accelerate-examples failes: "CUDA Exception: invalid handle" about accelerate HOT 6 CLOSED

hmm... sounds like you could be running the exact same as my main development machine, which is a little distressing. Do the examples from the CUDA SDK (from nvidia) work? What about the following in ghci?

ghci> :m +Foreign.CUDA.Driver
ghci> initialise []
ghci> props =<< device 0

from accelerate.

also, which version of ghc?

from accelerate.

Good to know that accelerate at least "should" work on my platform! I
am running ghc-7.0.3.

I am able to run the example programs from the CUDA SDK (in
/Developer/GPU Computing/C/bin/darwin/release) without any major
problems. Some of the programs give an out-of-memory error, but this
is fixable by turning down my screen resolution.

   ghci> :m +Foreign.CUDA.Driver
   ghci> initialise []
   ghci> props =<< device 0

Here's my output from those commands:

GHCi, version 7.0.3: http://www.haskell.org/ghc/ :? for help
Loading package ghc-prim ... linking ... done.
Loading package integer-gmp ... linking ... done.
Loading package base ... linking ... done.
Loading package ffi-1.0 ... linking ... done.
Prelude> :m +Foreign.CUDA.Driver
Prelude Foreign.CUDA.Driver> initialise []
Loading package extensible-exceptions-0.1.1.2 ... linking ... done.
Loading package bytestring-0.9.1.10 ... linking ... done.
Loading package cuda-0.3.2.2 ... linking ... done.
Prelude Foreign.CUDA.Driver> props =<< device 0
DeviceProperties {deviceName = "GeForce GT 120", computeCapability =
1.1, totalGlobalMem = 268107776, totalConstMem = 65536,
sharedMemPerBlock = 16384, regsPerBlock = 8192, warpSize = 32,
maxThreadsPerBlock = 512, maxBlockSize = (512,512,64), maxGridSize =
(65535,65535,1), maxTextureDim1D = 8192, maxTextureDim2D =
(65536,32768), maxTextureDim3D = (2048,2048,2048), clockRate =
1250000, multiProcessorCount = 4, memPitch = 2147483647,
textureAlignment = 256, computeMode = Default, deviceOverlap = True,
concurrentKernels = False, eccEnabled = False,
kernelExecTimeoutEnabled = True, integrated = False, canMapHostMemory
= True}

This seems to match what I get from running the SDK's deviceQuery program.

Any more ideas?

Thanks,
-Chris

Reply to this email directly or view it on GitHub:
https://github.com/mchakravarty/accelerate/issues/26#issuecomment-1778496

from accelerate.

Sorry for the late reply Chris, I should have mentioned earlier that I'm currently away and thus not working on Accelerate right now.

At any rate it seems that the haskell/cuda binding level is initialising the hardware okay, but that the reference to the execution context is being lost somewhere in accelerate. If you want to jump into the code, I'm happy to point you in the right directions, but otherwise it sounds like it will be difficult to diagnose the problem. I'll have to add some sort of verbose logging mechanism to accelerate he help pin down bug reports. hmm...

from accelerate.

I was able to fix the problem by upgrading to the lastest accelerate (0.9.0.0) from github, using the cuda-0.4 drivers and toolkit, and (probably most importantly) restarting my machine.

from accelerate.

Thanks for letting us know that this works now. It did seem strange given the similarity of our machines.

from accelerate.