emmtrix Code Generator

Our Solution for MATLAB® Code Generation

emmtrix Code Generator (eCG) translated MATLAB®, GNU Octave, or Scilab code into platform-independent and readable C or C++ code suitable for embedded processors. The generated code was easy to understand, prepared for parallelization, and could be adjusted to individual requirements. Automatically generated reports helped with the code certification process. In combination with emmtrix Parallel Studio, eCG enabled multicore programming or vectorization directly from *.m or *.sci script files. Furthermore, eCG worked hand in hand with emmtrix Model Code Generator to support C code generation from Simulink® models.

eCG Workflow

emmtrix Code Generator used *.m script files (MATLAB® or GNU Octave) or *.sce/sci script files (Scilab) as input and generated sequential C code according to C89, C99, or C11 standards as output. All generated source and header files could be directly compiled without dependencies on external libraries. A direct generation of executable files was also possible. When fixed-point data types were used, C++ templates were employed to represent them.

The generated C code was static and free of dynamic memory allocation, making it suitable for embedded systems and for parallelization tools like emmtrix Parallel Studio. Data dependencies could easily be analyzed. Matrices were represented by multi-dimensional arrays to enable loop transformations required for parallelization or vectorization.

Users could extend their input scripts with special eCG directives (similar to pragmas in C). These directives were treated as comments by the original tools, ensuring code compatibility. The special directives could be used to provide additional information in the output code, select different algorithms for built-in functions like FFT, or enable various code optimizations to improve performance, memory consumption, or parallelizability.

Features

Supported code generation compliant with ISO C90, C99, C11, C18, C++98, C++11, C++14, and C++17 standards
Performance & memory analysis
Highly comprehensive target-optimized C/C++ code
User-controlled cache and memory optimization
Bidirectional traceability via code generation reports
Embedded code generation without dynamic memory allocation

Interfacing custom C/C++ code or libraries
Automatable back-to-back tests for functional validation
Usable within continuous integration processes
Automatic floating-point to integer number conversion
Profiling-based performance analysis and visualization within GUI
Worst-case analysis
Optional array bound checking within generated code

Supported MATLAB®/GNU Octave/Scilab Language Features

N-dimensional arrays
Matrix operations
Processing of multiple input files
Variable-sized variables (growing arrays)
Subscripting, linear and logical indexing
Double-precision, single-precision, and integer data types

Complex numbers
Fixed-point arithmetic
Program control statements if, for, while, break, continue and switch/case
Arithmetic, relational, and logical operators
Function handles

Structures and array of structures
Global variables
Recursive function calls
Global and local functions
Functions with variable-length input/output arguments
Over 100 MATLAB®/GNU Octave/Scilab functions supported