[ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner

[damm315er]

.....This is worth 20KH/s on my 280X......from 343KHs to 363KH/s at 1020MHz clock
.....now somebody needs to find 20KH/s more for me.... 

change the XORBytesInPlace call from

Code:

	XORBytesInPlace(B + bufidx, input, BLAKE2S_OUT_SIZE);

to

Code:

      XORBytesInPlace(B + bufidx, input, bufidx);

and change the function itself to perform some byte alignment checking

Code:

//
// a bit of byte alignment checking goes a long ways...
//
void XORBytesInPlace(void *restrict dst, const void *restrict src, uint mod)
{
  switch(mod % 4)
  {
  case 0:
    #pragma unroll 2
    for(int i = 0; i < 4; i+=2)
    {
    	  ((uint2 *)dst)[i]   ^= ((uint2 *)src)[i]; 
     	  ((uint2 *)dst)[i+1] ^= ((uint2 *)src)[i+1];    
    }
    break;    

  case 2:  
    #pragma unroll 8
    for(int i = 0; i < 16; i+=2)
    {
    	  ((uchar2 *)dst)[i] ^= ((uchar2 *)src)[i]; 
    	  ((uchar2 *)dst)[i+1] ^= ((uchar2 *)src)[i+1]; 
    }
    break;

  default:
  #pragma unroll 8
   for(int i = 0; i < 31; i+=4)
   {
  	  ((uchar *)dst)[i] ^= ((uchar *)src)[i];
  	  ((uchar *)dst)[i+1] ^= ((uchar *)src)[i+1];
  	  ((uchar *)dst)[i+2] ^= ((uchar *)src)[i+2];
  	  ((uchar *)dst)[i+3] ^= ((uchar *)src)[i+3];   
    }
  }
}

This actually drops the hashrate by 5 kh/s on the 290's, but combined with the bobben2 mod increases by about 9 kh/s.

YMMV

[cat77]

Very interesting.   I get about 2% gain on 7950 and need to use (mod % 2) with the case statements adjusted accordingly.
My 280X gains almost 6% as is, but the gain difference between (mod % 2) and (mod % 4) is pretty small, like 1-2 KHs

My SMix call is a bit different, I simply put the sub-calls inline so it doesn't bother with ScratchpadStore and ScratchpadMix.
Perhaps this fits nicer into the core and needs less swapping.  

I have tried, unsuccessfully, to further streamline the SMix, but any other way I do it, its either all HW errors or vastly slower. Any guidance here would be appreciated.

Code:

void SMix(ulong16 *X, __global ulong16 *V, bool flag)
{
  int i = 0;
  int idx;

    while (i^256)
    {
    	  V[i++]   = X[0];
 	      V[i++]   = X[1];      
        neoscrypt_blkmix(X, flag);
    }
    do {      
        idx = (( (uint *)X)[48] & 0x7F) << 1;
 	      X[0] ^= V[idx];
 	      X[1] ^= V[idx+1]; 
        neoscrypt_blkmix(X, flag);    
    }   while (i-=2);
} 

 

[damm315er]

Very interesting.   I get about 2% gain on 7950 and need to use (mod % 2) with the case statements adjusted accordingly.
My 280X gains almost 6% as is, but the gain difference between (mod % 2) and (mod % 4) is pretty small, like 1-2 KHs

My SMix call is a bit different, I simply put the sub-calls inline so it doesn't bother with ScratchpadStore and ScratchpadMix.
Perhaps this fits nicer into the core and needs less swapping.  

I have tried, unsuccessfully, to further streamline the SMix, but any other way I do it, its either all HW errors or vastly slower. Any guidance here would be appreciated.

Code:

void SMix(ulong16 *X, __global ulong16 *V, bool flag)
{
  int i = 0;
  int idx;

    while (i^256)
    {
    	  V[i++]   = X[0];
 	      V[i++]   = X[1];      
        neoscrypt_blkmix(X, flag);
    }
    do {      
        idx = (( (uint *)X)[48] & 0x7F) << 1;
 	      X[0] ^= V[idx];
 	      X[1] ^= V[idx+1]; 
        neoscrypt_blkmix(X, flag);    
    }   while (i-=2);
} 

 

I gotta wonder if you touched on what the bottleneck is in wolf's kernel for the 290/290x, and that's why the speed dropped..

