After my previous post on SPICE modelling in Python, I need to use a good support example to go up to on the fly compilation in C++. This schema will also require some changes to support more than simple nodal analysis, so this now becomes Modified Nodal Analysis with state equations.
A few month ago, mystran published on KVR a small SPICE simulator for real-time processing. I liked the idea, the drawback being that the code is generic and not tailored like a static version of the optimizer. So I wondered if it was doable. But for this, I have to start from the basics and build from there. So let’s go.
ATK is updated to 2.3.0 with major fixes and code coverage improvement (see here). Lots of bugs were fixed during that effort and native build on embedded platforms was also fixed.
CMake builds on Linux don’t have to be installed before Python tests have to be ran. SIMD filters are now also easier to implement.
Recently, I moved to the finance industry. As usually when I start in a new domain, I look at the Python books for it. And Python for Finance from Yves Hilpisch is one of the most known ones.
ATK is updated to 2.2.0 with the major introduction of vectorized filters. This means that some filters (EQ for now) can use vectorization for maximum performance. More filters will be introduced later as well as the Python support. Vector lanes of size 4 and 8 are supported as well as instruction sets from SSE2 to AVX512.
This is also the first major release that officially supports the JUCE framework. This means that ATK can be added as modules (directly source code without requiring any binaries) in the Projucer. The caveat is that SIMD filters are not available in this configuration due to the requirement for CMake support to build the SIMD filters.
ATK is updated to 2.1.0 with a major refactoring of the Python wrappers and extensive testing of them. New filters were also added to support more complex pipelines (mute/solo and circular buffers for real-time spectrum displays) and Audio ToolKit provies now a CMake configuration file for easier integration in CMake projects.
ATK is updated to 2.0.0 with a major refactoring to ensure signed/unsigned consistency, new Adaptive module and EQ design. Complex-valued filters are also now available to allow simultaneous dual channel processes and advanced filters like complex LMS filters.
I’ve started working on porting some Python libraries to Python3, but I required using an old Visual Studio (2012) for which there is no Python3 version. In the end, I tried following this tutorial. The issue with the tutorial is that you are downloading the externals by hand. It is actually simpler to call get_externals.bat from the PCBuild folder.
Be aware that the solution is a little bit flawed. pylauncher is built in win32 mode in Release instead of x64. This has an impact on deployment.
Once this is done, I had to deploy the build to a proper location so that it is self contained. I inspired myself heavily from another tutorial by the same author, only adding 64 bits support in this gist.
Once this was done, time to build Boost.Python! To start, just compile bjam the usual way, don’t add Python options on the command line, this will utterly fail in Boost.Build. Then add in user-config.jam the following lines (with the proper folders):
using python : 3.4 : D:/Tools/Python-3.4.5/_INSTALL/python.exe : D:/Tools/Python-3.4.5/_INSTALL/include : D:/Tools/Python-3.4.5/_INSTALL/libs ;
This should build the debug and release mode with this line:
.\b2 --with-python --layout=versioned toolset=msvc-11.0 link=shared stage address-model=64
.\b2 --with-python --layout=versioned toolset=msvc-11.0 link=shared stage address-model=64 python-debugging=on