Lopper is a lightweight C++ template meta-programming framework for making vectorizing image-processing code easy, synthesized during Dropbox's 2016 hack week. It was inspired by many (more general) predecessors like Eigen and Halide, but the core focus was in enabling precise control over inlining of primitive computation. It can target platforms that support SSE (4.2) or NEON instructions. Note that Lopper does some very naughty things, like placing variables on the stack without being explicit about it, so take care when using it.
Lopper consists only of C++ header files, so no pre-compilation is necessary. Simply include "lopper/lopper.hpp" and compile your code with C++11-standard-compliant compiler. You will need CMake to build the unit test.
The code snippets below assume that using namespace lopper is in effect.
Images can be wrapped into an expression as follows:
auto a = Expr<1>(image); // image must be a 1-channel image.
auto rgb = ExprCache(Expr<3>(image)); // multi-channel images must first be wrapped by ExprCache.
auto r = rgb.get<0>();
One can combine expressions to form other expressions:
auto c = a + a * a;
auto d = c + 3;
auto e = c * d; // implicitly expand to (a + a * a) * ((a + a * a) + 3)
...
In general, every expression will be inlined during evaluation. In order to reuse values, use ExprCache; you must assign the resulting expression to a variable, owing to macro expansion that happens, and must first declare ExprPrepareContext as shown below, but it can be very powerful in controlling exactly what arithmetic happens in the evaluation.
ExprPrepareContext();
auto c = ExprCache(a + a * a);
auto d = c + 3;
auto e = c * d; // implicitly equivalent to c = a + a * a; e = c * (c + 3)
...
To trigger evaluation, use the ExprEval macro on an assignment operation if you haven't inlined anything:
auto a = Expr<1>(image1) + Expr<1>(image2);
ExprEval(Expr<1>(image3) = a * a);
On the other hand, you must use ExprEvalWithContext macro otherwise:
ExprPrepareContext();
auto a = ExprCache(Expr<1>(image1) + Expr<1>(image2));
ExprEvalWithContext(Expr<1>(image3) = a * a);
Lopper supports rudimentary index manipulation, without providing the full functional expansion (e.g. Halide). Any expression that's instantiated directly from a single-channel image can call reindex or offset.
Expr<1>(image1).offset(1, 0) - Expr<1>(image1); // expression for forward horizontal gradient
Expr<1>(image1).reindex([](int y) { return image1.getHeight() - 1 - y; }); // expression for flipping the image vertically.
See the unit tests for more examples.
If you plan to contribute a patch, please read the Contributor License Agreement at https://opensource.dropbox.com/cla/.
Unless otherwise noted:
Copyright (c) 2016 Dropbox, Inc.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Lopper was initially written by Jongmin Baek (jongmin@dropbox.com) with plenty of help and advice from Leonard Fink (leonard@dropbox.com), Lailin Chen (lailin@dropbox.com) and Ying Xiong (yxiong@dropbox.com).