vcl woes
#1
I tried setting up a vcl cluster of 7970's for trying out vclHashcat. Everything seems to work fine except vclHashcat claims that the Catalyst driver is too old, despite it's not. Per the instructions it's in fact too new - because it seems 12.8 is not supported on Ubuntu 12.10/Mint 14. But oclHashcat works just fine with the same drivers (Ubuntu supplied, I think 1016.4 means 12.10, right?). So why would it not work?


Code:
user@master:~$ vclrun ./vclHashcat-plus64.bin --gpu-temp-disable -m500 -a3 test.md5 -i -1 ?l?d?u ?1?1?1?1?1?1?1?1?1?1
oclHashcat-plus v0.09 by atom starting...

Hashes: 1 total, 1 unique salts, 1 unique digests
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Workload: 16 loops, 8 accel
Watchdog: Temperature abort trigger disabled
Watchdog: Temperature retain trigger disabled
Device #1: Tahiti, 2048MB, 0Mhz, 32MCU
Device #2: Tahiti, 2048MB, 0Mhz, 32MCU
Device #3: Tahiti, 2048MB, 0Mhz, 32MCU
Device #4: Tahiti, 2048MB, 0Mhz, 32MCU
Device #5: Tahiti, 2048MB, 0Mhz, 32MCU
Device #6: Tahiti, 2048MB, 0Mhz, 32MCU
Device #7: Tahiti, 2048MB, 0Mhz, 32MCU
Device #8: Tahiti, 2048MB, 0Mhz, 32MCU
Outdated GPU driver version detected, you need Catalyst v12.8 or higher.

user@master:~$ vclrun clinfo | grep Version
  Platform Version:                              OpenCL 1.1 0.0.0.0
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)
  Version:                                       OpenCL 1.2 AMD-APP (1016.4)

$ mpiexec -host node1 -host node2 -host node3 -host node4 -host node5 dmesg | grep 'module loaded - fglrx'
[    1.518911] [fglrx] module loaded - fglrx 9.0.2 [Sep 20 2012] with 1 minors
[    1.918176] [fglrx] module loaded - fglrx 9.0.2 [Sep 20 2012] with 1 minors
[    1.774890] [fglrx] module loaded - fglrx 9.0.2 [Sep 20 2012] with 2 minors
[    1.712370] [fglrx] module loaded - fglrx 9.0.2 [Sep 20 2012] with 2 minors
[    1.980531] [fglrx] module loaded - fglrx 9.0.2 [Sep 20 2012] with 2 minors

What's next, downgrade to an older distro? I think I'd rather just wait for new releases. Or is there something else I can do, or did wrong? I think I followed the wiki except for using Catalyst 12.10.
#2
you must use 12.8. No way round.
#3
VCL does not use Catalyst. Use --force with vclHashcat to ignore the message about the driver. Your compute nodes absolutely must be using 12.8.
#4
Downgrading to Ubuntu 12.04 and Catalyst 12.8 just made it worse. I posted in the other thread about very similar issues.
#5
Ubuntu 12.04 should work fine with cat.12.8. Can you provide more infos about the errors?

Ubuntu 12.10s kernel is not supported by cat 12.8.
#6
Since opencld seemed to bail silently, I concentrated on that and found out by trial'n'error that I could supply a -d option to it. I got this hint:

OPENCLD - Virtual OpenCL back-end service
Platform <AMD Accelerated Parallel Processing> was ignored because it is faulty:
If you still want to use it, modify the VCL configuration
No OpenCL devices found!

This is apparently not a vclHashcat problem but still, any ideas are welcome. Faulty? Without VCL, clinfo and oclHashcat works as expected. VCL has an option for "AMD version 2" or whatever it was called. I tried that but it made no difference.
#7
It's hard telling what you've done wrong. You ignored the wiki when you installed catalyst 12.10 at first, so who knows where else you deviated from the wiki. You also have oclHashcat installed on a compute node apparently, so there's another area where you've deviated from the wiki. Who knows what else you ignored.
#8
I never ignored the wiki, I just wasn't aware that Cat 12.8 does not support Ubuntu 12.10 and when I was faced with that fact I tried Cat 12.10 before starting over from scratch. Also, I simply can't use VCL 1.15 because only 1.16 is available for download.

Here's my understanding:

* master neither needs Catalyst nor X.
* nodes need Catalyst and X.

If this is correct, I see no reason why testing oclHashcat on a node, as part of tracking a problem down, would be an issue.
#9
My point is that there's no way for us to know what you have done in order to help troubleshoot your environment. But you did in fact deviate from the wiki when you used VCL 1.16. Nobody has gotten 1.16 to work. If you would have mentioned that you deviated from the wiki on the VCL version, I would have pointed you to a download link for VCL 1.15. So again, you ignored the wiki which makes it very difficult for us to help you.

VCL 1.17 is available for download now and should work with oclHashcat-plus without having to use vclHashcat-plus.
#10
Thanks, I'll try 1.17. So from now on there will be no separate vclHashcat?