Pages:
Author

Topic: [BBR] Boolberry GPU Miner Discussion - AMD & Nvidia - OpenCL & CUDA - page 21. (Read 82987 times)

full member
Activity: 212
Merit: 100
Added the release. It works on bbr.farm:1111 and makes about 80% of solo version.

1. I didn't change the get_whole_scratchpad so if it has a bug (looks like it has) it's better to contact the main branch developer as well.
2. m_job.difficulty shouldn't be zero. If it is - something's wrong with a pool. I added a return from the worker thread as it doesn't make any sense to search for nonce having wrong data.
updated github

Thanks. Can you post another windows release with this please? I don't have an environment ready to make a build and someone was asking me for it.
Edit: Actually this makes it so it never gets work it looks like.. unless it's just me
Weird. I was testing there, too. Just kept looping "Job difficulty is 0. Skipping the job." Guess it IS just me. What could be different?
hero member
Activity: 518
Merit: 500
Added the release. It works on bbr.farm:1111 and makes about 80% of solo version.

1. I didn't change the get_whole_scratchpad so if it has a bug (looks like it has) it's better to contact the main branch developer as well.
2. m_job.difficulty shouldn't be zero. If it is - something's wrong with a pool. I added a return from the worker thread as it doesn't make any sense to search for nonce having wrong data.
updated github

Thanks. Can you post another windows release with this please? I don't have an environment ready to make a build and someone was asking me for it.
Edit: Actually this makes it so it never gets work it looks like.. unless it's just me

Right on....

Any plans on making this working with CryptoNote multiminer instead?
mbk
member
Activity: 106
Merit: 10
Added the release. It works on bbr.farm:1111 and makes about 80% of solo version.

1. I didn't change the get_whole_scratchpad so if it has a bug (looks like it has) it's better to contact the main branch developer as well.
2. m_job.difficulty shouldn't be zero. If it is - something's wrong with a pool. I added a return from the worker thread as it doesn't make any sense to search for nonce having wrong data.
updated github

Thanks. Can you post another windows release with this please? I don't have an environment ready to make a build and someone was asking me for it.
Edit: Actually this makes it so it never gets work it looks like.. unless it's just me
full member
Activity: 212
Merit: 100
1. I didn't change the get_whole_scratchpad so if it has a bug (looks like it has) it's better to contact the main branch developer as well.
2. m_job.difficulty shouldn't be zero. If it is - something's wrong with a pool. I added a return from the worker thread as it doesn't make any sense to search for nonce having wrong data.
updated github

Thanks. Can you post another windows release with this please? I don't have an environment ready to make a build and someone was asking me for it.
Edit: Actually this makes it so it never gets work it looks like.. unless it's just me
legendary
Activity: 1064
Merit: 1000
is there any miner for cuda compute 1.2?
cuda developper kit dont install on my pc Cry Cry Cry
mbk
member
Activity: 106
Merit: 10
1. I didn't change the get_whole_scratchpad so if it has a bug (looks like it has) it's better to contact the main branch developer as well.
2. m_job.difficulty shouldn't be zero. If it is - something's wrong with a pool. I added a return from the worker thread as it doesn't make any sense to search for nonce having wrong data.
updated github
mbk, can you add m_scratchpad.clear(); at the beginning of the get_whole_scratchpad function. if the miner gets to a point where it wants to download the whole scratchpad again, it just appends instead of starts over.
Also, on my nvidia card it crashes because difficulty is zero at the start and it divides by that. I just put put a check for if zero, set to something else and it doesn't happen again.
full member
Activity: 212
Merit: 100
mbk, can you add m_scratchpad.clear(); at the beginning of the get_whole_scratchpad function. if the miner gets to a point where it wants to download the whole scratchpad again, it just appends instead of starts over.
Also, on my nvidia card it crashes because difficulty is zero at the start and it divides by that. I just put put a check for if zero, set to something else and it doesn't happen again.
member
Activity: 81
Merit: 1002
It was only the wind.
sr. member
Activity: 330
Merit: 252
Scratch that, it works!
710kh/s on a single 750Ti at stock clocks so far - saw 716 poolside!

NSFW
https://ottrbutt.com/tmp/bbrcudaminer.png

foxy! and I mean the numbers...  Cool
member
Activity: 81
Merit: 1002
It was only the wind.
Got a 42.6% increase on my 860M, which is basically a 750Ti. That took a while.

Proof: https://ottrbutt.com/tmp/bbropenclstock2.png vs. https://ottrbutt.com/tmp/bbropenclwolf2.png

wow! get it on... you do a lot of cool stuff at the moment (tsiv CN).
If your BBR version would be easy to compile for nvidia linux guys I would be more than happy.

