diff --git a/pyopencl/__init__.py b/pyopencl/__init__.py
index 2dca2218f9a89c1e9b39531e61f2f1155e05bb36..b211765148a7458ddf5230d696d46ce70a42c97b 100644
--- a/pyopencl/__init__.py
+++ b/pyopencl/__init__.py
@@ -285,17 +285,17 @@ def link_program(context, programs, options=[], devices=None):
def _add_functionality():
cls_to_info_cls = {
- _cl.Platform: (_cl.Platform.get_info, _cl.platform_info),
- _cl.Device: (_cl.Device.get_info, _cl.device_info),
- _cl.Context: (_cl.Context.get_info, _cl.context_info),
- _cl.CommandQueue: (_cl.CommandQueue.get_info, _cl.command_queue_info),
- _cl.Event: (_cl.Event.get_info, _cl.event_info),
- _cl.MemoryObjectHolder: (MemoryObjectHolder.get_info, _cl.mem_info),
- Image: (_cl.Image.get_image_info, _cl.image_info),
- Program: (Program.get_info, _cl.program_info),
- Kernel: (Kernel.get_info, _cl.kernel_info),
- _cl.Sampler: (Sampler.get_info, _cl.sampler_info),
- }
+ _cl.Platform: (_cl.Platform.get_info, _cl.platform_info),
+ _cl.Device: (_cl.Device.get_info, _cl.device_info),
+ _cl.Context: (_cl.Context.get_info, _cl.context_info),
+ _cl.CommandQueue: (_cl.CommandQueue.get_info, _cl.command_queue_info),
+ _cl.Event: (_cl.Event.get_info, _cl.event_info),
+ _cl.MemoryObjectHolder: (MemoryObjectHolder.get_info, _cl.mem_info),
+ Image: (_cl.Image.get_image_info, _cl.image_info),
+ Program: (Program.get_info, _cl.program_info),
+ Kernel: (Kernel.get_info, _cl.kernel_info),
+ _cl.Sampler: (Sampler.get_info, _cl.sampler_info),
+ }
def to_string(cls, value, default_format=None):
for name in dir(cls):
@@ -802,8 +802,6 @@ def create_some_context(interactive=None, answers=None):
if not platforms:
raise Error("no platforms found")
- elif len(platforms) == 1:
- platform, = platforms
else:
if not answers:
cc_print("Choose platform:")
diff --git a/pyopencl/array.py b/pyopencl/array.py
index d00253a4a0f7ad16b836ef23064c7f7236dae9fd..8d29b6c1d74db82d4230cb3f513a41e820bad7a8 100644
--- a/pyopencl/array.py
+++ b/pyopencl/array.py
@@ -1288,20 +1288,118 @@ class Array(object):
raise TypeError("unexpected keyword arguments: %s"
% kwargs.keys())
+ if order not in "CF":
+ raise ValueError("order must be either 'C' or 'F'")
+
# TODO: add more error-checking, perhaps
+
if isinstance(shape[0], tuple) or isinstance(shape[0], list):
shape = tuple(shape[0])
+ if -1 in shape:
+ shape = list(shape)
+ idx = shape.index(-1)
+ size = -reduce(lambda x, y: x * y, shape, 1)
+ shape[idx] = self.size // size
+ if any(s < 0 for s in shape):
+ raise ValueError("can only specify one unknown dimension")
+ shape = tuple(shape)
+
if shape == self.shape:
- return self
+ return self._new_with_changes(
+ data=self.base_data, offset=self.offset, shape=shape,
+ strides=self.strides)
- size = reduce(lambda x, y: x * y, shape, 1)
+ import operator
+ size = reduce(operator.mul, shape, 1)
if size != self.size:
raise ValueError("total size of new array must be unchanged")
+ # {{{ determine reshaped strides
+
+ # copied and translated from
+ # https://github.com/numpy/numpy/blob/4083883228d61a3b571dec640185b5a5d983bf59/numpy/core/src/multiarray/shape.c # noqa
+
+ newdims = shape
+ newnd = len(newdims)
+
+ # Remove axes with dimension 1 from the old array. They have no effect
+ # but would need special cases since their strides do not matter.
