DiabloMiner GPU Miner

[DiabloD3]

Sorry for that little info, i was late.

I am feeding it with this:
./DiabloMiner-Linux.sh -u workername -p password -l nofeemining.com

This worked fine with the release from Wed 06 Jul 2011 (modify date of DiabloMiner-Linux.sh)

Additional Info i am using SDK 2.4 and Catalyst 11.6


Old release:

Code:

[robert@RIG02 DiabloMiner.old]$ ./DiabloMiner-Linux.sh -u username -p password -l nofeemining.com
[7/18/11 10:54:34 PM] Started                                                
[7/18/11 10:54:34 PM] Connecting to: http://nofeemining.com:8332/            
[7/18/11 10:54:34 PM] Using AMD Accelerated Parallel Processing OpenCL 1.1 AMD-APP-SDK-v2.4 (595.10)
[7/18/11 10:54:37 PM] BFI_INT patching enabled, disabling hardware checking  
[7/18/11 10:54:37 PM] Added Cayman (#1) (24 CU, local work size of 256)      
[7/18/11 10:54:42 PM] BFI_INT patching enabled, disabling hardware checking  
[7/18/11 10:54:42 PM] Added Cayman (#2) (24 CU, local work size of 256)      
mhash: 407.6/459.3 | accept: 0 | reject: 0 | hw error: 21104

Newest Binary:

Code:

[robert@RIG02 DiabloMiner]$ ./DiabloMiner-Linux.sh -u username -p password -l nofeemining.com
Exception in thread "main" java.net.MalformedURLException: no protocol: nofeemining.com
    at java.net.URL.<init>(URL.java:583)
    at java.net.URL.<init>(URL.java:480)
    at java.net.URL.<init>(URL.java:429)
    at com.diablominer.DiabloMiner.DiabloMiner.execute(DiabloMiner.java:307)
    at com.diablominer.DiabloMiner.DiabloMiner.main(DiabloMiner.java:133)

Any Ideas where this hardware errors come from(I asume hw = hardware?)? When I install only one card into my rig than there is no problem (tested both of them).

-l requires an actual URL. You gave it only a domain, you probably want -o instead.

Since you're with two cards, and only one card does it, it sounds like you forgot to turn crossfire off. If you're on linux, run sudo aticonfig --initial --adapter=all -f and then restart X.

[nmat]

Here's another user who has lots of hardware errors: http://forum.bitcoin.org/index.php?topic=29960.0
Maybe related?

[macrum]

-l requires an actual URL. You gave it only a domain, you probably want -o instead.

Since you're with two cards, and only one card does it, it sounds like you forgot to turn crossfire off. If you're on linux, run sudo aticonfig --initial --adapter=all -f and then restart X.

I have tried -l with full http://rpc.nofeemining.com and i got the same error. But i can live with that. With -o Option and IP it works.
I indeed have Crossfire enabled.
I will play a little bit with my xorg.conf, thanks for the tip. 

[ancow]

I have tried -l with full http://rpc.nofeemining.com and i got the same error. But i can live with that. With -o Option and IP it works.

-o doesn't necessarily need an IP - you can just use the hostname as you did with the -l option above.

Just for the record: I'm not sure I remember correctly, but I believe when you use the -l option, you need to supply everything to it, username, password, protocol, hostname and port. But you shouldn't be getting a MalformedURLException supplying only protocol and hostname, if you do, there is likely a spelling error in there somewhere.

[DiabloD3]

I have tried -l with full http://rpc.nofeemining.com and i got the same error. But i can live with that. With -o Option and IP it works.

-o doesn't necessarily need an IP - you can just use the hostname as you did with the -l option above.

Just for the record: I'm not sure I remember correctly, but I believe when you use the -l option, you need to supply everything to it, username, password, protocol, hostname and port. But you shouldn't be getting a MalformedURLException supplying only protocol and hostname, if you do, there is likely a spelling error in there somewhere.

I use Java itself to parse the URL, and its pretty flexible. The only way it'll bitch is if you made a bad typo (such as, putting , instead of . and getting yourself stuck on the multi-pool code that wraps the URL parser).

