Confused Using scikits.cuda.cula

[llvll0hsen]

Hello all,

I want to use some of cula functionality like LU factorization or Matrix inverse but I have some problem regarding the pointer inputs. for example for doing LU factorization with scikits.cuda.cula.culaDeviceSgetrf(m, n, a, lda, ipiv) , one need to use pointer f "a" argument but there is no pointer in python explicitly(I know all variables in python are by ref) . So what should I do in this case? should I use ctype library to create python?

this is what I am trying to do:

import numpy as np
import scikits.cuda.cula as cula
import pycuda.gpuarray as gpuarray

cula.culaInitialize()

I create a square matrix for simplicity

a=np.array([[1,2,3,4],[6,7,8,9],[7,2,3,5],[2,4,5,6]])

n=b.shape[0]
ida=ipv=m

scikits.cuda.cula.culaDeviceSgetrf(m,n,a,n,n)

but I'm getting this error:

Traceback (most recent call last):
File "", line 1, in
File "/usr/local/lib/python2.7/dist-packages/scikits.cuda-0.042-py2.7.egg/scikits/cuda/cula.py", line 329, in culaDeviceSgetrf
status = _libcula.culaDeviceSgetrf(m, n, int(a), lda, int(ipiv))
TypeError: only length-1 arrays can be converted to Python scalars

and when I try
a_gpu = gpuarray.to_gpu(a)
scikits.cuda.cula.culaDeviceSgetrf(m,n,a_gpu,n,n) :

Traceback (most recent call last):
File "", line 1, in
File "/usr/local/lib/python2.7/dist-packages/scikits.cuda-0.042-py2.7.egg/scikits/cuda/cula.py", line 329, in culaDeviceSgetrf
status = _libcula.culaDeviceSgetrf(m, n, int(a), lda, int(ipiv))
TypeError: int() argument must be a string or a number, not 'GPUArray'

any solution ?

Thanks ,

Mohsen

[lebedov]

You can pass a_gpu.gpudata to the wrapper function; it gets cast to an integer that corresponds to the pointer value.

[llvll0hsen]

still not working :

cula.culaDeviceSgetrf(m,n,a_gpu.gpudata,n,n)
Traceback (most recent call last):
File "", line 1, in
File "/usr/local/lib/python2.7/dist-packages/scikits.cuda-0.042-py2.7.egg/scikits/cuda/cula.py", line 330, in culaDeviceSgetrf
culaCheckStatus(status)
File "/usr/local/lib/python2.7/dist-packages/scikits.cuda-0.042-py2.7.egg/scikits/cuda/cula.py", line 210, in culaCheckStatus
raise culaExceptionsstatus
scikits.cuda.cula.culaDataError: 3

[lebedov]

Can you please post the code that you are trying to run? The snippet in your initial posting is incomplete.

[llvll0hsen]

This is what I'm trying to do:

import pycuda.gpuarray as gpuarray
import pycuda.autoinit
import scikits.cuda.cula as cula

Can you run that on your system? I'm using cula 12 full version.

cula.culaInitialize()

a=np.array([[1,2,3,4],[6,7,8,9],[7,2,3,5],[2,4,5,6]])

n=m=ida=ipiv=a.shape[0]

a_gpu=gpuarray.to_gpu(a)

result=cula.culaDeviceSgetrf(m, n, a_gpu.gpudata, ida, ipiv)

print result

cula.culaShutdown()

[lebedov]

Thanks. You are getting the data error because the array a doesn't contain single precision floats. Also, the ipiv parameter must be a pointer to an integer array, not a scalar.

[llvll0hsen]

Thanks for your reply. Ok so for fixing the first problem I should make "a" like :

a=np.array([[1,2,3,4],[6,7,8,9],[7,2,3,5],[2,4,5,6]], dtype=np.float32)

but I dont know how to fix the ipiv. Actually I dont understand what is ipiv . can tell me more about it?

[lebedov]

ipiv contains the 1-offset indices of the pivots determined during the decomposition. You don't need to set ipiv; just allocate an empty int32 array in GPU memory and pass its gpudata attribute to the function. See the GETRF entry in the CULA Reference manual for more information.

[llvll0hsen]

I get "None" after that . this is my code(I think I allocate the empty array in wrong way ):

a=np.array([[1,2,3,4],[6,7,8,9],[7,2,3,5],[2,4,5,6]], dtype=np.float32)
n=m=ida=a.shape[0]
ipiv=np.empty(n,dtype=np.int32)
ipiv_gpu=gpuarray.to_gpu(ipiv)
a_gpu=gpuarray.to_gpu(a)
result=cula.culaDeviceSgetrf(m, n, a_gpu.gpudata, ida, ipiv_gpu.gpudata)
print result

---

when I define it like :
ipiv=np.array([],dtype=np.int32)

I get:

status = _libcula.culaDeviceSgetrf(m, n, int(a), lda, int(ipiv))
TypeError: int() argument must be a string or a number, not 'NoneType'

[lebedov]

Your code is correct; the CULA reference manual indicates that ipiv should have length min(m,n). Note also that the returned value of the function is its execution status, not the result of the computation.

[llvll0hsen]

yes It's working now.I expected the result to have the answer not status. Thanks alot!