further improved phatk_dia kernel for Phoenix + SDK 2.6 - 2012-01-13

[jh1523]

Are these changes in the phatk svn now? Compared to a svn checkout from 7/25 I see absolutely no change in hash rates on a 5850 or a 5870.

[ptfff]

thanks

[Diapolo]

Are these changes in the phatk svn now? Compared to a svn checkout from 7/25 I see absolutely no change in hash rates on a 5850 or a 5870.

I don't know about a phatk svn, sorry. This version here will and should give you higher performance compared to the stock phatk. Just do a diff between my file here and the one on the svn, to check for differences.

Dia

[jh1523]

Here's what I get when I do a diff between your kernel and the one from svn (diff kernel.diapolo kernel.svn):

Code:

79c79
< // state0subT1: state0subT1 = state0 - T1
---
> // T1substate0: T1substate0 = T1 - substate0
88c88
<                                               const uint PreVal4addT1, const uint state0subT1,
---
>                                               const uint PreVal4addT1, const uint T1substate0,
100c100
<       Vals[0] = W_3 + PreVal4addT1 + state0subT1;
---
>       Vals[0] = W_3 + PreVal4addT1 + T1substate0;
390c390
<               output[OUTPUT_SIZE] = output[OUTPUT_MASK & (W[3].x >> 2)] = W_3.x;
---
>               output[OUTPUT_SIZE] = output[(W[3].x >> 2) & OUTPUT_MASK] = W_3.x;
394c394
<               output[OUTPUT_SIZE] = output[OUTPUT_MASK & (W[3].y >> 2)] =  W_3.y;
---
>               output[OUTPUT_SIZE] = output[(W[3].y >> 2) & OUTPUT_MASK] =  W_3.y;
399c399
<               output[OUTPUT_SIZE] = output[OUTPUT_MASK & (W[3] >> 2)] = W_3;
---
>               output[OUTPUT_SIZE] = output[(W[3] >> 2) & OUTPUT_MASK] = W_3;

The repository is at http://svn3.xp-dev.com/svn/phoenix-miner/trunk/kernels/phatk

[Diapolo]

Here's what I get when I do a diff between your kernel and the one from svn (diff kernel.diapolo kernel.svn):

Code:

79c79
< // state0subT1: state0subT1 = state0 - T1
---
> // T1substate0: T1substate0 = T1 - substate0
88c88
<                                               const uint PreVal4addT1, const uint state0subT1,
---
>                                               const uint PreVal4addT1, const uint T1substate0,
100c100
<       Vals[0] = W_3 + PreVal4addT1 + state0subT1;
---
>       Vals[0] = W_3 + PreVal4addT1 + T1substate0;
390c390
<               output[OUTPUT_SIZE] = output[OUTPUT_MASK & (W[3].x >> 2)] = W_3.x;
---
>               output[OUTPUT_SIZE] = output[(W[3].x >> 2) & OUTPUT_MASK] = W_3.x;
394c394
<               output[OUTPUT_SIZE] = output[OUTPUT_MASK & (W[3].y >> 2)] =  W_3.y;
---
>               output[OUTPUT_SIZE] = output[(W[3].y >> 2) & OUTPUT_MASK] =  W_3.y;
399c399
<               output[OUTPUT_SIZE] = output[OUTPUT_MASK & (W[3] >> 2)] = W_3;
---
>               output[OUTPUT_SIZE] = output[(W[3] >> 2) & OUTPUT_MASK] = W_3;

The repository is at http://svn3.xp-dev.com/svn/phoenix-miner/trunk/kernels/phatk

Now it's clear, the one on svn is the same speed wise, but I uploaded mine again to rename a variable and reordered something for the output buffer. Nice that Jedi took my kernel as base for Phoenix, was not sure about that .

Dia

[Diapolo]

I'm working on a new version! The inputs came from the original Author of phatk, who released a version 2.0 of phatk (THANKS Phateus).
Currently my version IS slower, but I see this as a fair and cool competition, from which all of us will benefit in the end.

Dia

[jh1523]

*fingers crossed*

[shotgun]

I'm working on a new version! The inputs came from the original Author of phatk, who released a version 2.0 of phatk (THANKS Phateus).
Currently my version IS slower, but I see this as a fair and cool competition, from which all of us will benefit in the end.

Dia

where is the download link for version 2.0 of phatk?

[joulesbeef]

phatk 2.0 is in the phatk thread

but it sounds like you might want to wait til monday

Update:  Second new version with a few optimizations from Diapolo's Miner It appears to be somewhat broken, I will post an update this weekend

[cyrusfox]

Not working for me. phoenix 1.48, ubuntu 11.04 64bits.

[01/07/2011 14:45:48] Finding inner ELF...
[01/07/2011 14:45:48] Patching inner ELF...
[01/07/2011 14:45:48] Patching instructions...
[01/07/2011 14:45:48] BFI-patched 472 instructions...
[01/07/2011 14:45:48] Patch complete, returning to kernel...
[01/07/2011 14:45:48] Applied BFI_INT patch
[01/07/2011 14:45:48] FATAL kernel error: Failed to load OpenCL kernel!

Not working for me. phoenix 1.5, Linuxcoin.

