OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.1

[-ck]

Running 1.3.0 on Fedora 14 x86.. Had it hang last night, and no either awareness of the hang, and/or no attempt at recovery.. Here's what I saw on the screen:
Ideas?

Thread restart was temporarily disabled on 1.3.0 pending further changes. It's been reimplemented in 1.3.1 so please upgrade. Sometimes the GPU code just gets stuck and the GPU never returns, killing off the thread. That's why I put the thread watchdog in there but I had to disable it in 1.3.0 till more code went in to match architectural changes.

[loglow]

So it seems I can only run the program properly with my current directory set to /opt/cgminer-1.3.1/ for the time being.

I hate to bump my own post, but has anyone else experienced this? Am I missing something obvious? Thanks.

[zaytsev]

I hate to bump my own post, but has anyone else experienced this? Am I missing something obvious? Thanks.

Yes, I have the same problem, it tries to find the *.cl files in the current directory, should probably try to do basename argv[0] instead.

[Calibre]

Reboot?

Yes - that worked. All running fine now.

So the policy of "turn it off and turn it on again" actually works!

[loglow]

I hate to bump my own post, but has anyone else experienced this? Am I missing something obvious? Thanks.

Yes, I have the same problem, it tries to find the *.cl files in the current directory, should probably try to do basename argv[0] instead.

This leads me back to the issue that cgminer does not use ncurses output and segfaults on exit when run from /usr/local/bin/ which is where it installs itself (make install). Please correct me if I'm wrong, but the executable should look for the *.cl files in either its own directory (not the current user's directory) and/or using the system paths, no?

[EskimoBob]

2) Running the program from /usr/local/bin/ results in non-ncurses output (without flag) and the program segfaults on exit (ctrl-c) without displaying statistics.

If i use the screen and start the program, I can never exit properly and get the stats.
 
BTW, I did not run install on cgminer so it is not in /opt/ or   /usr/local/bin/

[ancow]

Hi,

for JSON a sensible multi-host config would look like this:

{
…
connections: [
   {url: "http://foo.example.org:8332", username: "bar", password: "baz"},
   {url: "http://bar.example.org:8332", username: "baz", password: "quux"}
],
…
}

did you get a multihost config to work?

I try to make a new example-conf.json but

I don't... cgminer says no server specified

This was an idea, not something that is actually implemented. Currently, you need to create individual JSON configs for each pool and pass them with -c parameters.

[de4l]

tried it on windows for kicks but get "Error: Getting Device IDs (num)"

[Tasty Champa]

I use a windows batch file to run it.

anyone windows can use it too.
pastes this into bat file you make inside the cgminer directory

Code:

@echo off
:A
Cls
cgminer -o http://pit.deepbit.net:8332 -u tasty@wtfiswrongwu.com -p passwurd -g 10 -w 256 -d 0 

I haven't been able to figure out kernel, work size GPU and CPU threads yet though.
I need to figure out how many threads my GPU has first though.

oh just figured out GPU threads and worksize. -g 10 -w 256

MAN, that works good!

[xcooling]

what gpus are you using ?

these are mine:

SLI Nvidia 9800gtx+ / SLI GTS250

Code:

del /F /Q *.bin
cgminer.exe -o http://server:port -u username -p pass -w 64 -t 0 -I 4

XFire 6970 / XFIRE 6950

Code:

del /F /Q *.bin
cgminer.exe -o http://server:port -u username -p pass -I 9 -t 0

[The00Dustin]

I'm not having much luck compiling this on Fedora 15.  This is where make craps out:

make[2]: Entering directory `/usr/src/cgminer-1.3.1'
gcc -DHAVE_CONFIG_H -I. -pthread -fno-strict-aliasing  -I./lib -I./lib  -DHAS_YASM -g -O2 -MT cgminer-main.o -MD -MP -MF .deps/cgminer-main.Tpo -c -o cgminer-main.o `test -f 'main.c' || echo './'`main.c
main.c: In function âload_configâ:
main.c:605:2: error: too many arguments to function âjson_load_fileâ
/usr/include/jansson.h:162:9: note: declared here
make[2]: *** [cgminer-main.o] Error 1
make[2]: Leaving directory `/usr/src/cgminer-1.3.1'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/usr/src/cgminer-1.3.1'
make: *** [all] Error 2


I looked at the MakeFile and if I edit it as follows:

-#JANSSON_INCLUDES =
-JANSSON_INCLUDES = -I$(top_srcdir)/compat/jansson
+JANSSON_INCLUDES =
+#JANSSON_INCLUDES = -I$(top_srcdir)/compat/jansson


then I can compile successfully, but will get this upon running cgminer:

[2011-07-20 12:41:07] Error: Getting Device IDs (num)


I am running an nVidia card at the moment, but poclbm and phoenix can each access it.  Also, for the record, 'rpm -qf /usr/include/jansson.h' returns these packages:

jansson-devel-1.2-1.fc13.x86_64
jansson-devel-1.2-1.fc13.i686


On a side note, cgminer seems to use a lot of CPU for GPU mining in Windows, and all found hashes get rejected.  Same nVidia card, and I know phoenix gets a lot of rejects too if I use BFI_INT.  I saw that BFI_INT is used in the original post, but I didn't see any way to disable it in the parameters.  Is it possible to disable it, or is this miner meant for ATI only?

