Would you like to test paceval.?
Demo Downloads
You can download various demo products from our homepage, e.g., reference implementations
-
for various C++ compilers
-
for mobile platforms
-
for microprocessors, microcontrollers
-
for Internet of things (IoT) solutions and devices
Free Software Development Kit
You can download the Software Development Kit for Windows, Apple, Linux, Android, ARM and RISC-V below.
System architects and software engineers can then easily test the product features and possibilities of paceval.
Be one of the first to get our new paceval. Software Development Kit 4.25.
The Software Development Kit includes demos and examples for Windows, Apple, Linux, ARM, RISC-V the latest compilers and earlier compiler versions.
You can download the installer for Windows here:
Download free ! paceval-Software Development Kit 4.25 (195 MB)
(Downloads – 7,767 : version 4.25)
If you are using other operating systems or processors. We have tailor-made solutions for all operating systems including Android and IBM POWER.
Or if you need a solution for a specific microcontroller like Microchip AVR, TI, STMicroelectronics, NXP Semiconductors or ASICs and FPGAs.
Don’t hesitate to contact us if you would like to license the paceval. library or the source code, send us an email at info@paceval.com.
Changes: (new) paceval. Software Development Kit 4.25
– new API function paceval_dGetMultipleComputationsResultsExt() now officially available
– updated APPLE and RISC-V libraries (the library for APPLE silicon is now the fastest mathematical engine available, see “5 ms” below 🚀)
– overall optimization of performance
– unlimited usage of multicore systems, e.g. 144 cores on Intel’s Xeon, 128 cores on AMD’s Epyc, 192 cores on ARM Ampere’s Altra or higher
– optimization for interval arithmetic (TINC) precision
The document “paceval’s 3-minute pitch” gives investors and customers a very brief summary of our solution.
You can download the document here:
Download paceval's 3-minute pitch.pdf (91.3 KB)(Downloads – 1,308)
document paceval’s 3-minute pitch
In a nutshell: The document describes the software library paceval. from paceval UG, which can perform complex mathematical calculations.
The document “paceval and artificial intelligence – Conversion of neural networks to closed-form expressions” describes the method of exporting any neural network into closed expressions using mathematical formulas. Closed expressions for the outputs of a neural network can also offer several advantages, making them an attractive option for certain applications.
You can download the document here:
Download paceval and artificial intelligence - Conversion of neural networks to closed-form expressions.pdf (761.3 KB)(Downloads – 378)
document paceval and artificial intelligence – Conversion of neural networks to closed-form expressions
In a nutshell: The paper explains simple methods for converting neural networks into closed mathematical expressions. This is particularly useful for systems that do not provide an interface for manipulating tensors. Closed expressions enable efficient computation, simplify integration with other mathematical models, and provide a transparent method for optimizing AI decisions.
The presentation “Mathematical Engine-Motivation” explains our motivation and the idea of creating a new type of engine for mathematics.
You can download the presentation here:
Download paceval-Mathematical Engine-Motivation.pdf (10.5 MB)(Downloads – 574)
presentation Mathematical Engine – Motivation
In a nutshell: This presentation explains what motivated the paceval UG to develop the paceval. mathematical engine. An ideal architecture of a mathematical engine should be based on the principles of transparency, traceability and continuous improveability.
The presentation “New efficient coprocessor in hardware for local intelligence” explains in an easy-to-understand manner the creation of a novel mathematical coprocessor in hardware based on paceval., which saves significant power compared to the solutions used today.
You can download the presentation here:
Download paceval-Vision paper-New efficient coprocessor in hardware for local intelligence.pdf (2.2 MB)(Downloads – 330)
presentation
New efficient coprocessor in hardware for local intelligence
Clearly explains the following questions:
- What hardware is needed for the parallelization approach?
In a nutshell: The approach of parallelizing computations in hardware requires only floating point units (FPUs) on a hardware module, such as a standard field programmable gate array (FPGA) from Intel/Altera or AMD/Xilinx. Processor cores are only required for the hardware driver. FPUs consume far less power than processor cores, are inexpensive to implement, and can be arranged in large numbers in a small space. - How does paceval. work internally?
In a nutshell: paceval. internally creates and processes linked lists of atomic calculations that represent the user’s mathematical expressions. Creating and processing an expression as a linked list offers many advantages over the usual approach, especially in terms of speed.
