# Misc Tools¶

## Misc tools¶

class sumpy.tools.BlockIndexRanges(cl_context, indices, ranges)[source]

Convenience class for working with blocks of a global array.

indices

A list of not necessarily continuous or increasing integers representing the indices of a global array. The individual blocks are delimited using ranges.

ranges

A list of nondecreasing integers used to index into indices. A block $$i$$ can be retrieved using indices[ranges[i]:ranges[i + 1]].

block_shape(i)[source]
get(queue=None)[source]
take(x, i)[source]

Return the subset of a global array x that is defined by the indices in block $$i$$.

class sumpy.tools.MatrixBlockIndexRanges(cl_context, row, col)[source]

Keep track of different ways to index into matrix blocks.

row

A BlockIndexRanges encapsulating row block indices.

col

A BlockIndexRanges encapsulating column block indices.

block_shape(i)[source]
block_take(x, i)[source]

Retrieve a block from a linear representation of the matrix blocks. A linear representation of the matrix blocks can be obtained, or should be consistent with

i = index.linear_row_indices()
j = index.linear_col_indices()
linear_blks = global_mat[i, j]

for k in range(index.nblocks):
assert np.allclose(index.block_take(linear_blks, k),
index.take(global_mat, k))

Parameters
Returns

requested block, reshaped into a 2D array.

get(queue=None)[source]

Transfer data to the host. Only the initial given data is transfered, not the arrays returned by linear_row_indices() and friends.

Returns

a copy of self in which all data lives on the host, i.e. all pyopencl.array.Array instances are replaces by numpy.ndarray instances.

linear_row_indices
take(x, i)[source]

Retrieve a block from a global matrix.

Parameters
Returns

requested block from the matrix.

# Installation¶

This command should install sumpy:

pip install sumpy


You may need to run this with sudo. If you don’t already have pip, run this beforehand:

curl -O https://raw.github.com/pypa/pip/master/contrib/get-pip.py
python get-pip.py


For a more manual installation, download the source, unpack it, and say:

python setup.py install


In addition, you need to have numpy installed.

# Usage¶

## Environment variables¶

Name

Purpose

SUMPY_FORCE_SYMBOLIC_BACKEND

Symbolic backend control, see Symbolic backends

SUMPY_NO_CACHE

If set, disables the on-disk cache

SUMPY_NO_OPT

If set, disables performance-oriented loopy transformations

## Symbolic backends¶

sumpy supports two symbolic backends: sympy and SymEngine. To use the SymEngine backend, ensure that the SymEngine library and the SymEngine Python bindings are installed.

By default, sumpy prefers using SymEngine but falls back to sympy if it detects that SymEngine is not installed. To force the use of a particular backend, set the environment variable SUMPY_FORCE_SYMBOLIC_BACKEND to symengine or sympy.

# User-visible Changes¶

## Version 2016.1¶

Note

This version is currently under development. You can get snapshots from sumpy’s git repository

• Initial release.

sumpy is licensed to you under the MIT/X Consortium license:

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Andreas Klöckner’s work on sumpy was supported in part by