It's simply a modification to mbk's OpenCL kernel. Nothing else needed.
mbk
member
Activity: 106
Merit: 10
Thanks for info. I found out Nvidia included libOpenCL.so in their driver not toolkit prior to cuda6. We need different cmake script to get them. I think it's OK to require cuda6 to make things simple. I'll add it to readme.

Looks like the OPENCL_LIBRARY is undefined. It should be in line 149. Edit FindOpenCL.cmake. Put debug messages after the find_library call.

mbk, you would not belive it - I could compile your miner!
I located the problem with the help of your debug messages. thanks... this was really helpful.

1) It is not possible to compile your miner with cuda 5.5 (as used in kopiemtu).
There is no libOpenCL.so as you expect in usr/local/cuda/lib64/
So if you don't know a workaround for cuda 5.5, it would be helpfull to warn people in the readme.

2) After switching to cuda 6.0 it was easy to compile a working miner.
I tried cuda 6.0 before you made your code compatible for the "older" openCL versions - but not again after the fix.  Undecided


sr. member
Activity: 330
Merit: 252
Looks like the OPENCL_LIBRARY is undefined. It should be in line 149. Edit FindOpenCL.cmake. Put debug messages after the find_library call.

mbk, you would not belive it - I could compile your miner!
I located the problem with the help of your debug messages. thanks... this was really helpful.

1) It is not possible to compile your miner with cuda 5.5 (as used in kopiemtu).
There is no libOpenCL.so as you expect in usr/local/cuda/lib64/
So if you don't know a workaround for cuda 5.5, it would be helpfull to warn people in the readme.

2) After switching to cuda 6.0 it was easy to compile a working miner.
I tried cuda 6.0 before you made your code compatible for the "older" openCL versions - but not again after the fix.  Undecided

newbie
Activity: 26
Merit: 0
Finally I managed to make it run. Basically I have deleted files, make git clone and re-compile the miner. This time I have used command make instead of make -j 5.

Few advices for installing on AMD/Linux. Don't use getPimp or Bamt. They are based upon old Debian. The only way to get required dendencies is by mixing stable and testing packages and that creates less than ideal environment.

Better way is:
1. unplugg all but first gpu  (otherwise Ubuntu fails to install)
2. install Ubuntu using stock drivers
3. google "unofficial AMD community" and get instructions how to install latest AMD driver
4. shutdown computer and plug in all gpus
5. re-execute sudo aticongif --adapter=all --init -f and reboot
6. install AMD APP SDK
7. install AMD ADL (if required, e.g. for Sgminer)
8. follow instructions on boolberry.com
9. before make command, execute export AMDAPPSDKROOT=/opt/AMDAPP
10. create folder, e.g. mkdir dr1
11. execute deamon using folder dr1, by doing that deamon creates json file
12. exit deamon
13. edit json file and change thread delay, make it 4000
14. create wallet using simplewallet and execurte command address to get wallet address
15. start deamon using folder dr1, wallet address and specifying desired number of mining threaths

basically you may follow any web guide for installation of mining rig, but you must unplug all but first gpu before initial Ubuntu installation

if you are sattisfied with AMD 13.12 drivers, you may install stock UBUNTU 13.10 and follow this guide

https://github.com/joshpatten/mining-bootstrap

afterwards proceed with instructions on boolberry.com

I am getting 811500Hs on Sapphire 7950 at stock frequencies drawing 148W from the wall.

mbk
member
Activity: 106
Merit: 10
It works on Windows as is but try to change --rpc-bind-port and --p2p-bind-port command line options:
Code:
./boolbd --data-dir dr1 --start-mining=1F9hbj8wp... --mining-threads=1 --rpc-bind-port=10102 --p2p-bind-port=10101
./boolbd --data-dir dr2 --start-mining=1F9hbj8w... --mining-threads=1 --rpc-bind-port=10112 --p2p-bind-port=10111
and so on. And don't forget to increment device_index starting from 0. Wait mining start before running next instance.
I have set thread_delay to 4000, but still does not work.

Is anybody else using latest beta 14.6 drivers?

Is there a way to manually start single instance with mining_threat=1 that would bind to 2nd or 3rd gpu? Perhapss using  "device_index" field? Because in thatt case I would manually start 5 instances of miner.
member
Activity: 81
Merit: 1002
It was only the wind.
Got a 42.6% increase on my 860M, which is basically a 750Ti. That took a while.

Proof: https://ottrbutt.com/tmp/bbropenclstock2.png vs. https://ottrbutt.com/tmp/bbropenclwolf2.png
newbie
Activity: 26
Merit: 0
I have set thread_delay to 4000, but still does not work.

