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

Quote from: cat77 on January 10, 2015, 04:16:31 AM

.....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);

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];
   }
  }
}

Later you said

Quote

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

When you used (mod %2, same as mod &1), what are the case statements inside XORBytesInPlace(void *restrict dst, const void *restrict src, uint mod)?