = _mm_xor_si128(xmm3, (chaining[3]));
|
||||
|
||||
/* store CV */
|
||||
chaining[0] = xmm0;
|
||||
chaining[1] = xmm1;
|
||||
chaining[2] = xmm2;
|
||||
chaining[3] = xmm3;
|
||||
|
||||
#ifdef IACA_TRACE
|
||||
IACA_END;
|
||||
#endif
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
void OF512(u64* h)
|
||||
{
|
||||
__m128i* const chaining = (__m128i*) h;
|
||||
static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
|
||||
static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
|
||||
static __m128i TEMP_MUL1[8];
|
||||
static __m128i TEMP_MUL2[8];
|
||||
static __m128i TEMP_MUL4;
|
||||
|
||||
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
|
||||
xmm8 = chaining[0];
|
||||
xmm10 = chaining[1];
|
||||
xmm12 = chaining[2];
|
||||
xmm14 = chaining[3];
|
||||
|
||||
/* there are now 2 rows of the CV in one xmm register */
|
||||
/* unpack to get 1 row of P (64 bit) into one half of an xmm register */
|
||||
/* result: the 8 input rows of P in xmm8 - xmm15 */
|
||||
Matrix_Transpose_O_B(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0);
|
||||
|
||||
/* compute the permutation P */
|
||||
/* result: the output of P(CV) in xmm8 - xmm15 */
|
||||
ROUNDS_P_Q();
|
||||
|
||||
/* unpack again to get two rows of P in one xmm register */
|
||||
/* result: P(CV) in xmm8, xmm10, xmm12, xmm14 */
|
||||
Matrix_Transpose_O_B_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);
|
||||
|
||||
/* xor CV to P output (feed-forward) */
|
||||
/* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
|
||||
xmm8 = _mm_xor_si128(xmm8, (chaining[0]));
|
||||
xmm10 = _mm_xor_si128(xmm10, (chaining[1]));
|
||||
xmm12 = _mm_xor_si128(xmm12, (chaining[2]));
|
||||
xmm14 = _mm_xor_si128(xmm14, (chaining[3]));
|
||||
|
||||
/* transform state back from row ordering into column ordering */
|
||||
/* result: final hash value in xmm9, xmm11 */
|
||||
Matrix_Transpose_A(xmm8, xmm10, xmm12, xmm14, xmm4, xmm9, xmm11, xmm0);
|
||||
VPERM_Transform(xmm9, xmm11, VPERM_OPT, xmm0, xmm1, xmm2, xmm3, xmm5, xmm6, xmm7);
|
||||
|
||||
/* we only need to return the truncated half of the state */
|
||||
chaining[2] = xmm9;
|
||||
chaining[3] = xmm11;
|
||||
|
||||
return;
|
||||
}//OF512()
|
||||
|
||||
|
||||
|
||||
@@ -16,48 +16,13 @@
|
||||
|
||||
#ifdef __AES__
|
||||
|
||||
#include "groestl-version.h"
|
||||
|
||||
#ifdef TASM
|
||||
#ifdef VAES
|
||||
#include "groestl-asm-aes.h"
|
||||
#else
|
||||
#ifdef VAVX
|
||||
#include "groestl-asm-avx.h"
|
||||
#else
|
||||
#ifdef VVPERM
|
||||
#include "groestl-asm-vperm.h"
|
||||
#else
|
||||
#error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
|
||||
#endif
|
||||
#endif
|
||||
#endif
|
||||
#else
|
||||
#ifdef TINTR
|
||||
#ifdef VAES
|
||||
#include "groestl-intr-aes.h"
|
||||
#else
|
||||
#ifdef VAVX
|
||||
#include "groestl-intr-avx.h"
|
||||
#else
|
||||
#ifdef VVPERM
|
||||
#include "groestl-intr-vperm.h"
|
||||
#else
|
||||
#error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
|
||||
#endif
|
||||
#endif
|
||||
#endif
|
||||
#else
|
||||
#error NO TYPE SPECIFIED (-DT[ASM/INTR])
|
||||
#endif
|
||||
#endif
|
||||
#include "groestl-intr-aes.h"
|
||||
|
||||
HashReturn_gr init_groestl( hashState_groestl* ctx, int hashlen )
|
||||
{
|
||||
int i;
|
||||
|
||||
ctx->hashlen = hashlen;
|
||||
SET_CONSTANTS();
|
||||
|
||||
if (ctx->chaining == NULL || ctx->buffer == NULL)
|
||||
return FAIL_GR;
|
||||
@@ -70,8 +35,6 @@ HashReturn_gr init_groestl( hashState_groestl* ctx, int hashlen )
|
||||
|
||||
// The only non-zero in the IV is len. It can be hard coded.
|
||||
ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
|
||||
// ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
|
||||
// INIT(ctx->chaining);
|
||||
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
@@ -92,8 +55,6 @@ HashReturn_gr reinit_groestl( hashState_groestl* ctx )
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
}
|
||||
ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
|
||||
// ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
|
||||
// INIT(ctx->chaining);
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
|
||||
@@ -109,7 +70,7 @@ HashReturn_gr reinit_groestl( hashState_groestl* ctx )
|
||||
// 5. Midstate will work at reduced impact than full hash, if total hash
|
||||
// (midstate + tail) is less than 1 block.
|
||||
// This, unfortunately, is the case with all current users.
|
||||
// 6. the morefull blocks the bigger the gain
|
||||
// 6. the more full blocks the bigger the gain
|
||||
|
||||
// use only for midstate precalc
|
||||
HashReturn_gr update_groestl( hashState_groestl* ctx, const void* input,
|
||||
@@ -143,12 +104,11 @@ HashReturn_gr update_groestl( hashState_groestl* ctx, const void* input,
|
||||
// deprecated do not use
|
||||
HashReturn_gr final_groestl( hashState_groestl* ctx, void* output )
|
||||
{
|
||||
const int len = (int)ctx->databitlen / 128; // bits to __m128i
|
||||
const int blocks = ctx->blk_count + 1; // adjust for final block
|
||||
|
||||
const int rem_ptr = ctx->rem_ptr; // end of data start of padding
|
||||
const int hashlen_m128i = ctx->hashlen / 16; // bytes to __m128i
|
||||
const int hash_offset = SIZE512 - hashlen_m128i; // where in buffer
|
||||
const int len = (int)ctx->databitlen / 128; // bits to __m128i
|
||||
const uint64_t blocks = ctx->blk_count + 1; // adjust for final block
|
||||
const int rem_ptr = ctx->rem_ptr; // end of data start of padding
|
||||
const int hashlen_m128i = ctx->hashlen / 16; // bytes to __m128i
|
||||
const int hash_offset = SIZE512 - hashlen_m128i; // where in buffer
|
||||
int i;
|
||||
|
||||
// first pad byte = 0x80, last pad byte = block count
|
||||
@@ -157,21 +117,18 @@ HashReturn_gr final_groestl( hashState_groestl* ctx, void* output )
|
||||
if ( rem_ptr == len - 1 )
|
||||
{
|
||||
// only 128 bits left in buffer, all padding at once
|
||||
ctx->buffer[rem_ptr] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
ctx->buffer[rem_ptr] = _mm_set_epi64x( blocks << 56, 0x80 );
|
||||
}
|
||||
else
|
||||
{
|
||||
// add first padding
|
||||
ctx->buffer[rem_ptr] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
ctx->buffer[rem_ptr] = m128_const_64( 0, 0x80 );
|
||||
// add zero padding
|
||||
for ( i = rem_ptr + 1; i < SIZE512 - 1; i++ )
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
|
||||
// add length padding, second last byte is zero unless blocks > 255
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks, blocks>>8, 0,0, 0,0,0,0,
|
||||
0, 0 ,0,0, 0,0,0,0 );
|
||||
ctx->buffer[i] = _mm_set_epi64x( blocks << 56, 0 );
|
||||
}
|
||||
|
||||
// digest final padding block and do output transform
|
||||
@@ -189,21 +146,20 @@ int groestl512_full( hashState_groestl* ctx, void* output,
|
||||
const void* input, uint64_t databitlen )
|
||||
{
|
||||
|
||||
int i;
|
||||
|
||||
ctx->hashlen = 64;
|
||||
SET_CONSTANTS();
|
||||
|
||||
for ( i = 0; i < SIZE512; i++ )
|
||||
{
|
||||
ctx->chaining[i] = _mm_setzero_si128();
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
}
|
||||
ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
int i;
|
||||
ctx->hashlen = 64;
|
||||
|
||||
for ( i = 0; i < SIZE512; i++ )
|
||||
{
|
||||
ctx->chaining[i] = _mm_setzero_si128();
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
}
|
||||
ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
|
||||
// --- update ---
|
||||
|
||||
const int len = (int)databitlen / 128;
|
||||
const int hashlen_m128i = ctx->hashlen / 16; // bytes to __m128i
|
||||
const int hash_offset = SIZE512 - hashlen_m128i;
|
||||
@@ -211,8 +167,6 @@ int groestl512_full( hashState_groestl* ctx, void* output,
|
||||
uint64_t blocks = len / SIZE512;
|
||||
__m128i* in = (__m128i*)input;
|
||||
|
||||
// --- update ---
|
||||
|
||||
// digest any full blocks, process directly from input
|
||||
for ( i = 0; i < blocks; i++ )
|
||||
TF1024( ctx->chaining, &in[ i * SIZE512 ] );
|
||||
@@ -231,26 +185,22 @@ int groestl512_full( hashState_groestl* ctx, void* output,
|
||||
if ( i == len -1 )
|
||||
{
|
||||
// only 128 bits left in buffer, all padding at once
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
ctx->buffer[i] = _mm_set_epi64x( blocks << 56, 0x80 );
|
||||
}
|
||||
else
|
||||
{
|
||||
// add first padding
|
||||
ctx->buffer[i] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
ctx->buffer[i] = m128_const_64( 0, 0x80 );
|
||||
// add zero padding
|
||||
for ( i += 1; i < SIZE512 - 1; i++ )
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
|
||||
// add length padding, second last byte is zero unless blocks > 255
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks, blocks>>8, 0,0, 0,0,0,0,
|
||||
0, 0 ,0,0, 0,0,0,0 );
|
||||
ctx->buffer[i] = _mm_set_epi64x( blocks << 56, 0 );
|
||||
}
|
||||
|
||||
// digest final padding block and do output transform
|
||||
TF1024( ctx->chaining, ctx->buffer );
|
||||
|
||||
OF1024( ctx->chaining );
|
||||
|
||||
// store hash result in output
|
||||
@@ -268,7 +218,7 @@ HashReturn_gr update_and_final_groestl( hashState_groestl* ctx, void* output,
|
||||
const int hashlen_m128i = ctx->hashlen / 16; // bytes to __m128i
|
||||
const int hash_offset = SIZE512 - hashlen_m128i;
|
||||
int rem = ctx->rem_ptr;
|
||||
int blocks = len / SIZE512;
|
||||
uint64_t blocks = len / SIZE512;
|
||||
__m128i* in = (__m128i*)input;
|
||||
int i;
|
||||
|
||||
@@ -292,26 +242,22 @@ HashReturn_gr update_and_final_groestl( hashState_groestl* ctx, void* output,
|
||||
if ( i == len -1 )
|
||||
{
|
||||
// only 128 bits left in buffer, all padding at once
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
ctx->buffer[i] = _mm_set_epi64x( blocks << 56, 0x80 );
|
||||
}
|
||||
else
|
||||
{
|
||||
// add first padding
|
||||
ctx->buffer[i] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
ctx->buffer[i] = m128_const_64( 0, 0x80 );
|
||||
// add zero padding
|
||||
for ( i += 1; i < SIZE512 - 1; i++ )
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
|
||||
// add length padding, second last byte is zero unless blocks > 255
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks, blocks>>8, 0,0, 0,0,0,0,
|
||||
0, 0 ,0,0, 0,0,0,0 );
|
||||
ctx->buffer[i] = _mm_set_epi64x( blocks << 56, 0 );
|
||||
}
|
||||
|
||||
// digest final padding block and do output transform
|
||||
TF1024( ctx->chaining, ctx->buffer );
|
||||
|
||||
OF1024( ctx->chaining );
|
||||
|
||||
// store hash result in output
|
||||
|
||||
@@ -13,41 +13,7 @@
|
||||
|
||||
#ifdef __AES__
|
||||
|
||||
#include "groestl-version.h"
|
||||
|
||||
#ifdef TASM
|
||||
#ifdef VAES
|
||||
#include "groestl256-asm-aes.h"
|
||||
#else
|
||||
#ifdef VAVX
|
||||
#include "groestl256-asm-avx.h"
|
||||
#else
|
||||
#ifdef VVPERM
|
||||
#include "groestl256-asm-vperm.h"
|
||||
#else
|
||||
#error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
|
||||
#endif
|
||||
#endif
|
||||
#endif
|
||||
#else
|
||||
#ifdef TINTR
|
||||
#ifdef VAES
|
||||
#include "groestl256-intr-aes.h"
|
||||
#else
|
||||
#ifdef VAVX
|
||||
#include "groestl256-intr-avx.h"
|
||||
#else
|
||||
#ifdef VVPERM
|
||||
#include "groestl256-intr-vperm.h"
|
||||
#else
|
||||
#error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
|
||||
#endif
|
||||
#endif
|
||||
#endif
|
||||
#else
|
||||
#error NO TYPE SPECIFIED (-DT[ASM/INTR])
|
||||
#endif
|
||||
#endif
|
||||
#include "groestl256-intr-aes.h"
|
||||
|
||||
/* initialise context */
|
||||
HashReturn_gr init_groestl256( hashState_groestl256* ctx, int hashlen )
|
||||
@@ -55,7 +21,6 @@ HashReturn_gr init_groestl256( hashState_groestl256* ctx, int hashlen )
|
||||
int i;
|
||||
|
||||
ctx->hashlen = hashlen;
|
||||
SET_CONSTANTS();
|
||||
|
||||
if (ctx->chaining == NULL || ctx->buffer == NULL)
|
||||
return FAIL_GR;
|
||||
@@ -249,6 +214,98 @@ HashReturn_gr update_and_final_groestl256( hashState_groestl256* ctx,
|
||||
return SUCCESS_GR;
|
||||
}
|
||||
|
||||
int groestl256_full( hashState_groestl256* ctx,
|
||||
void* output, const void* input, DataLength_gr databitlen )
|
||||
{
|
||||
int i;
|
||||
ctx->hashlen = 32;
|
||||
for ( i = 0; i < SIZE256; i++ )
|
||||
{
|
||||
ctx->chaining[i] = _mm_setzero_si128();
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
}
|
||||
((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
|
||||
INIT256( ctx->chaining );
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
|
||||
const int len = (int)databitlen / 128;
|
||||
const int hashlen_m128i = ctx->hashlen / 16; // bytes to __m128i
|
||||
const int hash_offset = SIZE256 - hashlen_m128i;
|
||||
int rem = ctx->rem_ptr;
|
||||
int blocks = len / SIZE256;
|
||||
__m128i* in = (__m128i*)input;
|
||||
|
||||
// --- update ---
|
||||
|
||||
// digest any full blocks, process directly from input
|
||||
for ( i = 0; i < blocks; i++ )
|
||||
TF512( ctx->chaining, &in[ i * SIZE256 ] );
|
||||
ctx->buf_ptr = blocks * SIZE256;
|
||||
|
||||
// cryptonight has 200 byte input, an odd number of __m128i
|
||||
// remainder is only 8 bytes, ie u64.
|
||||
if ( databitlen % 128 !=0 )
|
||||
{
|
||||
// must be cryptonight, copy 64 bits of data
|
||||
*(uint64_t*)(ctx->buffer) = *(uint64_t*)(&in[ ctx->buf_ptr ] );
|
||||
i = -1; // signal for odd length
|
||||
}
|
||||
else
|
||||
{
|
||||
// Copy any remaining data to buffer for final transform
|
||||
for ( i = 0; i < len % SIZE256; i++ )
|
||||
ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
|
||||
i += rem; // use i as rem_ptr in final
|
||||
}
|
||||
|
||||
//--- final ---
|
||||
|
||||
// adjust for final block
|
||||
blocks++;
|
||||
|
||||
if ( i == len - 1 )
|
||||
{
|
||||
// all padding at once
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks,blocks>>8,0,0, 0,0,0,0,
|
||||
0, 0,0,0, 0,0,0,0x80 );
|
||||
}
|
||||
else
|
||||
{
|
||||
if ( i == -1 )
|
||||
{
|
||||
// cryptonight odd length
|
||||
((uint64_t*)ctx->buffer)[ 1 ] = 0x80ull;
|
||||
// finish the block with zero and length padding as normal
|
||||
i = 0;
|
||||
}
|
||||
else
|
||||
{
|
||||
// add first padding
|
||||
ctx->buffer[i] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
|
||||
0,0,0,0, 0,0,0,0x80 );
|
||||
}
|
||||
// add zero padding
|
||||
for ( i += 1; i < SIZE256 - 1; i++ )
|
||||
ctx->buffer[i] = _mm_setzero_si128();
|
||||
// add length padding
|
||||
// cheat since we know the block count is trivial, good if block < 256
|
||||
ctx->buffer[i] = _mm_set_epi8( blocks,blocks>>8,0,0, 0,0,0,0,
|
||||
0, 0,0,0, 0,0,0,0 );
|
||||
}
|
||||
|
||||
// digest final padding block and do output transform
|
||||
TF512( ctx->chaining, ctx->buffer );
|
||||
OF512( ctx->chaining );
|
||||
|
||||
// store hash result in output
|
||||
for ( i = 0; i < hashlen_m128i; i++ )
|
||||
casti_m128i( output, i ) = ctx->chaining[ hash_offset + i ];
|
||||
|
||||
return SUCCESS_GR;
|
||||
}
|
||||
|
||||
|
||||
/* hash bit sequence */
|
||||
HashReturn_gr hash_groestl256(int hashbitlen,
|
||||
const BitSequence_gr* data,
|
||||
|
||||
@@ -115,4 +115,7 @@ HashReturn_gr hash_groestli256( int, const BitSequence_gr*, DataLength_gr,
|
||||
HashReturn_gr update_and_final_groestl256( hashState_groestl256*, void*,
|
||||
const void*, DataLength_gr );
|
||||
|
||||
int groestl256_full( hashState_groestl256* ctx,
|
||||
void* output, const void* input, DataLength_gr databitlen );
|
||||
|
||||
#endif /* __hash_h */
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "groestl-gate.h"
|
||||
|
||||
#if !defined(GROESTL_8WAY) && !defined(GROESTLX16R_4WAY)
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
@@ -88,4 +91,4 @@ int scanhash_groestl( struct work *work, uint32_t max_nonce,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -23,7 +23,6 @@ int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )
|
||||
int i;
|
||||
|
||||
ctx->hashlen = hashlen;
|
||||
SET_CONSTANTS();
|
||||
|
||||
if (ctx->chaining == NULL || ctx->buffer == NULL)
|
||||
return 1;
|
||||
@@ -36,9 +35,6 @@ int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )
|
||||
|
||||
// The only non-zero in the IV is len. It can be hard coded.
|
||||
ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
|
||||
// uint64_t len = U64BIG((uint64_t)LENGTH);
|
||||
// ctx->chaining[ COLS/2 -1 ] = _mm512_set4_epi64( len, 0, len, 0 );
|
||||
// INIT256_4way(ctx->chaining);
|
||||
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
@@ -46,6 +42,77 @@ int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )
|
||||
return 0;
|
||||
}
|
||||
|
||||
int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
|
||||
const void* input, uint64_t databitlen )
|
||||
{
|
||||
const int len = (int)databitlen / 128;
|
||||
const int hashlen_m128i = 32 / 16; // bytes to __m128i
|
||||
const int hash_offset = SIZE256 - hashlen_m128i;
|
||||
int rem = ctx->rem_ptr;
|
||||
int blocks = len / SIZE256;
|
||||
__m512i* in = (__m512i*)input;
|
||||
int i;
|
||||
|
||||
if (ctx->chaining == NULL || ctx->buffer == NULL)
|
||||
return 1;
|
||||
|
||||
for ( i = 0; i < SIZE256; i++ )
|
||||
{
|
||||
ctx->chaining[i] = m512_zero;
|
||||
ctx->buffer[i] = m512_zero;
|
||||
}
|
||||
|
||||
// The only non-zero in the IV is len. It can be hard coded.