[DiabloD3]

Update: Support expire time in X-Roll-NTime

[ancow]

System info:
Mac OS X v10.6.8
Darwin 10.8.0 64bit
Intel Core i7
AMD Radeon HD 6750M

Bug Log:

Code:

./DiabloMiner-OSx.sh -o swepool.net -r 8337 -p 3rdParty -u ancow.2 -w 256
[21.07.11 16:24:06] Started                                                  
[21.07.11 16:24:06] Connecting to: http://swepool.net:8337/                  
[21.07.11 16:24:06] Using Apple OpenCL 1.0 (Dec 26 2010 12:52:21)            
Invalid memory access of location 0x10e709000 rip=0x7fff8854bfe6

Segmentation fault

This happens every time with values for -w greater than 128, most of the time for values smaller than 128 and a lot of the time for -w 128.
I can't run this under gdb at the moment because bloody OS X needs admin privileges to debug and I don't have an admin acc on this machine...
Edit: this isn't exactly a useful backtrace, but here it is anyway:

Code:

Program received signal SIGTRAP, Trace/breakpoint trap.
0x00007fff5fc01028 in __dyld__dyld_start ()

[DiabloD3]

System info:
Mac OS X v10.6.8
Darwin 10.8.0 64bit
Intel Core i7
AMD Radeon HD 6750M

Bug Log:

Code:

./DiabloMiner-OSx.sh -o swepool.net -r 8337 -p 3rdParty -u ancow.2 -w 256
[21.07.11 16:24:06] Started                                                  
[21.07.11 16:24:06] Connecting to: http://swepool.net:8337/                  
[21.07.11 16:24:06] Using Apple OpenCL 1.0 (Dec 26 2010 12:52:21)            
Invalid memory access of location 0x10e709000 rip=0x7fff8854bfe6

Segmentation fault

This happens every time with values for -w greater than 128, most of the time for values smaller than 128 and a lot of the time for -w 128.
I can't run this under gdb at the moment because bloody OS X needs admin privileges to debug and I don't have an admin acc on this machine...
Edit: this isn't exactly a useful backtrace, but here it is anyway:

Code:

Program received signal SIGTRAP, Trace/breakpoint trap.
0x00007fff5fc01028 in __dyld__dyld_start ()

Driver bug, go report it to Apple.

[Operations]

Hello Diablo,

the current version don't work with the new p2pool (https://forum.bitcoin.org/index.php?topic=18313.0). If i'm starting the Diablo Miner i'm getting the following error and the Miner doesn't do anything.

johndoe@cluster ~/opt/DiabloMiner.bad $ ./DiabloMiner-Linux.sh -o 127.0.0.1 -r 9332 -v 2 -w 128
Exception in thread "DiabloMiner GetWorkAsync for 127.0.0.1" java.lang.NullPointerException
   at com.diablominer.DiabloMiner.DiabloMiner$NetworkState.doJSONRPC(DiabloMiner.java:684)
   at com.diablominer.DiabloMiner.DiabloMiner$NetworkState$GetWorkAsync.run(DiabloMiner.java:827)
   at java.lang.Thread.run(Thread.java:662)
[22.07.11 08:34:14] Started                                                 
[22.07.11 08:34:14] Connecting to: http://127.0.0.1:9332/                   
[22.07.11 08:34:14] Using AMD Accelerated Parallel Processing OpenCL 1.1 AMD-APP-SDK-v2.4 (595.10)
[22.07.11 08:34:18] BFI_INT patching enabled, disabling hardware checking   
[22.07.11 08:34:18] Added Redwood (#1) (5 CU, local work size of 128)       
mhash: 0,0/0,0 | accept: 0 | reject: 0 | hw error: 0

The version 2-4 days ago didn't have this issue.


Thanks.

Best Regards

Operations

[ned123]

Similar error with Deepbit. I'll post the code output if I need to...
Is it means something, version before this last one was a little slower (about 1-2%) than the version from 7/11 (still the fastest).

[ellipsis]

Commit 1bcbb3fd85178692941b5e168aecf71e755a3909 slows things down (5870 -v 2 -w 128).

