Add readthedocs config [skip ci] (#126)

* Add .readthedocs.yaml configuration

Fix some broken links in intro.rst (#119)

Update docs with information on conda package availability (#111)

* Update pyscamp docs with info about conda package.

* Update build instructions for windows CUDA builds. Add disclaimer that 32-bit builds are unsupported and non x86_64 configurations are untested.

* Add pointer to automated GPU benchmarks in documentation.

Cleanup to support conda packaging of pyscamp (#105)

* Prevent SCAMP from using non-MPL-licenced code from EIGEN.

* Add option to build pyscamp with external installs of pybind11 and eigen

* Moves CMake dependencies closer to where they are used in the project.

* Only configure the python module when building pyscamp.

Sickbay windows hardware emulation test for now.

Fix linker errors that can occur when building pyscamp shared libs. (#…

…100)

Clean up pyscamp build logic in setup.py (#96)

* Stop trying to detect visual studio in setup.py.
* Increases the required cmake version for pyscamp to be 3.15 or more.
* Move autoselection of CMAKE_GENERATOR_PLATFORM to inside CMakeLists.txt
* Pass CMAKE_GENERATOR_PLATFORM on Windows when compiling pyscamp. Remove it in SCAMP's CMakeLists.txt if it is not needed.
* Change recommendations for how to specify environment variables to set compilers/generators for SCAMP and pyscamp, This should be more aligned with the normal usage of cmake.
* Update documentation.

Apply optimizations to CPU Kernel reductions to improve performance o…

…n various toolchains. (#93)

* Optimize CPU Kernel Reductions for Various compilers.

Add cpu matrix reduction (#77)

* Make matrix summaries more accurate by allowing a floating point to represent the width of an output cell.

* Add test support for matrix summary profile types.

* Allow run_tests.py to use pyscamp with matrix summaries

* Update documentation to indicate that the matrix summary profile type is available when using CPUs only.

* Add matrix summary pyscamp test.

* Do not output a whitespace on the end of matrix summary output rows.

Dramatically improve performance of CPU kernels on less sophisticated…

… compilers (#78)

* Force inline CPU kernel methods

* Add warnings when compiling cpu kernels and inlining fails

* Optimize CPU kernels to improve the probability that loop vectorization occurs

* Only check for nan values in the CPU kernels if necessary.

* Clang will now output whether or not loops were vertorized during compilation.