int main ()
{
int i;
int in[1024];
int out[1024];
// grsiSET_CONSTANTS();
/* not really used */
work_restart = &workr;
/* not so random numer */
for (i=0;i<1024;i++)
{
in[i] = i;
}
/* check if typto trashing constants that are not constant */
quarkhash(out,in);
quarkhash(out,in);
/* show simple test */
for (i=0;i<32;i++)
{
printf("%x", out[i]);
}
printf("\n");
/* burn though many round of quarkhash */
for (i=0;i<40000;i++)
{
quarkhash(out,in);
quarkhash(in,out);
}
/* not really a usefull answer but but should never change */
for (i=0;i<32;i++){
printf("%x", out[i]);
}
printf("\n");
return 0;
}
void quarkcoin_regenhash(struct work *work)
{
uint32_t data[20];
char *scratchbuf;
uint32_t *nonce = (uint32_t *)(work->data + 76);
uint32_t *ohash = (uint32_t *)(work->hash);
be32enc_vect(data, (const uint32_t *)work->data, 19);
data[19] = htobe32(*nonce);
quarkhash(ohash, data);
}
bool scanhash_quarkcoin(struct thr_info *thr, const unsigned char __maybe_unused *pmidstate,
unsigned char *pdata, unsigned char __maybe_unused *phash1,
unsigned char __maybe_unused *phash, const unsigned char *ptarget,
uint32_t max_nonce, uint32_t *last_nonce, uint32_t n)
{
uint32_t *nonce = (uint32_t *)(pdata + 76);
char *scratchbuf;
uint32_t data[20];
uint32_t tmp_hash7;
uint32_t Htarg = le32toh(((const uint32_t *)ptarget)[7]);
bool ret = false;
be32enc_vect(data, (const uint32_t *)pdata, 19);
while(1) {
uint32_t ostate[8];
*nonce = ++n;
data[19] = (n);
quarkhash(ostate, data);
tmp_hash7 = (ostate[7]);
applog(LOG_INFO, "data7 %08lx",
(long unsigned int)data[7]);
if (unlikely(tmp_hash7 <= Htarg)) {
((uint32_t *)pdata)[19] = htobe32(n);
*last_nonce = n;
ret = true;
break;
}
if (unlikely((n >= max_nonce) || thr->work_restart)) {
*last_nonce = n;
break;
}
}
return ret;
}
/* Used externally as confirmation of correct OCL code */
int quarkcoin_test(unsigned char *pdata, const unsigned char *ptarget, uint32_t nonce)
{
uint32_t tmp_hash7, Htarg = le32toh(((const uint32_t *)ptarget)[7]);
uint32_t data[20], ohash[8];
//char *scratchbuf;
be32enc_vect(data, (const uint32_t *)pdata, 19);
data[19] = htobe32(nonce);
//scratchbuf = alloca(SCRATCHBUF_SIZE);
quarkhash(ohash, data);
tmp_hash7 = be32toh(ohash[7]);
applog(LOG_DEBUG, "htarget %08lx diff1 %08lx hash %08lx",
(long unsigned int)Htarg,
(long unsigned int)diff1targ,
(long unsigned int)tmp_hash7);
if (tmp_hash7 > diff1targ)
return -1;
if (tmp_hash7 > Htarg)
return 0;
return 1;
}
int scanhash_quark(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, uint64_t *hashes_done)
{
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t hash64[8] __attribute__((aligned(32)));
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 kk=0; kk < 32; kk++) {
be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]);
};
#ifdef DEBUG_ALGO
printf("[%d] Htarg=%X\n", thr_id, Htarg);
#endif
for (int m=0; m < sizeof(masks); m++) {
if (Htarg <= htmax[m]) {
uint32_t mask = masks[m];
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(hash64, &endiandata);
#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;
}
inline void quarkhash(void *state, const void *input)
{
/* shared temp space */
/* hash is really just 64bytes but it used to hold both hash and final round constants passed 64 */
DATA_ALIGN16(unsigned char hashbuf[128]);
DATA_ALIGN16(size_t hashptr);
DATA_ALIGN16(sph_u64 hashctA);
DATA_ALIGN16(sph_u64 hashctB);
//sph_keccak512_context ctx_keccak;
//sph_groestl512_context ctx_grs;
grsoState sts_grs;
//grssState sts_grs;
#ifdef SPEEDRUN
/* just run one of each */
int speedrun[] = {0, 1, 3, 4, 6, 7 };
#endif
int i;
DATA_ALIGN16(unsigned char hash[128]);
/* proably not needed */
memset(hash, 0, 128);
DECL_BLK;
BLK_I;
BLK_W;
#ifndef SPEEDRUN
/* this layout is so each "function" inlined just once */
/* i is not special rounds */
/* i+16 is special rounds */
for(i=0; i<9; i++) {
#else
for (i=0;i<6;i++) {
#endif
/* blake is split between 64byte hashes and the 80byte initial block */
//DECL_BLK;
#ifndef SPEEDRUN
switch (i+(16*((hash[0] & (uint32_t)(8)) == (uint32_t)(0))))
#else
switch (speedrun[i])
#endif
{
case 5 :
BLK_I;
BLK_U;
case 0:
case 16:
BLK_C;
break;
case 1:
case 17:
case 21: do {
DECL_BMW;
BMW_I;
BMW_U;
/* bmw compress uses some defines */
/* i havent gotten around to rewriting these */