|
||||
ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
|
||||
ctx->buf_ptr = 0;
|
||||
ctx->rem_ptr = 0;
|
||||
|
||||
// --- update ---
|
||||
|
||||
// digest any full blocks, process directly from input
|
||||
for ( i = 0; i < blocks; i++ )
|
||||
TF512_4way( ctx->chaining, &in[ i * SIZE256 ] );
|
||||
ctx->buf_ptr = blocks * SIZE256;
|
||||
|
||||
// copy any remaining data to buffer, it may already contain data
|
||||
// from a previous update for a midstate precalc
|
||||
for ( i = 0; i < len % SIZE256; i++ )
|
||||
ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
|
||||
i += rem; // use i as rem_ptr in final
|
||||
|
||||
//--- final ---
|
||||
|
||||
blocks++; // adjust for final block
|
||||
|
||||
if ( i == SIZE256 - 1 )
|
||||
{
|
||||
// only 1 vector left in buffer, all padding at once
|
||||
ctx->buffer[i] = m512_const2_64( (uint64_t)blocks << 56, 0x80 );
|
||||
}
|
||||
else
|
||||
{
|
||||
// add first padding
|
||||
ctx->buffer[i] = m512_const4_64( 0, 0x80, 0, 0x80 );
|
||||
// add zero padding
|
||||
for ( i += 1; i < SIZE256 - 1; i++ )
|
||||
ctx->buffer[i] = m512_zero;
|
||||
|
||||
// add length padding, second last byte is zero unless blocks > 255
|
||||
ctx->buffer[i] = m512_const2_64( (uint64_t)blocks << 56, 0 );
|
||||
}
|
||||
|
||||
// digest final padding block and do output transform
|
||||
TF512_4way( ctx->chaining, ctx->buffer );
|
||||
|
||||
OF512_4way( ctx->chaining );
|
||||
|
||||
// store hash result in output
|
||||
for ( i = 0; i < hashlen_m128i; i++ )
|
||||
casti_m512i( output, i ) = ctx->chaining[ hash_offset + i ];
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
int groestl256_4way_update_close( groestl256_4way_context* ctx, void* output,
|
||||
const void* input, uint64_t databitlen )
|
||||
{
|
||||
@@ -75,11 +142,11 @@ int groestl256_4way_update_close( groestl256_4way_context* ctx, void* output,
|
||||
blocks++; // adjust for final block
|
||||
|
||||
if ( i == SIZE256 - 1 )
|
||||
{
|
||||
{
|
||||
// only 1 vector left in buffer, all padding at once
|
||||
ctx->buffer[i] = m512_const1_128( _mm_set_epi8(
|
||||
blocks, blocks>>8,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80 ) );
|
||||
}
|
||||
}
|
||||
else
|
||||
{
|
||||
// add first padding
|
||||
|
||||
@@ -71,5 +71,8 @@ int groestl256_4way_init( groestl256_4way_context*, uint64_t );
|
||||
int groestl256_4way_update_close( groestl256_4way_context*, void*,
|
||||
const void*, uint64_t );
|
||||
|
||||
int groestl256_4way_full( groestl256_4way_context*, void*,
|
||||
const void*, uint64_t );
|
||||
|
||||
#endif
|
||||
#endif
|
||||
|
||||
@@ -14,17 +14,78 @@
|
||||
#include "groestl256-hash-4way.h"
|
||||
|
||||
#if defined(__VAES__)
|
||||
static const __m128i round_const_l0[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x7060504030201000, 0xffffffffffffffff },
|
||||
{ 0x7161514131211101, 0xffffffffffffffff },
|
||||
{ 0x7262524232221202, 0xffffffffffffffff },
|
||||
{ 0x7363534333231303, 0xffffffffffffffff },
|
||||
{ 0x7464544434241404, 0xffffffffffffffff },
|
||||
{ 0x7565554535251505, 0xffffffffffffffff },
|
||||
{ 0x7666564636261606, 0xffffffffffffffff },
|
||||
{ 0x7767574737271707, 0xffffffffffffffff },
|
||||
{ 0x7868584838281808, 0xffffffffffffffff },
|
||||
{ 0x7969594939291909, 0xffffffffffffffff }
|
||||
};
|
||||
|
||||
/* global constants */
|
||||
__m512i ROUND_CONST_Lx;
|
||||
__m512i ROUND_CONST_L0[ROUNDS512];
|
||||
__m512i ROUND_CONST_L7[ROUNDS512];
|
||||
//__m512i ROUND_CONST_P[ROUNDS1024];
|
||||
//__m512i ROUND_CONST_Q[ROUNDS1024];
|
||||
__m512i TRANSP_MASK;
|
||||
__m512i SUBSH_MASK[8];
|
||||
__m512i ALL_1B;
|
||||
__m512i ALL_FF;
|
||||
static const __m128i round_const_l7[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x0000000000000000, 0x8f9fafbfcfdfefff },
|
||||
{ 0x0000000000000000, 0x8e9eaebecedeeefe },
|
||||
{ 0x0000000000000000, 0x8d9dadbdcdddedfd },
|
||||
{ 0x0000000000000000, 0x8c9cacbcccdcecfc },
|
||||
{ 0x0000000000000000, 0x8b9babbbcbdbebfb },
|
||||
{ 0x0000000000000000, 0x8a9aaabacadaeafa },
|
||||
{ 0x0000000000000000, 0x8999a9b9c9d9e9f9 },
|
||||
{ 0x0000000000000000, 0x8898a8b8c8d8e8f8 },
|
||||
{ 0x0000000000000000, 0x8797a7b7c7d7e7f7 },
|
||||
{ 0x0000000000000000, 0x8696a6b6c6d6e6f6 }
|
||||
};
|
||||
|
||||
static const __m512i TRANSP_MASK = { 0x0d0509010c040800, 0x0f070b030e060a02,
|
||||
0x1d1519111c141810, 0x1f171b131e161a12,
|
||||
0x2d2529212c242820, 0x2f272b232e262a22,
|
||||
0x3d3539313c343830, 0x3f373b333e363a32 };
|
||||
|
||||
static const __m512i SUBSH_MASK0 = { 0x0c0f0104070b0e00, 0x03060a0d08020509,
|
||||
0x1c1f1114171b1e10, 0x13161a1d18121519,
|
||||
0x2c2f2124272b2e20, 0x23262a2d28222529,
|
||||
0x3c3f3134373b3e30, 0x33363a3d38323539 };
|
||||
|
||||
static const __m512i SUBSH_MASK1 = { 0x0e090205000d0801, 0x04070c0f0a03060b,
|
||||
0x1e191215101d1801, 0x14171c1f1a13161b,
|
||||
0x2e292225202d2821, 0x24272c2f2a23262b,
|
||||
0x3e393235303d3831, 0x34373c3f3a33363b };
|
||||
|
||||
static const __m512i SUBSH_MASK2 = { 0x080b0306010f0a02, 0x05000e090c04070d,
|
||||
0x181b1316111f1a12, 0x15101e191c14171d,
|
||||
0x282b2326212f2a22, 0x25202e292c24272d,
|
||||
0x383b3336313f3a32, 0x35303e393c34373d };
|
||||
|
||||
static const __m512i SUBSH_MASK3 = { 0x0a0d040702090c03, 0x0601080b0e05000f,
|
||||
0x1a1d141712191c13, 0x1611181b1e15101f,
|
||||
0x2a2d242722292c23, 0x2621282b2e25202f,
|
||||
0x3a3d343732393c33, 0x3631383b3e35303f };
|
||||
|
||||
static const __m512i SUBSH_MASK4 = { 0x0b0e0500030a0d04, 0x0702090c0f060108,
|
||||
0x1b1e1510131a1d14, 0x1712191c1f161118,
|
||||
0x2b2e2520232a2d24, 0x2722292c2f262128,
|
||||
0x3b3e3530333a3d34, 0x3732393c3f363138 };
|
||||
|
||||
static const __m512i SUBSH_MASK5 = { 0x0d080601040c0f05, 0x00030b0e0907020a,
|
||||
0x1d181611141c1f15, 0x10131b1e1917121a,
|
||||
0x2d282621242c2f25, 0x20232b2e2927222a,
|
||||
0x3d383631343c3f35, 0x30333b3e3937323a };
|
||||
|
||||
static const __m512i SUBSH_MASK6 = { 0x0f0a0702050e0906, 0x01040d080b00030c,
|
||||
0x1f1a1712151e1916, 0x11141d181b10131c,
|
||||
0x2f2a2722252e2926, 0x21242d282b20232c,
|
||||
0x3f3a3732353e3936, 0x31343d383b30333c };
|
||||
|
||||
static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
|
||||
0x191c101316181b17, 0x12151f1a1d11141e,
|
||||
0x292c202326282b27, 0x22252f2a2d21242e,
|
||||
0x393c303336383b37, 0x32353f3a3d31343e };
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
@@ -40,8 +101,6 @@ __m512i ALL_FF;
|
||||
i = _mm512_xor_si512(i, j);\
|
||||
}
|
||||
|
||||
/**/
|
||||
|
||||
/* Yet another implementation of MixBytes.
|
||||
This time we use the formulae (3) from the paper "Byte Slicing Groestl".
|
||||
Input: a0, ..., a7
|
||||
@@ -155,95 +214,36 @@ __m512i ALL_FF;
|
||||
b1 = _mm512_xor_si512(b1, a4);\
|
||||
}/*MixBytes*/
|
||||
|
||||
// calculate the round constants seperately and load at startup
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
ALL_1B = _mm512_set1_epi32( 0x1b1b1b1b );\
|
||||
TRANSP_MASK = _mm512_set_epi32( \
|
||||
0x3f373b33, 0x3e363a32, 0x3d353931, 0x3c343830, \
|
||||
0x2f272b23, 0x2e262a22, 0x2d252921, 0x2c242820, \
|
||||
0x1f171b13, 0x1e161a12, 0x1d151911, 0x1c141810, \
|
||||
0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800 ); \
|
||||
SUBSH_MASK[0] = _mm512_set_epi32( \
|
||||
0x33363a3d, 0x38323539, 0x3c3f3134, 0x373b3e30, \
|
||||
0x23262a2d, 0x28222529, 0x2c2f2124, 0x272b2e20, \
|
||||
0x13161a1d, 0x18121519, 0x1c1f1114, 0x171b1e10, \
|
||||
0x03060a0d, 0x08020509, 0x0c0f0104, 0x070b0e00 ); \
|
||||
SUBSH_MASK[1] = _mm512_set_epi32( \
|
||||
0x34373c3f, 0x3a33363b, 0x3e393235, 0x303d3831, \
|
||||
0x24272c2f, 0x2a23262b, 0x2e292225, 0x202d2821, \
|
||||
0x14171c1f, 0x1a13161b, 0x1e191215, 0x101d1801, \
|
||||
0x04070c0f, 0x0a03060b, 0x0e090205, 0x000d0801 );\
|
||||
SUBSH_MASK[2] = _mm512_set_epi32( \
|
||||
0x35303e39, 0x3c34373d, 0x383b3336, 0x313f3a32, \
|
||||
0x25202e29, 0x2c24272d, 0x282b2326, 0x212f2a22, \
|
||||
0x15101e19, 0x1c14171d, 0x181b1316, 0x111f1a12, \
|
||||
0x05000e09, 0x0c04070d, 0x080b0306, 0x010f0a02 );\
|
||||
SUBSH_MASK[3] = _mm512_set_epi32( \
|
||||
0x3631383b, 0x3e35303f, 0x3a3d3437, 0x32393c33, \
|
||||
0x2621282b, 0x2e25202f, 0x2a2d2427, 0x22292c23, \
|
||||
0x1611181b, 0x1e15101f, 0x1a1d1417, 0x12191c13, \
|
||||
0x0601080b, 0x0e05000f, 0x0a0d0407, 0x02090c03 );\
|
||||
SUBSH_MASK[4] = _mm512_set_epi32( \
|
||||
0x3732393c, 0x3f363138, 0x3b3e3530, 0x333a3d34, \
|
||||
0x2722292c, 0x2f262128, 0x2b2e2520, 0x232a2d24, \
|
||||
0x1712191c, 0x1f161118, 0x1b1e1510, 0x131a1d14, \
|
||||
0x0702090c, 0x0f060108, 0x0b0e0500, 0x030a0d04 );\
|
||||
SUBSH_MASK[5] = _mm512_set_epi32( \
|
||||
0x30333b3e, 0x3937323a, 0x3d383631, 0x343c3f35, \
|
||||
0x20232b2e, 0x2927222a, 0x2d282621, 0x242c2f25, \
|
||||
0x10131b1e, 0x1917121a, 0x1d181611, 0x141c1f15, \
|
||||
0x00030b0e, 0x0907020a, 0x0d080601, 0x040c0f05 );\
|
||||
SUBSH_MASK[6] = _mm512_set_epi32( \
|
||||
0x31343d38, 0x3b30333c, 0x3f3a3732, 0x353e3936, \
|
||||
0x21242d28, 0x2b20232c, 0x2f2a2722, 0x252e2926, \
|
||||
0x11141d18, 0x1b10131c, 0x1f1a1712, 0x151e1916, \
|
||||
0x01040d08, 0x0b00030c, 0x0f0a0702, 0x050e0906 );\
|
||||
SUBSH_MASK[7] = _mm512_set_epi32( \
|
||||
0x32353f3a, 0x3d31343e, 0x393c3033, 0x36383b37, \
|
||||
0x22252f2a, 0x2d21242e, 0x292c2023, 0x26282b27, \
|
||||
0x12151f1a, 0x1d11141e, 0x191c1013, 0x16181b17, \
|
||||
0x02050f0a, 0x0d01040e, 0x090c0003, 0x06080b07 );\
|
||||
for ( i = 0; i < ROUNDS512; i++ ) \
|
||||
{\
|
||||
ROUND_CONST_L0[i] = _mm512_set4_epi32( 0xffffffff, 0xffffffff, \
|
||||
0x70605040 ^ ( i * 0x01010101 ), 0x30201000 ^ ( i * 0x01010101 ) ); \
|
||||
ROUND_CONST_L7[i] = _mm512_set4_epi32( 0x8f9fafbf ^ ( i * 0x01010101 ), \
|
||||
0xcfdfefff ^ ( i * 0x01010101 ), 0x00000000, 0x00000000 ); \
|
||||
}\
|
||||
ROUND_CONST_Lx = _mm512_set4_epi32( 0xffffffff, 0xffffffff, \
|
||||
0x00000000, 0x00000000 ); \
|
||||
}while(0);\
|
||||
|
||||
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
|
||||
/* AddRoundConstant */\
|
||||
b1 = ROUND_CONST_Lx;\
|
||||
a0 = _mm512_xor_si512( a0, (ROUND_CONST_L0[i]) );\
|
||||
b1 = m512_const2_64( 0xffffffffffffffff, 0 ); \
|
||||
a0 = _mm512_xor_si512( a0, m512_const1_128( round_const_l0[i] ) );\
|
||||
a1 = _mm512_xor_si512( a1, b1 );\
|
||||
a2 = _mm512_xor_si512( a2, b1 );\
|
||||
a3 = _mm512_xor_si512( a3, b1 );\
|
||||
a4 = _mm512_xor_si512( a4, b1 );\
|
||||
a5 = _mm512_xor_si512( a5, b1 );\
|
||||
a6 = _mm512_xor_si512( a6, b1 );\
|
||||
a7 = _mm512_xor_si512( a7, (ROUND_CONST_L7[i]) );\
|
||||
a7 = _mm512_xor_si512( a7, m512_const1_128( round_const_l7[i] ) );\
|
||||
\
|
||||
/* ShiftBytes + SubBytes (interleaved) */\
|
||||
b0 = _mm512_xor_si512( b0, b0 );\
|
||||
a0 = _mm512_shuffle_epi8( a0, (SUBSH_MASK[0]) );\
|
||||
a0 = _mm512_shuffle_epi8( a0, SUBSH_MASK0 );\
|
||||
a0 = _mm512_aesenclast_epi128(a0, b0 );\
|
||||
a1 = _mm512_shuffle_epi8( a1, (SUBSH_MASK[1]) );\
|
||||
a1 = _mm512_shuffle_epi8( a1, SUBSH_MASK1 );\
|
||||
a1 = _mm512_aesenclast_epi128(a1, b0 );\
|
||||
a2 = _mm512_shuffle_epi8( a2, (SUBSH_MASK[2]) );\
|
||||
a2 = _mm512_shuffle_epi8( a2, SUBSH_MASK2 );\
|
||||
a2 = _mm512_aesenclast_epi128(a2, b0 );\
|
||||
a3 = _mm512_shuffle_epi8( a3, (SUBSH_MASK[3]) );\
|
||||
a3 = _mm512_shuffle_epi8( a3, SUBSH_MASK3 );\
|
||||
a3 = _mm512_aesenclast_epi128(a3, b0 );\
|
||||
a4 = _mm512_shuffle_epi8( a4, (SUBSH_MASK[4]) );\
|
||||
a4 = _mm512_shuffle_epi8( a4, SUBSH_MASK4 );\
|
||||
a4 = _mm512_aesenclast_epi128(a4, b0 );\
|
||||
a5 = _mm512_shuffle_epi8( a5, (SUBSH_MASK[5]) );\
|
||||
a5 = _mm512_shuffle_epi8( a5, SUBSH_MASK5 );\
|
||||
a5 = _mm512_aesenclast_epi128(a5, b0 );\
|
||||
a6 = _mm512_shuffle_epi8( a6, (SUBSH_MASK[6]) );\
|
||||
a6 = _mm512_shuffle_epi8( a6, SUBSH_MASK6 );\
|
||||
a6 = _mm512_aesenclast_epi128(a6, b0 );\
|
||||
a7 = _mm512_shuffle_epi8( a7, (SUBSH_MASK[7]) );\
|
||||
a7 = _mm512_shuffle_epi8( a7, SUBSH_MASK7 );\
|
||||
a7 = _mm512_aesenclast_epi128( a7, b0 );\
|
||||
\
|
||||
/* MixBytes */\
|
||||
@@ -390,29 +390,6 @@ __m512i ALL_FF;
|
||||
}/**/
|
||||
|
||||
|
||||
|
||||
void INIT256_4way( __m512i* chaining )
|
||||
{
|
||||
static __m512i xmm0, xmm2, xmm6, xmm7;
|
||||
static __m512i xmm12, xmm13, xmm14, xmm15;
|
||||
|
||||
/* load IV into registers xmm12 - xmm15 */
|
||||
xmm12 = chaining[0];
|
||||
xmm13 = chaining[1];
|
||||
xmm14 = chaining[2];
|
||||
xmm15 = chaining[3];
|
||||
|
||||
/* transform chaining value from column ordering into row ordering */
|
||||
/* we put two rows (64 bit) of the IV into one 128-bit XMM register */
|
||||
Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
|
||||
|
||||
/* store transposed IV */
|
||||
chaining[0] = xmm12;
|
||||
chaining[1] = xmm2;
|
||||
chaining[2] = xmm6;
|
||||
chaining[3] = xmm7;
|
||||
}
|
||||
|
||||
void TF512_4way( __m512i* chaining, __m512i* message )
|
||||
{
|
||||
static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
|
||||
|
||||
@@ -19,10 +19,6 @@
|
||||
|
||||
int groestl512_4way_init( groestl512_4way_context* ctx, uint64_t hashlen )
|
||||
{
|
||||
int i;
|
||||
|
||||
SET_CONSTANTS();
|
||||
|
||||
if (ctx->chaining == NULL || ctx->buffer == NULL)
|
||||
return 1;
|
||||
|
||||
@@ -99,7 +95,6 @@ int groestl512_4way_full( groestl512_4way_context* ctx, void* output,
|
||||
|
||||
// --- init ---
|
||||
|
||||
SET_CONSTANTS();
|
||||
memset_zero_512( ctx->chaining, SIZE512 );
|
||||
memset_zero_512( ctx->buffer, SIZE512 );
|
||||
ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );
|
||||
|
||||
@@ -15,16 +15,86 @@
|
||||
|
||||
#if defined(__VAES__)
|
||||
|
||||
/* global constants */
|
||||
__m512i ROUND_CONST_Lx;
|
||||
//__m128i ROUND_CONST_L0[ROUNDS512];
|
||||
//__m128i ROUND_CONST_L7[ROUNDS512];
|
||||
__m512i ROUND_CONST_P[ROUNDS1024];
|
||||
__m512i ROUND_CONST_Q[ROUNDS1024];
|
||||
__m512i TRANSP_MASK;
|
||||
__m512i SUBSH_MASK[8];
|
||||
__m512i ALL_1B;
|
||||
__m512i ALL_FF;
|
||||
static const __m128i round_const_p[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x7060504030201000, 0xf0e0d0c0b0a09080 },
|
||||
{ 0x7161514131211101, 0xf1e1d1c1b1a19181 },
|
||||
{ 0x7262524232221202, 0xf2e2d2c2b2a29282 },
|
||||
{ 0x7363534333231303, 0xf3e3d3c3b3a39383 },
|
||||
{ 0x7464544434241404, 0xf4e4d4c4b4a49484 },
|
||||
{ 0x7565554535251505, 0xf5e5d5c5b5a59585 },
|
||||
{ 0x7666564636261606, 0xf6e6d6c6b6a69686 },
|
||||
{ 0x7767574737271707, 0xf7e7d7c7b7a79787 },
|
||||
{ 0x7868584838281808, 0xf8e8d8c8b8a89888 },
|
||||
{ 0x7969594939291909, 0xf9e9d9c9b9a99989 },
|
||||
{ 0x7a6a5a4a3a2a1a0a, 0xfaeadacabaaa9a8a },
|
||||
{ 0x7b6b5b4b3b2b1b0b, 0xfbebdbcbbbab9b8b },
|
||||
{ 0x7c6c5c4c3c2c1c0c, 0xfcecdcccbcac9c8c },
|
||||
{ 0x7d6d5d4d3d2d1d0d, 0xfdedddcdbdad9d8d }
|
||||
};
|
||||
|
||||
static const __m128i round_const_q[] __attribute__ ((aligned (64))) =
|
||||
{
|
||||
{ 0x8f9fafbfcfdfefff, 0x0f1f2f3f4f5f6f7f },
|
||||
{ 0x8e9eaebecedeeefe, 0x0e1e2e3e4e5e6e7e },
|
||||
{ 0x8d9dadbdcdddedfd, 0x0d1d2d3d4d5d6d7d },
|
||||
{ 0x8c9cacbcccdcecfc, 0x0c1c2c3c4c5c6c7c },
|
||||
{ 0x8b9babbbcbdbebfb, 0x0b1b2b3b4b5b6b7b },
|
||||
{ 0x8a9aaabacadaeafa, 0x0a1a2a3a4a5a6a7a },
|
||||
{ 0x8999a9b9c9d9e9f9, 0x0919293949596979 },
|
||||
{ 0x8898a8b8c8d8e8f8, 0x0818283848586878 },
|
||||
{ 0x8797a7b7c7d7e7f7, 0x0717273747576777 },
|
||||
{ 0x8696a6b6c6d6e6f6, 0x0616263646566676 },
|
||||
{ 0x8595a5b5c5d5e5f5, 0x0515253545556575 },
|
||||
{ 0x8494a4b4c4d4e4f4, 0x0414243444546474 },
|
||||
{ 0x8393a3b3c3d3e3f3, 0x0313233343536373 },
|
||||
{ 0x8292a2b2c2d2e2f2, 0x0212223242526272 }
|
||||
};
|
||||
|
||||
static const __m512i TRANSP_MASK = { 0x0d0509010c040800, 0x0f070b030e060a02,
|
||||
0x1d1519111c141810, 0x1f171b131e161a12,
|
||||
0x2d2529212c242820, 0x2f272b232e262a22,
|
||||
0x3d3539313c343830, 0x3f373b333e363a32 };
|
||||
|
||||
static const __m512i SUBSH_MASK0 = { 0x0b0e0104070a0d00, 0x0306090c0f020508,
|
||||
0x1b1e1114171a1d10, 0x1316191c1f121518,
|
||||
0x2b2e2124272a2d20, 0x2326292c2f222528,
|
||||
0x3b3e3134373a3d30, 0x3336393c3f323538 };
|
||||
|
||||
static const __m512i SUBSH_MASK1 = { 0x0c0f0205080b0e01, 0x04070a0d00030609,
|
||||
0x1c1f1215181b1e11, 0x14171a1d10131619,
|
||||
0x2c2f2225282b2e21, 0x24272a2d20232629,
|
||||
0x3c3f3235383b3e31, 0x34373a3d30333639 };
|
||||
|
||||
static const __m512i SUBSH_MASK2 = { 0x0d000306090c0f02, 0x05080b0e0104070a,
|
||||
0x1d101316191c1f12, 0x15181b1e1114171a,
|
||||
0x2d202326292c2f22, 0x25282b2e2124272a,
|
||||
0x3d303336393c3f32, 0x35383b3e3134373a };
|
||||
|
||||
static const __m512i SUBSH_MASK3 = { 0x0e0104070a0d0003, 0x06090c0f0205080b,
|
||||
0x1e1114171a1d1013, 0x16191c1f1215181b,
|
||||
0x2e2124272a2d2023, 0x26292c2f2225282b,
|
||||
0x3e3134373a3d3033, 0x36393c3f3235383b };
|
||||
|
||||
static const __m512i SUBSH_MASK4 = { 0x0f0205080b0e0104, 0x070a0d000306090c,
|
||||
0x1f1215181b1e1114, 0x171a1d101316191c,
|
||||
0x2f2225282b2e2124, 0x272a2d202326292c,
|
||||
0x3f3235383b3e3134, 0x373a3d303336393c };
|
||||
|
||||
static const __m512i SUBSH_MASK5 = { 0x000306090c0f0205, 0x080b0e0104070a0d,
|
||||
0x101316191c1f1215, 0x181b1e1114171a1d,
|
||||
0x202326292c2f2225, 0x282b2e2124272a2d,