[ellipsis]

Exception in thread "DiabloMiner GetWorkAsync for 127.0.0.1" java.lang.NullPointerException
   at com.diablominer.DiabloMiner.DiabloMiner$NetworkState.doJSONRPC(DiabloMiner.java:684)

There is a missing null check on xRollNTime at line 681. Just make it look like the following:

if(xRollNTime != null && !"n".equalsIgnoreCase(xRollNTime)) {

[twmz]

I know you're pretty anti-Apple-OpenCL, but I get this error when trying to use DiabloMiner on OS X Lion (which has an updated version of OpenCL):

[7/22/11 6:15:36 AM] Using Apple OpenCL 1.1 (Jun 14 2011 23:31:14)           
[7/22/11 6:15:37 AM] Added ATI Radeon HD 5870 (#1) (20 CU, local work size of 256)
[7/22/11 6:15:41 AM] ERROR: [CL_INVALID_WORK_GROUP_SIZE] : OpenCL Error : clEnqueueNDRangeKernel failed: local_size[0] (256) != required local_size[0] (1024)
[7/22/11 6:15:41 AM] ERROR: Failed to queue kernel, error -54               

...followed by it crashing back to the command prompt.

In the event that that local_size is related to work size, I tried forcing the work size to 1024 (not sure if the error is talking about that or not), but that doesn't work either because DiabloMiner doesn't permit a work size larger than the detected max for the device.

[DiabloD3]

I know you're pretty anti-Apple-OpenCL, but I get this error when trying to use DiabloMiner on OS X Lion (which has an updated version of OpenCL):

[7/22/11 6:15:36 AM] Using Apple OpenCL 1.1 (Jun 14 2011 23:31:14)           
[7/22/11 6:15:37 AM] Added ATI Radeon HD 5870 (#1) (20 CU, local work size of 256)
[7/22/11 6:15:41 AM] ERROR: [CL_INVALID_WORK_GROUP_SIZE] : OpenCL Error : clEnqueueNDRangeKernel failed: local_size[0] (256) != required local_size[0] (1024)
[7/22/11 6:15:41 AM] ERROR: Failed to queue kernel, error -54               

...followed by it crashing back to the command prompt.

In the event that that local_size is related to work size, I tried forcing the work size to 1024 (not sure if the error is talking about that or not), but that doesn't work either because DiabloMiner doesn't permit a work size larger than the detected max for the device.

Wow, that is one hell of a driver bug. The error is saying -w should be 1024, but that is invalid for any Radeon manufactured. Not only that, it makes no sense.

Maybe it is quoting the error backwards, try -w 64.

[DiabloD3]

Exception in thread "DiabloMiner GetWorkAsync for 127.0.0.1" java.lang.NullPointerException
   at com.diablominer.DiabloMiner.DiabloMiner$NetworkState.doJSONRPC(DiabloMiner.java:684)

There is a missing null check on xRollNTime at line 681. Just make it look like the following:

if(xRollNTime != null && !"n".equalsIgnoreCase(xRollNTime)) {

Yeah, it seems theres a bug in iongchun's code to add expire time support and I didn't catch it because I don't regularly use any pools that don't support it.

For those wondering about Java, "string literal".equals(null) returns false instead of NullException.

I'll roll this fix out in a bit.

[twmz]

Wow, that is one hell of a driver bug. The error is saying -w should be 1024, but that is invalid for any Radeon manufactured. Not only that, it makes no sense.

Maybe it is quoting the error backwards, try -w 64.

-w 64 works to get passed that error, thanks. 

On the current version it still doesn't work, but it again goes back to the symptoms that Snow Leopard (massive reported hashrates and no actual found shares--I think you said this was because the kernel was causing errors but Apple's implementation was not reporting them as errors).

I tried an old version of DiabloMiner that doesn't have the kernel improvements from the last month or so and it runs and finds shares.

[DiabloD3]

Update: There, I fixed the bug.

And for future note, I'm going to treat all future bugs like this: If you're not using Eligius, it is not my problem.

[Operations]

The last update fixed the p2pool connection issue.

Thank you

[ellipsis]

And for future note, I'm going to treat all future bugs like this: If you're not using Eligius, it is not my problem.

You probably shouldn't advertise multipool support, then.

[DiabloD3]

And for future note, I'm going to treat all future bugs like this: If you're not using Eligius, it is not my problem.

You probably shouldn't advertise multipool support, then.

Eligius sometimes has multiple pool servers