+
+ olddims = []
+ oldstrides = []
+ for oi in range(len(self.shape)):
+ s = self.shape[oi]
+ if s != 1:
+ olddims.append(s)
+ oldstrides.append(self.strides[oi])
+
+ oldnd = len(olddims)
+
+ newstrides = [-1]*len(newdims)
+
+ # oi to oj and ni to nj give the axis ranges currently worked with
+ oi = 0
+ oj = 1
+ ni = 0
+ nj = 1
+ while ni < newnd and oi < oldnd:
+ np = newdims[ni]
+ op = olddims[oi]
+
+ while np != op:
+ if np < op:
+ # Misses trailing 1s, these are handled later
+ np *= newdims[nj]
+ nj += 1
+ else:
+ op *= olddims[oj]
+ oj += 1
+
+ # Check whether the original axes can be combined
+ for ok in range(oi, oj-1):
+ if order == "F":
+ if oldstrides[ok+1] != olddims[ok]*oldstrides[ok]:
+ raise ValueError("cannot reshape without copy")
+ else:
+ # C order
+ if (oldstrides[ok] != olddims[ok+1]*oldstrides[ok+1]):
+ raise ValueError("cannot reshape without copy")
+
+ # Calculate new strides for all axes currently worked with
+ if order == "F":
+ newstrides[ni] = oldstrides[oi]
+ for nk in xrange(ni+1, nj):
+ newstrides[nk] = newstrides[nk - 1]*newdims[nk - 1]
+ else:
+ # C order
+ newstrides[nj - 1] = oldstrides[oj - 1]
+ for nk in range(nj-1, ni, -1):
+ newstrides[nk - 1] = newstrides[nk]*newdims[nk]
+
+ ni = nj
+ nj += 1
+
+ oi = oj
+ oj += 1
+
+ # Set strides corresponding to trailing 1s of the new shape.
+ if ni >= 1:
+ last_stride = newstrides[ni - 1]
+ else:
+ last_stride = self.dtype.itemsize
+
+ if order == "F":
+ last_stride *= newdims[ni - 1]
+
+ for nk in range(ni, len(shape)):
+ newstrides[nk] = last_stride
+
+ # }}}
+
return self._new_with_changes(
data=self.base_data, offset=self.offset, shape=shape,
- strides=_make_strides(self.dtype.itemsize, shape, order))
+ strides=tuple(newstrides))
def ravel(self):
"""Returns flattened array containing the same data."""
diff --git a/pyopencl/ipython_ext.py b/pyopencl/ipython_ext.py
index 81fbcdf8669f01415966ec183f2677ed0d5f451b..bedbf77f039ba2f04280239ec13d274a10a3e82a 100644
--- a/pyopencl/ipython_ext.py
+++ b/pyopencl/ipython_ext.py
@@ -3,6 +3,7 @@ from __future__ import division
from IPython.core.magic import (magics_class, Magics, cell_magic, line_magic)
import pyopencl as cl
+import sys
def _try_to_utf8(text):
@@ -14,8 +15,10 @@ def _try_to_utf8(text):
@magics_class
class PyOpenCLMagics(Magics):
def _run_kernel(self, kernel, options):
- kernel = _try_to_utf8(kernel)
- options = _try_to_utf8(options).strip()
+ if sys.version_info < (3,):
+ kernel = _try_to_utf8(kernel)
+ options = _try_to_utf8(options).strip()
+
try:
ctx = self.shell.user_ns["cl_ctx"]
except KeyError:
@@ -34,37 +37,33 @@ class PyOpenCLMagics(Magics):
raise RuntimeError("unable to locate cl context, which must be "
"present in namespace as 'cl_ctx' or 'ctx'")
- prg = cl.Program(ctx, kernel).build(options=options)
+ prg = cl.Program(ctx, kernel).build(options=options.split())
for knl in prg.all_kernels():
self.shell.user_ns[knl.function_name] = knl
-
@cell_magic
def cl_kernel(self, line, cell):
kernel = cell
- opts, args = self.parse_options(line,'o:')
+ opts, args = self.parse_options(line, 'o:')