Yes I copied both files kernel.cl and __init__.py into kernels/phatk/

I get this when I try to start Phoenix(it will work fine with the original kernel files, 6850@930/830=250Mh/s, 6950@850/735= 340 Mh/s)

user@linuxcoin:/opt/miners/phoenix$ ./phoenix.py -u http://user:pass@bitcoinpool.com:8334/ -k phatk DEVICE=0 VECTORS AGGRESSION=7 -v FASTLOOP BFI_INT
[18/07/2011 09:27:56] Finding inner ELF...
[18/07/2011 09:27:56] Patching inner ELF...
[18/07/2011 09:27:56] Patching instructions...
[18/07/2011 09:27:56] BFI-patched 472 instructions...
[18/07/2011 09:27:56] Patch complete, returning to kernel...
[18/07/2011 09:27:56] Applied BFI_INT patch
[18/07/2011 09:27:56] FATAL kernel error: Failed to load OpenCL kernel!

Can anyone tell me what I am doing wrong? Thanks

It took setting up a second mining machine to figure out what I did wrong here. I had to type sudo the first time when I start the miner using the changed kernel file for some reason(really quite odd). Anyways my results
Card, prior hash, current hash
5770,212 Mhash to 229 Mhash(8%! improvement)
6950,341 Mhash to 344 Mhash

So if you are using linux and it gives you any sort of failure, try it again with super privileges, has worked everytime now. Thanks so much, I will make a humble donation.

[Sick_Boy]

with this kernel on my hd 5850 I reach 415MH/s, I was at 400 before with core at 1000MHz and memory at 350MHz

[Diapolo]

Hey guys, I'm making good progress for 69XX users and need a few to test my current version as I want to mature it before release.

- only 69XX users / users on VLIW4 design cards
- I need Win and Linux users
- I need SDK 2.1 / SDK 2.4 and SDK 2.5 users

Drop me a PM, if you want to participate and meet the requirements, I will then send you a DL link! If you have received the files the test results may be posted here and may be freely discussed, so that's no shitty NDA or something . I only want to make sure that the kernel works and can compete with phatk 2.1 .

I'm sorry to say, but 58XX users are better with phatk 2.1 for now :-/.

Regards,
Dia

[UrbanAdventurer]

Just updated. About +7Mh Thanks

[Diapolo]

New version was just released, it should be the fastest for 69XX cards:
Download version 2011-08-04 (pre-release): http://www.mediafire.com/?upwwud7kfyx7788

This is the preferred switch for Phoenix in order to achieve comparable performance:

Code:

-k phatk AGGRESSION=12 BFI_INT FASTLOOP=false VECTORS VECTORS2 WORKSIZE=256

or

Code:

-k phatk AGGRESSION=12 BFI_INT FASTLOOP=false VECTORS VECTORS2 WORKSIZE=128

Please test this version with SDK 2.4 / SDK 2.5! SDK 2.1 performance seems worse, but at least it should work. Report any errors and problems here and let me know what you think.
Have a look at your cards temperatures, I got a report, that they may be lower, which would be great .

Regards,
Dia

[Soak]

-k phatk VECTORS VECTORS2 BFI_INT WORKSIZE=128 AGGRESSION=11 DEVICE=1

Radeon 6970, 391 to 393. Nice

[st4rdust]

I modified the shipping phatk Kernel from Phoenix 1.50. I now get round about 13 MHash/s more on my 5830 (up from 310 to 323)!
The modified version seems to work, Hashes are accepted and no errors are observed so far.

Brilliant. I also run 5830s and got the same result as you, from ~310 to 322. Keep up the good work, myself and every other miner who uses phatk cannot overstate how much we appreciate your efforts.

[Clipse]

New version was just released, it should be the fastest for 69XX cards:
Download version 2011-08-04 (pre-release): http://www.mediafire.com/?upwwud7kfyx7788

This is the preferred switch for Phoenix in order to achieve comparable performance:

Code:

-k phatk AGGRESSION=12 BFI_INT FASTLOOP=false VECTORS VECTORS2 WORKSIZE=256

or

Code:

-k phatk AGGRESSION=12 BFI_INT FASTLOOP=false VECTORS VECTORS2 WORKSIZE=128

Please test this version with SDK 2.4 / SDK 2.5! SDK 2.1 performance seems worse, but at least it should work. Report any errors and problems here and let me know what you think.
Have a look at your cards temperatures, I got a report, that they may be lower, which would be great .

Regards,
Dia

Well I'll be damned, you do mention its for 69xx range but Im on 5850 and I got a boost of 9mh/s per card using the 08-04 pre-release versuin with VECTORS VECTORS4 , hot damn

Just to give everyone an idea of the gains:

stock phatk: 400mh maxed out (1000/180)
08-04 phatk: 440mh maxed out (1000/180)
08-04 phatk: 475mh maxed out (1100/180)
08-04 phatk: 500mh+ is my guess (1200/180) <---- This is in the pipeline, mostly just for interest sake I will be LN2 cooling it and running voltage at 1.4v card might die but 500mh will be sexy.

[littlekid]

Code:

-k phatk AGGRESSION=12 BFI_INT FASTLOOP=false VECTORS VECTORS2 WORKSIZE=256

My 5770@970/300 improve from 220 -> 223 MH/s
Cyclone 6850@1030/300 nothing changes, still 280 MH/s
not test ''VECTORS4'' switch yet
anyone have another configure switch for further improvement?

[Psychonautchn]

gotta try on my 5870, thx!

[Diapolo]

Updated 1st post kernel performance data with SDK 2.5 and KernelAnalyzer 1.9 Cal 11.7 profile.