Selecting the Right System Modeling Software March 18, 2024March 18, 2024 admin_mirabilis How to Select the Right System Modeling Software Selecting a system modeling software has become a project by and off itself. In an bygone era, the Architect or Systems Engineer would simply check if there is a support for a particular standard, or has a specific set of IP modeling blocks and a list of analysis. This is no longer the case. In today’s world, the selection of a system modeling software is not so simple. The selection process has gone three-Dimensional or 3D. First the single modeling environment must be a collaborative platform that will be used by the semiconductor, embedded system, software and network engineers. The second is that it must support digital, analog, protocols, task graphs and power systems. The third dimension is that it must support a host of experiments around latency, throughput, power consumption, battery lifecycle, failure analysis, efficiency and safety analysis. In the middle of this 3D Venn Diagram, there must be an environment that support multiple levels of abstraction, multi-mode models of computation and the infrastructure to support emerging technology- without knowledge of what they would be. The last one is challenging. Anyone can provide models of the processors and the interconnects of today. But the real solutions will provide a modeling infrastructure for the solutions of tomorrow. For example, today the hottest requirements are DNN and Chiplet. Tomorrow it might be an TCP/IP on top of UCIe. Being future-aware and future-relevant is the basic requirement of the system modeling solution. Collaboration of the future involves a platform where the OEM, Tier One, Semiconductor and software vendor can all work off a single specification. As the new semiconductor devices become more integrated with lots of computing resources such as CPU, GOU, DSP, DNN and Tensor, the integrator has to work closely with the software team to partition the applications across the reason. These highly efficient and power-saving System-on-Chip or SoCs are the enables of future products such as cars and aircrafts, At the same time, the system modeling solution becomes central to the correct implementation. As a wise man once said, “Eliminate all performance surprise before Integration”. Taking a look at the industry-standard VisualSim Architect, we find they have 40 different interconnects and interfaces- AMBA AHB, AMBA APB, AMBA AXI, AMBA CHI, NIC400, CMN600, CXL, Arteris NoC, PCI, PCIe6.0, UCIe, Ethernet, TSN, AVB, AFDX, UARTS, 1553B, memory Interleaving Crossbar, Network Interconnector, Fiberchannel, Tilelink, TTE, 10BaseT1S, Gateways, CAN buses and so many more. Similarly, VisualSim had over 100 different processor models from the sturdy Leon and Power to the latest ARM 9.0 and RISC-V with vector processing. Now, what happens if a new protocol is added. It will take 3-6 months to make it available. But what if you had an Interconnect generators and a new model could be built in 1-2 weeks. What if you had a new processor and needed a cycle-accurate architecture model by early next week? What if added a new memory controller or a new RTOS scheduler? What is the effort to build these? You do not want to visit a dozen vendor sites or call Application Engineers to grab your models. At that point, integration issues, individual support, compatibility and upgrades become an issue. What happens if your major vendor stops providing architecture models and tells you to use expensive Emulators? This is where an intelligent system modeling solution is the need of the hour. You need a model that will scale quickly as your requirements grow. It is also a methodology that supports anyone new coming along with an easy-to-learn mechanism. Also, you should be able to add other departments to the mix- say the thermal or the fluids team. The best system modeling solution requires a 4D solution- multiple levels of abstraction, large IP library that covers all aspects of engineering, multiple models of computation, and a curriculum that colleges and universities can adopt.