I can change the type to None, and the code runs fine but with warnings.
Python 3.5.1
Loo.py 2016.1.1
PyOpenCL 2016.1.1
For the sake of completeness, I have the entire output listed here.
Device: [<pyopencl.Device 'Intel(R) Iris(TM) Graphics 6100' on 'Apple' at
0x1024500>]
Traceback (most recent call last):
File "/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/__init__.py",
line 472, in wrapper
return getattr(obj, cache_dict_name)[key]
AttributeError: 'CompiledKernel' object has no attribute
'_memoize_dic_cl_kernel_info'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/__init__.py",
line 472, in wrapper
return getattr(obj, cache_dict_name)[key]
AttributeError: 'CompiledKernel' object has no attribute
'_memoize_dic_get_typed_and_scheduled_kernel'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "pocl.py", line 29, in <module>
evt, (out,) = knl(queue, a=a)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/kernel/__init__.py",
line 1126, in __call__
queue, **kwargs)
File "/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/compiled.py",
line 958, in __call__
frozenset(six.iteritems(arg_to_dtype)))
File "/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/__init__.py",
line 474, in wrapper
result = function(obj, *args, **kwargs)
File "/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/compiled.py",
line 847, in cl_kernel_info
kernel = self.get_typed_and_scheduled_kernel(arg_to_dtype_set)
File "/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/__init__.py",
line 474, in wrapper
result = function(obj, *args, **kwargs)
File "/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/compiled.py",
line 838, in get_typed_and_scheduled_kernel
kernel = preprocess_kernel(kernel)
File "/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/preprocess.py",
line 655, in preprocess_kernel
result = preprocess_cache[kernel]
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 421, in __getitem__
return self.fetch(key)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 320, in fetch
hexdigest_key = self.key_builder(key)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 190, in __call__
self.rec(key_hash, key)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 174, in rec
method(key_hash, self)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/kernel/__init__.py",
line 1201, in update_persistent_hash
key_builder.rec(key_hash, getattr(self, field_name))
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 174, in rec
method(key_hash, self)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/loopy/target/__init__.py",
line 38, in update_persistent_hash
key_builder.rec(key_hash, getattr(self, field_name))
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 182, in rec
method(key_hash, key)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 219, in update_for_tuple
self.rec(key_hash, obj_i)
File
"/Users/sam/anaconda3/lib/python3.5/site-packages/pytools/persistent_dict.py",
line 186, in rec
% type(key))
TypeError: unsupported type for persistent hash keying: <class
'pyopencl.cffi_cl.Device’>
> On Jun 19, 2016, at 4:41 PM, Andreas Kloeckner <li...@informa.tiker.net>
> wrote:
>
> Hi Sam,
>
> Samuel Miller <smill...@live.com> writes:
>> I’ve tried to implement the basic example on loo.py’s website with a
>> slight modification but have had some trouble getting it to work
>> right. I tried to add a pyOpenCLTarget for the kernel but it keeps
>> giving me the following error. I’m able to get it to work if I just
>> use an OpenCLTarget, but not with a PyOpenCLTarget. Am I missing
>> something or using loo.py wrong? I’m using Python 3.5 in OS X El
>> Capitan on a Macbook Pro with an Intel Iris 6100. I can include the
>> entire traceback if needed, but it’s about a page’s worth.
>
> Thanks for the report. What version of loopy do you have? What version
> of Python and PyOpenCL?
>
> At any rate, it is likely that you may be able to work around this by
> simply passing 'device=None' to the PyOpenCLTarget. This may produce
> some (harmless) warnings.
>
> Another option would be to use the current git version of loopy, where I
> can't reproduce this issue (to be released fairly soon).
>
> Hope that helps,
> Andreas
_______________________________________________
PyOpenCL mailing list
PyOpenCL@tiker.net
https://lists.tiker.net/listinfo/pyopencl