[error]

I'm not having much luck compiling this on Fedora 15.  This is where make craps out:

You have a conflict between jansson-1.2 which is included with Fedora and jansson-2.0 which comes with cgminer. Remove the jansson-devel packages and you'll be able to finish the build.

but will get this upon running cgminer:

[2011-07-20 12:41:07] Error: Getting Device IDs (num)


I am running an nVidia card at the moment, but poclbm and phoenix can each access it.

How did you install the NVIDIA drivers and CUDA toolkit?

[burp]

Current version (d987006cd82900ecb1e6d9d2743d832a595b0c43) segfaults for me with two pools and --load-balance --no-longpoll:

Code:

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7965114 in pthread_mutex_lock () from /lib/x86_64-linux-gnu/libpthread.so.0
(gdb) bt
#0  0x00007ffff7965114 in pthread_mutex_lock () from /lib/x86_64-linux-gnu/libpthread.so.0
#1  0x00000000004093c8 in tq_push (tq=0x0, data=<value optimized out>) at util.c:602
#2  0x000000000040971e in json_rpc_call (curl=0x9c2ae0, url=0x7fffffffe790 "http://bitparking.com:9098",
    userpass=0x64cbc0 "N1YWEnjSqNeScnzGPMysJxvqbWAEhXzWYS:x", rpc_req=<value optimized out>, probe=<value optimized out>, longpoll=false,
    pool=0x64da60) at util.c:352
#3  0x000000000040590e in pool_active (pool=0x64da60) at main.c:1467
#4  0x00000000004038a4 in main (argc=1, argv=<value optimized out>) at main.c:2767

It also seems that things get mixed up when one pool supports long polling, while the other one does not.

[-ck]

2) Running the program from /usr/local/bin/ results in non-ncurses output (without flag) and the program segfaults on exit (ctrl-c) without displaying statistics.

If i use the screen and start the program, I can never exit properly and get the stats.
 
BTW, I did not run install on cgminer so it is not in /opt/ or   /usr/local/bin/

It does not support "make install" just yet.

[-ck]

I use a windows batch file to run it.

anyone windows can use it too.
pastes this into bat file you make inside the cgminer directory

Code:

@echo off
:A
Cls
cgminer -o http://pit.deepbit.net:8332 -u tasty@wtfiswrongwu.com -p passwurd -g 10 -w 256 -d 0 

I haven't been able to figure out kernel, work size GPU and CPU threads yet though.
I need to figure out how many threads my GPU has first though.

oh just figured out GPU threads and worksize. -g 10 -w 256

MAN, that works good!

I suggest you don't give it a -g parameter at all. The default of 2 is often the best, but feel free to experiment.

[-ck]

On a side note, cgminer seems to use a lot of CPU for GPU mining in Windows, and all found hashes get rejected.  Same nVidia card, and I know phoenix gets a lot of rejects too if I use BFI_INT.  I saw that BFI_INT is used in the original post, but I didn't see any way to disable it in the parameters.  Is it possible to disable it, or is this miner meant for ATI only?

That should not be the case (the infinite rejects). There was an older version which had a bug with the nvidia implemented kernel and that would happen, but shouldn't happen with the latest version. cgminer only uses bfi int patching on cards that support it, and uses a separate kernel for nvidia cards.

Re the high cpu usage: Setting an intensity level usually uses a lot less CPU than letting it figure it out dynamically, but the windows build just uses a lot more cpu in general.

[ancow]

Compile attempt on a Mac with the latest changes from git:

Code:

gcc -DHAVE_CONFIG_H -I. -pthread -fno-strict-aliasing -I./compat/jansson -I./lib -I./lib   -O3 -Wall -MT cgminer-util.o -MD -MP -MF .deps/cgminer-util.Tpo -c -o cgminer-util.o `test -f 'util.c' || echo './'`util.c
util.c: In function 'json_rpc_call_sockopt_cb':
util.c:258: error: 'SOL_TCP' undeclared (first use in this function)
util.c:258: error: (Each undeclared identifier is reported only once
util.c:258: error: for each function it appears in.)
util.c:258: error: 'TCP_KEEPCNT' undeclared (first use in this function)
util.c:261: error: 'TCP_KEEPIDLE' undeclared (first use in this function)
util.c:264: error: 'TCP_KEEPINTVL' undeclared (first use in this function)
make[2]: *** [cgminer-util.o] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

[The00Dustin]

You have a conflict between jansson-1.2 which is included with Fedora and jansson-2.0 which comes with cgminer. Remove the jansson-devel packages and you'll be able to finish the build.

but will get this upon running cgminer:

[2011-07-20 12:41:07] Error: Getting Device IDs (num)


I am running an nVidia card at the moment, but poclbm and phoenix can each access it.

