CCminer palgin's mod (alexis78 fork)

[palgin]

Will start my own thread, can't guarantee I'd be able to keep it up-to-date, but anyway.

My mod now is based on alexis78 version, but lots of algos reworked.

Currently supported algos:

blake      
decred      
blakecoin
gostd  
vcash
blake2s  
keccak
lyra2  
lyra2v2  
skein  
skein2  
skunk  
nist5
quark
qubit
x11
c11
sib
x11evo
x13
x14
x15
x17
whirlpool
lbry
neoscrypt
sia
myr-gr
veltor

Latest releases (and pre-releases):

https://github.com/palginpav/ccminer/releases

[1715089809]

1715089809

[1715089809]

1715089809

[1715089809]

1715089809

[1715089809]

1715089809

[pallas]

watching this! thanks!
will be testing on linux as soon as sources are available.

[Kompik]

As I have already wrote in PM. I have tested two versions on the MSI GTX 1080, mining hush on Yiimp. power limit 65% ( 150W), +160 core clock (2088) , +537 Mhz ( 5049) memory.

Getting about 820 sols with v3, 730 sols with v2. Both stable, but on pool after 30 mins of mining still showing under 400 sols on pool for version 3. For v2 350 sols.

Also according to GPU-Z it seems that there is only 80-86% GPU load so running more instances would probably help?.

[palgin]

I appreciate any feedback, thank you! For me pool-side speed correlates in time with miner shown, but it crashes in 15-30 mins from start, not fully safe memory operations, need to check block-sync again and again and find another way to perform less operations with more stability. Anyway, I'm still learning and this is not practically used instrument at that moment 

[sp_]

How fast is it on the gtx 1070? The fastest private kernels are doing 550 with low power.

[maxim000]

on 1070 version 1 gives ~560 h/s
versions 2 and 3 gives 580 and 610 h/s, but often crashes , and pool hashrate same as version 1


oops, version 1 crashed too ))

[chrysophylax]

will be compiling and testing sometime this week if we can ...

new cards are coming this week also - so will check the miner on those also ...

#crysx

[Kompik]

On zpool I am getting the same hashrate as @miner, but v2 and v3 crash with illegal memory access error very fast ( usually lasts less than 2 mins), v1 works well so far.

v1 635 H/s with same settings as above.

[sp_]

[2017-05-08 18:19:35] NVML GPU monitoring enabled.
[2017-05-08 18:19:35] Starting on stratum+tcp://eu1-zcash.flypool.org:3333
[2017-05-08 18:19:36] NVAPI GPU monitoring enabled.
[2017-05-08 18:19:36] 1 miner thread started, using 'equihash' algorithm.
[2017-05-08 18:19:36] Failed to get extranonce2_size
[2017-05-08 18:19:36] Failed to get extranonce2_size
[2017-05-08 18:19:36] ...retry after 30 seconds
[2017-05-08 18:20:06] Failed to get extranonce2_size
[2017-05-08 18:20:18] Failed to get extranonce2_size
[2017-05-08 18:20:18] ...retry after 30 seconds

[palgin]

As I thought, ~15% of memory-related operations give wrong results (v1 inspection):

https://ibb.co/c6Eb85

Should replace unsafe memory operations another way, need some time to think about workaround for this.

[zjy]

Will start my own thread, can't guarantee I'd be able to keep it up-to-date.

My mod is 1:1 tpruvot's version, except timetravel10 and Equihash algos, maybe there'll be future mods to other algos as well. I won't release Equihash kernel source on GitHub until stable release

Latest releases (and pre-releases):

https://github.com/palginpav/ccminer/releases

Currently really need testers with different equipment to help me debug Equihash kernel momory issues, seems kinda unstable right now.

If you'd like to help, download latest pre-release and post your results here. Check all 3 versions, post speed without overclock, speed with overclock (with core+/mem+ values) and crash output (if occures).

Newbie acc, I know, but I'll try my best 

P.S: sorry for my bad English, not my primary language.

ubuntu 16.04 build error

make[2]: *** No rule to make target 'x11/bitcore.cu', needed by 'x11/bitcore.o'.  Stop.
make[2]: Leaving directory '/home/home2/btx'
Makefile:2113: recipe for target 'all-recursive' failed
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory '/home/home2/btx'
Makefile:631: recipe for target 'all' failed
make: *** [all] Error 2

[joaogti36]

v3 with some gtx 1070 gives about 530sols and after a few minutes... crashs.....

