MtxVec v6

Multicore math engine for science and engineering

For Delphi and C#, and now native C++

Delphi, C# or C++: deliver 10× faster code with AI conversion to MtxVec

Built on one math engine

MtxVec at the core — signal processing, statistics and data mining on top, plus a ready-made analyzer.

MtxVec

Multicore math engine for science and engineering

Data Miner

Artificial intelligence enabling components

Latest news

24 May 2026

Dew Lab Studio 2026 for Delphi and C++Builder

We are happy to announce the release of Dew Lab Studio 2026 for Delphi and C++Builder. This release is special, because it can easily qualify as the biggest feature improvement in the last decade and yes this was facilitated by the rise in the AI tool capabilities. It is the first time perhaps in history…

Read the article

One library for the whole numerical stack

Vectors, matrices, linear algebra, statistics and signal processing behind one object-oriented API.

Delphi, C# and
native C++

One engine in three languages: Embarcadero Delphi, .NET with C#, and native C++. C++Builder continues to use the Delphi packages.

Signal processing and statistics on the same engine

DSP Master and Stats Master build directly on MtxVec.

Why it is fast

10× where vectorized code starts
AVX-512 SIMD end to end — typically 2–3× beyond a vectorizing C++ compiler
32× on 32 cores — vectorized code scales linearly with thread count

Vectorized code

Every function uses the CPU's SIMD instruction sets, up to AVX-512, so one instruction processes many values.

Scales with every core

Per-thread memory pools remove lock contention, so vectorized code scales linearly with thread count.

Optimized primitives

Built on LAPACK and Intel's performance libraries, wrapped in a set of highly optimized primitives that every higher-level algorithm uses.

A faster scalar math library

For code that cannot be vectorized: IEEE 754-2008 compliant sin, cos, exp and ln for Win32 and Win64, about 2× faster than the .NET runtime and 7× faster than Delphi's built-ins. The vectorized versions keep a 5× lead.

Product features

Designed for large data sets, with complete vector and matrix arithmetic:

A comprehensive set of mathematical, signal-processing and statistical functions
Floating-point math accelerated with AVX, AVX2 and AVX-512
Vectorized code scales linearly with thread count — ideal for massively parallel systems
Compact, readable vector and matrix code
AI prior knowledge to migrate existing code to MtxVec patterns
Native 64-bit execution for memory-hungry applications
Shorter development: range checks and automatic memory management guard against a wide range of errors
Floating-point precision selectable at run time
No royalty fees on the compiled binaries you distribute

Having converted our models from Matlab, a speed increase of over 10x has been achieved. This core acceleration finally allows these models to be used in real-time automation control.

— Ned Scupolovic, Senior Systems Engineer, Hatch IAS (Embarcadero C++Builder)

Some of our customers

Partners

Embarcadero Technology Partner

Numerical library for Delphi, C# and C++ developers

Dew Research builds mathematical software for scientific and engineering computing, trusted by many customers. MtxVec for Delphi, C# and C++ is an alternative to Matlab, LabVIEW, O-Matrix and SciLab — a high-performance math, statistics and signal-processing library for:

Embarcadero Delphi and C++Builder — numerical libraries and components
.NET — C# and Visual Studio
Native C++