Indian startup Calligo leverages POSIT with RISC-V for high-performance computing

April 25, 2022 -- A significant advantage that India has in the semiconductor sector is the presence of several fast-growing fabless chip design startups. Either homegrown or with overseas investment, these new generation companies have been able to come up with solutions that promise to bring innovation to a wide range of verticals.

One such startup is the Bangalore-based Calligo Technologies, a nine-year-old company focusing on the development of products and solutions serving high-performance computing (HPC), big data, and AI/ML segments to global marquee enterprises.
What makes the company stand out is its approach to developing a co-processor capable of doing computations using the Posit Number System (PNS). Combining hardware and software, Calligo aims to offer value to the customer by enabling them to take full advantage of the computing power.

Why POSIT makes a difference

Since its development in 1985, IEEE 754 has been a popular technical standard for many hardware floating-point units. But despite all its benefits, this number system suffers from several weaknesses. In 2016, the American mathematician John Gustafson invented the POSIT number system that offered better mathematical accuracy and dynamic range.
“Unlike earlier forms of universal number (unum) arithmetic, POSITs do not require interval arithmetic or variable size operands – like floats, they round off if an answer is inexact,” explains Anantha Kinnal, chairman and managing director of Calligo. “However, they provide compelling advantages over floats – larger dynamic range, higher accuracy, identical results across systems, simpler hardware, simpler exception handling. POSITs never overflow to infinity or underflow to zero, and ‘Not-a-Number’ (NaN) indicates an action instead of a bit pattern.”

Posit Number Unit (PNU) takes less circuitry than IEEE float FPU. Using lower power and smaller silicon, POSIT operations per second (POPS) supported by a chip can be significantly higher than FLOPS using similar hardware resources. GPU accelerators and deep learning processors can do more per watt and dollar with POSITs while delivering superior quality.

How Calligo leverages POSIT

Kinnal points out that while Calligo uses POSIT, its strength is in providing a combined solution – software and hardware optimization along with the acceleration of application software, leveraging the full capability of the compute power deployed by each organization.

“Calligo Technologies has developed a co-processor Intellectual Property (IP) called POSIT Numeric Unit (PNU) capable of doing computations using the POSIT Number System,” Kinnal said. “Calligo has integrated this PNU co-processor with the RISC-V Processor, creating a highly power-efficient and computationally more accurate CPU core called CRISP-core (Calligo RISC-V with POSIT). Our multi-core System on Chip (SoC) called TUNGA will be capable of addressing the requirements of High-Performance Computing (HPC) & AI.”

Complementing this hardware, Calligo has modified RISC-V C/C+ compilers to generate POSIT-enabled executables without the need for any source-level changes, enabling Linux OS booting, compilation, and running C/C++ applications with POSIT-enabled executables.

The company has already demonstrated its solution at several global events – RISC-V Event in June 2019, HiPC Conference in Dec 2019, and VLSID Conference in Jan 2020.

Potential customers and partnerships

Calligo’s target customers are those who require high-performance computing (HPC). Until recently, HPCs were confined to scientific and engineering endeavors that require massive number crunching. But now, enterprises are deploying their own big data analytics as they seek to process and understand the ever-increasing torrents of data.
“Our CRISP core is a boon to any form of computing, as Posit Number System can potentially be alternate to Floating Point Number system,” Kinnal explained. “The initial market segments that we shall target are energy, oil & gas, scientific research & academics, life science & computational fluid dynamics, followed by other industries.”
The company is also keen on partnering with the other industry players that can add value to their initiative. Calligo’s initial solution will be an accelerator card that consists of CRISP and the associated system and software as a total solution. The software level improvisation will consist of code modernization – this involves parallelization and optimization at the source code and library levels.
“We will thus need to partner with OEMs to be able to certify our product co-functioning in their server systems, application software providers to get access to their source code for improvisation, and cloud service providers to be able to perform improvisation,” Kinnal said.

Growth potential and demand

High-performance computing is increasingly becoming one of the most sought-after technologies. According to Markets and Markets, the HPC market size is expected to grow from US$37.8 billion in 2020 to US$49.4 billion by 2025, at a CAGR of 5.5 percent. The main factor driving the growth is the ability of HPC solutions to process large volumes of data with speed and accuracy.
This bodes well for Calligo, which firmly believes in the future of HPC as an integral part of solutions to large computational needs. Several modern enterprises often push their computing infrastructure to its limits to derive value for their operations, leading to bottlenecks. HPC comes as the answer to this challenge, and companies like Calligo are well-placed to take advantage of it.