[palgin]

ubuntu 16.04 build error

make[2]: *** No rule to make target 'x11/bitcore.cu', needed by 'x11/bitcore.o'.  Stop.
make[2]: Leaving directory '/home/home2/btx'
Makefile:2113: recipe for target 'all-recursive' failed
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory '/home/home2/btx'
Makefile:631: recipe for target 'all' failed
make: *** [all] Error 2

Ooops, haven't mention that I'm new to github and can forget something here and there, change bitcore.cu to timetravel10.cu in makefile.am and this should be fine. Anyway, I'm working only on "windows" branch right now, can't guarantee you'll  be able to build this under Linux systems.

[Amph]

are you planning to boost some algo?,sweepstakecoin this new coin has the old JHA algo, which could see a boost i think

[palgin]

are you planning to boost some algo?,sweepstakecoin this new coin has the old JHA algo, which could see a boost i think

Maybe in future, currently I'm sitting counting days to my vacation trip and don't have enough work to do, so I decided to practice my skills and try adding something to crypto community, also I'm a miner myself (not so huge as SP, only 20x5 Nvidia cards farms).

[sp_]

[2017-05-08 19:17:38] accepted: 19/19 (diff 6.005), 258.94 H/s yes!
[2017-05-08 19:17:39] GPU #0: solver: CUDA error 'an illegal memory access was encountered' in func 'eq_cuda_context<RB, SM, SSM, THREADS, PACKER>::so
lve' line 2054
[2017-05-08 19:17:41] GPU #0: an illegal memory access was encountered
[2017-05-08 19:17:41] GPU #0: scanhash_equihash:188 all CUDA-capable devices are busy or unavailable
[2017-05-08 19:17:41] GPU #0: init: CUDA: failed to alloc memory

[sp_]

The results doesn't seem to validate on the pool.

When I run 4 instances I get high numbers on all the windows, but on the pool low numbers..
Low difficulty seem to make the miner more stable.

launch batfile:

:loop
ccminer -a equihash -o stratum+tcp://equihash.mine.zpool.ca:2142 -u 1CTiNJyoUmbdMRACtteRWXhGqtSETYd6Vd -i 20 -p d=32
goto loop

[palgin]

The results doesn't seem to validate on the pool.

When I run 4 instances I get high numbers on all the windows, but on the pool low numbers..
Low difficulty seem to make the miner more stable.

launch batfile:

:loop
ccminer -a equihash -o stratum+tcp://equihash.mine.zpool.ca:2142 -u 1CTiNJyoUmbdMRACtteRWXhGqtSETYd6Vd -i 20 -p d=32
goto loop

Thanks, I found the reason in unsafe shared mem operations I used in kernel, working on a solution, but it will take long to make it safe without huge affect on speed.

[djeZo]

What you most likely did to achieve high speed is to make buckets smaller. This can of course greatly speed up everything, and displayed iterations are high, but on pool, you will get slow speed. These buckets sizes are there for a reason and not something you can simply shed off and benefit.

I also don't understand why would you have pool issues; equihash protocol is standardized, every pool works if it follows official zcash stratum specifications.

Also, we have released the source under GPL v3 and you are obligated to disclose full source when distributing the bins...

[palgin]

What you most likely did to achieve high speed is to make buckets smaller. This can of course greatly speed up everything, and displayed iterations are high, but on pool, you will get slow speed. These buckets sizes are there for a reason and not something you can simply shed off and benefit.

I also don't understand why would you have pool issues; equihash protocol is standardized, every pool works if it follows official zcash stratum specifications.

Also, we have released the source under GPL v3 and you are obligated to disclose full source when distributing the bins...

Yes, you're right, that's exactly what I've done and found that won't work the right way, I'm still learning and practicing my skills, sorry for practicing on your work, which is highly appreciated by me. I've already removed the bins, think it's time to try to write my own kernels from the scratch, SKL and MKL versions.