|
||||
0x303336393c3f3235, 0x383b3e3134373a3d };
|
||||
|
||||
static const __m512i SUBSH_MASK6 = { 0x0104070a0d000306, 0x090c0f0205080b0e,
|
||||
0x1114171a1d101316, 0x191c1f1215181b1e,
|
||||
0x2124272a2d202326, 0x292c2f2225282b2e,
|
||||
0x3134373a3d303336, 0x393c3f3235383b3e };
|
||||
|
||||
static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
|
||||
0x16191c1f1215181b, 0x1e1114171a1d1013,
|
||||
0x26292c2f2225282b, 0x2e2124272a2d2023,
|
||||
0x36393c3f3235383b, 0x3e3134373a3d3033 };
|
||||
|
||||
#define tos(a) #a
|
||||
#define tostr(a) tos(a)
|
||||
@@ -155,69 +225,6 @@ __m512i ALL_FF;
|
||||
b1 = _mm512_xor_si512(b1, a4);\
|
||||
}/*MixBytes*/
|
||||
|
||||
// calculate the round constants seperately and load at startup
|
||||
|
||||
#define SET_CONSTANTS(){\
|
||||
ALL_FF = _mm512_set1_epi32( 0xffffffff );\
|
||||
ALL_1B = _mm512_set1_epi32( 0x1b1b1b1b );\
|
||||
TRANSP_MASK = _mm512_set_epi32( \
|
||||
0x3f373b33, 0x3e363a32, 0x3d353931, 0x3c343830, \
|
||||
0x2f272b23, 0x2e262a22, 0x2d252921, 0x2c242820, \
|
||||
0x1f171b13, 0x1e161a12, 0x1d151911, 0x1c141810, \
|
||||
0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800 ); \
|
||||
SUBSH_MASK[0] = _mm512_set_epi32( \
|
||||
0x3336393c, 0x3f323538, 0x3b3e3134, 0x373a3d30, \
|
||||
0x2326292c, 0x2f222528, 0x2b2e2124, 0x272a2d20, \
|
||||
0x1316191c, 0x1f121518, 0x1b1e1114, 0x171a1d10, \
|
||||
0x0306090c, 0x0f020508, 0x0b0e0104, 0x070a0d00 ); \
|
||||
SUBSH_MASK[1] = _mm512_set_epi32( \
|
||||
0x34373a3d, 0x30333639, 0x3c3f3235, 0x383b3e31, \
|
||||
0x24272a2d, 0x20232629, 0x2c2f2225, 0x282b2e21, \
|
||||
0x14171a1d, 0x10131619, 0x1c1f1215, 0x181b1e11, \
|
||||
0x04070a0d, 0x00030609, 0x0c0f0205, 0x080b0e01 ); \
|
||||
SUBSH_MASK[2] = _mm512_set_epi32( \
|
||||
0x35383b3e, 0x3134373a, 0x3d303336, 0x393c3f32, \
|
||||
0x25282b2e, 0x2124272a, 0x2d202326, 0x292c2f22, \
|
||||
0x15181b1e, 0x1114171a, 0x1d101316, 0x191c1f12, \
|
||||
0x05080b0e, 0x0104070a, 0x0d000306, 0x090c0f02 ); \
|
||||
SUBSH_MASK[3] = _mm512_set_epi32( \
|
||||
0x36393c3f, 0x3235383b, 0x3e313437, 0x3a3d3033, \
|
||||
0x26292c2f, 0x2225282b, 0x2e212427, 0x2a2d2023, \
|
||||
0x16191c1f, 0x1215181b, 0x1e111417, 0x1a1d1013, \
|
||||
0x06090c0f, 0x0205080b, 0x0e010407, 0x0a0d0003 ); \
|
||||
SUBSH_MASK[4] = _mm512_set_epi32( \
|
||||
0x373a3d30, 0x3336393c, 0x3f323538, 0x3b3e3134, \
|
||||
0x272a2d20, 0x2326292c, 0x2f222528, 0x2b2e2124, \
|
||||
0x171a1d10, 0x1316191c, 0x1f121518, 0x1b1e1114, \
|
||||
0x070a0d00, 0x0306090c, 0x0f020508, 0x0b0e0104 ); \
|
||||
SUBSH_MASK[5] = _mm512_set_epi32( \
|
||||
0x383b3e31, 0x34373a3d, 0x30333639, 0x3c3f3235, \
|
||||
0x282b2e21, 0x24272a2d, 0x20232629, 0x2c2f2225, \
|
||||
0x181b1e11, 0x14171a1d, 0x10131619, 0x1c1f1215, \
|
||||
0x080b0e01, 0x04070a0d, 0x00030609, 0x0c0f0205 ); \
|
||||
SUBSH_MASK[6] = _mm512_set_epi32( \
|
||||
0x393c3f32, 0x35383b3e, 0x3134373a, 0x3d303336, \
|
||||
0x292c2f22, 0x25282b2e, 0x2124272a, 0x2d202326, \
|
||||
0x191c1f12, 0x15181b1e, 0x1114171a, 0x1d101316, \
|
||||
0x090c0f02, 0x05080b0e, 0x0104070a, 0x0d000306 ); \
|
||||
SUBSH_MASK[7] = _mm512_set_epi32( \
|
||||
0x3e313437, 0x3a3d3033, 0x36393c3f, 0x3235383b, \
|
||||
0x2e212427, 0x2a2d2023, 0x26292c2f, 0x2225282b, \
|
||||
0x1e111417, 0x1a1d1013, 0x16191c1f, 0x1215181b, \
|
||||
0x0e010407, 0x0a0d0003, 0x06090c0f, 0x0205080b ); \
|
||||
for( i = 0; i < ROUNDS1024; i++ ) \
|
||||
{ \
|
||||
ROUND_CONST_P[i] = _mm512_set4_epi32( 0xf0e0d0c0 ^ (i * 0x01010101), \
|
||||
0xb0a09080 ^ (i * 0x01010101), \
|
||||
0x70605040 ^ (i * 0x01010101), \
|
||||
0x30201000 ^ (i * 0x01010101) ); \
|
||||
ROUND_CONST_Q[i] = _mm512_set4_epi32( 0x0f1f2f3f ^ (i * 0x01010101), \
|
||||
0x4f5f6f7f ^ (i * 0x01010101), \
|
||||
0x8f9fafbf ^ (i * 0x01010101), \
|
||||
0xcfdfefff ^ (i * 0x01010101));\
|
||||
} \
|
||||
}while(0);\
|
||||
|
||||
/* one round
|
||||
* a0-a7 = input rows
|
||||
* b0-b7 = output rows
|
||||
@@ -242,30 +249,32 @@ __m512i ALL_FF;
|
||||
for ( round_counter = 0; round_counter < 14; round_counter += 2 ) \
|
||||
{ \
|
||||
/* AddRoundConstant P1024 */\
|
||||
xmm8 = _mm512_xor_si512( xmm8, ( ROUND_CONST_P[ round_counter ] ) );\
|
||||
xmm8 = _mm512_xor_si512( xmm8, m512_const1_128( \
|
||||
casti_m128i( round_const_p, round_counter ) ) ); \
|
||||
/* ShiftBytes P1024 + pre-AESENCLAST */\
|
||||
xmm8 = _mm512_shuffle_epi8( xmm8, ( SUBSH_MASK[0] ) );\
|
||||
xmm9 = _mm512_shuffle_epi8( xmm9, ( SUBSH_MASK[1] ) );\
|
||||
xmm10 = _mm512_shuffle_epi8( xmm10, ( SUBSH_MASK[2] ) );\
|
||||
xmm11 = _mm512_shuffle_epi8( xmm11, ( SUBSH_MASK[3] ) );\
|
||||
xmm12 = _mm512_shuffle_epi8( xmm12, ( SUBSH_MASK[4] ) );\
|
||||
xmm13 = _mm512_shuffle_epi8( xmm13, ( SUBSH_MASK[5] ) );\
|
||||
xmm14 = _mm512_shuffle_epi8( xmm14, ( SUBSH_MASK[6] ) );\
|
||||
xmm15 = _mm512_shuffle_epi8( xmm15, ( SUBSH_MASK[7] ) );\
|
||||
xmm8 = _mm512_shuffle_epi8( xmm8, SUBSH_MASK0 ); \
|
||||
xmm9 = _mm512_shuffle_epi8( xmm9, SUBSH_MASK1 );\
|
||||
xmm10 = _mm512_shuffle_epi8( xmm10, SUBSH_MASK2 );\
|
||||
xmm11 = _mm512_shuffle_epi8( xmm11, SUBSH_MASK3 );\
|
||||
xmm12 = _mm512_shuffle_epi8( xmm12, SUBSH_MASK4 );\
|
||||
xmm13 = _mm512_shuffle_epi8( xmm13, SUBSH_MASK5 );\
|
||||
xmm14 = _mm512_shuffle_epi8( xmm14, SUBSH_MASK6 );\
|
||||
xmm15 = _mm512_shuffle_epi8( xmm15, SUBSH_MASK7 );\
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
\
|
||||
/* AddRoundConstant P1024 */\
|
||||
xmm0 = _mm512_xor_si512( xmm0, ( ROUND_CONST_P[ round_counter+1 ] ) );\
|
||||
xmm0 = _mm512_xor_si512( xmm0, m512_const1_128( \
|
||||
casti_m128i( round_const_p, round_counter+1 ) ) ); \
|
||||
/* ShiftBytes P1024 + pre-AESENCLAST */\
|
||||
xmm0 = _mm512_shuffle_epi8( xmm0, ( SUBSH_MASK[0] ) );\
|
||||
xmm1 = _mm512_shuffle_epi8( xmm1, ( SUBSH_MASK[1] ) );\
|
||||
xmm2 = _mm512_shuffle_epi8( xmm2, ( SUBSH_MASK[2] ) );\
|
||||
xmm3 = _mm512_shuffle_epi8( xmm3, ( SUBSH_MASK[3] ) );\
|
||||
xmm4 = _mm512_shuffle_epi8( xmm4, ( SUBSH_MASK[4] ) );\
|
||||
xmm5 = _mm512_shuffle_epi8( xmm5, ( SUBSH_MASK[5] ) );\
|
||||
xmm6 = _mm512_shuffle_epi8( xmm6, ( SUBSH_MASK[6] ) );\
|
||||
xmm7 = _mm512_shuffle_epi8( xmm7, ( SUBSH_MASK[7] ) );\
|
||||
xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK0 );\
|
||||
xmm1 = _mm512_shuffle_epi8( xmm1, SUBSH_MASK1 );\
|
||||
xmm2 = _mm512_shuffle_epi8( xmm2, SUBSH_MASK2 );\
|
||||
xmm3 = _mm512_shuffle_epi8( xmm3, SUBSH_MASK3 );\
|
||||
xmm4 = _mm512_shuffle_epi8( xmm4, SUBSH_MASK4 );\
|
||||
xmm5 = _mm512_shuffle_epi8( xmm5, SUBSH_MASK5 );\
|
||||
xmm6 = _mm512_shuffle_epi8( xmm6, SUBSH_MASK6 );\
|
||||
xmm7 = _mm512_shuffle_epi8( xmm7, SUBSH_MASK7 );\
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
}\
|
||||
@@ -284,16 +293,17 @@ __m512i ALL_FF;
|
||||
xmm12 = _mm512_xor_si512( xmm12, xmm1 );\
|
||||
xmm13 = _mm512_xor_si512( xmm13, xmm1 );\
|
||||
xmm14 = _mm512_xor_si512( xmm14, xmm1 );\
|
||||
xmm15 = _mm512_xor_si512( xmm15, ( ROUND_CONST_Q[ round_counter ] ) );\
|
||||
xmm15 = _mm512_xor_si512( xmm15, m512_const1_128( \
|
||||
casti_m128i( round_const_q, round_counter ) ) ); \
|
||||
/* ShiftBytes Q1024 + pre-AESENCLAST */\
|
||||
xmm8 = _mm512_shuffle_epi8( xmm8, ( SUBSH_MASK[1] ) );\
|
||||
xmm9 = _mm512_shuffle_epi8( xmm9, ( SUBSH_MASK[3] ) );\
|
||||
xmm10 = _mm512_shuffle_epi8( xmm10, ( SUBSH_MASK[5] ) );\
|
||||
xmm11 = _mm512_shuffle_epi8( xmm11, ( SUBSH_MASK[7] ) );\
|
||||
xmm12 = _mm512_shuffle_epi8( xmm12, ( SUBSH_MASK[0] ) );\
|
||||
xmm13 = _mm512_shuffle_epi8( xmm13, ( SUBSH_MASK[2] ) );\
|
||||
xmm14 = _mm512_shuffle_epi8( xmm14, ( SUBSH_MASK[4] ) );\
|
||||
xmm15 = _mm512_shuffle_epi8( xmm15, ( SUBSH_MASK[6] ) );\
|
||||
xmm8 = _mm512_shuffle_epi8( xmm8, SUBSH_MASK1 );\
|
||||
xmm9 = _mm512_shuffle_epi8( xmm9, SUBSH_MASK3 );\
|
||||
xmm10 = _mm512_shuffle_epi8( xmm10, SUBSH_MASK5 );\
|
||||
xmm11 = _mm512_shuffle_epi8( xmm11, SUBSH_MASK7 );\
|
||||
xmm12 = _mm512_shuffle_epi8( xmm12, SUBSH_MASK0 );\
|
||||
xmm13 = _mm512_shuffle_epi8( xmm13, SUBSH_MASK2 );\
|
||||
xmm14 = _mm512_shuffle_epi8( xmm14, SUBSH_MASK4 );\
|
||||
xmm15 = _mm512_shuffle_epi8( xmm15, SUBSH_MASK6 );\
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
|
||||
\
|
||||
@@ -306,16 +316,17 @@ __m512i ALL_FF;
|
||||
xmm4 = _mm512_xor_si512( xmm4, xmm9 );\
|
||||
xmm5 = _mm512_xor_si512( xmm5, xmm9 );\
|
||||
xmm6 = _mm512_xor_si512( xmm6, xmm9 );\
|
||||
xmm7 = _mm512_xor_si512( xmm7, ( ROUND_CONST_Q[ round_counter+1 ] ) );\
|
||||
xmm7 = _mm512_xor_si512( xmm7, m512_const1_128( \
|
||||
casti_m128i( round_const_q, round_counter+1 ) ) ); \
|
||||
/* ShiftBytes Q1024 + pre-AESENCLAST */\
|
||||
xmm0 = _mm512_shuffle_epi8( xmm0, ( SUBSH_MASK[1] ) );\
|
||||
xmm1 = _mm512_shuffle_epi8( xmm1, ( SUBSH_MASK[3] ) );\
|
||||
xmm2 = _mm512_shuffle_epi8( xmm2, ( SUBSH_MASK[5] ) );\
|
||||
xmm3 = _mm512_shuffle_epi8( xmm3, ( SUBSH_MASK[7] ) );\
|
||||
xmm4 = _mm512_shuffle_epi8( xmm4, ( SUBSH_MASK[0] ) );\
|
||||
xmm5 = _mm512_shuffle_epi8( xmm5, ( SUBSH_MASK[2] ) );\
|
||||
xmm6 = _mm512_shuffle_epi8( xmm6, ( SUBSH_MASK[4] ) );\
|
||||
xmm7 = _mm512_shuffle_epi8( xmm7, ( SUBSH_MASK[6] ) );\
|
||||
xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK1 );\
|
||||
xmm1 = _mm512_shuffle_epi8( xmm1, SUBSH_MASK3 );\
|
||||
xmm2 = _mm512_shuffle_epi8( xmm2, SUBSH_MASK5 );\
|
||||
xmm3 = _mm512_shuffle_epi8( xmm3, SUBSH_MASK7 );\
|
||||
xmm4 = _mm512_shuffle_epi8( xmm4, SUBSH_MASK0 );\
|
||||
xmm5 = _mm512_shuffle_epi8( xmm5, SUBSH_MASK2 );\
|
||||
xmm6 = _mm512_shuffle_epi8( xmm6, SUBSH_MASK4 );\
|
||||
xmm7 = _mm512_shuffle_epi8( xmm7, SUBSH_MASK6 );\
|
||||
/* SubBytes + MixBytes */\
|
||||
SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
|
||||
}\
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "myrgr-gate.h"
|
||||
|
||||
#if !defined(MYRGR_8WAY) && !defined(MYRGR_4WAY)
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
@@ -86,3 +89,4 @@ int scanhash_myriad( struct work *work, uint32_t max_nonce,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -35,6 +35,8 @@
|
||||
|
||||
#include "sph_groestl.h"
|
||||
|
||||
#if !defined(__AES__)
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C"{
|
||||
#endif
|
||||
@@ -3116,4 +3118,6 @@ sph_groestl512_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
|
||||
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
|
||||
#endif // !AES
|
||||
#endif
|
||||
|
||||
@@ -42,6 +42,7 @@ extern "C"{
|
||||
#include <stddef.h>
|
||||
#include "algo/sha/sph_types.h"
|
||||
|
||||
#if !defined(__AES__)
|
||||
/**
|
||||
* Output size (in bits) for Groestl-224.
|
||||
*/
|
||||
@@ -326,4 +327,5 @@ void sph_groestl512_addbits_and_close(
|
||||
}
|
||||
#endif
|
||||
|
||||
#endif // !AES
|
||||
#endif
|
||||
|
||||
@@ -1,156 +0,0 @@
|
||||
#include "algo-gate-api.h"
|
||||
#include <stdio.h>
|
||||
#include <string.h>
|
||||
#include <openssl/sha.h>
|
||||
#include <stdint.h>
|
||||
#include <stdlib.h>
|
||||
#include "sph_hefty1.h"
|
||||
#include "algo/luffa/sph_luffa.h"
|
||||
#include "algo/fugue/sph_fugue.h"
|
||||
#include "algo/skein/sph_skein.h"
|
||||
#include "algo/whirlpool/sph_whirlpool.h"
|
||||
#include "algo/shabal/sph_shabal.h"
|
||||
#include "algo/echo/sph_echo.h"
|
||||
#include "algo/hamsi/sph_hamsi.h"
|
||||
#include "algo/luffa/luffa_for_sse2.h"
|
||||
#ifdef __AES__
|
||||
#include "algo/echo/aes_ni/hash_api.h"
|
||||
#endif
|
||||
|
||||
void bastionhash(void *output, const void *input)
|
||||
{
|
||||
unsigned char hash[64] __attribute__ ((aligned (64)));
|
||||
|
||||
#ifdef __AES__
|
||||
hashState_echo ctx_echo;
|
||||
#else
|
||||
sph_echo512_context ctx_echo;
|
||||
#endif
|
||||
hashState_luffa ctx_luffa;
|
||||
sph_fugue512_context ctx_fugue;
|
||||
sph_whirlpool_context ctx_whirlpool;
|
||||
sph_shabal512_context ctx_shabal;
|
||||
sph_hamsi512_context ctx_hamsi;
|
||||
sph_skein512_context ctx_skein;
|
||||
|
||||
HEFTY1(input, 80, hash);
|
||||
|
||||
init_luffa( &ctx_luffa, 512 );
|
||||
update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
|
||||
(const BitSequence*)hash, 64 );
|
||||
|
||||
if (hash[0] & 0x8)
|
||||
{
|
||||
sph_fugue512_init(&ctx_fugue);
|
||||
sph_fugue512(&ctx_fugue, hash, 64);
|
||||
sph_fugue512_close(&ctx_fugue, hash);
|
||||
} else {
|
||||
sph_skein512_init( &ctx_skein );
|
||||
sph_skein512( &ctx_skein, hash, 64 );
|
||||
sph_skein512_close( &ctx_skein, hash );
|
||||
}
|
||||
|
||||
sph_whirlpool_init(&ctx_whirlpool);
|
||||
sph_whirlpool(&ctx_whirlpool, hash, 64);
|
||||
sph_whirlpool_close(&ctx_whirlpool, hash);
|
||||
|
||||
sph_fugue512_init(&ctx_fugue);
|
||||
sph_fugue512(&ctx_fugue, hash, 64);
|
||||
sph_fugue512_close(&ctx_fugue, hash);
|
||||
|
||||
if (hash[0] & 0x8)
|
||||
{
|
||||
#ifdef __AES__
|
||||
init_echo( &ctx_echo, 512 );
|
||||
update_final_echo ( &ctx_echo,(BitSequence*)hash,
|
||||
(const BitSequence*)hash, 512 );
|
||||
#else
|
||||
sph_echo512_init(&ctx_echo);
|
||||
sph_echo512(&ctx_echo, hash, 64);
|
||||
sph_echo512_close(&ctx_echo, hash);
|
||||
#endif
|
||||
} else {
|
||||
init_luffa( &ctx_luffa, 512 );
|
||||
update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
|
||||
(const BitSequence*)hash, 64 );
|
||||
}
|
||||
|
||||
sph_shabal512_init(&ctx_shabal);
|
||||
sph_shabal512(&ctx_shabal, hash, 64);
|
||||
sph_shabal512_close(&ctx_shabal, hash);
|
||||
|
||||
sph_skein512_init( &ctx_skein );
|
||||
sph_skein512( &ctx_skein, hash, 64 );
|
||||
sph_skein512_close( &ctx_skein, hash );
|
||||
|
||||
if (hash[0] & 0x8)
|
||||
{
|
||||
sph_shabal512_init(&ctx_shabal);
|
||||
sph_shabal512(&ctx_shabal, hash, 64);
|
||||
sph_shabal512_close(&ctx_shabal, hash);
|
||||
} else {
|
||||
sph_whirlpool_init(&ctx_whirlpool);
|
||||
sph_whirlpool(&ctx_whirlpool, hash, 64);
|
||||
sph_whirlpool_close(&ctx_whirlpool, hash);
|
||||
}
|
||||
|
||||
sph_shabal512_init(&ctx_shabal);
|
||||
sph_shabal512(&ctx_shabal, hash, 64);
|
||||
sph_shabal512_close(&ctx_shabal, hash);
|
||||
|
||||
if (hash[0] & 0x8)
|
||||
{
|
||||
sph_hamsi512_init(&ctx_hamsi);
|
||||
sph_hamsi512(&ctx_hamsi, hash, 64);
|
||||
sph_hamsi512_close(&ctx_hamsi, hash);
|
||||
} else {
|
||||
init_luffa( &ctx_luffa, 512 );
|
||||
update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
|
||||
(const BitSequence*)hash, 64 );
|
||||
}
|
||||
|
||||
memcpy(output, hash, 32);
|
||||
}
|
||||
|
||||
int scanhash_bastion( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr)
|
||||
{
|
||||
uint32_t _ALIGN(64) hash32[8];
|
||||
uint32_t _ALIGN(64) endiandata[20];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
|
||||
uint32_t n = first_nonce;
|
||||
|
||||
for (int i=0; i < 19; i++)
|
||||
be32enc(&endiandata[i], pdata[i]);
|
||||
|
||||
do {
|
||||
be32enc(&endiandata[19], n);
|
||||
bastionhash(hash32, endiandata);
|
||||
if (hash32[7] < Htarg && fulltest(hash32, ptarget)) {
|
||||
pdata[19] = n;
|
||||
submit_solution( work, hash32, mythr );
|
||||
}
|
||||
n++;
|
||||
|
||||
} while (n < max_nonce && !work_restart[thr_id].restart);
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
bool register_bastion_algo( algo_gate_t* gate )
|
||||
{
|
||||
gate->optimizations = SSE2_OPT | AES_OPT;
|
||||
gate->scanhash = (void*)&scanhash_bastion;
|
||||
gate->hash = (void*)&bastionhash;
|
||||
return true;
|
||||
};
|
||||
|
||||
@@ -1,111 +0,0 @@
|
||||
#include <string.h>
|
||||
#include <openssl/sha.h>
|
||||
#include <stdint.h>
|
||||
|
||||
#include "algo-gate-api.h"
|
||||
#include "sph_hefty1.h"
|
||||
#include "algo/keccak/sph_keccak.h"
|
||||
#include "algo/blake/sph_blake.h"
|
||||
#include "algo/groestl/sph_groestl.h"
|
||||
|
||||
/* Combines top 64-bits from each hash into a single hash */
|
||||
static void combine_hashes(uint32_t *out, uint32_t *hash1, uint32_t *hash2, uint32_t *hash3, uint32_t *hash4)
|
||||
{
|
||||
uint32_t *hash[4] = { hash1, hash2, hash3, hash4 };
|
||||
|
||||
/* Transpose first 64 bits of each hash into out */
|
||||
memset(out, 0, 32);
|
||||
int bits = 0;
|
||||
for (unsigned int i = 7; i >= 6; i--) {
|
||||
for (uint32_t mask = 0x80000000; mask; mask >>= 1) {
|
||||
for (unsigned int k = 0; k < 4; k++) {
|
||||
out[(255 - bits)/32] <<= 1;
|
||||
if ((hash[k][i] & mask) != 0)
|
||||
out[(255 - bits)/32] |= 1;
|
||||
bits++;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
extern void heavyhash(unsigned char* output, const unsigned char* input, int len)
|
||||
{
|
||||
unsigned char hash1[32];
|
||||
HEFTY1(input, len, hash1);
|
||||
|
||||
// HEFTY1 is new, so take an extra security measure to eliminate
|
||||
// * the possiblity of collisions:
|
||||
// *
|
||||
// * Hash(x) = SHA256(x + HEFTY1(x))
|
||||
// *
|
||||
// * N.B. '+' is concatenation.
|
||||
//
|
||||
unsigned char hash2[32];;
|
||||
SHA256_CTX ctx;
|
||||
SHA256_Init(&ctx);
|
||||
SHA256_Update(&ctx, input, len);
|
||||
SHA256_Update(&ctx, hash1, sizeof(hash1));
|
||||
SHA256_Final(hash2, &ctx);
|
||||
|
||||
// * Additional security: Do not rely on a single cryptographic hash
|
||||
// * function. Instead, combine the outputs of 4 of the most secure
|
||||
// * cryptographic hash functions-- SHA256, KECCAK512, GROESTL512
|
||||
// * and BLAKE512.
