Bitcoin Forum

Other => Beginners & Help => Topic started by: LOG123 on September 01, 2011, 04:02:31 PM



Title: Poclbm/Phatk speedup!
Post by: LOG123 on September 01, 2011, 04:02:31 PM
I've heard some people are getting quite a few coins from hacks like this, so, here goes.

I've found that if you go into your poclbm kernel.cl (only tested on mine  ;D) and change
Code:
#define Ma2(x, y, z) ((y & z) | (x & (y | z)))
To
Code:
#define Ma2(x, y, z) amd_bytealign((z^y), (x), (y))
I get a ~3mhash speedup!
all of my hashes are getting accepted, so I don't know if there are any problems :P
My bitcoin address is
Code:
1M8Av4rzfyMMMrP5nqUmmhKHhZKB1U7Zux
Feel free to donate if this works for you  ;D


Title: Re: Poclbm/Phatk speedup!
Post by: LOG123 on September 02, 2011, 03:29:20 PM
bump so people can see this  ;D


Title: Re: Poclbm/Phatk speedup!
Post by: LOG123 on September 02, 2011, 04:33:04 PM
I just want to point out that the poster did not find this, rethaw did, and detailed it in his post here:

http://bitcointalk.org/index.php?topic=23067.0

So if you are going to donate to anyone, I'd say let it go to the person that originally discovered this.

I might be wrong about this, but I'm pretty sure a lot of updated miner's have already incorporated this as well, but check for yourself, rethaw's instructions are pretty detailed.

I have been running with the modification myself for about a month, no different in stales but a slight bump in MH/S

--------------EDIT-----------------

So I might be wrong about who originally came up with this, because the post referenced to in the one I linked is by someone else again, but anyway, this has been documented before.

If you actually looked at what I did, it's  different from that other modification. I'm not even sure if it really increases speed, or if I'm just having a lucky day with my miner :P
I changed
Code:
#define Ma2(x, y, z) ((y & z) | (x & (y | z)))
to
Code:
#define Ma2(x, y, z) amd_bytealign((z^y), (x), (y))

The other modification is
Code:
  #define Ma(x, y, z) amd_bytealign((y), (x | z), (z & x))
Code:
   #define Ma(x, y, z) amd_bytealign( (z^x), (y), (x) ) 
EDIT: Leaving this up so nobody asks the same thing again


Title: Re: Poclbm/Phatk speedup!
Post by: echris1 on September 02, 2011, 06:17:21 PM
Oops, sorry about that, yeah I do feel a bit foolish now, guess that is why I never got into any kind of coding, lack of attention to detail =)

Anyway, I deleted my post, sorry again.



Title: Re: Poclbm/Phatk speedup!
Post by: LOG123 on September 02, 2011, 06:28:06 PM
That's alright  ;D
Also, have you tried this tweak? And has it improved your mhash rate at all?


Title: Re: Poclbm/Phatk speedup!
Post by: echris1 on September 02, 2011, 07:06:45 PM
I am using phatk2, which doesn't seem to have that line in it.