#define M(x) sph_dec64le_aligned(data + 8 * (x))
#define H(x) (h[x])
#define dH(x) (dh[x])
BMW_C;
#undef M
#undef H
#undef dH
} while(0); continue;;
case 2:
#ifdef CHEAT
/* blake and bmw are almost free
* if luck means that groestl will now be run twice
* give up and try another nonce */
return;
#endif
case 19:
case 3: do {
GRS_I;
GRS_U;
GRS_C;
//grsoInit(&sts_grs);
//grsoUpdateq(&sts_grs, (char*)hash);
//grsoFinal(&sts_grs, (char*)hash);
//grssInit(&sts_grs,512);
//grssUpdate(&sts_grs, (char*)hash,128*8*4);
//grssFinal(&sts_grs, (char*)hash);
} while(0); continue;
case 4:
case 20:
case 24: do {
DECL_JH;
/* JH ended up being really short */
JH_H;
} while(0); continue;
case 6:
case 22:
case 8: do {
DECL_KEC;
KEC_I;
KEC_U;
KEC_C;
} while(0); continue;
case 18:
case 7:
case 23: do {
DECL_SKN;
SKN_I;
SKN_U;
SKN_C; /* is a magintue faster than others, done */
} while(0); continue;
default:
/* bad things happend, i counted to potato */
abort();
}
/* only blake shouuld get here without continue */
/* blake finishs from top split */
//BLK_C;
}
asm volatile ("emms");
memcpy(state, hash, 32);
}
int scanhash_quark(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t hash64[8] __attribute__((aligned(32)));
uint32_t endiandata[32];
//we need bigendian data...
//lessons learned: do NOT endianchange directly in pdata, this will all proof-of-works be considered as stale from minerd....
int kk=0;
for (; kk < 32; kk++)
{
be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]);
};
// if (opt_debug)
// {
// applog(LOG_DEBUG, "Thr: %02d, firstN: %08x, maxN: %08x, ToDo: %d", thr_id, first_nonce, max_nonce, max_nonce-first_nonce);
// }
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(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;
}
int scanhash_quark(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, uint64_t *hashes_done)
{
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t hash64[8] __attribute__((aligned(32)));
uint32_t endiandata[32];
//char testdata[] = {"\x70\x00\x00\x00\x5d\x38\x5b\xa1\x14\xd0\x79\x97\x0b\x29\xa9\x41\x8f\xd0\x54\x9e\x7d\x68\xa9\x5c\x7f\x16\x86\x21\xa3\x14\x20\x10\x00\x00\x00\x00\x57\x85\x86\xd1\x49\xfd\x07\xb2\x2f\x3a\x8a\x34\x7c\x51\x6d\xe7\x05\x2f\x03\x4d\x2b\x76\xff\x68\xe0\xd6\xec\xff\x9b\x77\xa4\x54\x89\xe3\xfd\x51\x17\x32\x01\x1d\xf0\x73\x10\x00"};
//we need bigendian data...
//lessons learned: do NOT endianchange directly in pdata, this will all proof-of-works be considered as stale from minerd....
int kk=0;
for (; kk < 32; kk++)
{
be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]);
};
// if (opt_debug)
// {
// applog(LOG_DEBUG, "Thr: %02d, firstN: %08x, maxN: %08x, ToDo: %d", thr_id, first_nonce, max_nonce, max_nonce-first_nonce);
// }
/* I'm to lazy to put the loop in an inline function... so dirty copy'n'paste.... */
/* i know that i could set a variable, but i don't know how the compiler will optimize it, not that then the cpu needs to load the value *everytime* in a register */
if (ptarget[7]==0) {
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(hash64, &endiandata);
if (((hash64[7]&0xFFFFFFFF)==0) &&
fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
} while (n < max_nonce && !work_restart[thr_id].restart);
}
else if (ptarget[7]<=0xF)
{
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(hash64, &endiandata);
if (((hash64[7]&0xFFFFFFF0)==0) &&
fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
} while (n < max_nonce && !work_restart[thr_id].restart);
}
else if (ptarget[7]<=0xFF)
{
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(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);
}
else if (ptarget[7]<=0xFFF)
{
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(hash64, &endiandata);
if (((hash64[7]&0xFFFFF000)==0) &&
fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
} while (n < max_nonce && !work_restart[thr_id].restart);
}
else if (ptarget[7]<=0xFFFF)
{
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(hash64, &endiandata);
if (((hash64[7]&0xFFFF0000)==0) &&
fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
} while (n < max_nonce && !work_restart[thr_id].restart);
}
else
{
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
quarkhash(hash64, &endiandata);
if (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;
}