FYI this change is found already in both the phatk and poclbm kernels in linux coin.
Ah, thank you -- just the confirmation I was looking for, without having to muddle with my OS
Topic: 3% faster mining with phoenix+phatk, diablo, or poclbm for everyone (Read 39142 times)
October 06, 2011, 01:11:56 PM
Ah, thank you -- just the confirmation I was looking for, without having to muddle with my OS
July 31, 2011, 02:31:10 PM
FYI this change is found already in both the phatk and poclbm kernels in linux coin.
July 19, 2011, 09:16:36 AM
Does this tweak/mod apply to the latest GUIMiner? When I installed GUIMiner, I did not create a new miner, GUIMiner is using what ever it uses when you fire it up [Default] tab.
Could you add explicit instructions " I am guessing I am using POCLBM?".
From what I see, there are instructions for phatk's.
But then you say
I opened bitcoinminer.cl and did not find
Could you add explicit instructions " I am guessing I am using POCLBM?".
From what I see, there are instructions for phatk's.
But then you say
Quote
Works also for POCLBM, just need to edit bitcoinminer.cl and change very same line.
I opened bitcoinminer.cl and did not find
Quote
#define Ma(x, y, z) amd_bytealign((y), (x | z), (z & x))
July 11, 2011, 01:37:27 AM
Diapolo I have been following your thread. You have put a lot of work into it and I look forward to further testing. That being said, including these kernel mods makes it difficult to tell how successful your changes are (lack of control). If your modifications are truly beneficial they will be included in the mainline.
Also, I believe the original developers are not getting the attention they deserve. These kernel mods and experimentation are welcomed by the community, but lets not forget who put in the original effort!
Also, I believe the original developers are not getting the attention they deserve. These kernel mods and experimentation are welcomed by the community, but lets not forget who put in the original effort!
I would say mods are "truly beneficial", if they lower the needed ALU OPs to process the kernel. This can be checked via AMD APP KernelAnalyzer and is what I try to do. You are absolutely right, that we should not forget the ones, who created a kernel in it's basic version. I couldn't have done this, that's for sure! But in the end we are all interested in the same ... to calculate BTC faster or more efficient, right?
Dia
July 11, 2011, 01:15:39 AM
Diapolo I have been following your thread. You have put a lot of work into it and I look forward to further testing. That being said, including these kernel mods makes it difficult to tell how successful your changes are (lack of control). If your modifications are truly beneficial they will be included in the mainline.
Also, I believe the original developers are not getting the attention they deserve. These kernel mods and experimentation are welcomed by the community, but lets not forget who put in the original effort!
Also, I believe the original developers are not getting the attention they deserve. These kernel mods and experimentation are welcomed by the community, but lets not forget who put in the original effort!
July 11, 2011, 12:57:42 AM
As will DiabloMiner
I would just like to restate, if you would like to get these tweaks your best bet is to just update your miner. The big 3 have all been updated at this point and you are better off not editing source unless you have to. The developers are very responsive to modifications in the kernel that can be shown to improve efficiency.
That would state, one should trust only "the big 3" when it comes to kernel kernel updates, which I think is NOT true
.You are right, it's harder to edit the OpenCL kernel for yourself, but there is no reason in not trying out modified / optimized kernels (like mine)!
Dia
July 10, 2011, 07:55:04 PM
As will DiabloMiner
I would just like to restate, if you would like to get these tweaks your best bet is to just update your miner. The big 3 have all been updated at this point and you are better off not editing source unless you have to. The developers are very responsive to modifications in the kernel that can be shown to improve efficiency.
July 10, 2011, 02:29:44 PM
Win7, dual 6970s, Guiminer w/ poclbm.
Went from ~ 409Mhash per card to 423Mhash per card.
Went from ~ 409Mhash per card to 423Mhash per card.