[semajjames]

Try memclock 1500, or if that is not stable, try 1250.

AMD gpus  have memory  timings set at 125MHz intervals, always ending at even 125MHz. Like this:

875-1000
1001-1125
1126-1250
1251-1375
1376-1500
1501-1625

 mining is mostly about random access latency, not about sequential reads. You get best random access speed at the high end of each range, like 1250 or 1500.

It's very likely that 1250MHz is faster than 1400MHz, because 1400MHz is so close to the start of the range and 1250 is exactly at the end..

So these figures :-

875-1000
1001-1125
1126-1250
1251-1375
1376-1500
1501-1625

are these only for a gpu with stock engine of 875 or would these figures still be correct for a gpu with stock engine of 947 ect ?

[DragonSlayer]

I am getting an error while compiling in VS2013.

error LNK1104: cannot open file 'jansson.lib'

Anyone know what I am doing wrong?

Thanks

[yudhistira]

whats wrong with this, please help

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

sgminer.exe --kernel bitblock -o stratum+tcp://stratum.westhash.com:3336 -u 12tyqFRW384n27ytyM77edUignkpGDDbZ9 -p d=0.001 -I 18 --worksize 64 -g 2 --gpu-powertune 20 --gpu-engine 1130 --gpu-memclock 1500 --lookup-gap 2 --auto-fan --gpu-fan 40-70 --temp-cutoff 85 --temp-overheat 80

[nicehash]

whats wrong with this, please help

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

sgminer.exe --kernel bitblock -o stratum+tcp://stratum.westhash.com:3336 -u 12tyqFRW384n27ytyM77edUignkpGDDbZ9 -p d=0.001 -I 18 --worksize 64 -g 2 --gpu-powertune 20 --gpu-engine 1130 --gpu-memclock 1500 --lookup-gap 2 --auto-fan --gpu-fan 40-70 --temp-cutoff 85 --temp-overheat 80

Port 3336 is for X11 algorithm, use "--kernel darkcoin-mod".

[drr0ss]

I am getting an error while compiling in VS2013.

error LNK1104: cannot open file 'jansson.lib'

Anyone know what I am doing wrong?

Thanks

Compile jansson project first.

[poiuty]

Wolf0, please make a bin for HD 7790 (bonaire) (x11). Thx.
https://www.reddit.com/r/DRKCoin/comments/2o1yoz/rewritten_x11_binaries/

[Tiger78]

Wolf0, please make a bin for HD 7790 (bonaire) (x11). Thx.
https://www.reddit.com/r/DRKCoin/comments/2o1yoz/rewritten_x11_binaries/

+1

[kopam]

So what are the best numbers you guys get for x13 on 7950 and 280x ?

Thanks !

[Txslady]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

[semajjames]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

are you sure you connected to the  pools  correct port ?

i think the miner loads by defult ckolivas.cl when the wrong kernel is selected for the coin

show your .conf or .bat file

[Txslady]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

are you sure you connected to the  pools  correct port ?

i think the miner loads by defult ckolivas.cl when the wrong kernel is selected for the coin

show your .conf or .bat file

I am using a cmd for a test.  All other algos work fine.
sgminer -k Lyra2 -o stratum+tcp://yaamp.com:4433 -u 1GDh3dYW1HPho3SF4Wb7qnUjeedh2spkhk -p xx --intensity 15 --scan-time 3 --gpu-platform 1

[DragonSlayer]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

are you sure you connected to the  pools  correct port ?

i think the miner loads by defult ckolivas.cl when the wrong kernel is selected for the coin

show your .conf or .bat file

I am using a cmd for a test.  All other algos work fine.
sgminer -k Lyra2 -o stratum+tcp://yaamp.com:4433 -u 1GDh3dYW1HPho3SF4Wb7qnUjeedh2spkhk -p xx --intensity 15 --scan-time 3 --gpu-platform 1

Should be Lyra2RE.

I am not sure which version 5.1 being used.  I do not think lyra2re is in master or developer version yet.