build_options = opts.get('o', '')
self._run_kernel(kernel, build_options)
-
def _load_kernel_and_options(self, line):
- opts, args = self.parse_options(line,'o:f:')
+ opts, args = self.parse_options(line, 'o:f:')
build_options = opts.get('o')
kernel = self.shell.find_user_code(opts.get('f') or args)
return kernel, build_options
-
@line_magic
def cl_kernel_from_file(self, line):
kernel, build_options = self._load_kernel_and_options(line)
self._run_kernel(kernel, build_options)
-
@line_magic
def cl_load_edit_kernel(self, line):
kernel, build_options = self._load_kernel_and_options(line)
diff --git a/setup.py b/setup.py
index 86a6bdad7e6f570048796aa19d885ab8609e4ff3..b6f26b30f046b0575b27b31531635f307b6d7626 100644
--- a/setup.py
+++ b/setup.py
@@ -1,5 +1,14 @@
#!/usr/bin/env python
# -*- coding: latin-1 -*-
+from setuptools.command.build_ext import build_ext as _build_ext
+
+class build_ext(_build_ext):
+ def finalize_options(self):
+ _build_ext.finalize_options(self)
+ # Prevent numpy from thinking it is still in its setup process:
+ __builtins__.__NUMPY_SETUP__ = False
+ import numpy
+ self.include_dirs.append(numpy.get_include())
__copyright__ = """
Copyright (C) 2009-14 Andreas Kloeckner
@@ -222,6 +231,7 @@ def main():
],
install_requires=[
+ "numpy",
"pytools>=2014.2",
"pytest>=2",
"decorator>=3.2.0",
diff --git a/test/test_array.py b/test/test_array.py
index eb29b9b6cac86e134da49293b975f6c2a90644fc..acb82ec8876b0b96867379f9c129a24c3d15ccb9 100644
--- a/test/test_array.py
+++ b/test/test_array.py
@@ -148,7 +148,7 @@ def test_mix_complex(ctx_factory):
err = la.norm(host_result-dev_result)/la.norm(host_result)
print(err)
- correct = err < 1e-5
+ correct = err < 1e-4
if not correct:
print(host_result)
print(dev_result)
@@ -719,6 +719,24 @@ def test_view_and_strides(ctx_factory):
assert (y.get() == X.get()[:3, :5]).all()
+def test_meshmode_view(ctx_factory):
+ context = ctx_factory()
+ queue = cl.CommandQueue(context)
+
+ n = 2
+ result = cl.array.empty(queue, (2, n*6), np.float32)
+
+ def view(z):
+ return z[..., n*3:n*6].reshape(z.shape[:-1] + (n, 3))
+
+ result = result.with_queue(queue)
+ result.fill(0)
+ view(result)[0].fill(1)
+ view(result)[1].fill(1)
+ x = result.get()
+ assert (view(x) == 1).all()
+
+
def test_event_management(ctx_factory):
context = ctx_factory()
queue = cl.CommandQueue(context)
@@ -758,6 +776,28 @@ def test_event_management(ctx_factory):
assert len(x.events) < 100
+def test_reshape(ctx_factory):
+ context = ctx_factory()
+ queue = cl.CommandQueue(context)
+
+ a = np.arange(128).reshape(8, 16).astype(np.float32)
+ a_dev = cl_array.to_device(queue, a)
+
+ # different ways to specify the shape
+ a_dev.reshape(4, 32)
+ a_dev.reshape((4, 32))
+ a_dev.reshape([4, 32])
+
+ # using -1 as unknown dimension
+ assert a_dev.reshape(-1, 32).shape == (4, 32)
+ assert a_dev.reshape((32, -1)).shape == (32, 4)
+ assert a_dev.reshape(((8, -1, 4))).shape == (8, 4, 4)
+
+ import pytest
+ with pytest.raises(ValueError):
+ a_dev.reshape(-1, -1, 4)
+
+
if __name__ == "__main__":
# make sure that import failures get reported, instead of skipping the
# tests.