Big picture: While everyone who closely follows the technology industry realizes that there is an important connection between today’s latest semiconductor chips and the performance of software applications, a new, deeper level of connection seems to be evolving between the two. From cloud-based chip emulation kits to AI infrastructure enhancements and new tools for building applications that run over network connections, we seem to be entering a new era of silicon-centric software optimizations that are being created – unexpectedly, unexpectedly – by chip manufacturers. themselves.
Intel’s first On event is an update to the company’s former IDF, or Intel Developer Forum, where the company highlighted the importance of software optimization in a series of announcements. Collectively, these announcements, including the launch of the updated Developer Zone website – once again highlighted Intel’s commitment to putting developers of all types and levels at the forefront.
Intel’s focus on its oneAPI platform underscores the increasingly important role that software tools play in getting the most out of today’s advances in silicon. OneAPI, first introduced last year, aims to greatly simplify the process of writing applications that can use both x86 processors and GPUs and other types of accelerators through an open unified programming environment. The upcoming 2022 version is expected to be released next month with over 900 new features, including things like the C ++ / SYCL / Fortran unified compiler and Data Parallel Python support.
With the increasing complexity of processors and GPUs, it’s easy to see why new developer tools are needed to not only make better use of these individual components, but also to take advantage of the capabilities that are made possible by the combination of these chips. While Intel has always had a robust set of in-house software developers and built cutting-edge tools such as compilers that play an important role in the development of applications and other software, the company has focused on a tiny elite segment of the entire developer market. To achieve greater impact, it is important for a company to create software tools that can be used by a much wider range of developers. As Intel itself put it, the company wants to “meet with developers wherever they are” about skills and experience.
As a result, Intel is doing things like increasing the number of oneAPI centers of excellence to provide more places where people can learn the skills they need to make the best use of oneAPI. In addition, the company announced that it is developing new acceleration engines and new optimized versions of popular AI toolkits designed to make the best use of its future data center-focused Intel Xeon Scalable processor.
The company also talked about cloud-based chip emulation environments that could enable a wider range of programmers to create applications for different types of accelerators without having to physically access them.
Obviously, many of these developments are due to the increased complexity of the various individual chip architectures. While architectural improvements in these semiconductors have clearly improved overall performance, even the most experienced programmers are unlikely to be able to keep track of all the functions involved.
Add to that the ability to create applications that can use multiple types of chips, and well, it’s easy to imagine how quickly things can get overwhelming. It’s also easy to see why few applications really take full advantage of modern processors, GPUs, and other chips. Conversely, it’s not hard to see why many applications (especially those that don’t get updated on a regular basis) don’t work as well as they could with some newer chips.
Taking a step back, what is interesting about this Intel effort is that it appears to be a logical continuation of other similar actions recently announced by Arm and Apple, two other silicon-focused companies (each in its own way). At last week’s Arm Developer Summit, the company highlighted its own efforts to build cloud-native hardware with virtual chips through its Total Solutions for IoT program and its new Centauri project. As with Intel, Arm’s goal is to expand the pool of developers who can potentially write applications for its architecture.
In terms of chip-centric software optimization, early benchmark data seems to suggest that the latest Arm-based Apple M1 Pro and M1 Pro Max chips offer extremely impressive results with Apple’s own apps that have no doubt been optimized for these new chips. … However, for other types of unoptimized applications and workloads, the results appear to be slightly more moderate. It just emphasizes that increased software optimization is required to achieve the best possible performance for a given chip architecture.
Now, some might argue that the hardware-independent (or at least hardware-ambivalent) nature of Intel oneAPI is in direct conflict with the carefully tuned chip-specific software optimizations that include some of Intel’s other new designs. At a basic level, however, they all highlight a significant portion of the low-level software enhancements that increasingly complex new chip architectures (such as new 12th Gen Intel Core processors) require to achieve the best possible performance.
Simply put, it is becoming increasingly difficult to achieve the clean, ubiquitous performance gains that we have enjoyed with advances in hardware chips. As a result, increasingly sophisticated and well-optimized software will be required to keep computing performance at the level we are accustomed to. Seen in this light, Intel’s recent efforts are likely to bring more attention to the platform in the short term, but will have an even more significant impact over time.
Bob O’Donnell – Founder and Principal Analyst TECHnalysis Research, LLC a technology consulting company that provides strategic consulting and market research services to the high-tech industry and the professional financial community. You can follow him on Twitter @bobodtech…