[platinum4]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

are you sure you connected to the  pools  correct port ?

i think the miner loads by defult ckolivas.cl when the wrong kernel is selected for the coin

show your .conf or .bat file

I am using a cmd for a test.  All other algos work fine.
sgminer -k Lyra2 -o stratum+tcp://yaamp.com:4433 -u 1GDh3dYW1HPho3SF4Wb7qnUjeedh2spkhk -p xx --intensity 15 --scan-time 3 --gpu-platform 1

Should be Lyra2RE.

Yep or you can grab the one from badman74's signature, he aliased the algorithm to accept lowercased variants.

[colin012]

I had an idea for speeding up nonce and "attempt" generation.

Let me qualify this first: I am brand new to the nitty gritty technical details of mining. I am in no way an expert.

Now that that is out of the way, I am presuming that the nonce serves as a seed to a PRNG which builds the data to be hashed and checked against the difficulty. I will call this data the "attempt string" because I don't know what the proper term is.

Anyways, so far, the fastest PRNG I have ever heard of is the SFMT. It is open source and written in c. It is optimized to take advantage of certain features of the CPU to generate random numbers as fast as possible. It may be worth while to check out because it could speed up nonce generation and attempt string generation. It is possible that this could improve the overall performance of your mining software.

Now, on to a question. NXT has recently released the MS or Monetary System. Some coins released on the MS can be "minted" using SHA-256, Scrypt, or Keccak25. Minting is almost the same thing as mining with the difference that it works like a transaction on the NXT blockchain and thus costs a transaction fee to the account of the minter.

There is talk about making a GPU/ASIC minting software... The only thing is, the Devs at NXT only know Java and we all know that it is faster to run mining or "minting" software natively rather than over the JVM.

It would not be that hard to adapt your software to mint MSCoins. All it would take is native transaction signing and some code to submit that to the NXT API server. Plus, of course, removal of unnecessary code and features.

Being the first GPU minting software for MSCoins would certainly be a noteworthy accomplishment!

[Txslady]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

are you sure you connected to the  pools  correct port ?

i think the miner loads by defult ckolivas.cl when the wrong kernel is selected for the coin

show your .conf or .bat file

I am using a cmd for a test.  All other algos work fine.
sgminer -k Lyra2 -o stratum+tcp://yaamp.com:4433 -u 1GDh3dYW1HPho3SF4Wb7qnUjeedh2spkhk -p xx --intensity 15 --scan-time 3 --gpu-platform 1

Should be Lyra2RE.

I am not sure which version 5.1 being used.  I do not think lyra2re is in master or developer version yet.

I am using 5.1.0.  I tried the ones from badman74 with no luck.  Guess it is no available right now, hopefully in the future.

[badman74]

Need help please.... I am running 5.1 and when I try to run lyra2 I get "low difficulty share" or lyra2RE I get "share above target".  I added --difficulty-multiplier 2 but have had no luck on fixing the issue.  Lyra worked fine on sgminer 4.2 so I am not sure what the issue maybe. I also noticed when I run the file it initializes kernel ckolivas.cl instead of lyra2.cl or lyra2RE.cl  All help is appreciated.

are you sure you connected to the  pools  correct port ?

i think the miner loads by defult ckolivas.cl when the wrong kernel is selected for the coin

show your .conf or .bat file

I am using a cmd for a test.  All other algos work fine.
sgminer -k Lyra2 -o stratum+tcp://yaamp.com:4433 -u 1GDh3dYW1HPho3SF4Wb7qnUjeedh2spkhk -p xx --intensity 15 --scan-time 3 --gpu-platform 1

Should be Lyra2RE.

I am not sure which version 5.1 being used.  I do not think lyra2re is in master or developer version yet.

I am using 5.1.0.  I tried the ones from badman74 with no luck.  Guess it is no available right now, hopefully in the future.

try --algorithm lyra2re with my version

[platinum4]

try --algorithm lyra2re with my version

Hey badman74, you know as always I'm appreciative as hell for the binary wrap-ups you've done.  Can you tell me if the version in your signature is later than 4.2.2-308, and where I can find the version on github in the code?  I remember looking for it during the 5.1.0-dev bump but not really able to find it in yours.  Just want to know what build we're on for it.

Also, anybody who can implement scrypt-jane in this sgminer5 would be a badass!