The presentation “Mathematical Engine-Our solution” describes the implementation of our patent protected mathematical engine paceval. in detail.
You can download the presentation here:
Download paceval-Mathematical Engine-Our solution.pdf (2.8 MB)(Downloads – 520)
Changes: (new) added: paceval. a new coprocessor
presentation
Mathematical Engine in Software, Hardware and in the Cloud – Our solution
Clearly explains the following questions:
- What is paceval.?
In a nutshell: paceval. is a unique, easily integrable mathematical engine that enables product innovation and increases the productivity of software development and maintenance. - How paceval. helps with software development and product development?
In a nutshell: paceval. reduces programming effort, maintenance and service for calculations. This reduces downtime to a minimum. With paceval., innovative ideas are possible without restrictions, as complex calculations become feasible.
The document “Vision paper-“Mathematics is everywhere” Enabling sustainable distributed and decentralized mathematics” explains how you can easily realize a setup for AI that scales to your needs and reduces the environmental impact, too.
You can download the document here:
Download paceval-Vision paper-“Mathematics is everywhere” Enabling sustainable distributed and decentralized mathematics.pdf (1.5 MB)(Downloads – 442)
document
Vision paper-“Mathematics is everywhere” Enabling sustainable distributed and decentralized mathematics
In a nutshell: The vision paper aims to enable sustainable, distributed and decentralized mathematics with a cloud-based mathematical engine.
The document “paceval with Python-the Mathematical Engine as a Service (e.g. for multi-party computations)” explains how you can use the Python module, see https://pypi.org/project/paceval.
You can download the document here:
Download paceval with Python-the Mathematical Engine as a Service (e.g. for multi-party computations).pdf (189.2 KB)(Downloads – 936)
document
paceval with Python-the Mathematical Engine as a Service (e.g. for multi-party computations)
Use our high-performance engine server with Python for free !
In a nutshell: The document describes the use of mathematical engines for IoT solutions that require complex mathematical calculations but do not have the necessary computing power or are battery-operated.
The documents “paceval-service-a Linux server for x64 processors (Intel and AMD)” and “paceval-service-a Linux server for ARM64 processors (includes Raspberry Pi and APPLE silicon M1/M2)” explain how you can get and start the paceval-service with Docker.
The Docker Container Image paceval-service is a Mathematical Engine as a Service, e.g. for multi-party computations.
This service is easy to use, for free and the foundation of the Eclipse Diafanis solution.
You can download the documents here:
Download document-paceval-service-a Linux server for x64 processors (Intel and AMD).pdf (239.1 KB)
(Downloads – 422)
Download document-paceval-service-a Linux server for ARM64 processors (includes Raspberry Pi and APPLE silicon) (239.2 KB)
(Downloads – 377)
documents
paceval-service-a Linux server for x64 processors (Intel and AMD)
and
paceval-service-a Linux server for ARM64 processors (includes Raspberry Pi and APPLE silicon)
In a nutshell: The document describes the paceval-service, which performs mathematical calculations on a remote computer or server.
The document “Vision paper-Faster mathematics with even less power consumption implemented on a mathematical engine in hardware” explains how to create
a product in hardware based on our
product paceval. with an USB FPGA.
You can download the document here:
Download paceval-Vision paper-Faster mathematics with even less power consumption implemented on a mathematical engine in hardware.pdf (1.6 MB)(Downloads – 412)
document
Vision paper-Faster mathematics with even less power consumption implemented on a mathematical engine in hardware
In a nutshell: Although paceval. already has a small binary file, impressive performance and low power consumption, the paceval UG wants to further improve this status.
The document “Vision Paper-Add capabilities of complex and precise mathematical functions to a database” explains how wide data can be analyzed with AI / ML and paceval.
You can download the document here:
Download paceval-Vision Paper-Add capabilities of complex and precise mathematical functions to a database.pdf (2.1 MB)(Downloads – 473)
document
Vision Paper-Add capabilities of complex and precise mathematical functions to a database
In a nutshell: Artificial intelligence models with their representative mathematical functions can be easily used in SQL expressions and further optimized there.
OTHER PRESENTATIONS
To download our presentations, send us an email at info@paceval.com.
presentation
Revolutionary Approach to Implement Transparent Algorithms in Software for Machine Learning Models
We’ve moved all of the other downloadable items to the Login section.
If you want to increase your productivity and need access, send us an email at info@paceval.com.