No. The kernel computes correctly only the 4 first bytes. It's confusing, because there is a code in BitcoinMiner.cl BelowOrEquals() which checks 8 bytes - this produces better assembler for some reason, at least in my setup. It can be replaced with 'if (H == 0)' (but it was slower). That's exactly why the targets are hard coded to difficulty of 1 (00000000 FFFF0000).
If you mean that this:
#ifdef VECTORS
if (belowOrEquals(H.x, targetH, G.x, targetG))
{
output[OUTPUT_SIZE] = output[nonce.x & OUTPUT_MASK] = nonce.x;
}
else if (belowOrEquals(H.y, targetH, G.y, targetG))
{
output[OUTPUT_SIZE] = output[nonce.y & OUTPUT_MASK] = nonce.y;
}
#else
if (belowOrEquals(H, targetH, G, targetG))
{
output[OUTPUT_SIZE] = output[nonce & OUTPUT_MASK] = nonce;
}
#endif
}
is faster than this:
#ifdef VECTORS
if (H.x == 0)
{
output[OUTPUT_SIZE] = output[nonce.x & OUTPUT_MASK] = nonce.x;
}
else if (H.y == 0)
{
output[OUTPUT_SIZE] = output[nonce.y & OUTPUT_MASK] = nonce.y;
}
#else
if (H == 0)
{
output[OUTPUT_SIZE] = output[nonce & OUTPUT_MASK] = nonce;
}
#endif
then I got very different results in my tests on Windows 7 x64 & HD5850.
With belowEquals I get ~270Mhps
With the H == 0 version I get ~275Mhps
That's a 1.8% speed increase.
Actually you are right, but in a different way - because of this I should use hard coded kernel target of 00000000 FFFFFFFF in order to not lose 1 thousandth of a percent of all valid difficulty=1 candidates. I'll do this with the next release.
I'd just stop the target from being sent to the GPU and do that H==0 thing. I see no logical reason that it would be slower. There might be illogical reasons, though :-)
Why you check for 'lower' with a 'greater than' operator? Where did the '0x80000000' came from?
I guess I didn't explain that part that well. I chose 0x80000000 to make it easy to determine if bytes 4-7 were actually calculated correctly. I edited BitcoinMiner.cl to only accept hashes below 0x80000000. In the Python code, "> 0x80000000" (should've been >= though) is checking if that part of the hash is over 0x80000000 and warns the user if it is, i.e. if the GPU calculated the hash wrong. But none of that matters since it wasn't even supposed to calculate bytes 4-7 right.
Is this H==0 mod. going to go into official version of poclbm miner on github ?
Cheers.
