cbuchner1
Not working with Tesla C2050
Hi, Thanks for the work on X11. I am trying to mine Darkcoin with Tesla C2050, Win 7 x64. It seems to me that the newest release 1.0 is using CPUs rather than GPUs. I did see some hash rate (very high), but no shares accepted in the pool.
The Tesla C2050 is Compute Capability 2.0, and that has been dropped for latest CCMiner releases IIRC.
a lot of algorithms in ccMiner wouldn't support Compute 2.0 at all (some rewrite/redesign needed)
2014-05-11 16:45 GMT+02:00 whuplus [email protected]:
Thanks Christian, just solved it by compiling with Compute 2.0.
— Reply to this email directly or view it on GitHubhttps://github.com/cbuchner1/ccminer/issues/6#issuecomment-42772390 .
Interesting. How did you compile it exactly? I tried running it at compute_20,sm_20, tried changing the MAXWELL_OR_FERMI constant, but it keeps erroring out with a message about Cuda Toolkit 5.5 not liking it. Will I need an older toolkit? BTW, just out curiosity, what's your hashrate at x11? ;)
Not at all. I just dropped support for Jackpotcoin, Animecoin and Quarkcoin, and deleted two 'compactionTest.cu' files.
Currently 1.3MH/s with the x64 release build (compute_20,sm_20).
whuplus, where did you disable those algorithms? I can't find the options for configure anywhere, in fact I can't even find any build documentation at all!
Actually, I rewrote "cpu-miner.c" line 866-918 and removed some sources files that cause compiling errors.
Oh yeah? Would you possibly be so kind as to publish a fork of this? (or at very least a patch file :) )
Sure, but I am not familiar with Github and it's very late. I'll try it tomorrow. ^_^
This is what I did for compute 2.x @ my fork: https://github.com/KBomba/ccminer-KBomba/commit/f91269c1d02e46e3172b30bfc1bf7474b03345d6 Many thanks for whuplus!
Thanks, compiling now! I had to rip out a few things from the Makefile, seeing as you're using an IDE to compile the code, only understandable.
Hey fellas, I'm running into same issue with the Makefile since I'm trying to compile for Ubuntu, I get the error below. cdp1337, can you explain what you ripped out to help it compile? I'm targeting x11 on Nvidia 560Ti's as my end goal. Thanks so much!
make[2]: *** No rule to make target JHA/cuda_jha_compactionTest.cu', needed byJHA/cuda_jha_compactionTest.o'. Stop.
p.s. I'm happy to throw a donation at whoever can help me compile this on Ubuntu 13.10, I'm sure it's an easy fix but I don't have time to learn all the make file and figure it out. $100 in BTC sound fair?
More info on this, I need it to compile for Computer Capability 2.1 since I'm using Nvidia 560ti's. I was trying to build KBomba's release: https://github.com/KBomba/ccminer-KBomba/releases/tag/v1.0.2x initially, but whatever will work best for my needs I'm totally open to options. Thoughts?
When compiling on Linux, the key is to modify Makefile.am. Alter the places where it says compute_35, sm_35 and compute_30, sm_30 into compute_21, sm_21
remove the all compactionTest.cu modules from Makefile.am
comment out the code lines where the (now missing) functions in compactionTest are called.
The Quark, Anime, Jackpotcoin algorithms will not work afterwards.
X11 and others remain functional.
2014-06-05 18:00 GMT+02:00 Daniel Driscoll [email protected]:
More info on this, I need it to compile for Computer Capability 2.1 since I'm using Nvidia 560ti's. I was trying to build KBomba's release: https://github.com/KBomba/ccminer-KBomba/releases/tag/v1.0.2x initially, but whatever will work best for my needs I'm totally open to options. Thoughts?
— Reply to this email directly or view it on GitHub https://github.com/cbuchner1/ccminer/issues/6#issuecomment-45238424.