|
||||
|
||||
|
||||
uint32_t hash3[16];
|
||||
sph_keccak512_context keccakCtx;
|
||||
sph_keccak512_init(&keccakCtx);
|
||||
sph_keccak512(&keccakCtx, input, len);
|
||||
sph_keccak512(&keccakCtx, hash1, sizeof(hash1));
|
||||
sph_keccak512_close(&keccakCtx, (void *)&hash3);
|
||||
|
||||
uint32_t hash4[16];
|
||||
sph_groestl512_context groestlCtx;
|
||||
sph_groestl512_init(&groestlCtx);
|
||||
sph_groestl512(&groestlCtx, input, len);
|
||||
sph_groestl512(&groestlCtx, hash1, sizeof(hash1));
|
||||
sph_groestl512_close(&groestlCtx, (void *)&hash4);
|
||||
|
||||
uint32_t hash5[16];
|
||||
sph_blake512_context blakeCtx;
|
||||
sph_blake512_init(&blakeCtx);
|
||||
sph_blake512(&blakeCtx, input, len);
|
||||
sph_blake512(&blakeCtx, (unsigned char *)&hash1, sizeof(hash1));
|
||||
sph_blake512_close(&blakeCtx, (void *)&hash5);
|
||||
|
||||
uint32_t *final = (uint32_t *)output;
|
||||
combine_hashes(final, (uint32_t *)hash2, hash3, hash4, hash5);
|
||||
|
||||
}
|
||||
|
||||
int scanhash_heavy( uint32_t *pdata, const uint32_t *ptarget,
|
||||
uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr)
|
||||
{
|
||||
uint32_t hash[8];
|
||||
uint32_t start_nonce = pdata[19];
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
|
||||
do {
|
||||
heavyhash((unsigned char *)hash, (unsigned char *)pdata, 80);
|
||||
|
||||
if (hash[7] <= ptarget[7]) {
|
||||
if (fulltest(hash, ptarget)) {
|
||||
*hashes_done = pdata[19] - start_nonce;
|
||||
return 1;
|
||||
break;
|
||||
}
|
||||
}
|
||||
pdata[19]++;
|
||||
} while (pdata[19] < max_nonce && !work_restart[thr_id].restart);
|
||||
*hashes_done = pdata[19] - start_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
bool register_heavy_algo( algo_gate_t* gate )
|
||||
{
|
||||
gate->scanhash = (void*)&scanhash_heavy;
|
||||
gate->hash = (void*)&heavyhash;
|
||||
return true;
|
||||
};
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "jha-gate.h"
|
||||
|
||||
#if !defined(JHA_8WAY) && !defined(JHA_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -133,3 +136,4 @@ int scanhash_jha( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -28,30 +28,32 @@ int scanhash_keccak_8way( struct work *work, uint32_t max_nonce,
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
__m512i *noncev = (__m512i*)vdata + 9; // aligned
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
int thr_id = mythr->id;
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
mm512_bswap32_intrlv80_8x64( vdata, pdata );
|
||||
*noncev = mm512_intrlv_blend_32(
|
||||
_mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
|
||||
n+3, 0, n+2, 0, n+1, 0, n , 0 ), *noncev );
|
||||
do {
|
||||
*noncev = mm512_intrlv_blend_32( mm512_bswap_32(
|
||||
_mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
|
||||
n+3, 0, n+2, 0, n+1, 0, n , 0 ) ), *noncev );
|
||||
|
||||
keccakhash_8way( hash, vdata );
|
||||
|
||||
for ( int lane = 0; lane < 8; lane++ )
|
||||
if ( hash7[ lane<<1 ] <= Htarg )
|
||||
if unlikely( hash7[ lane<<1 ] <= Htarg && !bench )
|
||||
{
|
||||
extr_lane_8x64( lane_hash, hash, lane, 256 );
|
||||
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
|
||||
if ( valid_hash( lane_hash, ptarget ) )
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm512_add_epi32( *noncev,
|
||||
m512_const1_64( 0x0000000800000000 ) );
|
||||
n += 8;
|
||||
|
||||
} while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
|
||||
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
@@ -79,29 +81,30 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
__m256i *noncev = (__m256i*)vdata + 9; // aligned
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
int thr_id = mythr->id;
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
mm256_bswap32_intrlv80_4x64( vdata, pdata );
|
||||
*noncev = mm256_intrlv_blend_32(
|
||||
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
|
||||
do {
|
||||
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
|
||||
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
|
||||
|
||||
keccakhash_4way( hash, vdata );
|
||||
|
||||
for ( int lane = 0; lane < 4; lane++ )
|
||||
if ( hash7[ lane<<1 ] <= Htarg )
|
||||
if unlikely( hash7[ lane<<1 ] <= Htarg && !bench )
|
||||
{
|
||||
extr_lane_4x64( lane_hash, hash, lane, 256 );
|
||||
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
|
||||
if ( valid_hash( lane_hash, ptarget ))
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm256_add_epi32( *noncev,
|
||||
m256_const1_64( 0x0000000400000000 ) );
|
||||
n += 4;
|
||||
|
||||
} while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
|
||||
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
|
||||
@@ -1,5 +1,9 @@
|
||||
#include "keccak-gate.h"
|
||||
#include "sph_keccak.h"
|
||||
|
||||
int hard_coded_eb = 1;
|
||||
|
||||
// KECCAK
|
||||
|
||||
bool register_keccak_algo( algo_gate_t* gate )
|
||||
{
|
||||
@@ -19,6 +23,8 @@ bool register_keccak_algo( algo_gate_t* gate )
|
||||
return true;
|
||||
};
|
||||
|
||||
// KECCAKC
|
||||
|
||||
bool register_keccakc_algo( algo_gate_t* gate )
|
||||
{
|
||||
gate->optimizations = AVX2_OPT | AVX512_OPT;
|
||||
@@ -37,3 +43,50 @@ bool register_keccakc_algo( algo_gate_t* gate )
|
||||
return true;
|
||||
};
|
||||
|
||||
// SHA3D
|
||||
|
||||
void sha3d( void *state, const void *input, int len )
|
||||
{
|
||||
uint32_t _ALIGN(64) buffer[16], hash[16];
|
||||
sph_keccak_context ctx_keccak;
|
||||
|
||||
sph_keccak256_init( &ctx_keccak );
|
||||
sph_keccak256 ( &ctx_keccak, input, len );
|
||||
sph_keccak256_close( &ctx_keccak, (void*) buffer );
|
||||
|
||||
sph_keccak256_init( &ctx_keccak );
|
||||
sph_keccak256 ( &ctx_keccak, buffer, 32 );
|
||||
sph_keccak256_close( &ctx_keccak, (void*) hash );
|
||||
|
||||
memcpy(state, hash, 32);
|
||||
}
|
||||
|
||||
void sha3d_gen_merkle_root( char* merkle_root, struct stratum_ctx* sctx )
|
||||
{
|
||||
sha3d( merkle_root, sctx->job.coinbase, (int) sctx->job.coinbase_size );
|
||||
for ( int i = 0; i < sctx->job.merkle_count; i++ )
|
||||
{
|
||||
memcpy( merkle_root + 32, sctx->job.merkle[i], 32 );
|
||||
sha256d( merkle_root, merkle_root, 64 );
|
||||
}
|
||||
}
|
||||
|
||||
bool register_sha3d_algo( algo_gate_t* gate )
|
||||
{
|
||||
hard_coded_eb = 6;
|
||||
// opt_extranonce = false;
|
||||
gate->optimizations = AVX2_OPT | AVX512_OPT;
|
||||
gate->gen_merkle_root = (void*)&sha3d_gen_merkle_root;
|
||||
#if defined (KECCAK_8WAY)
|
||||
gate->scanhash = (void*)&scanhash_sha3d_8way;
|
||||
gate->hash = (void*)&sha3d_hash_8way;
|
||||
#elif defined (KECCAK_4WAY)
|
||||
gate->scanhash = (void*)&scanhash_sha3d_4way;
|
||||
gate->hash = (void*)&sha3d_hash_4way;
|
||||
#else
|
||||
gate->scanhash = (void*)&scanhash_sha3d;
|
||||
gate->hash = (void*)&sha3d_hash;
|
||||
#endif
|
||||
return true;
|
||||
};
|
||||
|
||||
|
||||
@@ -10,24 +10,37 @@
|
||||
#define KECCAK_4WAY 1
|
||||
#endif
|
||||
|
||||
extern int hard_coded_eb;
|
||||
|
||||
#if defined(KECCAK_8WAY)
|
||||
|
||||
void keccakhash_8way( void *state, const void *input );
|
||||
int scanhash_keccak_8way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
|
||||
void sha3d_hash_8way( void *state, const void *input );
|
||||
int scanhash_sha3d_8way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
|
||||
#elif defined(KECCAK_4WAY)
|
||||
|
||||
void keccakhash_4way( void *state, const void *input );
|
||||
int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
|
||||
void sha3d_hash_4way( void *state, const void *input );
|
||||
int scanhash_sha3d_4way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
|
||||
#else
|
||||
|
||||
void keccakhash( void *state, const void *input );
|
||||
int scanhash_keccak( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
|
||||
#endif
|
||||
void sha3d_hash( void *state, const void *input );
|
||||
int scanhash_sha3d( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
|
||||
#endif
|
||||
#endif
|
||||
|
||||
@@ -1,6 +1,9 @@
|
||||
#include <stddef.h>
|
||||
#include <stdint.h>
|
||||
#include "keccak-hash-4way.h"
|
||||
#include "keccak-gate.h"
|
||||
|
||||
#if defined(__AVX2__)
|
||||
|
||||
static const uint64_t RC[] = {
|
||||
0x0000000000000001, 0x0000000000008082,
|
||||
@@ -168,7 +171,7 @@ static void keccak64_8way_close( keccak64_ctx_m512i *kc, void *dst,
|
||||
size_t j;
|
||||
size_t m512_len = byte_len >> 3;
|
||||
|
||||
eb = 0x100 >> 8;
|
||||
eb = hard_coded_eb;
|
||||
if ( kc->ptr == (lim - 8) )
|
||||
{
|
||||
const uint64_t t = eb | 0x8000000000000000;
|
||||
@@ -238,7 +241,7 @@ keccak512_8way_close(void *cc, void *dst)
|
||||
|
||||
#endif // AVX512
|
||||
|
||||
#if defined(__AVX2__)
|
||||
// AVX2
|
||||
|
||||
#define INPUT_BUF(size) do { \
|
||||
size_t j; \
|
||||
@@ -349,7 +352,7 @@ static void keccak64_close( keccak64_ctx_m256i *kc, void *dst, size_t byte_len,
|
||||
size_t j;
|
||||
size_t m256_len = byte_len >> 3;
|
||||
|
||||
eb = 0x100 >> 8;
|
||||
eb = hard_coded_eb;
|
||||
if ( kc->ptr == (lim - 8) )
|
||||
{
|
||||
const uint64_t t = eb | 0x8000000000000000;
|
||||
|
||||
@@ -1,4 +1,6 @@
|
||||
#include "algo-gate-api.h"
|
||||
#include "keccak-gate.h"
|
||||
|
||||
#if !defined(KECCAK_8WAY) && !defined(KECCAK_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
@@ -18,36 +20,35 @@ void keccakhash(void *state, const void *input)
|
||||
memcpy(state, hash, 32);
|
||||
}
|
||||
|
||||
int scanhash_keccak( struct work *work,
|
||||
uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr )
|
||||
int scanhash_keccak( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t n = pdata[19] - 1;
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
//const uint32_t Htarg = ptarget[7];
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
uint32_t _ALIGN(64) hash64[8];
|
||||
uint32_t _ALIGN(64) endiandata[32];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t n = pdata[19];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce;
|
||||
const int thr_id = mythr->id;
|
||||
|
||||
uint32_t _ALIGN(32) hash64[8];
|
||||
uint32_t endiandata[32];
|
||||
for ( int i=0; i < 19; i++ )
|
||||
be32enc( &endiandata[i], pdata[i] );
|
||||
|
||||
for (int i=0; i < 19; i++)
|
||||
be32enc(&endiandata[i], pdata[i]);
|
||||
do {
|
||||
be32enc( &endiandata[19], n );
|
||||
keccakhash( hash64, endiandata );
|
||||
if ( valid_hash( hash64, ptarget ) && !opt_benchmark )
|
||||
{
|
||||
pdata[19] = n;
|
||||
submit_solution( work, hash64, mythr );
|
||||
}
|
||||
n++;
|
||||
} while ( n < last_nonce && !work_restart[thr_id].restart );
|
||||
|
||||
do {
|
||||
|
||||
pdata[19] = ++n;
|
||||
be32enc(&endiandata[19], n);
|
||||
keccakhash(hash64, endiandata);
|
||||
if (((hash64[7]&0xFFFFFF00)==0) &&
|
||||
fulltest(hash64, ptarget)) {
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
return true;
|
||||
}
|
||||
} while (n < max_nonce && !work_restart[thr_id].restart);
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
*hashes_done = n - first_nonce;
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
126
algo/keccak/sha3d-4way.c
Normal file
126
algo/keccak/sha3d-4way.c
Normal file
@@ -0,0 +1,126 @@
|
||||
#include "keccak-gate.h"
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
#include "sph_keccak.h"
|
||||
#include "keccak-hash-4way.h"
|
||||
|
||||
#if defined(KECCAK_8WAY)
|
||||
|
||||
void sha3d_hash_8way(void *state, const void *input)
|
||||
{
|
||||
uint32_t buffer[16*8] __attribute__ ((aligned (128)));
|
||||
keccak256_8way_context ctx;
|
||||
|
||||
keccak256_8way_init( &ctx );
|
||||
keccak256_8way_update( &ctx, input, 80 );
|
||||
keccak256_8way_close( &ctx, buffer );
|
||||
|
||||
keccak256_8way_init( &ctx );
|
||||
keccak256_8way_update( &ctx, buffer, 32 );
|
||||
keccak256_8way_close( &ctx, state );
|
||||
}
|
||||
|
||||
int scanhash_sha3d_8way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t vdata[24*8] __attribute__ ((aligned (128)));
|
||||
uint32_t hash[16*8] __attribute__ ((aligned (64)));
|
||||
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
|
||||
uint32_t *hash7 = &(hash[49]); // 3*16+1
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t n = pdata[19];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 8;
|
||||
__m512i *noncev = (__m512i*)vdata + 9; // aligned
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
mm512_bswap32_intrlv80_8x64( vdata, pdata );
|
||||
*noncev = mm512_intrlv_blend_32(
|
||||
_mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
|
||||
n+3, 0, n+2, 0, n+1, 0, n , 0 ), *noncev );
|
||||
do {
|
||||
sha3d_hash_8way( hash, vdata );
|
||||
|
||||
for ( int lane = 0; lane < 8; lane++ )
|
||||
if ( unlikely( hash7[ lane<<1 ] <= Htarg && !bench ) )
|
||||
{
|
||||
extr_lane_8x64( lane_hash, hash, lane, 256 );
|
||||
if ( valid_hash( lane_hash, ptarget ) )
|
||||
{
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm512_add_epi32( *noncev,
|
||||
m512_const1_64( 0x0000000800000000 ) );
|
||||
n += 8;
|
||||
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#elif defined(KECCAK_4WAY)
|
||||
|
||||
void sha3d_hash_4way(void *state, const void *input)
|
||||
{
|
||||
uint32_t buffer[16*4] __attribute__ ((aligned (64)));
|
||||
keccak256_4way_context ctx;
|
||||
|
||||
keccak256_4way_init( &ctx );
|
||||
keccak256_4way_update( &ctx, input, 80 );
|
||||
keccak256_4way_close( &ctx, buffer );
|
||||
|
||||
keccak256_4way_init( &ctx );
|
||||
keccak256_4way_update( &ctx, buffer, 32 );
|
||||
keccak256_4way_close( &ctx, state );
|
||||
}
|
||||
|
||||
int scanhash_sha3d_4way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
|
||||
uint32_t hash[16*4] __attribute__ ((aligned (32)));
|
||||
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
|
||||
uint32_t *hash7 = &(hash[25]); // 3*8+1
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t n = pdata[19];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 4;
|
||||
__m256i *noncev = (__m256i*)vdata + 9; // aligned
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
mm256_bswap32_intrlv80_4x64( vdata, pdata );
|
||||
*noncev = mm256_intrlv_blend_32(
|
||||
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
|
||||
do {
|
||||
sha3d_hash_4way( hash, vdata );
|
||||
|
||||
for ( int lane = 0; lane < 4; lane++ )
|
||||
if ( unlikely( hash7[ lane<<1 ] <= Htarg && !bench ) )
|
||||
{
|
||||
extr_lane_4x64( lane_hash, hash, lane, 256 );
|
||||
if ( valid_hash( lane_hash, ptarget ) )
|
||||
{
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm256_add_epi32( *noncev,
|
||||
m256_const1_64( 0x0000000400000000 ) );
|
||||
n += 4;
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
54
algo/keccak/sha3d.c
Normal file
54
algo/keccak/sha3d.c
Normal file
@@ -0,0 +1,54 @@
|
||||
#include "keccak-gate.h"
|
||||
|
||||
#if !defined(KECCAK_8WAY) && !defined(KECCAK_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
#include "sph_keccak.h"
|
||||
|
||||
void sha3d_hash(void *state, const void *input)
|
||||
{
|
||||
uint32_t buffer[16];
|
||||
sph_keccak256_context ctx_keccak;
|
||||
|
||||
sph_keccak256_init( &ctx_keccak );
|
||||
sph_keccak256 ( &ctx_keccak, input, 80 );
|
||||
sph_keccak256_close( &ctx_keccak, buffer );
|
||||
sph_keccak256_init( &ctx_keccak );
|
||||
sph_keccak256 ( &ctx_keccak, buffer, 32 );
|
||||
sph_keccak256_close( &ctx_keccak, state );
|
||||
}
|
||||
|
||||
int scanhash_sha3d( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t _ALIGN(64) hash64[8];
|
||||
uint32_t _ALIGN(64) endiandata[32];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t n = pdata[19];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce;
|
||||
const int thr_id = mythr->id;
|
||||
|
||||
for ( int i=0; i < 19; i++ )
|
||||
be32enc( &endiandata[i], pdata[i] );
|
||||
|
||||
do {
|
||||
be32enc( &endiandata[19], n );
|
||||
sha3d_hash( hash64, endiandata );
|
||||
if ( valid_hash( hash64, ptarget ) && !opt_benchmark )
|
||||
{
|
||||
pdata[19] = n;
|
||||
submit_solution( work, hash64, mythr );
|
||||
}
|
||||
n++;
|
||||
} while ( n < last_nonce && !work_restart[thr_id].restart );
|
||||
|
||||
*hashes_done = n - first_nonce;
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
@@ -32,8 +32,8 @@
|
||||
|
||||
#include <stddef.h>
|
||||
#include <string.h>
|
||||
|
||||
#include "sph_keccak.h"
|
||||
#include "keccak-gate.h"
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C"{
|
||||
@@ -1616,7 +1616,7 @@ keccak_core(sph_keccak_context *kc, const void *data, size_t len, size_t lim)
|
||||
} u; \
|
||||
size_t j; \
|
||||
\
|
||||
eb = (0x100 | (ub & 0xFF)) >> (8 - n); \
|
||||
eb = hard_coded_eb; \
|
||||
if (kc->ptr == (lim - 1)) { \
|
||||
if (n == 7) { \
|
||||
u.tmp[0] = eb; \
|
||||
|
||||
@@ -1,63 +0,0 @@
|
||||
#include "algo-gate-api.h"
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
#include <stdio.h>
|
||||
|
||||
#include "sph_luffa.h"
|
||||
|
||||
void luffahash(void *output, const void *input)
|
||||
{
|
||||
unsigned char _ALIGN(128) hash[64];
|
||||
sph_luffa512_context ctx_luffa;
|
||||
|
||||
sph_luffa512_init(&ctx_luffa);
|
||||
sph_luffa512 (&ctx_luffa, input, 80);
|
||||
sph_luffa512_close(&ctx_luffa, (void*) hash);
|
||||
|
||||
memcpy(output, hash, 32);
|
||||
}
|
||||
|
||||
int scanhash_luffa(int thr_id, struct work *work,
|
||||
uint32_t max_nonce, uint64_t *hashes_done)
|
||||
{
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
|
||||
uint32_t _ALIGN(64) hash64[8];
|
||||
uint32_t _ALIGN(64) endiandata[20];
|
||||
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
|
||||
uint32_t n = first_nonce;
|
||||
|
||||
for (int i=0; i < 19; i++)
|
||||
be32enc(&endiandata[i], pdata[i]);
|
||||
|
||||
do {
|
||||
be32enc(&endiandata[19], n);
|
||||
luffahash(hash64, endiandata);
|
||||
if (hash64[7] < Htarg && fulltest(hash64, ptarget)) {
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
return true;
|
||||
}
|
||||
n++;
|
||||
|
||||
} while (n < max_nonce && !work_restart[thr_id].restart);
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
bool register_luffa_algo( algo_gate_t* gate )
|
||||
{
|
||||
gate->scanhash = (void*)&scanhash_luffa;
|
||||
gate->hash = (void*)&luffahash;
|
||||
return true;
|
||||
};
|
||||
|
||||
@@ -344,17 +344,12 @@ HashReturn update_and_final_luffa( hashState_luffa *state, BitSequence* output,
|
||||
|
||||
// 16 byte partial block exists for 80 byte len
|
||||
if ( state->rembytes )
|
||||
{
|
||||
// padding of partial block
|
||||
rnd512( state, _mm_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0x80,0,0,0 ),
|
||||
// padding of partial block
|
||||
rnd512( state, m128_const_64( 0, 0x80000000 ),
|
||||
mm128_bswap_32( cast_m128i( data ) ) );
|
||||
}
|
||||
else
|
||||
{
|
||||
// empty pad block
|
||||
rnd512( state, _mm_setzero_si128(),
|
||||
_mm_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0x80,0,0,0 ) );
|
||||
}
|
||||
// empty pad block
|
||||
rnd512( state, m128_zero, m128_const_64( 0, 0x80000000 ) );
|
||||
|
||||
finalization512( state, (uint32*) output );
|
||||
if ( state->hashbitlen > 512 )
|
||||
@@ -363,6 +358,56 @@ HashReturn update_and_final_luffa( hashState_luffa *state, BitSequence* output,
|
||||
return SUCCESS;
|
||||
}
|
||||
|
||||
|
||||
int luffa_full( hashState_luffa *state, BitSequence* output, int hashbitlen,
|
||||
const BitSequence* data, size_t inlen )
|
||||
{
|
||||
// Optimized for integrals of 16 bytes, good for 64 and 80 byte len
|
||||
int i;
|
||||
state->hashbitlen = hashbitlen;
|
||||
/* set the lower 32 bits to '1' */
|
||||
MASK= _mm_set_epi32(0x00000000, 0x00000000, 0x00000000, 0xffffffff);
|
||||
/* set all bits to '1' */
|
||||
ALLONE = _mm_set_epi32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff);
|
||||
/* set the 32-bit round constant values to the 128-bit data field */
|
||||
for ( i=0; i<32; i++ )
|
||||
CNS128[i] = _mm_load_si128( (__m128i*)&CNS_INIT[i*4] );
|
||||
for ( i=0; i<10; i++ )
|
||||
state->chainv[i] = _mm_load_si128( (__m128i*)&IV[i*4] );
|
||||
memset(state->buffer, 0, sizeof state->buffer );
|
||||
|
||||
// update
|
||||
|
||||
int blocks = (int)( inlen / 32 );
|
||||
state->rembytes = inlen % 32;
|
||||
|
||||
// full blocks
|
||||
for ( i = 0; i < blocks; i++ )
|
||||
{
|
||||
rnd512( state, mm128_bswap_32( casti_m128i( data, 1 ) ),
|
||||
mm128_bswap_32( casti_m128i( data, 0 ) ) );
|
||||
data += MSG_BLOCK_BYTE_LEN;
|
||||
}
|
||||
|
||||
// final
|
||||
|
||||
// 16 byte partial block exists for 80 byte len
|
||||
if ( state->rembytes )
|
||||
// padding of partial block
|
||||
rnd512( state, m128_const_64( 0, 0x80000000 ),
|
||||
mm128_bswap_32( cast_m128i( data ) ) );
|
||||
else
|
||||
// empty pad block
|
||||
rnd512( state, m128_zero, m128_const_64( 0, 0x80000000 ) );
|
||||
|
||||
finalization512( state, (uint32*) output );
|
||||
if ( state->hashbitlen > 512 )
|
||||
finalization512( state, (uint32*)( output+128 ) );
|
||||
|
||||
return SUCCESS;
|
||||
}
|
||||
|
||||
|
||||
/***************************************************/
|
||||
/* Round function */
|
||||
/* state: hash context */
|
||||
|
||||
@@ -1,3 +1,6 @@
|
||||
#if !defined(LUFFA_FOR_SSE2_H__)
|
||||
#define LUFFA_FOR_SSE2_H__ 1
|
||||
|
||||
/*
|
||||
* luffa_for_sse2.h
|
||||
* Version 2.0 (Sep 15th 2009)
|
||||
@@ -48,8 +51,6 @@
|
||||
typedef struct {
|
||||
uint32 buffer[8] __attribute((aligned(32)));
|
||||
__m128i chainv[10] __attribute((aligned(32))); /* Chaining values */
|
||||
// uint64 bitlen[2]; /* Message length in bits */
|
||||
// uint32 rembitlen; /* Length of buffer data to be hashed */
|
||||
int hashbitlen;
|
||||
int rembytes;
|
||||
} hashState_luffa;
|
||||
@@ -65,5 +66,6 @@ HashReturn final_luffa( hashState_luffa *state, BitSequence *hashval );
|
||||
HashReturn update_and_final_luffa( hashState_luffa *state, BitSequence* output,
|
||||
const BitSequence* data, size_t inlen );
|
||||
|
||||
|
||||
|
||||
int luffa_full( hashState_luffa *state, BitSequence* output, int hashbitlen,
|
||||
const BitSequence* data, size_t inlen );
|
||||
#endif // LUFFA_FOR_SSE2_H___
|
||||
|
||||
@@ -115,9 +115,8 @@ void allium_16way_hash( void *state, const void *input )
|
||||
intrlv_4x128( vhashA, hash0, hash1, hash2, hash3, 256 );
|
||||
intrlv_4x128( vhashB, hash4, hash5, hash6, hash7, 256 );
|
||||
|
||||
cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
|
||||
cube_4way_full( &ctx.cube, vhashA, 256, vhashA, 32 );
|
||||
cube_4way_full( &ctx.cube, vhashB, 256, vhashB, 32 );
|
||||
|
||||
dintrlv_4x128( hash0, hash1, hash2, hash3, vhashA, 256 );
|
||||
dintrlv_4x128( hash4, hash5, hash6, hash7, vhashB, 256 );
|
||||
@@ -125,10 +124,8 @@ void allium_16way_hash( void *state, const void *input )
|
||||
intrlv_4x128( vhashA, hash8, hash9, hash10, hash11, 256 );
|
||||
intrlv_4x128( vhashB, hash12, hash13, hash14, hash15, 256 );
|
||||
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
|
||||
cube_4way_full( &ctx.cube, vhashA, 256, vhashA, 32 );
|
||||
cube_4way_full( &ctx.cube, vhashB, 256, vhashB, 32 );
|
||||
|
||||
dintrlv_4x128( hash8, hash9, hash10, hash11, vhashA, 256 );
|
||||
dintrlv_4x128( hash12, hash13, hash14, hash15, vhashB, 256 );
|
||||
@@ -169,7 +166,6 @@ void allium_16way_hash( void *state, const void *input )
|
||||
skein256_8way_update( &ctx.skein, vhashB, 32 );
|
||||
skein256_8way_close( &ctx.skein, vhashB );
|
||||
|
||||
|
||||
dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
|
||||
vhashA, 256 );
|
||||
dintrlv_8x64( hash8, hash9, hash10, hash11, hash12, hash13, hash14, hash15,
|
||||
@@ -179,77 +175,43 @@ void allium_16way_hash( void *state, const void *input )
|
||||
|
||||
intrlv_4x128( vhash, hash0, hash1, hash2, hash3, 256 );
|
||||
|
||||
groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
|
||||
groestl256_4way_full( &ctx.groestl, vhash, vhash, 256 );
|
||||
|
||||
dintrlv_4x128( state, state+32, state+64, state+96, vhash, 256 );
|
||||
intrlv_4x128( vhash, hash4, hash5, hash6, hash7, 256 );
|
||||
|
||||
groestl256_4way_init( &ctx.groestl, 32 );
|
||||
groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
|
||||
groestl256_4way_full( &ctx.groestl, vhash, vhash, 256 );
|
||||
|
||||
dintrlv_4x128( state+128, state+160, state+192, state+224, vhash, 256 );
|
||||
intrlv_4x128( vhash, hash8, hash9, hash10, hash11, 256 );
|
||||
|
||||
groestl256_4way_init( &ctx.groestl, 32 );
|
||||
groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
|
||||
groestl256_4way_full( &ctx.groestl, vhash, vhash, 256 );
|
||||
|
||||
dintrlv_4x128( state+256, state+288, state+320, state+352, vhash, 256 );
|
||||
intrlv_4x128( vhash, hash12, hash13, hash14, hash15, 256 );
|
||||
|
||||
groestl256_4way_init( &ctx.groestl, 32 );
|
||||
groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
|
||||
groestl256_4way_full( &ctx.groestl, vhash, vhash, 256 );
|
||||
|
||||
dintrlv_4x128( state+384, state+416, state+448, state+480, vhash, 256 );
|
||||
|
||||
#else
|
||||
|
||||
update_and_final_groestl256( &ctx.groestl, state, hash0, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+32, hash1, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+64, hash2, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+96, hash3, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+128, hash4, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+160, hash5, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+192, hash6, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+224, hash7, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+256, hash8, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+288, hash9, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+320, hash10, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+352, hash11, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+384, hash12, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+416, hash13, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+448, hash14, 256 );
|
||||
memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+480, hash15, 256 );
|
||||
|
||||
groestl256_full( &ctx.groestl, state, hash0, 256 );
|
||||
groestl256_full( &ctx.groestl, state+32, hash1, 256 );
|
||||
groestl256_full( &ctx.groestl, state+64, hash2, 256 );
|
||||
groestl256_full( &ctx.groestl, state+96, hash3, 256 );
|
||||
groestl256_full( &ctx.groestl, state+128, hash4, 256 );
|
||||
groestl256_full( &ctx.groestl, state+160, hash5, 256 );
|
||||
groestl256_full( &ctx.groestl, state+192, hash6, 256 );
|
||||
groestl256_full( &ctx.groestl, state+224, hash7, 256 );
|
||||
groestl256_full( &ctx.groestl, state+256, hash8, 256 );
|
||||
groestl256_full( &ctx.groestl, state+288, hash9, 256 );
|
||||
groestl256_full( &ctx.groestl, state+320, hash10, 256 );
|
||||
groestl256_full( &ctx.groestl, state+352, hash11, 256 );
|
||||
groestl256_full( &ctx.groestl, state+384, hash12, 256 );
|
||||
groestl256_full( &ctx.groestl, state+416, hash13, 256 );
|
||||
groestl256_full( &ctx.groestl, state+448, hash14, 256 );
|
||||
groestl256_full( &ctx.groestl, state+480, hash15, 256 );
|
||||
#endif
|
||||
}
|
||||
|
||||
@@ -263,37 +225,32 @@ int scanhash_allium_16way( struct work *work, uint32_t max_nonce,
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
uint32_t n = first_nonce;
|
||||
const uint32_t last_nonce = max_nonce - 16;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
__m512i *noncev = (__m512i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
if ( opt_benchmark )
|
||||
( (uint32_t*)ptarget )[7] = 0x0000ff;
|
||||
if ( bench ) ( (uint32_t*)ptarget )[7] = 0x0000ff;
|
||||
|
||||
mm512_bswap32_intrlv80_16x32( vdata, pdata );
|
||||
*noncev = _mm512_set_epi32( n+15, n+14, n+13, n+12, n+11, n+10, n+ 9, n+ 8,
|
||||
n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n +1, n );
|
||||
|
||||
blake256_16way_init( &allium_16way_ctx.blake );
|
||||
blake256_16way_update( &allium_16way_ctx.blake, vdata, 64 );
|
||||
|
||||
do {
|
||||
*noncev = mm512_bswap_32( _mm512_set_epi32( n+15, n+14, n+13, n+12,
|
||||
n+11, n+10, n+ 9, n+ 8,
|
||||
n+ 7, n+ 6, n+ 5, n+ 4,
|
||||
n+ 3, n+ 2, n +1, n ) );
|
||||
|
||||
allium_16way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int lane = 0; lane < 16; lane++ ) if ( (hash+(lane<<3))[7] <= Htarg )
|
||||
for ( int lane = 0; lane < 16; lane++ )
|
||||
if ( unlikely( valid_hash( hash+(lane<<3), ptarget ) && !bench ) )
|
||||
{
|
||||
if ( fulltest( hash+(lane<<3), ptarget ) && !opt_benchmark )
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
submit_lane_solution( work, hash+(lane<<3), mythr, lane );
|
||||
}
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, hash+(lane<<3), mythr, lane );
|
||||
}
|
||||
*noncev = _mm512_add_epi32( *noncev, m512_const1_32( 16 ) );
|
||||
n += 16;
|
||||
} while ( (n < last_nonce) && !work_restart[thr_id].restart);
|
||||
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
@@ -320,18 +277,18 @@ bool init_allium_8way_ctx()
|
||||
return true;
|
||||
}
|
||||
|
||||
void allium_8way_hash( void *state, const void *input )
|
||||
void allium_8way_hash( void *hash, const void *input )
|
||||
{
|
||||
uint32_t vhashA[8*8] __attribute__ ((aligned (64)));
|
||||
uint32_t vhashB[8*8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash0[8] __attribute__ ((aligned (32)));
|
||||
uint32_t hash1[8] __attribute__ ((aligned (32)));
|
||||
uint32_t hash2[8] __attribute__ ((aligned (32)));
|
||||
uint32_t hash3[8] __attribute__ ((aligned (32)));
|
||||
uint32_t hash4[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash5[8] __attribute__ ((aligned (32)));
|
||||
uint32_t hash6[8] __attribute__ ((aligned (32)));
|
||||
uint32_t hash7[8] __attribute__ ((aligned (32)));
|
||||
uint64_t vhashA[4*8] __attribute__ ((aligned (64)));
|
||||
uint64_t vhashB[4*8] __attribute__ ((aligned (64)));
|
||||
uint64_t *hash0 = (uint64_t*)hash;
|
||||
uint64_t *hash1 = (uint64_t*)hash+ 4;
|
||||
uint64_t *hash2 = (uint64_t*)hash+ 8;
|
||||
uint64_t *hash3 = (uint64_t*)hash+12;
|
||||
uint64_t *hash4 = (uint64_t*)hash+16;
|
||||
uint64_t *hash5 = (uint64_t*)hash+20;
|
||||
uint64_t *hash6 = (uint64_t*)hash+24;
|
||||
uint64_t *hash7 = (uint64_t*)hash+28;
|
||||
allium_8way_ctx_holder ctx __attribute__ ((aligned (64)));
|
||||
|
||||
memcpy( &ctx, &allium_8way_ctx, sizeof(allium_8way_ctx) );
|
||||
@@ -398,69 +355,52 @@ void allium_8way_hash( void *state, const void *input )
|
||||
dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, 256 );
|
||||
dintrlv_4x64( hash4, hash5, hash6, hash7, vhashB, 256 );
|
||||
|
||||
update_and_final_groestl256( &ctx.groestl, state, hash0, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+32, hash1, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+64, hash2, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+96, hash3, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+128, hash4, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+160, hash5, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+192, hash6, 256 );
|
||||
memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
|
||||
sizeof(hashState_groestl256) );
|
||||
update_and_final_groestl256( &ctx.groestl, state+224, hash7, 256 );
|
||||
groestl256_full( &ctx.groestl, hash0, hash0, 256 );
|
||||
groestl256_full( &ctx.groestl, hash1, hash1, 256 );
|
||||
groestl256_full( &ctx.groestl, hash2, hash2, 256 );
|
||||
groestl256_full( &ctx.groestl, hash3, hash3, 256 );
|
||||
groestl256_full( &ctx.groestl, hash4, hash4, 256 );
|
||||
groestl256_full( &ctx.groestl, hash5, hash5, 256 );
|
||||
groestl256_full( &ctx.groestl, hash6, hash6, 256 );
|
||||
groestl256_full( &ctx.groestl, hash7, hash7, 256 );
|
||||
}
|
||||
|
||||
int scanhash_allium_8way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t hash[8*8] __attribute__ ((aligned (64)));
|
||||
uint64_t hash[4*8] __attribute__ ((aligned (64)));
|
||||
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint64_t *ptarget = (uint64_t*)work->target;
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 8;
|
||||
uint32_t n = first_nonce;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
__m256i *noncev = (__m256i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id;
|
||||
|
||||
if ( opt_benchmark )
|
||||
( (uint32_t*)ptarget )[7] = 0x0000ff;
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
mm256_bswap32_intrlv80_8x32( vdata, pdata );
|
||||
*noncev = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
|
||||
|
||||
blake256_8way_init( &allium_8way_ctx.blake );
|
||||
blake256_8way_update( &allium_8way_ctx.blake, vdata, 64 );
|
||||
|
||||
do {
|
||||
*noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
|
||||
n+3, n+2, n+1, n ) );
|
||||
|
||||
allium_8way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int lane = 0; lane < 8; lane++ ) if ( (hash+(lane<<3))[7] <= Htarg )
|
||||
for ( int lane = 0; lane < 8; lane++ )
|
||||
{
|
||||
if ( fulltest( hash+(lane<<3), ptarget ) && !opt_benchmark )
|
||||
const uint64_t *lane_hash = hash + (lane<<2);
|
||||
if ( unlikely( valid_hash( lane_hash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
submit_lane_solution( work, hash+(lane<<3), mythr, lane );
|
||||
}
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
n += 8;
|
||||
} while ( (n < last_nonce) && !work_restart[thr_id].restart);
|
||||
|
||||
*noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
|
||||
} while ( likely( (n <= last_nonce) && !work_restart[thr_id].restart ) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "lyra2-gate.h"
|
||||
|
||||
#if !( defined(ALLIUM_16WAY) || defined(ALLIUM_8WAY) || defined(ALLIUM_4WAY) )
|
||||
|
||||
#include <memory.h>
|
||||
#include "algo/blake/sph_blake.h"
|
||||
#include "algo/keccak/sph_keccak.h"
|
||||
@@ -107,3 +110,4 @@ int scanhash_allium( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -194,7 +194,7 @@ bool register_allium_algo( algo_gate_t* gate )
|
||||
|
||||
/////////////////////////////////////////
|
||||
|
||||
bool phi2_has_roots;
|
||||
bool phi2_has_roots = false;
|
||||
bool phi2_use_roots = false;
|
||||
|
||||
int phi2_get_work_data_size() { return phi2_use_roots ? 144 : 128; }
|
||||
@@ -220,7 +220,7 @@ void phi2_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
|
||||
// Assemble block header
|
||||
algo_gate.build_block_header( g_work, le32dec( sctx->job.version ),
|
||||
(uint32_t*) sctx->job.prevhash, (uint32_t*) merkle_tree,
|
||||
le32dec( sctx->job.ntime ), le32dec(sctx->job.nbits) );
|
||||
le32dec( sctx->job.ntime ), le32dec(sctx->job.nbits), NULL );
|
||||
for ( t = 0; t < 16; t++ )
|
||||
g_work->data[ 20+t ] = ((uint32_t*)sctx->job.extra)[t];
|
||||
}
|
||||
|
||||
@@ -75,7 +75,6 @@ int scanhash_lyra2rev2_4way( struct work *work, uint32_t max_nonce,
|
||||
bool init_lyra2rev2_4way_ctx();
|
||||
|
||||
#else
|
||||
|
||||
void lyra2rev2_hash( void *state, const void *input );
|
||||
int scanhash_lyra2rev2( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr );
|
||||
@@ -189,7 +188,7 @@ bool init_allium_ctx();
|
||||
// #define PHI2_4WAY
|
||||
#endif
|
||||
|
||||
bool phi2_has_roots;
|
||||
extern bool phi2_has_roots;
|
||||
|
||||
bool register_phi2_algo( algo_gate_t* gate );
|
||||
#if defined(PHI2_4WAY)
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "lyra2-gate.h"
|
||||
|
||||
#if !( defined(LYRA2H_8WAY) || defined(LYRA2H_4WAY) )
|
||||
|
||||
#include <memory.h>
|
||||
#include <mm_malloc.h>
|
||||
#include "lyra2.h"
|
||||
@@ -71,3 +74,4 @@ int scanhash_lyra2h( struct work *work, uint32_t max_nonce,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "lyra2-gate.h"
|
||||
|
||||
#if !( defined(LYRA2REV2_16WAY) || defined(LYRA2REV2_8WAY) || defined(LYRA2REV2_4WAY) )
|
||||
|
||||
#include <memory.h>
|
||||
#include "algo/blake/sph_blake.h"
|
||||
#include "algo/cubehash/sph_cubehash.h"
|
||||
@@ -107,4 +110,4 @@ int scanhash_lyra2rev2( struct work *work,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -79,19 +79,16 @@ void lyra2rev3_16way_hash( void *state, const void *input )
|
||||
dintrlv_2x256( hash14, hash15, vhash, 256 );
|
||||
|
||||
intrlv_4x128( vhash, hash0, hash1, hash2, hash3, 256 );
|
||||
cube_4way_update_close( &ctx.cube, vhash, vhash, 32 );
|
||||
cube_4way_full( &ctx.cube, vhash, 256, vhash, 32 );
|
||||
dintrlv_4x128( hash0, hash1, hash2, hash3, vhash, 256 );
|
||||
intrlv_4x128( vhash, hash4, hash5, hash6, hash7, 256 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhash, vhash, 32 );
|
||||
cube_4way_full( &ctx.cube, vhash, 256, vhash, 32 );
|
||||
dintrlv_4x128( hash4, hash5, hash6, hash7, vhash, 256 );
|
||||
intrlv_4x128( vhash, hash8, hash9, hash10, hash11, 256 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhash, vhash, 32 );
|
||||
cube_4way_full( &ctx.cube, vhash, 256, vhash, 32 );
|
||||
dintrlv_4x128( hash8, hash9, hash10, hash11, vhash, 256 );
|
||||
intrlv_4x128( vhash, hash12, hash13, hash14, hash15, 256 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhash, vhash, 32 );
|
||||
cube_4way_full( &ctx.cube, vhash, 256, vhash, 32 );
|
||||
dintrlv_4x128( hash12, hash13, hash14, hash15, vhash, 256 );
|
||||
|
||||
intrlv_2x256( vhash, hash0, hash1, 256 );
|
||||
@@ -224,21 +221,14 @@ void lyra2rev3_8way_hash( void *state, const void *input )
|
||||
LYRA2REV3( l2v3_wholeMatrix, hash6, 32, hash6, 32, hash6, 32, 1, 4, 4 );
|
||||
LYRA2REV3( l2v3_wholeMatrix, hash7, 32, hash7, 32, hash7, 32, 1, 4, 4 );
|
||||
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash4, (const byte*) hash4, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash5, (const byte*) hash5, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash6, (const byte*) hash6, 32 );
|
||||
cubehashInit( &ctx.cube, 256, 16, 32 );
|
||||
cubehashUpdateDigest( &ctx.cube, (byte*) hash7, (const byte*) hash7, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash1, 256, (const byte*) hash1, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash2, 256, (const byte*) hash2, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash3, 256, (const byte*) hash3, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash4, 256, (const byte*) hash4, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash5, 256, (const byte*) hash5, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash6, 256, (const byte*) hash6, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash7, 256, (const byte*) hash7, 32 );
|
||||
|
||||
LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
|
||||
LYRA2REV3( l2v3_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
|
||||
@@ -265,25 +255,24 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
|
||||
uint32_t *hash7 = &hash[7<<3];
|
||||
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
|
||||
uint32_t *pdata = work->data;
|
||||
const uint32_t *ptarget = work->target;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 8;
|
||||
uint32_t n = first_nonce;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
__m256i *noncev = (__m256i*)vdata + 19; // aligned
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
if ( opt_benchmark ) ( (uint32_t*)ptarget )[7] = 0x0000ff;
|
||||
if ( bench ) ptarget[7] = 0x0000ff;
|
||||
|
||||
mm256_bswap32_intrlv80_8x32( vdata, pdata );
|
||||
|
||||
*noncev = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
|
||||
blake256_8way_init( &l2v3_8way_ctx.blake );
|
||||
blake256_8way_update( &l2v3_8way_ctx.blake, vdata, 64 );
|
||||
|
||||
do
|
||||
{
|
||||
*noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
|
||||
n+3, n+2, n+1, n ) );
|
||||
|
||||
lyra2rev3_8way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
@@ -291,15 +280,17 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
|
||||
if ( unlikely( hash7[lane] <= Htarg ) )
|
||||
{
|
||||
extr_lane_8x32( lane_hash, hash, lane, 256 );
|
||||
if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
|
||||
if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
|
||||
n += 8;
|
||||
} while ( likely( (n < max_nonce-8) && !work_restart[thr_id].restart ) );
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "lyra2-gate.h"
|
||||
|
||||
#if !( defined(LYRA2REV3_16WAY) || defined(LYRA2REV3_8WAY) || defined(LYRA2REV3_4WAY) )
|
||||
|
||||
#include <memory.h>
|
||||
#include "algo/blake/sph_blake.h"
|
||||
#include "algo/cubehash/sph_cubehash.h"
|
||||
@@ -96,4 +99,4 @@ int scanhash_lyra2rev3( struct work *work,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -97,41 +97,42 @@ void lyra2z_16way_hash( void *state, const void *input )
|
||||
int scanhash_lyra2z_16way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t hash[8*16] __attribute__ ((aligned (128)));
|
||||
uint64_t hash[4*16] __attribute__ ((aligned (128)));
|
||||
uint32_t vdata[20*16] __attribute__ ((aligned (64)));
