|
No, I couldn't think of an implementation for radix sort in a hashing algorithm that couldn't be done more quickly and with less memory than quicksort, and also any way to really effectively insert it into the hashing algorithm and be easily invertible.
I'll pm you later about the other implementation, it's still FPGA and ASIC hard. I wish I had the time to work on it more or the money to pay a coder to implement it.
|
|
|
|
|
I own one of these cards, voltage unlocked but runs HOT, the only way I could get it to run cool mining LTC was to rip the fans out and strap 120mm fans on it :/
|
|
|
|
@tacotime Thanks for the link ! 2. SHA256 followed by BLAKE256 followed by keccak256 (SHA3-256) for the crypt algorithm, to enhance circuit size in ASICs without strongly affecting hash speed.
I prefer to "enhance" like this : 2. Grøstl256 followed by keccak256 (SHA3-256) for the crypt algorithm, to enhance circuit size in ASICs without strongly affecting hash speed.
That approach was naive, although I figured out another solution to it. |
|
|
|
|
Ignore cgminer and go with whatever the temperature in gpu-z is. That is correct. You can also tell if the voltage changed by changes in temperature.
|
|
|
|
|
Thread concurrency is too low, increase thread concurrency
|
|
|
|
This is the scrypt core from reaper's implementation, void scrypt_core(uint4 X[8], __global uint4*restrict lookup)
{
shittify(X);
const uint zSIZE = 8;
const uint ySIZE = (1024/LOOKUP_GAP+(1024%LOOKUP_GAP>0));
const uint xSIZE = CONCURRENT_THREADS;
uint x = get_global_id(0)%xSIZE;
for(uint y=0; y<1024/LOOKUP_GAP; ++y)
{
#pragma unroll
for(uint z=0; z<zSIZE; ++z)
lookup[CO] = X[z];
for(uint i=0; i<LOOKUP_GAP; ++i)
salsa(X);
}
#if (LOOKUP_GAP != 2) && (LOOKUP_GAP != 4) && (LOOKUP_GAP != 8)
{
uint y = (1024/LOOKUP_GAP);
#pragma unroll
for(uint z=0; z<zSIZE; ++z)
lookup[CO] = X[z];
for(uint i=0; i<1024%LOOKUP_GAP; ++i)
salsa(X);
}
#endif
for (uint i=0; i<1024; ++i)
{
uint4 V[8];
uint j = X[7].x & 0x3FF;
uint y = (j/LOOKUP_GAP);
#pragma unroll
for(uint z=0; z<zSIZE; ++z)
V[z] = lookup[CO];
#if (LOOKUP_GAP == 2)
if (j&1)
salsa(V);
#else
uint val = j%LOOKUP_GAP;
for (uint z=0; z<val; ++z)
salsa(V);
#endif
#pragma unroll
for(uint z=0; z<zSIZE; ++z)
X[z] ^= V[z];
salsa(X);
}
unshittify(X);
} It looks like the scratchpad is regenerated extensively for cases of LOOKUP_GAP != 2 but not for LOOKUP_GAP == 2, in which we As compared to uint val = j%LOOKUP_GAP;
for (uint z=0; z<val; ++z)
salsa(V); The first loop is a simplication of the bottom loop for this specific case, and it looks like it only needs one run of salsa to keep moving. |
|
|
|
|
Hi,
I meant to dump the contents of scrypt-poclbm.ini between a bracketed "code" and "/code" in your post (just put brackets around those), I can't see anything in the post you gave above.
|
|
|
|
|
The problem is mainly in transaction times... as this gets worse it could slow to one block an hour and take forever to confirm. This would effectively halt the chain if it became severe enough, as no one can transfer their coins to anyone else in a reasonable amount of time.
You need to hard code a fix for the difficulty algorithm if you want to correct this.
|
|
|
|
Why is terracoin difficulty jumping around 3x back and forth? Anyone else see a pattern here? Looks like some big miners are exploiting the fast difficulty adjustment. I guess so much for the supposed feature 'miner jump resonance free' Someone please make a chart to trace back when this all started height difficulty
96000 5001
96030 8254
96060 2671
96090 10580
96120 3830
96150 9622
96180 3270
96210 8391
96240 3029
96270 9203
96300 2330
96330 8029
Was kind of waiting for this, I published that python exploit script a while ago and I figured someone would jump on it... It's good to see that someone's put one hour of my hard work into use I guess. |
|
|
|
|
Come mine litecoins with us.
It costs 100x more power to mine 1 USD in LTC as compared to BTC with an ASIC. Get your money's worth -- there's a reason it's more profitable to miner LTC than BTC right now.
|
|
|
|
|
most likely you will get more kh/s by using --gpu-memclock 1200
|
|
|
|
Thank you!!  I can't see the output of your poclbm-scrypt.ini file above, you might want to try posting it with the [/code] tags |
|
|
|
How do you guys usually do the short? Just cut it out of the ribbon and then connect them with a wire?
You see the picture in the OP ? This red thing is a wire, Use a solid core cat5 braid and insert in behind the pin of either the extender or the slot itself. How do I keep it from making contact with the other nearby pins? |
|
|
|
|
ltcmine.ru is up, been accepting shares throughout the attack too
Someone's trying to game the difficulty, looks like it's not working all that well as many pools are surviving, coinotron and notroll are all still accepting shares, all the other pools are still giving/accepting work
|
|
|
|
|
Okay. For some reason those are all default settings entered for user, password, host, and port. Can you select File --> Save in guiminer and then dump the contents of C:\Users\you\AppData\Roaming\poclbm\poclbm_scrypt.ini here? You can remove your miner passwords if they are saved there
|
|
|
|
|
How do you guys usually do the short? Just cut it out of the ribbon and then connect them with a wire?
|
|
|
|
|
Have you tried high thread concurrencies with high intensities? eg 21712 and intensity 20
|
|
|
|
|
Hi,
From the above the problem is with your credentials. Do you have a registered account at ltcmine.ru and worker credentials (-u user -p password should be your username and password for worker)?
|
|
|
|
|
Show Posts