Yeah, what they said. If I get time soon I'll put up a fork of my changes, but my rig that handled the nvidia compiling jobs got commandeered for use in the NAS, so instead I'm stuck with an ancient celeron processor for testing with :/
Well, I think you lost me at the "comment out code lines" part, because after adjusting as you described I get this, and can't seem to find the debug line that prints it:
gcc -std=gnu99 -DHAVE_CONFIG_H -I. -msse2 -fopenmp -pthread -fno-strict-aliasing -I./compat/jansson -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -g -O2 -MT ccminer-sha2.o -MD -MP -MF .deps/ccminer-sha2.Tpo -c -o ccminer-sha2.o test -f 'sha2.c' || echo './'sha2.c
mv -f .deps/ccminer-sha2.Tpo .deps/ccminer-sha2.Po
nvcc -g -O2 -I . -Xptxas "-abi=no -v" -gencode=arch=compute_35,code="sm_35,compute_35" -gencode=arch=compute_21,code="sm_21,compute_21" --maxrregcount=80 --ptxas-options=-v -I./compat/jansson -o heavy/heavy.o -c heavy/heavy.cu
nvcc fatal : Unsupported gpu architecture 'compute_21'
make[2]: *** [heavy/heavy.o] Error 255
make[2]: Leaving directory /home/lemma/cc102' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory/home/lemma/cc102'
make: *** [all] Error 2
hmm it would appear compute_21 is not a valid compute specification, use compute_20 and sm_21
Christian
2014-06-06 6:38 GMT+02:00 Daniel Driscoll [email protected]:
Well, I think you lost me at the "comment out code lines" part, because after adjusting as you described I get this, and can't seem to find the debug line that prints it:
gcc -std=gnu99 -DHAVE_CONFIG_H -I. -msse2 -fopenmp -pthread -fno-strict-aliasing -I./compat/jansson -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -g -O2 -MT ccminer-sha2.o -MD -MP -MF .deps/ccminer-sha2.Tpo -c -o ccminer-sha2.o test -f 'sha2.c' || echo './' sha2.c mv -f .deps/ccminer-sha2.Tpo .deps/ccminer-sha2.Po nvcc -g -O2 -I . -Xptxas "-abi=no -v" -gencode=arch=compute_35,code="sm_35,compute_35" -gencode=arch=compute_21,code="sm_21,compute_21" --maxrregcount=80 --ptxas-options=-v -I./compat/jansson -o heavy/heavy.o -c heavy/heavy.cu nvcc fatal : Unsupported gpu architecture 'compute_21' make[2]: *** [heavy/heavy.o] Error 255 make[2]: Leaving directory /home/lemma/cc102' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory/home/lemma/cc102' make: *** [all] Error 2
— Reply to this email directly or view it on GitHub https://github.com/cbuchner1/ccminer/issues/6#issuecomment-45302298.
Awesome, that did the trick. Now I've just gotten figure out where to comment out these function calls, will let you know when I do.
Finally figured it out, removing the compilation includes for quark, JHA, etc threw me for a bit. Thanks for the quick responses Christian, donation coming your way.
I am having a hard time figure out how to modify Makefile.am to compile for Compute 2.0. I am still not sure which includes I am supposed to take out, maybe I am completely misunderstanding what I need to do. I keep getting "error: identifier "__shfl" is undefined"
__shfl requires Compute 3.0
here's an older fork that removes the algorithms requiring Compute 3.0
https://github.com/KBomba/ccminer-KBomba
Christian
2014-09-04 2:24 GMT+02:00 Cameron Bravo [email protected]:
I am having a hard time figure out how to modify Makefile.am to compile for Compute 2.0. I am still not sure which includes I am supposed to take out, maybe I am completely misunderstanding what I need to do. I keep getting "error: identifier "__shfl" is undefined"
— Reply to this email directly or view it on GitHub https://github.com/cbuchner1/ccminer/issues/6#issuecomment-54388399.
Has it moved (__shfl requires Compute 3.0 here's an older fork that removes the algorithms requiring Compute 3.0 https://github.com/KBomba/ccminer-KBomba Christian) ??
Can anyone help meplease to get sources to be able to compile to compute_20 and sm_21, where can i find sources or bin for x86_64 linux ?
Yeah, I do have the older fork already, I just can not figure out how to modify the Makefile.am to get it to compile for a Compute 2.0 device.
On Mon, Sep 22, 2014 at 3:20 AM, nreal2 [email protected] wrote:
Or can someone post that Makefile.am that does the trick please?
— Reply to this email directly or view it on GitHub https://github.com/cbuchner1/ccminer/issues/6#issuecomment-56354397.
I could try that, but I suspect it doesnt compile I spend whole day trying but no luck, have those "older" cards here ...and no windowses never
2014-09-22 20:46 GMT+03:00 Cameron Bravo [email protected]:
Yeah, I do have the older fork already, I just can not figure out how to modify the Makefile.am to get it to compile for a Compute 2.0 device.
On Mon, Sep 22, 2014 at 3:20 AM, nreal2 [email protected] wrote:
Or can someone post that Makefile.am that does the trick please?
— Reply to this email directly or view it on GitHub https://github.com/cbuchner1/ccminer/issues/6#issuecomment-56354397.
— Reply to this email directly or view it on GitHub https://github.com/cbuchner1/ccminer/issues/6#issuecomment-56411130.