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
uint32_t n = first_nonce;
|
||||
const uint32_t last_nonce = max_nonce - 16;
|
||||
__m512i *noncev = (__m512i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
if ( opt_benchmark )
|
||||
ptarget[7] = 0x0000ff;
|
||||
if ( bench ) ptarget[7] = 0x0000ff;
|
||||
|
||||
mm512_bswap32_intrlv80_16x32( vdata, pdata );
|
||||
*noncev = _mm512_set_epi32( n+15, n+14, n+13, n+12, n+11, n+10, n+ 9, n+ 8,
|
||||
n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n +1, n );
|
||||
lyra2z_16way_midstate( vdata );
|
||||
|
||||
do {
|
||||
*noncev = mm512_bswap_32( _mm512_set_epi32( n+15, n+14, n+13, n+12,
|
||||
n+11, n+10, n+ 9, n+ 8,
|
||||
n+ 7, n+ 6, n+ 5, n+ 4,
|
||||
n+ 3, n+ 2, n+ 1, n ) );
|
||||
lyra2z_16way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int i = 0; i < 16; i++ )
|
||||
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget )
|
||||
&& !opt_benchmark )
|
||||
for ( int lane = 0; lane < 16; lane++ )
|
||||
{
|
||||
pdata[19] = n+i;
|
||||
submit_lane_solution( work, hash+(i<<3), mythr, i );
|
||||
const uint64_t *lane_hash = hash + (lane<<2);
|
||||
if ( unlikely( valid_hash( lane_hash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm512_add_epi32( *noncev, m512_const1_32( 16 ) );
|
||||
n += 16;
|
||||
} while ( (n < max_nonce-16) && !work_restart[thr_id].restart);
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
@@ -195,39 +196,40 @@ void lyra2z_8way_hash( void *state, const void *input )
|
||||
int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t hash[8*8] __attribute__ ((aligned (64)));
|
||||
uint64_t hash[4*8] __attribute__ ((aligned (64)));
|
||||
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 8;
|
||||
uint32_t n = first_nonce;
|
||||
__m256i *noncev = (__m256i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
if ( opt_benchmark )
|
||||
ptarget[7] = 0x0000ff;
|
||||
if ( bench ) ptarget[7] = 0x0000ff;
|
||||
|
||||
mm256_bswap32_intrlv80_8x32( vdata, pdata );
|
||||
*noncev = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
|
||||
lyra2z_8way_midstate( vdata );
|
||||
|
||||
do {
|
||||
*noncev = mm256_bswap_32(
|
||||
_mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n ) );
|
||||
lyra2z_8way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int i = 0; i < 8; i++ )
|
||||
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget )
|
||||
&& !opt_benchmark )
|
||||
for ( int lane = 0; lane < 8; lane++ )
|
||||
{
|
||||
pdata[19] = n+i;
|
||||
submit_lane_solution( work, hash+(i<<3), mythr, i );
|
||||
const uint64_t *lane_hash = hash + (lane<<2);
|
||||
if ( unlikely( valid_hash( lane_hash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
|
||||
n += 8;
|
||||
} while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
@@ -274,39 +276,40 @@ void lyra2z_4way_hash( void *state, const void *input )
|
||||
int scanhash_lyra2z_4way( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t hash[8*4] __attribute__ ((aligned (64)));
|
||||
uint64_t hash[4*4] __attribute__ ((aligned (64)));
|
||||
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 4;
|
||||
uint32_t n = first_nonce;
|
||||
__m128i *noncev = (__m128i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
if ( opt_benchmark )
|
||||
ptarget[7] = 0x0000ff;
|
||||
if ( bench ) ptarget[7] = 0x0000ff;
|
||||
|
||||
mm128_bswap32_intrlv80_4x32( vdata, pdata );
|
||||
*noncev = _mm_set_epi32( n+3, n+2, n+1, n );
|
||||
lyra2z_4way_midstate( vdata );
|
||||
|
||||
do {
|
||||
*noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
|
||||
|
||||
lyra2z_4way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int i = 0; i < 4; i++ )
|
||||
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget )
|
||||
&& !opt_benchmark )
|
||||
for ( int lane = 0; lane < 4; lane++ )
|
||||
{
|
||||
pdata[19] = n+i;
|
||||
submit_lane_solution( work, hash+(i<<3), mythr, i );
|
||||
const uint64_t *lane_hash = hash + (lane<<2);
|
||||
if ( unlikely( valid_hash( lane_hash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm_add_epi32( *noncev, m128_const1_32( 4 ) );
|
||||
n += 4;
|
||||
} while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,6 +1,9 @@
|
||||
#include <memory.h>
|
||||
#include <mm_malloc.h>
|
||||
#include "lyra2-gate.h"
|
||||
|
||||
#if !( defined(LYRA2Z_16WAY) || defined(LYRA2Z_8WAY) || defined(LYRA2Z_4WAY) )
|
||||
|
||||
#include "lyra2.h"
|
||||
#include "algo/blake/sph_blake.h"
|
||||
#include "simd-utils.h"
|
||||
@@ -80,4 +83,4 @@ int scanhash_lyra2z( struct work *work, uint32_t max_nonce,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -96,32 +96,30 @@ int scanhash_phi2( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t _ALIGN(128) hash[8];
|
||||
uint32_t _ALIGN(128) endiandata[36];
|
||||
uint32_t _ALIGN(128) edata[36];
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
uint32_t n = first_nonce;
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
|
||||
if(opt_benchmark){
|
||||
ptarget[7] = 0x00ff;
|
||||
}
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
if( bench ) ptarget[7] = 0x00ff;
|
||||
|
||||
phi2_has_roots = false;
|
||||
for ( int i=0; i < 36; i++ )
|
||||
|
||||
for ( int i = 0; i < 36; i++ )
|
||||
{
|
||||
be32enc(&endiandata[i], pdata[i]);
|
||||
be32enc( &edata[i], pdata[i] );
|
||||
if ( i >= 20 && pdata[i] ) phi2_has_roots = true;
|
||||
}
|
||||
|
||||
do {
|
||||
be32enc( &endiandata[19], n );
|
||||
phi2_hash( hash, endiandata );
|
||||
if ( hash[7] < Htarg )
|
||||
if ( fulltest( hash, ptarget ) && !opt_benchmark )
|
||||
edata[19] = n;
|
||||
phi2_hash( hash, edata );
|
||||
if ( valid_hash( hash, ptarget ) && !opt_benchmark )
|
||||
{
|
||||
pdata[19] = n;
|
||||
be32enc( pdata+19, n );
|
||||
submit_solution( work, hash, mythr );
|
||||
}
|
||||
n++;
|
||||
|
||||
@@ -89,6 +89,9 @@ inline void initState( uint64_t State[/*16*/] )
|
||||
*
|
||||
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's G function
|
||||
*/
|
||||
|
||||
#if !defined(__AVX512F__) && !defined(__AVX2__) && !defined(__SSE2__)
|
||||
|
||||
inline static void blake2bLyra( uint64_t *v )
|
||||
{
|
||||
ROUND_LYRA(0);
|
||||
@@ -114,6 +117,8 @@ inline static void reducedBlake2bLyra( uint64_t *v )
|
||||
ROUND_LYRA(0);
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
/**
|
||||
* Performs a squeeze operation, using Blake2b's G function as the
|
||||
* internal permutation
|
||||
|
||||
@@ -171,7 +171,6 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
|
||||
LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \
|
||||
LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7)
|
||||
|
||||
|
||||
#endif // AVX2 else SSE2
|
||||
|
||||
// Scalar
|
||||
@@ -200,7 +199,6 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
|
||||
G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
|
||||
G(r,7,v[ 3],v[ 4],v[ 9],v[14]);
|
||||
|
||||
|
||||
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
|
||||
|
||||
union _ovly_512
|
||||
|
||||
@@ -267,8 +267,13 @@ int scanhash_m7m_hash( struct work* work, uint64_t max_nonce,
|
||||
SHA256_Final( (unsigned char*) hash, &ctxf_sha256 );
|
||||
}
|
||||
|
||||
if ( unlikely( hash[7] <= ptarget[7] ) )
|
||||
if ( likely( fulltest( hash, ptarget ) && !opt_benchmark ) )
|
||||
|
||||
if ( unlikely( valid_hash( (uint64_t*)hash, (uint64_t*)ptarget )
|
||||
&& !opt_benchmark ) )
|
||||
|
||||
|
||||
// if ( unlikely( hash[7] <= ptarget[7] ) )
|
||||
// if ( likely( fulltest( hash, ptarget ) && !opt_benchmark ) )
|
||||
{
|
||||
if ( opt_debug )
|
||||
{
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "nist5-gate.h"
|
||||
|
||||
#if !defined(NIST5_8WAY) && !defined(NIST5_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -105,13 +108,4 @@ int scanhash_nist5( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
/*
|
||||
bool register_nist5_algo( algo_gate_t* gate )
|
||||
{
|
||||
gate->optimizations = SSE2_OPT | AES_OPT;
|
||||
init_nist5_ctx();
|
||||
gate->scanhash = (void*)&scanhash_nist5;
|
||||
gate->hash = (void*)&nist5hash;
|
||||
return true;
|
||||
};
|
||||
*/
|
||||
#endif
|
||||
|
||||
@@ -154,14 +154,13 @@ int scanhash_zr5( struct work *work, uint32_t max_nonce,
|
||||
}
|
||||
|
||||
void zr5_get_new_work( struct work* work, struct work* g_work, int thr_id,
|
||||
uint32_t* end_nonce_ptr, bool clean_job )
|
||||
uint32_t* end_nonce_ptr )
|
||||
{
|
||||
// ignore POK in first word
|
||||
// const int nonce_i = 19;
|
||||
const int wkcmp_sz = 72; // (19-1) * sizeof(uint32_t)
|
||||
uint32_t *nonceptr = algo_gate.get_nonceptr( work->data );
|
||||
if ( memcmp( &work->data[1], &g_work->data[1], wkcmp_sz )
|
||||
&& ( clean_job || ( *nonceptr >= *end_nonce_ptr ) ) )
|
||||
|| ( *nonceptr >= *end_nonce_ptr ) )
|
||||
{
|
||||
work_free( work );
|
||||
work_copy( work, g_work );
|
||||
|
||||
@@ -1,5 +1,8 @@
|
||||
#include "cpuminer-config.h"
|
||||
#include "anime-gate.h"
|
||||
|
||||
#if !defined(ANIME_8WAY) && !defined(ANIME_4WAY)
|
||||
|
||||
#include <stdio.h>
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
@@ -169,4 +172,4 @@ int scanhash_anime( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "hmq1725-gate.h"
|
||||
|
||||
#if !defined(HMQ1725_8WAY) && !defined(HMQ1725_4WAY)
|
||||
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
#include "algo/blake/sph_blake.h"
|
||||
@@ -7,10 +10,7 @@
|
||||
#include "algo/jh/sph_jh.h"
|
||||
#include "algo/keccak/sph_keccak.h"
|
||||
#include "algo/skein/sph_skein.h"
|
||||
#include "algo/luffa/sph_luffa.h"
|
||||
#include "algo/cubehash/sph_cubehash.h"
|
||||
#include "algo/shavite/sph_shavite.h"
|
||||
#include "algo/simd/sph_simd.h"
|
||||
#include "algo/echo/sph_echo.h"
|
||||
#include "algo/hamsi/sph_hamsi.h"
|
||||
#include "algo/fugue/sph_fugue.h"
|
||||
@@ -21,6 +21,9 @@
|
||||
#if defined(__AES__)
|
||||
#include "algo/groestl/aes_ni/hash-groestl.h"
|
||||
#include "algo/echo/aes_ni/hash_api.h"
|
||||
#else
|
||||
#include "algo/groestl/sph_groestl.h"
|
||||
#include "algo/echo/sph_echo.h"
|
||||
#endif
|
||||
#include "algo/luffa/luffa_for_sse2.h"
|
||||
#include "algo/cubehash/cubehash_sse2.h"
|
||||
@@ -392,3 +395,4 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -1,5 +1,8 @@
|
||||
#include "cpuminer-config.h"
|
||||
#include "quark-gate.h"
|
||||
|
||||
#if !defined(QUARK_8WAY) && !defined(QUARK_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -137,4 +140,4 @@ int scanhash_quark( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "deep-gate.h"
|
||||
|
||||
#if !defined(DEEP_8WAY) && !defined(DEEP_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -114,4 +117,4 @@ int scanhash_deep( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "qubit-gate.h"
|
||||
|
||||
#if !defined(QUBIT_8WAY) && !defined(QUBIT_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -126,4 +129,4 @@ int scanhash_qubit( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "lbry-gate.h"
|
||||
|
||||
#if !defined(LBRY_16WAY) && !defined(LBRY_8WAY) && !defined(LBRY_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -100,3 +103,4 @@ int scanhash_lbry( struct work *work, uint32_t max_nonce,
|
||||
pdata[27] = n;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -1,505 +0,0 @@
|
||||
/*
|
||||
* Copyright 2009 Colin Percival, 2011 ArtForz, 2011-2014 pooler, 2015 Jordan Earls
|
||||
* All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* 1. Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* 2. Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
|
||||
*/
|
||||
|
||||
#include "cpuminer-config.h"
|
||||
#include "algo-gate-api.h"
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
|
||||
#define BLOCK_HEADER_SIZE 80
|
||||
|
||||
// windows
|
||||
#ifndef htobe32
|
||||
#define htobe32(x) ((uint32_t)htonl((uint32_t)(x)))
|
||||
#endif
|
||||
|
||||
#ifdef _MSC_VER
|
||||
#define ROTL(a, b) _rotl(a,b)
|
||||
#define ROTR(a, b) _rotr(a,b)
|
||||
#else
|
||||
#define ROTL(a, b) (((a) << b) | ((a) >> (32 - b)))
|
||||
#define ROTR(a, b) ((a >> b) | (a << (32 - b)))
|
||||
#endif
|
||||
|
||||
#if defined(_MSC_VER) && defined(_M_X64)
|
||||
#define _VECTOR __vectorcall
|
||||
#include <intrin.h>
|
||||
//#include <emmintrin.h> //SSE2
|
||||
//#include <pmmintrin.h> //SSE3
|
||||
//#include <tmmintrin.h> //SSSE3
|
||||
//#include <smmintrin.h> //SSE4.1
|
||||
//#include <nmmintrin.h> //SSE4.2
|
||||
//#include <ammintrin.h> //SSE4A
|
||||
//#include <wmmintrin.h> //AES
|
||||
//#include <immintrin.h> //AVX
|
||||
#define OPT_COMPATIBLE
|
||||
#elif defined(__GNUC__) && defined(__x86_64__)
|
||||
#include <x86intrin.h>
|
||||
#define _VECTOR
|
||||
#endif
|
||||
|
||||
static __thread char *scratchbuf;
|
||||
|
||||
#ifdef OPT_COMPATIBLE
|
||||
static void _VECTOR xor_salsa8(__m128i B[4], const __m128i Bx[4], int i)
|
||||
{
|
||||
__m128i X0, X1, X2, X3;
|
||||
|
||||
if (i <= 128) {
|
||||
// a xor 0 = a
|
||||
X0 = B[0] = Bx[0];
|
||||
X1 = B[1] = Bx[1];
|
||||
X2 = B[2] = Bx[2];
|
||||
X3 = B[3] = Bx[3];
|
||||
} else {
|
||||
X0 = B[0] = _mm_xor_si128(B[0], Bx[0]);
|
||||
X1 = B[1] = _mm_xor_si128(B[1], Bx[1]);
|
||||
X2 = B[2] = _mm_xor_si128(B[2], Bx[2]);
|
||||
X3 = B[3] = _mm_xor_si128(B[3], Bx[3]);
|
||||
}
|
||||
|
||||
for (i = 0; i < 4; i++) {
|
||||
/* Operate on columns. */
|
||||
X1.m128i_u32[0] ^= ROTL(X0.m128i_u32[0] + X3.m128i_u32[0], 7);
|
||||
X2.m128i_u32[1] ^= ROTL(X1.m128i_u32[1] + X0.m128i_u32[1], 7);
|
||||
X3.m128i_u32[2] ^= ROTL(X2.m128i_u32[2] + X1.m128i_u32[2], 7);
|
||||
X0.m128i_u32[3] ^= ROTL(X3.m128i_u32[3] + X2.m128i_u32[3], 7);
|
||||
|
||||
X2.m128i_u32[0] ^= ROTL(X1.m128i_u32[0] + X0.m128i_u32[0], 9);
|
||||
X3.m128i_u32[1] ^= ROTL(X2.m128i_u32[1] + X1.m128i_u32[1], 9);
|
||||
X0.m128i_u32[2] ^= ROTL(X3.m128i_u32[2] + X2.m128i_u32[2], 9);
|
||||
X1.m128i_u32[3] ^= ROTL(X0.m128i_u32[3] + X3.m128i_u32[3], 9);
|
||||
|
||||
X3.m128i_u32[0] ^= ROTL(X2.m128i_u32[0] + X1.m128i_u32[0], 13);
|
||||
X0.m128i_u32[1] ^= ROTL(X3.m128i_u32[1] + X2.m128i_u32[1], 13);
|
||||
X1.m128i_u32[2] ^= ROTL(X0.m128i_u32[2] + X3.m128i_u32[2], 13);
|
||||
X2.m128i_u32[3] ^= ROTL(X1.m128i_u32[3] + X0.m128i_u32[3], 13);
|
||||
|
||||
X0.m128i_u32[0] ^= ROTL(X3.m128i_u32[0] + X2.m128i_u32[0], 18);
|
||||
X1.m128i_u32[1] ^= ROTL(X0.m128i_u32[1] + X3.m128i_u32[1], 18);
|
||||
X2.m128i_u32[2] ^= ROTL(X1.m128i_u32[2] + X0.m128i_u32[2], 18);
|
||||
X3.m128i_u32[3] ^= ROTL(X2.m128i_u32[3] + X1.m128i_u32[3], 18);
|
||||
|
||||
/* Operate on rows. */
|
||||
X0.m128i_u32[1] ^= ROTL(X0.m128i_u32[0] + X0.m128i_u32[3], 7); X1.m128i_u32[2] ^= ROTL(X1.m128i_u32[1] + X1.m128i_u32[0], 7);
|
||||
X2.m128i_u32[3] ^= ROTL(X2.m128i_u32[2] + X2.m128i_u32[1], 7); X3.m128i_u32[0] ^= ROTL(X3.m128i_u32[3] + X3.m128i_u32[2], 7);
|
||||
X0.m128i_u32[2] ^= ROTL(X0.m128i_u32[1] + X0.m128i_u32[0], 9); X1.m128i_u32[3] ^= ROTL(X1.m128i_u32[2] + X1.m128i_u32[1], 9);
|
||||
X2.m128i_u32[0] ^= ROTL(X2.m128i_u32[3] + X2.m128i_u32[2], 9); X3.m128i_u32[1] ^= ROTL(X3.m128i_u32[0] + X3.m128i_u32[3], 9);
|
||||
|
||||
X0.m128i_u32[3] ^= ROTL(X0.m128i_u32[2] + X0.m128i_u32[1], 13); X1.m128i_u32[0] ^= ROTL(X1.m128i_u32[3] + X1.m128i_u32[2], 13);
|
||||
X2.m128i_u32[1] ^= ROTL(X2.m128i_u32[0] + X2.m128i_u32[3], 13); X3.m128i_u32[2] ^= ROTL(X3.m128i_u32[1] + X3.m128i_u32[0], 13);
|
||||
X0.m128i_u32[0] ^= ROTL(X0.m128i_u32[3] + X0.m128i_u32[2], 18); X1.m128i_u32[1] ^= ROTL(X1.m128i_u32[0] + X1.m128i_u32[3], 18);
|
||||
X2.m128i_u32[2] ^= ROTL(X2.m128i_u32[1] + X2.m128i_u32[0], 18); X3.m128i_u32[3] ^= ROTL(X3.m128i_u32[2] + X3.m128i_u32[1], 18);
|
||||
}
|
||||
|
||||
B[0] = _mm_add_epi32(B[0], X0);
|
||||
B[1] = _mm_add_epi32(B[1], X1);
|
||||
B[2] = _mm_add_epi32(B[2], X2);
|
||||
B[3] = _mm_add_epi32(B[3], X3);
|
||||
}
|
||||
|
||||
#else
|
||||
|
||||
static inline void xor_salsa8(uint32_t B[16], const uint32_t Bx[16], int i)
|
||||
{
|
||||
uint32_t x00,x01,x02,x03,x04,x05,x06,x07,x08,x09,x10,x11,x12,x13,x14,x15;
|
||||
|
||||
if (i <= 128) {
|
||||
// a xor 0 = a
|
||||
x00 = B[ 0] = Bx[ 0]; x01 = B[ 1] = Bx[ 1]; x02 = B[ 2] = Bx[ 2]; x03 = B[ 3] = Bx[ 3];
|
||||
x04 = B[ 4] = Bx[ 4]; x05 = B[ 5] = Bx[ 5]; x06 = B[ 6] = Bx[ 6]; x07 = B[ 7] = Bx[ 7];
|
||||
x08 = B[ 8] = Bx[ 8]; x09 = B[ 9] = Bx[ 9]; x10 = B[10] = Bx[10]; x11 = B[11] = Bx[11];
|
||||
x12 = B[12] = Bx[12]; x13 = B[13] = Bx[13]; x14 = B[14] = Bx[14]; x15 = B[15] = Bx[15];
|
||||
} else {
|
||||
x00 = (B[ 0] ^= Bx[ 0]);
|
||||
x01 = (B[ 1] ^= Bx[ 1]);
|
||||
x02 = (B[ 2] ^= Bx[ 2]);
|
||||
x03 = (B[ 3] ^= Bx[ 3]);
|
||||
x04 = (B[ 4] ^= Bx[ 4]);
|
||||
x05 = (B[ 5] ^= Bx[ 5]);
|
||||
x06 = (B[ 6] ^= Bx[ 6]);
|
||||
x07 = (B[ 7] ^= Bx[ 7]);
|
||||
x08 = (B[ 8] ^= Bx[ 8]);
|
||||
x09 = (B[ 9] ^= Bx[ 9]);
|
||||
x10 = (B[10] ^= Bx[10]);
|
||||
x11 = (B[11] ^= Bx[11]);
|
||||
x12 = (B[12] ^= Bx[12]);
|
||||
x13 = (B[13] ^= Bx[13]);
|
||||
x14 = (B[14] ^= Bx[14]);
|
||||
x15 = (B[15] ^= Bx[15]);
|
||||
}
|
||||
|
||||
for (i = 0; i < 8; i += 2) {
|
||||
/* Operate on columns. */
|
||||
x04 ^= ROTL(x00 + x12, 7); x09 ^= ROTL(x05 + x01, 7);
|
||||
x14 ^= ROTL(x10 + x06, 7); x03 ^= ROTL(x15 + x11, 7);
|
||||
|
||||
x08 ^= ROTL(x04 + x00, 9); x13 ^= ROTL(x09 + x05, 9);
|
||||
x02 ^= ROTL(x14 + x10, 9); x07 ^= ROTL(x03 + x15, 9);
|
||||
|
||||
x12 ^= ROTL(x08 + x04, 13); x01 ^= ROTL(x13 + x09, 13);
|
||||
x06 ^= ROTL(x02 + x14, 13); x11 ^= ROTL(x07 + x03, 13);
|
||||
|
||||
x00 ^= ROTL(x12 + x08, 18); x05 ^= ROTL(x01 + x13, 18);
|
||||
x10 ^= ROTL(x06 + x02, 18); x15 ^= ROTL(x11 + x07, 18);
|
||||
|
||||
/* Operate on rows. */
|
||||
x01 ^= ROTL(x00 + x03, 7); x06 ^= ROTL(x05 + x04, 7);
|
||||
x11 ^= ROTL(x10 + x09, 7); x12 ^= ROTL(x15 + x14, 7);
|
||||
|
||||
x02 ^= ROTL(x01 + x00, 9); x07 ^= ROTL(x06 + x05, 9);
|
||||
x08 ^= ROTL(x11 + x10, 9); x13 ^= ROTL(x12 + x15, 9);
|
||||
|
||||
x03 ^= ROTL(x02 + x01, 13); x04 ^= ROTL(x07 + x06, 13);
|
||||
x09 ^= ROTL(x08 + x11, 13); x14 ^= ROTL(x13 + x12, 13);
|
||||
|
||||
x00 ^= ROTL(x03 + x02, 18); x05 ^= ROTL(x04 + x07, 18);
|
||||
x10 ^= ROTL(x09 + x08, 18); x15 ^= ROTL(x14 + x13, 18);
|
||||
}
|
||||
B[ 0] += x00;
|
||||
B[ 1] += x01;
|
||||
B[ 2] += x02;
|
||||
B[ 3] += x03;
|
||||
B[ 4] += x04;
|
||||
B[ 5] += x05;
|
||||
B[ 6] += x06;
|
||||
B[ 7] += x07;
|
||||
B[ 8] += x08;
|
||||
B[ 9] += x09;
|
||||
B[10] += x10;
|
||||
B[11] += x11;
|
||||
B[12] += x12;
|
||||
B[13] += x13;
|
||||
B[14] += x14;
|
||||
B[15] += x15;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
static const uint32_t sha256_k[64] = {
|
||||
0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
|
||||
0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
|
||||
0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
|
||||
0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
|
||||
0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
|
||||
0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
|
||||
0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
|
||||
0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
|
||||
0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
|
||||
0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
|
||||
0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
|
||||
0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
|
||||
0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
|
||||
0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
|
||||
0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
|
||||
0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
|
||||
};
|
||||
|
||||
/* Elementary functions used by SHA256 */
|
||||
#define Ch(x, y, z) ((x & (y ^ z)) ^ z)
|
||||
#define Maj(x, y, z) ((x & (y | z)) | (y & z))
|
||||
#define S0(x) (ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22))
|
||||
#define S1(x) (ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25))
|
||||
#define s0(x) (ROTR(x, 7) ^ ROTR(x, 18) ^ (x >> 3))
|
||||
#define s1(x) (ROTR(x, 17) ^ ROTR(x, 19) ^ (x >> 10))
|
||||
|
||||
/* SHA256 round function */
|
||||
#define RND(a, b, c, d, e, f, g, h, k) \
|
||||
do { \
|
||||
t0 = h + S1(e) + Ch(e, f, g) + k; \
|
||||
t1 = S0(a) + Maj(a, b, c); \
|
||||
d += t0; \
|
||||
h = t0 + t1; \
|
||||
} while (0)
|
||||
|
||||
/* Adjusted round function for rotating state */
|
||||
