Cryptohaze.com

It's having trouble detecting my 3 GTX580s, when running in debug it says "Got 0 CUDA devices!" The older Multiforcer detects them with no problem as does other CUDA programs we have installed.

Here's the debug output
./New-Multiforcer -h NTLM -c charsets/charsetall -f test_hashes/Hashes-NTLM-Full.txt --nocpu --debug

Code: Select all

./New-Multiforcer: /usr/lib/libOpenCL.so.1: no version information available (required by ./New-Multiforcer)
Got 0 CUDA devices!
Got 3 OpenCL GPUs
Device 0: OpenCL, p:0, d:0
Device 1: OpenCL, p:0, d:1
Device 2: OpenCL, p:0, d:2
MFND: Setting hash name NTLM
MFND: Setting password len 1
MFND: Status: Starting pw len 1
MFND: Setting password len 1
MFND: WU status: 0/1
MFND: Hash status: 0/13
MFND: Total crack rate 0.00 /s
MFND: Total rate: 0.00
MFND: Status: Td 0: CID 17767.
MFND: Status: Td 1: CID 39017.
MFND: Status: Td 2: CID 9158.
Error compiling program!
Error: (GRTOpenCL.cpp:764)Program build failure
Could not get build log: Invalid value


If I try to disable opencl I get this:
./New-Multiforcer -h NTLM -c charsets/charsetall -f test_hashes/Hashes-NTLM-Full.txt --nocpu --noopencl --debug

Code: Select all

./New-Multiforcer: /usr/lib/libOpenCL.so.1: no version information available (required by ./New-Multiforcer)
Got 0 CUDA devices!
MFND: Setting hash name NTLM
MFND: Setting password len 1
MFND: Status: Starting pw len 1
terminate called after throwing an instance of 'std::out_of_range'
  what():  vector::_M_range_check
Aborted


Very odd.

What does ShowOpenCLConfig (or whatever that damned utility is called) say?

It sees them too:

Code: Select all

./ShowConfig-OpenCL: /usr/lib/libOpenCL.so.1: no version information available (required by ./ShowConfig-OpenCL)
Platform ID 0:
CL_PLATFORM_NAME: NVIDIA CUDA
CL_PLATFORM_VENDOR: NVIDIA Corporation
CL_PLATFORM_VERSION: OpenCL 1.1 CUDA 4.1.1


Platform 0:
Device ID 0:
CL_DEVICE_NAME: GeForce GTX 580
CL_DEVICE_VENDOR: NVIDIA Corporation
CL_DEVICE_AVAILABLE: Yes
CL_DEVICE_MAX_COMPUTE_UNITS: 16
CL_DEVICE_TYPE: CL_DEVICE_TYPE_GPU
CL_DEVICE_MAX_CLOCK_FREQUENCY: 1594 MHz
CL_DEVICE_GLOBAL_MEM_SIZE: 1535 MB
CL_DEVICE_LOCAL_MEM_SIZE: 48 KB
CL_DEVICE_LOCAL_MEM_TYPE: CL_LOCAL
CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 64 KB


Device ID 1:
CL_DEVICE_NAME: GeForce GTX 580
CL_DEVICE_VENDOR: NVIDIA Corporation
CL_DEVICE_AVAILABLE: Yes
CL_DEVICE_MAX_COMPUTE_UNITS: 16
CL_DEVICE_TYPE: CL_DEVICE_TYPE_GPU
CL_DEVICE_MAX_CLOCK_FREQUENCY: 1594 MHz
CL_DEVICE_GLOBAL_MEM_SIZE: 1535 MB
CL_DEVICE_LOCAL_MEM_SIZE: 48 KB
CL_DEVICE_LOCAL_MEM_TYPE: CL_LOCAL
CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 64 KB


Device ID 2:
CL_DEVICE_NAME: GeForce GTX 580
CL_DEVICE_VENDOR: NVIDIA Corporation
CL_DEVICE_AVAILABLE: Yes
CL_DEVICE_MAX_COMPUTE_UNITS: 16
CL_DEVICE_TYPE: CL_DEVICE_TYPE_GPU
CL_DEVICE_MAX_CLOCK_FREQUENCY: 1594 MHz
CL_DEVICE_GLOBAL_MEM_SIZE: 1535 MB
CL_DEVICE_LOCAL_MEM_SIZE: 48 KB
CL_DEVICE_LOCAL_MEM_TYPE: CL_LOCAL
CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 64 KB


What driver version are you running, and is it the latest?

I see "CUDA 4.1.1" in your debug output. CUDA 4.2 is current, so I think you're on an older driver.

What version of Cryptohaze-Multiforcer are you using? I believe the latest version is built against 4.2, but older versions are not.

The MFN tools are built against the newest CUDA framework - should be against 4.2. If you've got older drivers that don't support the 4.2 environment, this could happen.

The code that checks for CUDA device count is the same between the Cryptohaze-Multiforcer and the New-Multiforcer, so I'm guessing it's something odd about your environment.

Try the latest driver release & see if it behaves. If not, we can troubleshoot more.

CUDA 4.2 did the trick, which makes sense as to why my home desktop and our work PW Cracking server didn't.

Picch wrote:CUDA 4.2 did the trick, which makes sense as to why my home desktop and our work PW Cracking server didn't.

Sweet.