PyCUDA March 2009

pycuda@tiker.net

17 participants
36 discussions

Out of memory (and possibly other errors) when running test_driver on Quadro NVS 140M

by Philipp Lies

Hi, first of all: nice piece of work Andreas! Now to my problem: I installed CUDA 2.1 on my T61 laptop computer with a Nvidia Quadro NVS 140M graphics card w/ 128MB memory. git version of PyCUDA (0.93 beta I assume) installs fine but I get some errors when I try to run tests. First test_driver: /opt/PyCUDA/pycuda-git/test$python test_driver.py ...E....Eterminate called after throwing an instance of 'cuda::error' what(): cuMemFree failed: invalid context Aborted Or runnung gpuarray speed test: /opt/PyCUDA/pycuda-git/test$python undistributed/test_gpuarray_speed.py 1024 2048 4096 8192 16384 32768 65536 131072 262144 524288 1048576 2097152 Traceback (most recent call last): File "undistributed/test_gpuarray_speed.py", line 83, in <module> main() File "undistributed/test_gpuarray_speed.py", line 27, in main b = gpuarray.zeros((size,), dtype=numpy.float32) File "/usr/lib/python2.5/site-packages/pycuda-0.93beta-py2.5-linux-x86_64.egg/pycuda/gpuarray.py", line 409, in zeros result = GPUArray(shape, dtype, stream, allocator) File "/usr/lib/python2.5/site-packages/pycuda-0.93beta-py2.5-linux-x86_64.egg/pycuda/gpuarray.py", line 75, in __init__ self.gpudata = self.allocator(self.size * self.dtype.itemsize) pycuda._driver.MemoryError: cuMemAlloc failed: out of memory ----------------------------------------------------------- PyCUDA WARNING: I'm being asked to destroy a context that's part of the current context stack. ----------------------------------------------------------- I will pick the next lower active context from the context stack. Since this choice is happening at an unspecified point in time, your code may be making false assumptions about which context is active at what point. Call Context.pop() to avoid this warning. ----------------------------------------------------------- If Python is terminating abnormally (eg. exiting upon an unhandled exception), you may ignore this. ----------------------------------------------------------- test_gpuarray only threw some warnings: python test_gpuarray.py /usr/lib/python2.5/site-packages/pycuda-0.93beta-py2.5-linux-x86_64.egg/pycuda/gpuarray.py:498: UserWarning: behavior change: arange guessed dtype other than float32. suggest specifying explicit dtype. warn("behavior change: arange guessed dtype other than float32. " .................. ---------------------------------------------------------------------- Ran 18 tests in 14.756s OK Here are some debug information you might need: In [61]: pycuda.autoinit.device.get_attributes() Out[62]: {pycuda._driver.device_attribute.MAX_THREADS_PER_BLOCK: 512, pycuda._driver.device_attribute.MAX_BLOCK_DIM_X: 512, pycuda._driver.device_attribute.MAX_BLOCK_DIM_Y: 512, pycuda._driver.device_attribute.MAX_BLOCK_DIM_Z: 64, pycuda._driver.device_attribute.MAX_GRID_DIM_X: 65535, pycuda._driver.device_attribute.MAX_GRID_DIM_Y: 65535, pycuda._driver.device_attribute.MAX_GRID_DIM_Z: 1, pycuda._driver.device_attribute.MAX_SHARED_MEMORY_PER_BLOCK: 16384, pycuda._driver.device_attribute.TOTAL_CONSTANT_MEMORY: 65536, pycuda._driver.device_attribute.WARP_SIZE: 32, pycuda._driver.device_attribute.MAX_PITCH: 262144, pycuda._driver.device_attribute.MAX_REGISTERS_PER_BLOCK: 8192, pycuda._driver.device_attribute.CLOCK_RATE: 800000, pycuda._driver.device_attribute.TEXTURE_ALIGNMENT: 256, pycuda._driver.device_attribute.GPU_OVERLAP: 1, pycuda._driver.device_attribute.MULTIPROCESSOR_COUNT: 2} In [63]: pycuda.autoinit.device.compute_capability() Out[63]: (1, 1) In [65]: pycuda.autoinit.device.total_memory() Out[65]: 133496832 In [66]: pycuda.autoinit.device.make_context() terminate called after throwing an instance of 'cuda::error' what(): cuMemFree failed: invalid context Aborted All the other tests were fine, so PyCUDA works nicely. Could you please check where the "invalid context" error comes from? And maybe add a memory checker to your tests so we low-mem GPU user are not discriminated ;-). Keep up the good work! Philipp

11 years, 8 months

cuda profiler

by Nicholas Tung

Hi Andreas, Is there any support for functions like the CUDA profiler, which currently will output the number of non-coalesced loads, etc.? Are these functions available through the driver API? Thanks, Nicholas

12 years

Re: [PyCuda] cuda profiler

by Nicholas Tung

I'm not sure, I saw similar behavior with empty headers when I did a sys.exit() in some places, though in others it worked fine. I'll tell you if I find out more. I did a tail -f and it looks like the file is only zeroed once, so it might be waiting for a context pop call to write everything; I'm not sure. btw, /usr/local/cuda/doc/CUDA_Profiler_*.txt has more options / details than what Andreas listed. regards, Nicholas I took care of that fact when I did it :P . I was able to fix the > problem though. I dont know why, but the error occured when I replaced > > /import pycuda.driver as cuda > import //pycuda.autoinit > / > > *with * > > > /import pycuda.driver as cuda > import pycuda.autoinit > > cuda.init() > assert cuda.Device.count() >= 1 > > dev = cuda.Device(0) > ctx = dev.make_context()/ > > Is this a bug or did I miss something in the above? > > Siddhardh >

12 years

test_ldexp in test_math failed

by tianyicui＠gmail.com

The test failed because math.ldexp expect an int or long as second argument. Otherwise TypeError raised. Change the line in test_math.py from self.assert_(math.ldexp(a[i], a2[i]) == b[i]) to self.assert_(math.ldexp(a[i], int(a2[i])) == b[i]) will get it passed. PyCUDA: git version Python: 2.6.1

12 years

Re: [PyCuda] Variables changing on transfers between host and device?

by Nicholas Tung

try result = numpy.array([0], dtype=numpy.float32) regards, Nicholas

12 years

Variables changing on transfers between host and device?

by Siddhardh chandra

Hi, I tried to following code because I was getting very weird errors in my main program. I create a variable ed which is initialized to zero on host and then copied over to the device and back to host without any manipulation. For some reason unknown to me, the value of the variable changes. Also, is there a better way to copy the variable back other than creating a numpy array for it? If I try copying it to a simple variable, it says "TypeError: expected a writeable buffer object". / 1 import pycuda.driver as cuda 2 import pycuda.autoinit 3 4 cuda.init() 5 assert cuda.Device.count() >= 1 6 7 dev = cuda.Device(0) 8 ctx = dev.make_context() 9 10 import numpy 11 12 ed = numpy.float32(0.0) 13 14 edd = cuda.mem_alloc(ed.nbytes) 15 16 cuda.memcpy_htod(edd, ed) 17 18 result = numpy.array([0.0]) 19 cuda.memcpy_dtoh(result, edd) 20 print ed 21 print result / *On running the program, I get the following output* / msid ~/python $ python cudaTest.py 0.0 [-0.00476747] ----------------------------------------------------------- PyCuda WARNING: I'm being asked to destroy a context that's part of the current context stack. ----------------------------------------------------------- I will pick the next lower active context from the context stack. Since this choice is happening at an unspecified point in time, your code may be making false assumptions about which context is active at what point. Call Context.pop() to avoid this warning. ----------------------------------------------------------- If Python is terminating abnormally (eg. exiting upon an unhandled exception), you may ignore this. ----------------------------------------------------------- terminate called after throwing an instance of 'cuda::error' what(): cuCtxPushCurrent failed: invalid value Aborted / Could someone please tell me why this is happening(pardon me if this is common knowledge) Thank you. Siddhardh

12 years

Re: [PyCuda] Acme, le nouveau bug d'OMiSCID

by J-Pascal Mercier

Salut > Mouai, en regardant le code, a priori, cela n'arrive que > si l'entête ne contient pas une taille qui serait réelle... Ce qui me semble, serais une attaque peu complexe et efficace!!! A patcher absolument. > Non, c'est le retassement en cas où il y avait des données > résiduelles (paquet suivant) à la fin du traitement du paquet > courant. Je ne dis pas qu'il n'y a pas un bug. Faudra que tu > me file un exe qui fasse planter systématiquement. pas vraiment d'exec, c'est du code 100% python, mais je peux te filer quelque chose si tu veux. a demain

12 years

Python 2.6 md5 --> hashlib

by Michael Freitas

I keep getting warning for the deprecated md5 library in python 2.6. /opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/site-packages/pycuda-0.93beta-py2.6-macosx-10.5-i386.egg/pycuda/compiler.py:23: DeprecationWarning: the md5 module is deprecated; use hashlib instead import md5 I lifted a patch off another site. if you change line 23+ in compiler.py to the following, you can get rid of the warnings (sorry for not generating a diff). I compiled and tested. It seems to work fine. # import md5 # checksum = md5.new() try: import hashlib checksum = hashlib.md5() except ImportError: # for Python << 2.5 import md5 checksum = md5.new() PyCuda 0.93 (~ 1 week old from git), boost 1.38 python 2.6.1 OSX 10.5.6 -- Dr. Michael A. Freitas Research Assistant Professor Ohio State University Medical Center 906 Biomedical Research Tower 460 West 12th Avenue Columbus, OH 43210 email (OSU): freitas.5(a)osu.edu phone: 614-688-8432 fax: 614-688-8675

12 years

PyCuda starting examples:

by William King

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Ok, for someone new to python(but very familiar with perl, java, php) and new to CUDA. I have been able to setup my test enviroment and get the test_driver.py and all to run properly. I would like to help document(while I learn) how to use PyCuda to build libraries of algorithms to run in a GPU enviroment. My eventual goal is to be able to integrate PyCuda into a distributed computing management system such as Boinc. I'm willing to document my steps if someone can help figure out the CUDA/PyCUDA/Python side of things. - -William -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAknBATUACgkQZ63Azu7rahqZlACfaPJkmlesR6nn97jR/QtjCP+G v1cAn1nwHM/VwVt4D96YHRq5Qg6Kb36M =/94f -----END PGP SIGNATURE-----

12 years

time_kernel argument in pycuda.driver.Function question

by Ahmed Fasih

Hi, I'm invoking a kernel like using pycuda.driver.Out and using the time_kernel=True argument to return the time in seconds. Does this time include the device-to-host copy time? Thanks! I'm happy to report a preliminary 165x speed up over our existing radar imaging implementation thanks to PyCUDA, developed in 5 days.

12 years

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

PyCUDA March 2009