#define RNDr(S, W, i) \
|
||||
RND(S[(64 - i) % 8], S[(65 - i) % 8], \
|
||||
S[(66 - i) % 8], S[(67 - i) % 8], \
|
||||
S[(68 - i) % 8], S[(69 - i) % 8], \
|
||||
S[(70 - i) % 8], S[(71 - i) % 8], \
|
||||
W[i] + sha256_k[i])
|
||||
|
||||
|
||||
static void sha256_transform_volatile(uint32_t *state, uint32_t *block)
|
||||
{
|
||||
uint32_t* W=block; //note: block needs to be a mutable 64 int32_t
|
||||
uint32_t S[8];
|
||||
uint32_t t0, t1;
|
||||
int i;
|
||||
|
||||
for (i = 16; i < 64; i += 2) {
|
||||
W[i] = s1(W[i - 2]) + W[i - 7] + s0(W[i - 15]) + W[i - 16];
|
||||
W[i+1] = s1(W[i - 1]) + W[i - 6] + s0(W[i - 14]) + W[i - 15];
|
||||
}
|
||||
|
||||
/* 2. Initialize working variables. */
|
||||
memcpy(S, state, 32);
|
||||
|
||||
/* 3. Mix. */
|
||||
RNDr(S, W, 0);
|
||||
RNDr(S, W, 1);
|
||||
RNDr(S, W, 2);
|
||||
RNDr(S, W, 3);
|
||||
RNDr(S, W, 4);
|
||||
RNDr(S, W, 5);
|
||||
RNDr(S, W, 6);
|
||||
RNDr(S, W, 7);
|
||||
RNDr(S, W, 8);
|
||||
RNDr(S, W, 9);
|
||||
RNDr(S, W, 10);
|
||||
RNDr(S, W, 11);
|
||||
RNDr(S, W, 12);
|
||||
RNDr(S, W, 13);
|
||||
RNDr(S, W, 14);
|
||||
RNDr(S, W, 15);
|
||||
RNDr(S, W, 16);
|
||||
RNDr(S, W, 17);
|
||||
RNDr(S, W, 18);
|
||||
RNDr(S, W, 19);
|
||||
RNDr(S, W, 20);
|
||||
RNDr(S, W, 21);
|
||||
RNDr(S, W, 22);
|
||||
RNDr(S, W, 23);
|
||||
RNDr(S, W, 24);
|
||||
RNDr(S, W, 25);
|
||||
RNDr(S, W, 26);
|
||||
RNDr(S, W, 27);
|
||||
RNDr(S, W, 28);
|
||||
RNDr(S, W, 29);
|
||||
RNDr(S, W, 30);
|
||||
RNDr(S, W, 31);
|
||||
RNDr(S, W, 32);
|
||||
RNDr(S, W, 33);
|
||||
RNDr(S, W, 34);
|
||||
RNDr(S, W, 35);
|
||||
RNDr(S, W, 36);
|
||||
RNDr(S, W, 37);
|
||||
RNDr(S, W, 38);
|
||||
RNDr(S, W, 39);
|
||||
RNDr(S, W, 40);
|
||||
RNDr(S, W, 41);
|
||||
RNDr(S, W, 42);
|
||||
RNDr(S, W, 43);
|
||||
RNDr(S, W, 44);
|
||||
RNDr(S, W, 45);
|
||||
RNDr(S, W, 46);
|
||||
RNDr(S, W, 47);
|
||||
RNDr(S, W, 48);
|
||||
RNDr(S, W, 49);
|
||||
RNDr(S, W, 50);
|
||||
RNDr(S, W, 51);
|
||||
RNDr(S, W, 52);
|
||||
RNDr(S, W, 53);
|
||||
RNDr(S, W, 54);
|
||||
RNDr(S, W, 55);
|
||||
RNDr(S, W, 56);
|
||||
RNDr(S, W, 57);
|
||||
RNDr(S, W, 58);
|
||||
RNDr(S, W, 59);
|
||||
RNDr(S, W, 60);
|
||||
RNDr(S, W, 61);
|
||||
RNDr(S, W, 62);
|
||||
RNDr(S, W, 63);
|
||||
|
||||
/* 4. Mix local working variables into global state */
|
||||
for (i = 0; i < 8; i++)
|
||||
state[i] += S[i];
|
||||
}
|
||||
|
||||
// standard sha256 hash
|
||||
#if 1
|
||||
static void sha256_hash(unsigned char *hash, const unsigned char *data, int len)
|
||||
{
|
||||
uint32_t _ALIGN(64) S[16];
|
||||
uint32_t _ALIGN(64) T[64];
|
||||
int i, r;
|
||||
|
||||
sha256_init(S);
|
||||
for (r = len; r > -9; r -= 64) {
|
||||
if (r < 64)
|
||||
memset(T, 0, 64);
|
||||
memcpy(T, data + len - r, r > 64 ? 64 : (r < 0 ? 0 : r));
|
||||
if (r >= 0 && r < 64)
|
||||
((unsigned char *)T)[r] = 0x80;
|
||||
for (i = 0; i < 16; i++)
|
||||
T[i] = be32dec(T + i);
|
||||
if (r < 56)
|
||||
T[15] = 8 * len;
|
||||
//sha256_transform(S, T, 0);
|
||||
sha256_transform_volatile(S, T);
|
||||
}
|
||||
for (i = 0; i < 8; i++)
|
||||
be32enc((uint32_t *)hash + i, S[i]);
|
||||
}
|
||||
#else
|
||||
#include <openssl/sha.h>
|
||||
static void sha256_hash(unsigned char *hash, const unsigned char *data, int len)
|
||||
{
|
||||
SHA256_CTX ctx;
|
||||
SHA256_Init(&ctx);
|
||||
SHA256_Update(&ctx, data, len);
|
||||
SHA256_Final(hash, &ctx);
|
||||
}
|
||||
#endif
|
||||
|
||||
// hash exactly 64 bytes (ie, sha256 block size)
|
||||
static void sha256_hash512(uint32_t *hash, const uint32_t *data)
|
||||
{
|
||||
uint32_t _ALIGN(64) S[16];
|
||||
uint32_t _ALIGN(64) T[64];
|
||||
uchar _ALIGN(64) E[64*4] = { 0 };
|
||||
int i;
|
||||
|
||||
sha256_init(S);
|
||||
|
||||
for (i = 0; i < 16; i++)
|
||||
T[i] = be32dec(&data[i]);
|
||||
sha256_transform_volatile(S, T);
|
||||
|
||||
E[3] = 0x80;
|
||||
E[61] = 0x02; // T[15] = 8 * 64 => 0x200;
|
||||
sha256_transform_volatile(S, (uint32_t*)E);
|
||||
|
||||
for (i = 0; i < 8; i++)
|
||||
be32enc(&hash[i], S[i]);
|
||||
}
|
||||
|
||||
void pluck_hash(uint32_t *hash, const uint32_t *data, uchar *hashbuffer, const int N)
|
||||
{
|
||||
int size = N * 1024;
|
||||
sha256_hash(hashbuffer, (void*)data, BLOCK_HEADER_SIZE);
|
||||
memset(&hashbuffer[32], 0, 32);
|
||||
|
||||
for(int i = 64; i < size - 32; i += 32)
|
||||
{
|
||||
uint32_t _ALIGN(64) randseed[16];
|
||||
uint32_t _ALIGN(64) randbuffer[16];
|
||||
uint32_t _ALIGN(64) joint[16];
|
||||
//i-4 because we use integers for all references against this, and we don't want to go 3 bytes over the defined area
|
||||
//we could use size here, but then it's probable to use 0 as the value in most cases
|
||||
int randmax = i - 4;
|
||||
|
||||
//setup randbuffer to be an array of random indexes
|
||||
memcpy(randseed, &hashbuffer[i - 64], 64);
|
||||
|
||||
if(i > 128) memcpy(randbuffer, &hashbuffer[i - 128], 64);
|
||||
//else memset(randbuffer, 0, 64);
|
||||
|
||||
xor_salsa8((void*)randbuffer, (void*)randseed, i);
|
||||
memcpy(joint, &hashbuffer[i - 32], 32);
|
||||
|
||||
//use the last hash value as the seed
|
||||
for (int j = 32; j < 64; j += 4)
|
||||
{
|
||||
//every other time, change to next random index
|
||||
//randmax - 32 as otherwise we go beyond memory that's already been written to
|
||||
uint32_t rand = randbuffer[(j - 32) >> 2] % (randmax - 32);
|
||||
joint[j >> 2] = *((uint32_t *)&hashbuffer[rand]);
|
||||
}
|
||||
|
||||
sha256_hash512((uint32_t*) &hashbuffer[i], joint);
|
||||
|
||||
//setup randbuffer to be an array of random indexes
|
||||
//use last hash value and previous hash value(post-mixing)
|
||||
memcpy(randseed, &hashbuffer[i - 32], 64);
|
||||
|
||||
if(i > 128) memcpy(randbuffer, &hashbuffer[i - 128], 64);
|
||||
//else memset(randbuffer, 0, 64);
|
||||
|
||||
xor_salsa8((void*)randbuffer, (void*)randseed, i);
|
||||
|
||||
//use the last hash value as the seed
|
||||
for (int j = 0; j < 32; j += 2)
|
||||
{
|
||||
uint32_t rand = randbuffer[j >> 1] % randmax;
|
||||
*((uint32_t *)(hashbuffer + rand)) = *((uint32_t *)(hashbuffer + j + randmax));
|
||||
}
|
||||
}
|
||||
|
||||
memcpy(hash, hashbuffer, 32);
|
||||
}
|
||||
|
||||
int scanhash_pluck( struct work *work, uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t _ALIGN(64) endiandata[20];
|
||||
uint32_t _ALIGN(64) hash[8];
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
volatile uint8_t *restart = &(work_restart[thr_id].restart);
|
||||
uint32_t n = first_nonce;
|
||||
|
||||
|
||||
if (opt_benchmark)
|
||||
((uint32_t*)ptarget)[7] = 0x0ffff;
|
||||
|
||||
for (int i=0; i < 19; i++)
|
||||
be32enc(&endiandata[i], pdata[i]);
|
||||
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
do {
|
||||
//be32enc(&endiandata[19], n);
|
||||
endiandata[19] = n;
|
||||
pluck_hash(hash, endiandata, scratchbuf, opt_pluck_n);
|
||||
|
||||
if (hash[7] <= Htarg && fulltest(hash, ptarget))
|
||||
{
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = htobe32(endiandata[19]);
|
||||
return 1;
|
||||
}
|
||||
n++;
|
||||
} while (n < max_nonce && !(*restart));
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
bool pluck_miner_thread_init( int thr_id )
|
||||
{
|
||||
scratchbuf = malloc( 128 * 1024 );
|
||||
if ( scratchbuf )
|
||||
return true;
|
||||
applog( LOG_ERR, "Thread %u: Pluck buffer allocation failed", thr_id );
|
||||
return false;
|
||||
}
|
||||
|
||||
bool register_pluck_algo( algo_gate_t* gate )
|
||||
{
|
||||
algo_not_tested();
|
||||
gate->miner_thread_init = (void*)&pluck_miner_thread_init;
|
||||
gate->scanhash = (void*)&scanhash_pluck;
|
||||
gate->hash = (void*)&pluck_hash;
|
||||
opt_target_factor = 65536.0;
|
||||
return true;
|
||||
};
|
||||
|
||||
|
||||
@@ -28,46 +28,10 @@
|
||||
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
|
||||
#include "sysendian.h"
|
||||
|
||||
#include "sha256_p.h"
|
||||
#include "simd-utils.h"
|
||||
#include "hmac-sha256-hash.h"
|
||||
#include "compat.h"
|
||||
|
||||
|
||||
/* Elementary functions used by SHA256 */
|
||||
#define Ch(x, y, z) ((x & (y ^ z)) ^ z)
|
||||
#define Maj(x, y, z) ((x & (y | z)) | (y & z))
|
||||
#define SHR(x, n) (x >> n)
|
||||
#define ROTR(x, n) ((x >> n) | (x << (32 - n)))
|
||||
#define S0(x) (ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22))
|
||||
#define S1(x) (ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25))
|
||||
#define s0(x) (ROTR(x, 7) ^ ROTR(x, 18) ^ SHR(x, 3))
|
||||
#define s1(x) (ROTR(x, 17) ^ ROTR(x, 19) ^ SHR(x, 10))
|
||||
|
||||
/* SHA256 round function */
|
||||
#define RND(a, b, c, d, e, f, g, h, k) \
|
||||
t0 = h + S1(e) + Ch(e, f, g) + k; \
|
||||
t1 = S0(a) + Maj(a, b, c); \
|
||||
d += t0; \
|
||||
h = t0 + t1;
|
||||
|
||||
/* Adjusted round function for rotating state */
|
||||
#define RNDr(S, W, i, k) \
|
||||
RND(S[(64 - i) % 8], S[(65 - i) % 8], \
|
||||
S[(66 - i) % 8], S[(67 - i) % 8], \
|
||||
S[(68 - i) % 8], S[(69 - i) % 8], \
|
||||
S[(70 - i) % 8], S[(71 - i) % 8], \
|
||||
W[i] + k)
|
||||
|
||||
/*
|
||||
static unsigned char PAD[64] = {
|
||||
0x80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
||||
};
|
||||
*/
|
||||
/**
|
||||
* SHA256_Buf(in, len, digest):
|
||||
* Compute the SHA256 hash of ${len} bytes from ${in} and write it to ${digest}.
|
||||
@@ -76,9 +40,9 @@ void
|
||||
SHA256_Buf( const void * in, size_t len, uint8_t digest[32] )
|
||||
{
|
||||
SHA256_CTX ctx;
|
||||
SHA256_Init( &ctx );
|
||||
SHA256_Update( &ctx, in, len );
|
||||
SHA256_Final( digest, &ctx );
|
||||
SHA256_Init( &ctx );
|
||||
SHA256_Update( &ctx, in, len );
|
||||
SHA256_Final( digest, &ctx );
|
||||
}
|
||||
|
||||
/**
|
||||
@@ -87,19 +51,18 @@ SHA256_Buf( const void * in, size_t len, uint8_t digest[32] )
|
||||
* length ${Klen}, and write the result to ${digest}.
|
||||
*/
|
||||
void
|
||||
HMAC_SHA256_Buf(const void * K, size_t Klen, const void * in, size_t len,
|
||||
uint8_t digest[32])
|
||||
HMAC_SHA256_Buf( const void *K, size_t Klen, const void *in, size_t len,
|
||||
uint8_t digest[32])
|
||||
{
|
||||
HMAC_SHA256_CTX ctx;
|
||||
|
||||
HMAC_SHA256_Init( &ctx, K, Klen );
|
||||
HMAC_SHA256_Update( &ctx, in, len );
|
||||
HMAC_SHA256_Final( digest, &ctx );
|
||||
HMAC_SHA256_CTX ctx;
|
||||
HMAC_SHA256_Init( &ctx, K, Klen );
|
||||
HMAC_SHA256_Update( &ctx, in, len );
|
||||
HMAC_SHA256_Final( digest, &ctx );
|
||||
}
|
||||
|
||||
/* Initialize an HMAC-SHA256 operation with the given key. */
|
||||
void
|
||||
HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
|
||||
HMAC_SHA256_Init( HMAC_SHA256_CTX *ctx, const void *_K, size_t Klen )
|
||||
{
|
||||
unsigned char pad[64];
|
||||
unsigned char khash[32];
|
||||
@@ -107,7 +70,8 @@ HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
|
||||
size_t i;
|
||||
|
||||
/* If Klen > 64, the key is really SHA256(K). */
|
||||
if (Klen > 64) {
|
||||
if ( Klen > 64 )
|
||||
{
|
||||
SHA256_Init( &ctx->ictx );
|
||||
SHA256_Update( &ctx->ictx, K, Klen );
|
||||
SHA256_Final( khash, &ctx->ictx );
|
||||
@@ -116,7 +80,7 @@ HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
|
||||
}
|
||||
|
||||
/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
|
||||
SHA256_Init( &ctx->ictx );
|
||||
SHA256_Init( &ctx->ictx );
|
||||
memset( pad, 0x36, 64 );
|
||||
for ( i = 0; i < Klen; i++ )
|
||||
pad[i] ^= K[i];
|
||||
@@ -128,23 +92,19 @@ HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
|
||||
for ( i = 0; i < Klen; i++ )
|
||||
pad[i] ^= K[i];
|
||||
SHA256_Update( &ctx->octx, pad, 64 );
|
||||
|
||||
/* Clean the stack. */
|
||||
//memset(khash, 0, 32);
|
||||
}
|
||||
|
||||
/* Add bytes to the HMAC-SHA256 operation. */
|
||||
void
|
||||
HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void *in, size_t len)
|
||||
HMAC_SHA256_Update( HMAC_SHA256_CTX *ctx, const void *in, size_t len )
|
||||
{
|
||||
|
||||
/* Feed data to the inner SHA256 operation. */
|
||||
SHA256_Update( &ctx->ictx, in, len );
|
||||
}
|
||||
|
||||
/* Finish an HMAC-SHA256 operation. */
|
||||
void
|
||||
HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx )
|
||||
HMAC_SHA256_Final( unsigned char digest[32], HMAC_SHA256_CTX *ctx )
|
||||
{
|
||||
unsigned char ihash[32];
|
||||
|
||||
@@ -156,9 +116,6 @@ HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx )
|
||||
|
||||
/* Finish the outer SHA256 operation. */
|
||||
SHA256_Final( digest, &ctx->octx );
|
||||
|
||||
/* Clean the stack. */
|
||||
//memset(ihash, 0, 32);
|
||||
}
|
||||
|
||||
/**
|
||||
@@ -167,52 +124,51 @@ HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx )
|
||||
* write the output to buf. The value dkLen must be at most 32 * (2^32 - 1).
|
||||
*/
|
||||
void
|
||||
PBKDF2_SHA256(const uint8_t * passwd, size_t passwdlen, const uint8_t * salt,
|
||||
size_t saltlen, uint64_t c, uint8_t * buf, size_t dkLen)
|
||||
PBKDF2_SHA256( const uint8_t *passwd, size_t passwdlen, const uint8_t *salt,
|
||||
size_t saltlen, uint64_t c, uint8_t *buf, size_t dkLen )
|
||||
{
|
||||
HMAC_SHA256_CTX PShctx, hctx;
|
||||
uint8_t _ALIGN(128) T[32];
|
||||
uint8_t _ALIGN(128) U[32];
|
||||
uint8_t ivec[4];
|
||||
uint32_t ivec;
|
||||
size_t i, clen;
|
||||
uint64_t j;
|
||||
int k;
|
||||
|
||||
/* Compute HMAC state after processing P and S. */
|
||||
HMAC_SHA256_Init(&PShctx, passwd, passwdlen);
|
||||
HMAC_SHA256_Update(&PShctx, salt, saltlen);
|
||||
HMAC_SHA256_Init( &PShctx, passwd, passwdlen );
|
||||
HMAC_SHA256_Update( &PShctx, salt, saltlen );
|
||||
|
||||
/* Iterate through the blocks. */
|
||||
for (i = 0; i * 32 < dkLen; i++) {
|
||||
for ( i = 0; i * 32 < dkLen; i++ )
|
||||
{
|
||||
/* Generate INT(i + 1). */
|
||||
be32enc(ivec, (uint32_t)(i + 1));
|
||||
ivec = bswap_32( i+1 );
|
||||
|
||||
/* Compute U_1 = PRF(P, S || INT(i)). */
|
||||
memcpy(&hctx, &PShctx, sizeof(HMAC_SHA256_CTX));
|
||||
HMAC_SHA256_Update(&hctx, ivec, 4);
|
||||
HMAC_SHA256_Final(U, &hctx);
|
||||
memcpy( &hctx, &PShctx, sizeof(HMAC_SHA256_CTX) );
|
||||
HMAC_SHA256_Update( &hctx, &ivec, 4 );
|
||||
HMAC_SHA256_Final( U, &hctx );
|
||||
|
||||
/* T_i = U_1 ... */
|
||||
memcpy(T, U, 32);
|
||||
memcpy( T, U, 32 );
|
||||
|
||||
for (j = 2; j <= c; j++) {
|
||||
for ( j = 2; j <= c; j++ )
|
||||
{
|
||||
/* Compute U_j. */
|
||||
HMAC_SHA256_Init(&hctx, passwd, passwdlen);
|
||||
HMAC_SHA256_Update(&hctx, U, 32);
|
||||
HMAC_SHA256_Final(U, &hctx);
|
||||
HMAC_SHA256_Init( &hctx, passwd, passwdlen );
|
||||
HMAC_SHA256_Update( &hctx, U, 32 );
|
||||
HMAC_SHA256_Final( U, &hctx );
|
||||
|
||||
/* ... xor U_j ... */
|
||||
for (k = 0; k < 32; k++)
|
||||
for ( k = 0; k < 32; k++ )
|
||||
T[k] ^= U[k];
|
||||
}
|
||||
|
||||
/* Copy as many bytes as necessary into buf. */
|
||||
clen = dkLen - i * 32;
|
||||
if (clen > 32)
|
||||
if ( clen > 32 )
|
||||
clen = 32;
|
||||
memcpy(&buf[i * 32], T, clen);
|
||||
memcpy( &buf[i * 32], T, clen );
|
||||
}
|
||||
|
||||
/* Clean PShctx, since we never called _Final on it. */
|
||||
//memset(&PShctx, 0, sizeof(HMAC_SHA256_CTX_Y));
|
||||
}
|
||||
@@ -26,23 +26,24 @@
|
||||
* $FreeBSD: src/lib/libmd/sha256_Y.h,v 1.2 2006/01/17 15:35:56 phk Exp $
|
||||
*/
|
||||
|
||||
#ifndef _SHA256_H_
|
||||
#define _SHA256_H_
|
||||
#ifndef HMAC_SHA256_H__
|
||||
#define HMAC_SHA256_H__
|
||||
|
||||
#include <sys/types.h>
|
||||
#include <stdint.h>
|
||||
#include <openssl/sha.h>
|
||||
|
||||
typedef struct HMAC_SHA256Context {
|
||||
SHA256_CTX ictx;
|
||||
SHA256_CTX octx;
|
||||
typedef struct HMAC_SHA256Context
|
||||
{
|
||||
SHA256_CTX ictx;
|
||||
SHA256_CTX octx;
|
||||
} HMAC_SHA256_CTX;
|
||||
|
||||
void SHA256_Buf( const void * in, size_t len, uint8_t digest[32] );
|
||||
void SHA256_Buf( const void *, size_t len, uint8_t digest[32] );
|
||||
void HMAC_SHA256_Init( HMAC_SHA256_CTX *, const void *, size_t );
|
||||
void HMAC_SHA256_Update( HMAC_SHA256_CTX *, const void *, size_t );
|
||||
void HMAC_SHA256_Final( unsigned char [32], HMAC_SHA256_CTX * );
|
||||
void HMAC_SHA256_Buf( const void * K, size_t Klen, const void * in,
|
||||
void HMAC_SHA256_Buf( const void *, size_t Klen, const void *,
|
||||
size_t len, uint8_t digest[32] );
|
||||
|
||||
/**
|
||||
@@ -53,4 +54,4 @@ void HMAC_SHA256_Buf( const void * K, size_t Klen, const void * in,
|
||||
void PBKDF2_SHA256( const uint8_t *, size_t, const uint8_t *, size_t,
|
||||
uint64_t, uint8_t *, size_t);
|
||||
|
||||
#endif /* !_SHA256_H_ */
|
||||
#endif // HMAC_SHA256_H__
|
||||
@@ -94,6 +94,37 @@ static const uint32_t K256[64] =
|
||||
_mm_xor_si128( _mm_xor_si128( \
|
||||
mm128_ror_32(x, 17), mm128_ror_32(x, 19) ), _mm_srli_epi32(x, 10) )
|
||||
|
||||
#define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
|
||||
do { \
|
||||
__m128i K = _mm_set1_epi32( K256[( (j)+(i) )] ); \
|
||||
__m128i T1 = mm128_ror_32( E, 14 ); \
|
||||
__m128i T2 = mm128_ror_32( A, 9 ); \
|
||||
__m128i T3 = _mm_xor_si128( F, G ); \
|
||||
__m128i T4 = _mm_or_si128( A, B ); \
|
||||
__m128i T5 = _mm_and_si128( A, B ); \
|
||||
K = _mm_add_epi32( K, W[i] ); \
|
||||
T1 = _mm_xor_si128( T1, E ); \
|
||||
T2 = _mm_xor_si128( T2, A ); \
|
||||
T3 = _mm_and_si128( T3, E ); \
|
||||
T4 = _mm_and_si128( T4, C ); \
|
||||
K = _mm_add_epi32( H, K ); \
|
||||
T1 = mm128_ror_32( T1, 5 ); \
|
||||
T2 = mm128_ror_32( T2, 11 ); \
|
||||
T3 = _mm_xor_si128( T3, G ); \
|
||||
T4 = _mm_or_si128( T4, T5 ); \
|
||||
T1 = _mm_xor_si128( T1, E ); \
|
||||
T2 = _mm_xor_si128( T2, A ); \
|
||||
T1 = mm128_ror_32( T1, 6 ); \
|
||||
T2 = mm128_ror_32( T2, 2 ); \
|
||||
T1 = _mm_add_epi32( T1, T3 ); \
|
||||
T2 = _mm_add_epi32( T2, T4 ); \
|
||||
T1 = _mm_add_epi32( T1, K ); \
|
||||
H = _mm_add_epi32( T1, T2 ); \
|
||||
D = _mm_add_epi32( D, T1 ); \
|
||||
} while (0)
|
||||
|
||||
|
||||
/*
|
||||
#define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
|
||||
do { \
|
||||
__m128i T1, T2; \
|
||||
@@ -104,6 +135,8 @@ do { \
|
||||
D = _mm_add_epi32( D, T1 ); \
|
||||
H = _mm_add_epi32( T1, T2 ); \
|
||||
} while (0)
|
||||
*/
|
||||
|
||||
|
||||
static void
|
||||
sha256_4way_round( sha256_4way_context *ctx, __m128i *in, __m128i r[8] )
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "sha256t-gate.h"
|
||||
|
||||
#if !defined(SHA256T_16WAY) && !defined(SHA256T_8WAY) && !defined(SHA256T_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -102,3 +105,4 @@ int scanhash_sha256q( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "sha256t-gate.h"
|
||||
|
||||
#if !defined(SHA256T_16WAY) && !defined(SHA256T_8WAY) && !defined(SHA256T_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -98,3 +101,5 @@ int scanhash_sha256t( struct work *work, uint32_t max_nonce,
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
@@ -319,7 +319,7 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
|
||||
|
||||
// SHA-512 4 way 64 bit
|
||||
|
||||
|
||||
/*
|
||||
#define CH(X, Y, Z) \
|
||||
_mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( Y, Z ), X ), Z )
|
||||
|
||||
@@ -327,6 +327,15 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
|
||||
_mm256_or_si256( _mm256_and_si256( X, Y ), \
|
||||
_mm256_and_si256( _mm256_or_si256( X, Y ), Z ) )
|
||||
|
||||
#define BSG5_0(x) \
|
||||
mm256_ror_64( _mm256_xor_si256( mm256_ror_64( \
|
||||
_mm256_xor_si256( mm256_ror_64( x, 5 ), x ), 6 ), x ), 28 )
|
||||
|
||||
#define BSG5_1(x) \
|
||||
mm256_ror_64( _mm256_xor_si256( mm256_ror_64( \
|
||||
_mm256_xor_si256( mm256_ror_64( x, 23 ), x ), 4 ), x ), 14 )
|
||||
*/
|
||||
/*
|
||||
#define BSG5_0(x) \
|
||||
_mm256_xor_si256( _mm256_xor_si256( \
|
||||
mm256_ror_64(x, 28), mm256_ror_64(x, 34) ), mm256_ror_64(x, 39) )
|
||||
@@ -334,7 +343,8 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
|
||||
#define BSG5_1(x) \
|
||||
_mm256_xor_si256( _mm256_xor_si256( \
|
||||
mm256_ror_64(x, 14), mm256_ror_64(x, 18) ), mm256_ror_64(x, 41) )
|
||||
|
||||
*/
|
||||
/*
|
||||
#define SSG5_0(x) \
|
||||
_mm256_xor_si256( _mm256_xor_si256( \
|
||||
mm256_ror_64(x, 1), mm256_ror_64(x, 8) ), _mm256_srli_epi64(x, 7) )
|
||||
@@ -342,7 +352,7 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
|
||||
#define SSG5_1(x) \
|
||||
_mm256_xor_si256( _mm256_xor_si256( \
|
||||
mm256_ror_64(x, 19), mm256_ror_64(x, 61) ), _mm256_srli_epi64(x, 6) )
|
||||
|
||||
*/
|
||||
// Interleave SSG0 & SSG1 for better throughput.