July 10, 2011, 03:22:59 AM
Question, if you have to use the BFI_INT flag for this tweak to work with the Phoenix Phatk kernel what flag do you use to make use of it with POCLBM? I'm running the latest version of GUIMiner with the default OpenCL POCLBM miner and extra flags "-v -w128 -f2". I applied this kernel tweak to the modified phatk.cl file that is in the main directory for GUIMiner and it seemed to give me a boost in hash rate but is it actually working without an additional flag?
poclbm will automatically do BFI_INT if your hardware is capable.
As will DiabloMiner
July 09, 2011, 01:36:45 PM
Question, if you have to use the BFI_INT flag for this tweak to work with the Phoenix Phatk kernel what flag do you use to make use of it with POCLBM? I'm running the latest version of GUIMiner with the default OpenCL POCLBM miner and extra flags "-v -w128 -f2". I applied this kernel tweak to the modified phatk.cl file that is in the main directory for GUIMiner and it seemed to give me a boost in hash rate but is it actually working without an additional flag?
poclbm will automatically do BFI_INT if your hardware is capable.
July 09, 2011, 12:51:02 PM
well with this boost and the -f2 tag that i tried via burning's post i have a boost of 10 mhash which is 3.3%
You applied the tweak to the phatk.cl file in the main part of the GUIMiner folder didn't you? Because I don't see the bitminer.cl file included with the latest update to GUIMiner any more so I'm guessing that is due to it being replaced with the modified phatk kernel.
July 09, 2011, 12:46:53 PM
Question, if you have to use the BFI_INT flag for this tweak to work with the Phoenix Phatk kernel what flag do you use to make use of it with POCLBM? I'm running the latest version of GUIMiner with the default OpenCL POCLBM miner and extra flags "-v -w128 -f2". I applied this kernel tweak to the modified phatk.cl file that is in the main directory for GUIMiner and it seemed to give me a boost in hash rate but is it actually working without an additional flag?
July 09, 2011, 09:44:35 AM
This actually slows down my poor little 5670 
94==>92
What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
94==>92What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
If you're on one, you should be. OTOH, it shouldn't be slowing it down if you dont have BFI_INT enabled
lol? All 5xxx and 6xxx do.
July 09, 2011, 09:16:07 AM
This actually slows down my poor little 5670 94==>92
What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
94==>92What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
If you're on one, you should be. OTOH, it shouldn't be slowing it down if you dont have BFI_INT enabled
July 09, 2011, 03:43:12 AM
All who tried this kernel, the MA-function patch from this thread is included in my modified phatk kernel.
You are able to run it with SDK 2.1, too ... so give it a try.
http://forum.bitcoin.org/index.php?topic=25860.0
YES, this seems to be an ad
.
Dia
You are able to run it with SDK 2.1, too ... so give it a try
.http://forum.bitcoin.org/index.php?topic=25860.0
YES, this seems to be an ad
.Dia
July 09, 2011, 12:35:12 AM
This actually slows down my poor little 5670 94==>92
What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
94==>92What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
If you're on one, you should be. OTOH, it shouldn't be slowing it down if you dont have BFI_INT enabled
July 08, 2011, 09:17:51 PM
This actually slows down my poor little 5670 94==>92
What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
94==>92What is meant by "This will ONLY WORK if you're running with BFI_INT"? Maybe that's my problem
July 08, 2011, 08:58:23 PM
Thanks a ton!
July 06, 2011, 08:40:27 AM
I was getting 272mh/s with my 5830. After implimenting this change, I'm getting 278 or 279. Not too shabby
July 05, 2011, 09:30:30 PM
Got a few mhash on each of my systems, except on my dual 5870 system... Didn't really see a change.
This may just be me or a coincidence, but my single card system showed bigger gains than my multiple card systems. Also my more heavily overclocked systems showed less gains than my less overclocked systems.
Thanks a bunch.
This may just be me or a coincidence, but my single card system showed bigger gains than my multiple card systems. Also my more heavily overclocked systems showed less gains than my less overclocked systems.
Thanks a bunch.
Jump to: