Hi all,
So, I have found ways to pass the numerical constants in, but cannot figure out
how to allocate memory such that my code complies.
blockwidth = numpy .shape(array)[1] #Want the block width to be the number of
columns in the array.
maxblock = cuda .Device.get_attribute( cuda .Context.get_device(), cuda
.device_attribute.MAX_THREADS_PER_BLOCK)
blockheight =int( maxblock / blockwidth )
gridx = cuda .Device.get_attribute( cuda .Context.get_device(), cuda
.device_attribute.MAX_GRID_DIM_X)
gridy = cuda .Device.get_attribute( cuda .Context.get_device(), cuda
.device_attribute.MAX_GRID_DIM_Y)
a_ gpu = cuda . mem _ alloc (array. nbytes )
cuda . memcpy _ htod (a_ gpu , array)
mod = pycuda .compiler. SourceModule ("""
__device__ int convrt _(integer *i__, doublereal * gdlat , doublereal *
gdalt , doublereal * gclat , doublereal * rkm )
omitting function code. . .
__global__ void thefunk ( doublereal *in)
{int holder;
int thread = threadIdx .y* blockDim .x;
int block= blockIdx .x* blockDim .x* blockDim .y;
int grid= blockIdx .y* gridDim .x* blockDim .x* blockDim .y;
int idx = thread+block+grid;
holder = (int)in[ idx ];
int *fix = &holder;
convrt _(fix, &in[ idx +1], &in[ idx +2], &in[ idx +3], &in[ idx +4]);
}
""")
func = mod.get_function(" thefunk ")
func (a_ gpu , block=( blockwidth , blockheight ,1), grid=( gridx , gridy ))
Gives the error:
File " fortry . py ", line 128, in compute
func (a_ gpu , block=( blockwidth , blockheight ,1), grid=( gridx , gridy ))
File "/ usr /local/ midas
/baseline/2010/python-2.6.2/pkg/lib/python2.6/site-packages/ pycuda
-0.94.2-py2.6-linux-x86_64.egg/ pycuda /driver. py ", line 179, in
function_call
func .launch_grid(*grid)
pycuda ._driver. LaunchError : cuLaunchGrid failed: launch out of resources
Where have I gone astray?
Thanks,
-drp
----- Original Message -----
----- Original Message -----
From: "Andreas Kloeckner " <lists@ informa . tiker .net>
To: uteboy @comcast.net, pycuda @ tiker .net
Sent: Wednesday, June 22, 2011 1:55:24 PM
Subject: Re: [ PyCUDA ] issues with device_attribute class
These are numerical constants that need to be passed to functions such
as Device.get_attribute().
HTH ,
Andreas
_______________________________________________
PyCUDA mailing list
PyCUDA @ tiker .net
http ://lists. tiker .net/ listinfo / pycuda
_______________________________________________
PyCUDA mailing list
PyCUDA@tiker.net
http://lists.tiker.net/listinfo/pycuda