|
||||
// return ssg0(w0) + ssg1(w1)
|
||||
static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
|
||||
@@ -361,7 +371,7 @@ static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
|
||||
return _mm256_add_epi64( w0a, w1a );
|
||||
}
|
||||
|
||||
|
||||
/*
|
||||
#define SSG512x2_0( w0, w1, i ) do \
|
||||
{ \
|
||||
__m256i X0a, X1a, X0b, X1b; \
|
||||
@@ -391,7 +401,51 @@ static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
|
||||
w0 = _mm256_xor_si256( X0a, X0b ); \
|
||||
w1 = _mm256_xor_si256( X1a, X1b ); \
|
||||
} while(0)
|
||||
*/
|
||||
|
||||
#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
|
||||
do { \
|
||||
__m256i K = _mm256_set1_epi64x( K512[ i ] ); \
|
||||
__m256i T1 = mm256_ror_64( E, 23 ); \
|
||||
__m256i T2 = mm256_ror_64( A, 5 ); \
|
||||
__m256i T3 = _mm256_xor_si256( F, G ); \
|
||||
__m256i T4 = _mm256_or_si256( A, B ); \
|
||||
__m256i T5 = _mm256_and_si256( A, B ); \
|
||||
K = _mm256_add_epi64( K, W[i] ); \
|
||||
T1 = _mm256_xor_si256( T1, E ); \
|
||||
T2 = _mm256_xor_si256( T2, A ); \
|
||||
T3 = _mm256_and_si256( T3, E ); \
|
||||
T4 = _mm256_and_si256( T4, C ); \
|
||||
K = _mm256_add_epi64( H, K ); \
|
||||
T1 = mm256_ror_64( T1, 4 ); \
|
||||
T2 = mm256_ror_64( T2, 6 ); \
|
||||
T3 = _mm256_xor_si256( T3, G ); \
|
||||
T4 = _mm256_or_si256( T4, T5 ); \
|
||||
T1 = _mm256_xor_si256( T1, E ); \
|
||||
T2 = _mm256_xor_si256( T2, A ); \
|
||||
T1 = mm256_ror_64( T1, 14 ); \
|
||||
T2 = mm256_ror_64( T2, 28 ); \
|
||||
T1 = _mm256_add_epi64( T1, T3 ); \
|
||||
T2 = _mm256_add_epi64( T2, T4 ); \
|
||||
T1 = _mm256_add_epi64( T1, K ); \
|
||||
H = _mm256_add_epi64( T1, T2 ); \
|
||||
D = _mm256_add_epi64( D, T1 ); \
|
||||
} while (0)
|
||||
|
||||
/*
|
||||
#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
|
||||
do { \
|
||||
__m256i K = _mm256_add_epi64( W[i], _mm256_set1_epi64x( K512[ i ] ) ); \
|
||||
__m256i T1 = BSG5_1(E); \
|
||||
__m256i T2 = BSG5_0(A); \
|
||||
T1 = mm256_add4_64( T1, H, CH(E, F, G), K ); \
|
||||
T2 = _mm256_add_epi64( T2, MAJ(A, B, C) ); \
|
||||
D = _mm256_add_epi64( D, T1 ); \
|
||||
H = _mm256_add_epi64( T1, T2 ); \
|
||||
} while (0)
|
||||
*/
|
||||
|
||||
/*
|
||||
#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
|
||||
do { \
|
||||
__m256i T1, T2; \
|
||||
@@ -402,7 +456,7 @@ do { \
|
||||
D = _mm256_add_epi64( D, T1 ); \
|
||||
H = _mm256_add_epi64( T1, T2 ); \
|
||||
} while (0)
|
||||
|
||||
*/
|
||||
|
||||
static void
|
||||
sha512_4way_round( sha512_4way_context *ctx, __m256i *in, __m256i r[8] )
|
||||
|
||||
124
algo/simd/nist.c
124
algo/simd/nist.c
@@ -360,18 +360,116 @@ HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
|
||||
return SUCCESS;
|
||||
}
|
||||
|
||||
int simd_full( hashState_sd *state, BitSequence *hashval,
|
||||
const BitSequence *data, DataLength databitlen )
|
||||
{
|
||||
|
||||
|
||||
/*HashReturn Hash(int hashbitlen, const BitSequence *data, DataLength databitlen,
|
||||
BitSequence *hashval) {
|
||||
hashState_sd s;
|
||||
HashReturn r;
|
||||
r = Init(&s, hashbitlen);
|
||||
if (r != SUCCESS)
|
||||
return r;
|
||||
r = Update(&s, data, databitlen);
|
||||
if (r != SUCCESS)
|
||||
return r;
|
||||
r = Final(&s, hashval);
|
||||
return r;
|
||||
InitIV( state, 512, IV_512 );
|
||||
|
||||
int current, i;
|
||||
unsigned int bs = state->blocksize;
|
||||
static int align = -1;
|
||||
BitSequence out[64];
|
||||
int isshort = 1;
|
||||
u64 l;
|
||||
|
||||
if (align == -1)
|
||||
align = RequiredAlignment();
|
||||
|
||||
#ifdef HAS_64
|
||||
current = state->count & (bs - 1);
|
||||
#else
|
||||
current = state->count_low & (bs - 1);
|
||||
#endif
|
||||
|
||||
if ( current & 7 )
|
||||
{
|
||||
// The number of hashed bits is not a multiple of 8.
|
||||
// Very painfull to implement and not required by the NIST API.
|
||||
return FAIL;
|
||||
}
|
||||
|
||||
while ( databitlen > 0 )
|
||||
{
|
||||
if ( IS_ALIGNED(data,align) && current == 0 && databitlen >= bs )
|
||||
{
|
||||
// We can hash the data directly from the input buffer.
|
||||
SIMD_Compress(state, data, 0);
|
||||
databitlen -= bs;
|
||||
data += bs/8;
|
||||
IncreaseCounter(state, bs);
|
||||
}
|
||||
else
|
||||
{
|
||||
// Copy a chunk of data to the buffer
|
||||
unsigned int len = bs - current;
|
||||
if ( databitlen < len )
|
||||
{
|
||||
memcpy( state->buffer+current/8, data, (databitlen+7)/8 );
|
||||
IncreaseCounter( state, databitlen );
|
||||
break;
|
||||
}
|
||||
else
|
||||
{
|
||||
memcpy( state->buffer+current/8, data, len/8 );
|
||||
IncreaseCounter( state,len );
|
||||
databitlen -= len;
|
||||
data += len/8;
|
||||
current = 0;
|
||||
SIMD_Compress( state, state->buffer, 0 );
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
current = state->count & (state->blocksize - 1);
|
||||
|
||||
// If there is still some data in the buffer, hash it
|
||||
if ( current )
|
||||
{
|
||||
// We first need to zero out the end of the buffer.
|
||||
if ( current & 7 )
|
||||
{
|
||||
BitSequence mask = 0xff >> ( current & 7 );
|
||||
state->buffer[current/8] &= ~mask;
|
||||
}
|
||||
current = ( current+7 ) / 8;
|
||||
memset( state->buffer+current, 0, state->blocksize/8 - current );
|
||||
SIMD_Compress( state, state->buffer, 0 );
|
||||
}
|
||||
|
||||
//* Input the message length as the last block
|
||||
memset( state->buffer, 0, state->blocksize / 8 );
|
||||
l = state->count;
|
||||
for ( i=0; i<8; i++ )
|
||||
{
|
||||
state->buffer[i] = l & 0xff;
|
||||
l >>= 8;
|
||||
}
|
||||
if ( state->count < 16384 )
|
||||
isshort = 2;
|
||||
|
||||
SIMD_Compress( state, state->buffer, isshort );
|
||||
|
||||
// Decode the 32-bit words into a BitSequence
|
||||
for ( i=0; i < 2*state->n_feistels; i++ )
|
||||
{
|
||||
u32 x = state->A[i];
|
||||
out[4*i ] = x & 0xff;
|
||||
x >>= 8;
|
||||
out[4*i+1] = x & 0xff;
|
||||
x >>= 8;
|
||||
out[4*i+2] = x & 0xff;
|
||||
x >>= 8;
|
||||
out[4*i+3] = x & 0xff;
|
||||
}
|
||||
|
||||
memcpy( hashval, out, state->hashbitlen / 8 );
|
||||
if ( state->hashbitlen % 8 )
|
||||
{
|
||||
BitSequence mask = 0xff << ( 8 - (state->hashbitlen % 8) );
|
||||
hashval[state->hashbitlen/8 + 1] = out[state->hashbitlen/8 + 1] & mask;
|
||||
}
|
||||
return SUCCESS;
|
||||
}
|
||||
*/
|
||||
|
||||
|
||||
@@ -47,8 +47,8 @@ HashReturn final_sd(hashState_sd *state, BitSequence *hashval);
|
||||
HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
|
||||
const BitSequence *data, DataLength databitlen );
|
||||
|
||||
//HashReturn Hash(int hashbitlen, const BitSequence *data, DataLength databitlen,
|
||||
// BitSequence *hashval);
|
||||
int simd_full( hashState_sd *state, BitSequence *hashval,
|
||||
const BitSequence *data, DataLength databitlen );
|
||||
|
||||
/*
|
||||
* Internal API
|
||||
|
||||
@@ -727,7 +727,7 @@ skein_big_core_4way( skein512_4way_context *sc, const void *data,
|
||||
{
|
||||
memcpy_256( buf + (ptr>>3), vdata, len>>3 );
|
||||
sc->ptr = ptr + len;
|
||||
return;
|
||||
if ( ptr < buf_size ) return;
|
||||
}
|
||||
|
||||
READ_STATE_BIG( sc );
|
||||
@@ -745,6 +745,8 @@ skein_big_core_4way( skein512_4way_context *sc, const void *data,
|
||||
clen = buf_size - ptr;
|
||||
if ( clen > len )
|
||||
clen = len;
|
||||
len -= clen;
|
||||
if ( len == 0 ) break;
|
||||
memcpy_256( buf + (ptr>>3), vdata, clen>>3 );
|
||||
ptr += clen;
|
||||
vdata += (clen>>3);
|
||||
@@ -769,9 +771,12 @@ skein_big_close_4way( skein512_4way_context *sc, unsigned ub, unsigned n,
|
||||
|
||||
READ_STATE_BIG(sc);
|
||||
|
||||
memset_zero_256( buf + (ptr>>3), (buf_size - ptr) >> 3 );
|
||||
et = 352 + ((bcount == 0) << 7);
|
||||
UBI_BIG_4WAY( et, ptr );
|
||||
if ( ptr )
|
||||
{
|
||||
memset_zero_256( buf + (ptr>>3), (buf_size - ptr) >> 3 );
|
||||
et = 352 + ((bcount == 0) << 7);
|
||||
UBI_BIG_4WAY( et, ptr );
|
||||
}
|
||||
|
||||
memset_zero_256( buf, buf_size >> 3 );
|
||||
bcount = 0;
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "algo-gate-api.h"
|
||||
|
||||
#if !defined(SKEIN_8WAY) && !defined(SKEIN_4WAY)
|
||||
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
#include "sph_skein.h"
|
||||
@@ -52,4 +55,4 @@ int scanhash_skein( struct work *work, uint32_t max_nonce,
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "skein-gate.h"
|
||||
|
||||
#if !defined(SKEIN_8WAY) && !defined(SKEIN_4WAY)
|
||||
|
||||
#include <string.h>
|
||||
#include <stdint.h>
|
||||
|
||||
@@ -66,4 +69,4 @@ int scanhash_skein2( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
#endif
|
||||
|
||||
@@ -252,12 +252,6 @@ SPH_XCAT(HASH, _addbits_and_close)(void *cc,
|
||||
current = (unsigned)sc->count_low & (SPH_BLEN - 1U);
|
||||
#endif
|
||||
|
||||
//uint64_t *b= (uint64_t*)sc->buf;
|
||||
//uint64_t *s= (uint64_t*)sc->state;
|
||||
// printf("Sptr 1= %u\n",current);
|
||||
// printf("SBuf %016llx %016llx %016llx %016llx\n", b[0], b[1], b[2], b[3] );
|
||||
// printf("SBuf %016llx %016llx %016llx %016llx\n", b[4], b[5], b[6], b[7] );
|
||||
|
||||
#ifdef PW01
|
||||
sc->buf[current ++] = (0x100 | (ub & 0xFF)) >> (8 - n);
|
||||
#else
|
||||
@@ -269,10 +263,6 @@ SPH_XCAT(HASH, _addbits_and_close)(void *cc,
|
||||
}
|
||||
#endif
|
||||
|
||||
// printf("Sptr 2= %u\n",current);
|
||||
// printf("SBuf %016llx %016llx %016llx %016llx\n", b[0], b[1], b[2], b[3] );
|
||||
// printf("SBuf %016llx %016llx %016llx %016llx\n", b[4], b[5], b[6], b[7] );
|
||||
|
||||
if (current > SPH_MAXPAD) {
|
||||
memset(sc->buf + current, 0, SPH_BLEN - current);
|
||||
RFUN(sc->buf, SPH_VAL);
|
||||
@@ -333,16 +323,8 @@ SPH_XCAT(HASH, _addbits_and_close)(void *cc,
|
||||
#endif
|
||||
#endif
|
||||
|
||||
// printf("Sptr 3= %u\n",current);
|
||||
// printf("SBuf %016llx %016llx %016llx %016llx\n", b[0], b[1], b[2], b[3] );
|
||||
// printf("SBuf %016llx %016llx %016llx %016llx\n", b[4], b[5], b[6], b[7] );
|
||||
|
||||
RFUN(sc->buf, SPH_VAL);
|
||||
|
||||
// printf("Sptr after= %u\n",current);
|
||||
// printf("SState %016llx %016llx %016llx %016llx\n", s[0], s[1], s[2], s[3] );
|
||||
// printf("SState %016llx %016llx %016llx %016llx\n", s[4], s[5], s[6], s[7] );
|
||||
|
||||
#ifdef SPH_NO_OUTPUT
|
||||
(void)dst;
|
||||
(void)rnum;
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "c11-gate.h"
|
||||
|
||||
#if !defined(C11_8WAY) && !defined(C11_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -9,9 +12,6 @@
|
||||
#include "algo/keccak/sph_keccak.h"
|
||||
#include "algo/skein/sph_skein.h"
|
||||
#include "algo/shavite/sph_shavite.h"
|
||||
#include "algo/luffa/sph_luffa.h"
|
||||
#include "algo/cubehash/sph_cubehash.h"
|
||||
#include "algo/simd/sph_simd.h"
|
||||
#include "algo/luffa/luffa_for_sse2.h"
|
||||
#include "algo/cubehash/cubehash_sse2.h"
|
||||
#include "algo/simd/nist.h"
|
||||
@@ -149,3 +149,4 @@ int scanhash_c11( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
131
algo/x11/fresh.c
131
algo/x11/fresh.c
@@ -1,131 +0,0 @@
|
||||
#include "algo-gate-api.h"
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
#include <stdio.h>
|
||||
|
||||
#include "algo/shavite/sph_shavite.h"
|
||||
#include "algo/simd/sph_simd.h"
|
||||
#include "algo/echo/sph_echo.h"
|
||||
|
||||
//#define DEBUG_ALGO
|
||||
|
||||
extern void freshhash(void* output, const void* input, uint32_t len)
|
||||
{
|
||||
unsigned char hash[128]; // uint32_t hashA[16], hashB[16];
|
||||
#define hashA hash
|
||||
#define hashB hash+64
|
||||
|
||||
sph_shavite512_context ctx_shavite;
|
||||
sph_simd512_context ctx_simd;
|
||||
sph_echo512_context ctx_echo;
|
||||
|
||||
sph_shavite512_init(&ctx_shavite);
|
||||
sph_shavite512(&ctx_shavite, input, len);
|
||||
sph_shavite512_close(&ctx_shavite, hashA);
|
||||
|
||||
sph_simd512_init(&ctx_simd);
|
||||
sph_simd512(&ctx_simd, hashA, 64);
|
||||
sph_simd512_close(&ctx_simd, hashB);
|
||||
|
||||
sph_shavite512_init(&ctx_shavite);
|
||||
sph_shavite512(&ctx_shavite, hashB, 64);
|
||||
sph_shavite512_close(&ctx_shavite, hashA);
|
||||
|
||||
sph_simd512_init(&ctx_simd);
|
||||
sph_simd512(&ctx_simd, hashA, 64);
|
||||
sph_simd512_close(&ctx_simd, hashB);
|
||||
|
||||
sph_echo512_init(&ctx_echo);
|
||||
sph_echo512(&ctx_echo, hashB, 64);
|
||||
sph_echo512_close(&ctx_echo, hashA);
|
||||
|
||||
memcpy(output, hash, 32);
|
||||
}
|
||||
|
||||
int scanhash_fresh( struct work *work,
|
||||
uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr)
|
||||
{
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
uint32_t len = 80;
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
|
||||
uint32_t n = pdata[19] - 1;
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
#ifdef _MSC_VER
|
||||
uint32_t __declspec(align(32)) hash64[8];
|
||||
#else
|
||||
uint32_t hash64[8] __attribute__((aligned(32)));
|
||||
#endif
|
||||
uint32_t endiandata[32];
|
||||
|
||||
uint64_t htmax[] = {
|
||||
0,
|
||||
0xF,
|
||||
0xFF,
|
||||
0xFFF,
|
||||
0xFFFF,
|
||||
0x10000000
|
||||
};
|
||||
uint32_t masks[] = {
|
||||
0xFFFFFFFF,
|
||||
0xFFFFFFF0,
|
||||
0xFFFFFF00,
|
||||
0xFFFFF000,
|
||||
0xFFFF0000,
|
||||
0
|
||||
};
|
||||
|
||||
// we need bigendian data...
|
||||
for (int k = 0; k < 19; k++)
|
||||
be32enc(&endiandata[k], pdata[k]);
|
||||
|
||||
#ifdef DEBUG_ALGO
|
||||
if (Htarg != 0)
|
||||
printf("[%d] Htarg=%X\n", thr_id, Htarg);
|
||||
#endif
|
||||
for (int m=0; m < 6; m++) {
|
||||
if (Htarg <= htmax[m]) {
|
||||
uint32_t mask = masks[m];
|
||||
do {
|
||||
pdata[19] = ++n;
|
||||
be32enc(&endiandata[19], n);
|
||||
freshhash(hash64, endiandata, len);
|
||||
#ifndef DEBUG_ALGO
|
||||
if ((!(hash64[7] & mask)) && fulltest(hash64, ptarget)) {
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
return true;
|
||||
}
|
||||
#else
|
||||
if (!(n % 0x1000) && !thr_id) printf(".");
|
||||
if (!(hash64[7] & mask)) {
|
||||
printf("[%d]",thr_id);
|
||||
if (fulltest(hash64, ptarget)) {
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
return true;
|
||||
}
|
||||
}
|
||||
#endif
|
||||
} while (n < max_nonce && !work_restart[thr_id].restart);
|
||||
// see blake.c if else to understand the loop on htmax => mask
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
pdata[19] = n;
|
||||
return 0;
|
||||
}
|
||||
|
||||
bool register_fresh_algo( algo_gate_t* gate )
|
||||
{
|
||||
algo_not_tested();
|
||||
gate->scanhash = (void*)&scanhash_fresh;
|
||||
gate->hash = (void*)&freshhash;
|
||||
opt_target_factor = 256.0;
|
||||
return true;
|
||||
};
|
||||
|
||||
@@ -1,5 +1,7 @@
|
||||
#include "timetravel-gate.h"
|
||||
|
||||
#if !defined(TIMETRAVEL_8WAY) && !defined(TIMETRAVEL_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -290,4 +292,4 @@ int scanhash_timetravel( struct work *work, uint32_t max_nonce,
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
#endif
|
||||
|
||||
@@ -1,4 +1,7 @@
|
||||
#include "timetravel10-gate.h"
|
||||
|
||||
#if !defined(TIMETRAVEL10_8WAY) && !defined(TIMETRAVEL10_4WAY)
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
@@ -329,3 +332,4 @@ int scanhash_timetravel10( struct work *work, uint32_t max_nonce,
|
||||
*hashes_done = pdata[19] - first_nonce + 1;
|
||||
return 0;
|
||||
}
|
||||
#endif
|
||||
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user