CCminer(SP-MOD) Modded GPU kernels.

[scryptr]

LINUX COMPILE ERROR --

I also get a compile error:



Something about conflicting types in line 696 of scrypt.c and line 354 of miner.h?

I can hardly wait to try it, sp_ .  Thanks for your work!       --scryptr

[sp_]

The autotune function in scrypt needs cuda compute 2.1 to work. I think this is removed in the latest cuda kit. Downlad cuda 6.5.

I added more mulitthreading tweaking to all the kernals. quark with -g 3 is 6-8MHASH on the 750ti standard clocked in benchmark mode. with no cpu validation errors.
On the pool the hashrate is lower, but the cpu is near 100%, so I think I need to add some cpu sleeping

Looks like I have managed to remove the does not validate on the cpu..

edit: got one does not validate after 5 minutes of hashing on the pool

 (quark)

[DougB62]

The autotune function in scrypt needs cuda compute 2.1 to work. I think this is removed in the latest cuda kit. Downlad cuda 6.5.

I added more mulitthreading tweaking to all the kernals. quark with -g 3 is 6-8MHASH on the 750ti standard clocked in benchmark mode. with no cpu validation errors.
On the pool the hashrate is lower, but the cpu is near 100%, so I think I need to add some cpu sleeping

Using 6.5 here... but getting build error

[KL0nLutiy]

The autotune function in scrypt needs cuda compute 2.1 to work. I think this is removed in the latest cuda kit. Downlad cuda 6.5.

I added more mulitthreading tweaking to all the kernals. quark with -g 3 is 6-8MHASH on the 750ti standard clocked in benchmark mode. with no cpu validation errors.
On the pool the hashrate is lower, but the cpu is near 100%, so I think I need to add some cpu sleeping

Using 6.5 here... but getting build error

Delete compute_52,sm_52 in ccminer.vcxproj if you have gtx 750 ti

[arpika]

LINUX COMPILE ERROR --

I also get a compile error:

http://s1.postimg.org/basdqn6cv/error.png

Something about conflicting types in line 696 of scrypt.c and line 354 of miner.h?

I can hardly wait to try it, sp_ .  Thanks for your work!       --scryptr

Getting the same build error (linux, CUDA 6.5)

[scryptr]

LINUX COMPILE ERROR --

I also get a compile error:



Something about conflicting types in line 696 of scrypt.c and line 354 of miner.h?

I can hardly wait to try it, sp_ .  Thanks for your work!       --scryptr

Getting the same build error (linux, CUDA 6.5)

JUST NOW --

SP_  just finished depositing a newer build on GitHub, but I still get the same conflict error when attempting to build it.  I do have Cuda 6.5.       --scryptr

[sp_]

LINUX COMPILE ERROR --

I also get a compile error:



Something about conflicting types in line 696 of scrypt.c and line 354 of miner.h?

I can hardly wait to try it, sp_ .  Thanks for your work!       --scryptr

Getting the same build error (linux, CUDA 6.5)

JUST NOW --

SP_  just finished depositing a newer build on GitHub, but I still get the same conflict error when attempting to build it.  I do have Cuda 6.5.       --scryptr

You are not building the latest source... Submitted a fix for the intensity parameter.

With -g 3 quark can run -i 22 without any problems. This is because the memory is fragmented, so the gpu doesn't need to allocate a big continuous block.

I get higher rates on the pool now  

[scryptr]

LINUX COMPILE ERROR --

I also get a compile error:



Something about conflicting types in line 696 of scrypt.c and line 354 of miner.h?

I can hardly wait to try it, sp_ .  Thanks for your work!       --scryptr

Getting the same build error (linux, CUDA 6.5)

JUST NOW --

SP_  just finished depositing a newer build on GitHub, but I still get the same conflict error when attempting to build it.  I do have Cuda 6.5.       --scryptr

You are not building the latest source... Submitted a fix for the intensity parameter.

With -g 3 quark can run -i 22 without any problems. This is because the memory is fragmented, so the gpu doesn't need to allocate a big continuous block.

I get higher rates on the pool now  

SAME ERROR --

I just pulled from Git and attempted to compile, and the "type conflict" is still there.  I did see your Intensity fix comments. That is 3 commits today?       --scryptr

[sp_]

Just submitted commit number 6 today. you have the wrong version of the file ccminer.cpp. Get this file from the head and build

[scryptr]

Just submitted commit number 6 today. you have the wrong version of the file ccminer.cpp. Get this file from the head and build

SP_ --

I must be misunderstanding.  I did a clean git clone into an empty direcory and ran build.sh.  I see your most recent comments.  However, the "type conflict" is still there, the build fails.  Here is my command line for the git, executed in an empty build directory:

       "git clone https://github.com/sp-hash/ccminer ."


If you mean "pull it from your head", I am afraid you are out of reach.  Any use of pliers or suction would likely cause irrepairable damage, and your golden eggs would cease.       --scryptr

[sp_]

hehe. sorry, I corrected it now. The makefile was including an old file

[scryptr]

hehe. sorry, I corrected it now. The makefile was including an old file

UNDEFINED REFERENCE --

Another error exists:



Thanks for your attention to this.         --scryptr

[KL0nLutiy]

With the last build from git also a lot of does not validate on cpu with -g 3 on 750 ti on quark and other algos too.

[sp_]

Yes there still are some. Try with -g 2.

[KL0nLutiy]

Yes there still are some. Try with -g 2.

Does not validate on cpu also with -g 2

[sp_]

I reverted the quark syncronization code. Seems to be bether now. Let the miner run for a while with vardiff.
 
It looks like the "does not validate..." comes more often on low difficulty shares.

[scryptr]

CCMINER -g 3 -i 22 --

It works now.  The build was a success:


1) 6x750ti FTW Rig, command line "ccminer -a quark -g 3 -i 22 -o URL -u USER -p PSWD"


2) 6x750ti FTW Rig, command line "ccminer -a quark -g 3 -i 22 -o URL -u USER -p PSWD"

I used your suggested configuration to achieve this result.  The "CPU..." errors settle down after a few minutes.  Some accepted hashes are 70Mh/s+.  I'd say that the rig is averaging 55Mh/s, up from 36.4Mh/s yesteday.  That is about 2x290X hash rate improvement from yesterday.       --scryptr  

[sp_]

I will let my test rig run overnight and check if it crashes.

-g 3 -i 22 !!

Verified higher rates at the pool.

The miner is working fine in benchmark mode without the stratum connection, perhaps you should try to solomine? Until I figure out what's wrong.

Don't forget to donate guys. I am to lazy to make a percentage miner right now.


Your support is my motivation.

[scryptr]

I will let my test rig run overnight and check if it crashes.

-g 3 -i 22 !!

Verified higher rates at the pool.

The miner is working fine in benchmark mode without the stratum connection, perhaps you should try to solomine? Until I figure out what's wrong.

Don't forget to donate guys. I am to lazy to make a percentage miner right now.


Your support is my motivation.

SP_ --

THANKS!  I flipped another two-bits your way.  Please continue to tune up this miner!       --scryptr

[AliMan]

For me it crashes when trying to run all cards together in a single miner, running individually is fine, but I don't see any 25% boost or anything like that.