Computing-centric companies worldwide are relentlessly building or improving high-performance computing solutions for highly compute-intensive problems. These organizations are seeking new processor platforms that can be leveraged to deliver maximum performance to end-users on algorithms of interest, with the spotlight on AI. In spite of profound environmental concerns around the humongous energy usage by Information and Communication Technology around the globe (estimated to rise two-fold from the current 1,500 TWH by 2020), the data centers for back-end high-performance computing for AI and other applications have no option for massively parallel high-performance processing. With all guns blazing at the tripartite hurdle of power-hungry generic CPUs & GPUs, hard-to-use FPGA based custom accelerators, and financially infeasible ASICs, Morphing Machines creates customizable massively parallel low power processor platforms to replace current generation conventional processors.

Morphing Machines is a bootstrapped, closely-held fab-less Semiconductor Company originally incorporated in the IISc Bangalore incubation by its Alumni & faculty – Dr. Chandan Haldar (MD & CEO), Prof. S. K. Nandy (Founder & Honorary Chief Scientist), and Dr. Ranjani Narayan (CTO & Director), who have decades of experience with top global computer technology

corporations. This Bangalore-based company offers a deep portfolio of ‘reconfigurable’ silicon cores and software IPs including point IPs such as AES and elliptic curve cryptography IPs, reconfigurable router-based Network-on-Chip IPs, single-precision and double-precision Floating Point Unit IPs, and custom compilers for many-core processors. Current IP core product offerings from Morphing Machines include IP cores for cryptographic applications, FFT computation, and floating-point computation.

Morphing Machines’ path-breaking patented product ‘REDEFINE’ is a scalable, customizable, reconfigurable, power-optimizing, massively parallel processor and SOC platform with distributed-memory and message-passing architecture

With deep expertise and experience in algorithm analysis, architecture, system design, implementation, verification & validation of semi & full-custom chips, and other software tools for parallel, distributed & embedded systems, the company has world-class in-house capability in problem analysis and algorithm characterization, rapid reference implementation in C/C++, and optimization for hardware realization. Further, it includes compilers for targeting custom hardware accelerators, high-level synthesis using advanced functional hardware description languages, along with design & optimization of special purpose application for specific custom functional units.

Defining ‘REDEFINE’
Morphing Machines’ path-breaking patented product ‘REDEFINE’ is a scalable, customizable,

reconfigurable, power-optimizing, massively parallel processor and SOC platform with distributed-memory and message-passing architecture. Its hardware platform comprises a RISC-V ISA based multi-core compute fabric interconnected by a high-speed Network-on-Chip which unleashes the power of data-flow driven computing and multi-level parallelism (DLP, ILP, & TLP) to massively accelerate a variety of compute-intensive algorithms including Deep Learning, Machine Learning, AI, Numerical Computation, Big Data Analytics, Block-Chain, and Reactive Control Systems. New custom hardware acceleration units can also be configured post-silicon by dynamic runtime configuration of special reconfigurable function units. Right-sized virtual cores are dynamically created at runtime by partitioning the reconfigurable REDEFINE compute fabric to optimally match and execute compiler-generated executable software kernel structures to deliver maximum algorithmic performance at minimum power.

Attracting International Attention
Featured as one of the top 50 Finalist (Top 6 in Data and AI Track) at ‘Hello Tomorrow 2016’ global deep technology startup summit at Paris and in Gartner Research Cool Vendors 2011 report, Morphing Machines has established a small but steady IP licensing revenue stream from customers akin to The Safran Group. Moreover, Morphing Machines’ REDEFINE technology was demonstrated in the 4th RISC-V Workshop at MIT, Cambridge in 2016. “Once the ASIC fabricated initial prototype REDEFINE Neuro Core SOC is available, we’ll announce wider availability of real REDEFINE hardware for Indian & global customers in the form of a compact accelerator board named the REDEFINE Machine Learning and AI Platform,” reveals Dr. Chandan.