How did you install the NVIDIA drivers and CUDA toolkit?

Will removing the jansson-devel packages be beneficial at all compared to the workaround I described (editing the makefile)?  Regarding the NVIDIA drivers, I remember doing something with akmod-nvidia, but that may have been prior to mining, I downloaded the propietary binary package at some point as well, but I may have only installed it on an older version.  I think I installed the CUDA toolkit by extracting it per NVIDIA's instructions, but I don't think it is used.  I had trouble with it throwing an error about gcc 4.5 and greater not being supported trying to compile something a while back and couldn't figure out how to get it to recognize that I was using gcc34 as a compiler.  Also, I think the gpu mining compiled fine in this case because I also installed the ATI SDK, so I think it used the OpenCL drivers from the ATI SDK and not the CUDA toolkit.  Come to think of it, the working phoenix and poclbm came from git and weren't compiled by me, but I'd still expect them to use the same OpenCL driver.

[The00Dustin]

On a side note, cgminer seems to use a lot of CPU for GPU mining in Windows, and all found hashes get rejected.  Same nVidia card, and I know phoenix gets a lot of rejects too if I use BFI_INT.  I saw that BFI_INT is used in the original post, but I didn't see any way to disable it in the parameters.  Is it possible to disable it, or is this miner meant for ATI only?

That should not be the case (the infinite rejects). There was an older version which had a bug with the nvidia implemented kernel and that would happen, but shouldn't happen with the latest version. cgminer only uses bfi int patching on cards that support it, and uses a separate kernel for nvidia cards.

Re the high cpu usage: Setting an intensity level usually uses a lot less CPU than letting it figure it out dynamically, but the windows build just uses a lot more cpu in general.

These cards weren't purchased for mining...  They are 9500GT cards (I know they won't do much, I am just experimenting), I experienced infinite rejects using the windows binary which I downloaded (as opposed to compiling myself), and was unable to test on Linux thus far (because it couldn't recognize the card), but when I use BFI_INT on phoenix, I also get infinite rejects, so I just figured it might be related, could be the age of the card and a lack of testing against it or something.

EDIT:  I see 100% of one core used by poclbm and phoenix in Linux, so if this does less than that when I get it working, that will be a plus.  rpcminer-cuda uses practically no cpu in Windows, but these cards don't get along with it (computer actually performs better with cgminer using more than 100% of one core [I've seen it up to 67% of two cores] than rpcminer-cuda with aggression set to 3.  I may try out the intensity level just to see, though.

[error]

You have a conflict between jansson-1.2 which is included with Fedora and jansson-2.0 which comes with cgminer. Remove the jansson-devel packages and you'll be able to finish the build.

but will get this upon running cgminer:

[2011-07-20 12:41:07] Error: Getting Device IDs (num)


I am running an nVidia card at the moment, but poclbm and phoenix can each access it.

How did you install the NVIDIA drivers and CUDA toolkit?

Will removing the jansson-devel packages be beneficial at all compared to the workaround I described (editing the makefile)?  Regarding the NVIDIA drivers, I remember doing something with akmod-nvidia, but that may have been prior to mining, I downloaded the propietary binary package at some point as well, but I may have only installed it on an older version.  I think I installed the CUDA toolkit by extracting it per NVIDIA's instructions, but I don't think it is used.  I had trouble with it throwing an error about gcc 4.5 and greater not being supported trying to compile something a while back and couldn't figure out how to get it to recognize that I was using gcc34 as a compiler.  Also, I think the gpu mining compiled fine in this case because I also installed the ATI SDK, so I think it used the OpenCL drivers from the ATI SDK and not the CUDA toolkit.  Come to think of it, the working phoenix and poclbm came from git and weren't compiled by me, but I'd still expect them to use the same OpenCL driver.

If you installed the NVIDIA drivers from RPM Fusion (including xorg-x11-drv-nvidia-devel package) and the CUDA toolkit from NVIDIA in the default /usr/local/cuda, then this should get you going:

Code:

CFLAGS=-I/usr/local/cuda/include LDFLAGS=-L/usr/lib64/nvidia ./configure
make clean
make