Is anybody else using latest beta 14.6 drivers?

Is there a way to manually start single instance with mining_threat=1 that would bind to 2nd or 3rd gpu? Perhapss using  "device_index" field? Because in thatt case I would manually start 5 instances of miner.
newbie
Activity: 29
Merit: 0
newbie
Activity: 26
Merit: 0
I am trying to execute following:

./boolbd  --start-mining=1F9hbj8w... --mining-threads=1
./boolbd --data-dir dr1 --start-mining=1F9hbj8wp... --mining-threads=1
./boolbd --data-dir dr2 --start-mining=1F9hbj8w... --mining-threads=1
./boolbd --data-dir dr3 --start-mining=1F9hbj8....  --mining-threads=1
./boolbd --data-dir dr4 --start-mining=1F9hb... --mining-threads=1

first instance works, but second do not, I am getting following error message

2014-Jul-27 01:55:55.783175 ERROR /home/klemen/boolberry-opencl/contrib/epee/include/net/abstract_tcp_server2.inl:475[bool epee::net_utils::boosted_tcp_server::init_server(uint32_t, std::string) [with t_protocol_handler = epee::levin::async_protocol_handler >; uint32_t = unsigned int; std::string = std::basic_string]]Exception at [boosted_tcp_server::init_server], what=bind: Address already in use
2014-Jul-27 01:55:55.783220 ERROR /home/klemen/boolberry-opencl/src/p2p/net_node.inl:262[bool nodetool::node_server::init(const boost::program_options::variables_map&) [with t_payload_net_handler = currency::t_currency_protocol_handler]]Failed to bind server
2014-Jul-27 01:55:55.783245 ERROR /home/klemen/boolberry-opencl/src/daemon/daemon.cpp:150[int main(int, char**)]Failed to initialize p2p server.



also setting   "thread_delay": 2000, and using 5 mining threads do not work


{
  "current_extra_message_index": 0,
  "device_index": 0,
  "difficulty": 16777215,
  "donation_decision": true,
  "donation_decision_made": false,
  "donation_descision": true,
  "donation_percent_opencl": "2.0",
  "kernel": "wild_keccak.cl",
  "kernel_type": 0,
  "platform_index": 0,
  "thread_delay": 2000,
  "work_size": 262144
}


mbk
member
Activity: 106
Merit: 10
newbie
Activity: 26
Merit: 0
after reading this important info:

Copy *.cl kernels from src/cl to executable's directory.

I am getting following:

2014-Jul-26 23:22:00.628664 Miner thread was started
2014-Jul-26 23:22:00.628745 [miner 0][OCL 0/0] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:22:00.734255 Miner thread was started [1]
2014-Jul-26 23:22:00.734388 [miner 1][OCL 0/1] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:22:00.734422 [miner 1][OCL 0/1] error:  CL_INVALID_VALUE
2014-Jul-26 23:22:00.734445 [miner 1][OCL 0/1] platform index must me between 0 and 4294967295
2014-Jul-26 23:22:00.828791 Miner thread was started [2]
2014-Jul-26 23:22:00.828856 [miner 2][OCL 0/2] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:22:00.828879 [miner 2][OCL 0/2] error:  CL_INVALID_VALUE
2014-Jul-26 23:22:00.828897 [miner 2][OCL 0/2] platform index must me between 0 and 4294967295
2014-Jul-26 23:22:00.936981 Miner thread was started [3]
2014-Jul-26 23:22:00.937955 [miner 3][OCL 0/3] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:22:00.938108 [miner 3][OCL 0/3] device index must me between 0 and 0
2014-Jul-26 23:22:01.028934 [RPC1]Mining has started with 5 threads, good luck!
2014-Jul-26 23:22:01.037096 Miner thread was started [4]
2014-Jul-26 23:22:01.037270 [miner 4][OCL 0/4] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:22:01.037396 [miner 4][OCL 0/4] device index must me between 0 and 0
hr: 0, efficiency: 0% (shares 0/0, blocks 0/0)
2014-Jul-26 23:22:01.591695 [miner 0][OCL 0/0] using platform: AMD Accelerated Parallel Processing
2014-Jul-26 23:22:01.591750 [miner 0][OCL 0/0] using device:Tahiti
2014-Jul-26 23:22:01.591772 [miner 0]Share target (diff 16777215) 0000010000010000
hr: 349234, efficiency: 0% (shares 0/0, blocks 0/0)
hr: 494446, efficiency: 0% (shares 0/0, blocks 0/0)
hr: 567364, efficiency: 0% (shares 0/0, blocks 0/0)

It seams that first gpu is working and other are not.
Pages:
Jump to: