Questions
https://bugs.libre-soc.org/show_bug.cgi?id=1095
Introduce yourself and your company.
I am Luke Kenneth Casson Leighton, lead of the Libre-SOC project. For the NGI Search project we created a consortium of the following companies: - Red Semiconductor, lead by David Calderwood - VectorCamp, represented by Konstantinos Margaritis - VanTosh, lead by Toshaan Bharvani
Why is SVP64 relevant to Next Generation Internet Search?
Everyone thinks the Internet is a Software-only environment. Reality is that Software can only run on hardware, and we are addressing the fundamental performance and power-efficiency of the Hardware, by verifying that low-level memory, string, regular-expression and Machine Learning algorithms run optimally on SVP64.
The biggest differentiator however is that unlike say Intel or ARM, if we find that SVP64 is sub-optimal we will take the opportunity provided us by NGI Search to investigate how to actually improve the Instruction Set Architecture.
The bottom line is that hardware-enhanced performance benefits everyone's algorithms, with performance and energy improvements.
Who are the individual consortium members, and what are their specialisations?
- Red Semiconductor is a fabless semiconductor company which will implement working Silicon deploying the SVP64 Enhancements to the Power ISA, making developer kits available to prospective IoT/EDGE customers, the FOSSHW and NGI hardware communities alike
- VectorCamp is a provider of SIMD Vectorization and Software optimization and Training Services.
- VanTosh is an ISV with expertise in porting to the POWER platform, and an MSP for running your private cloud based multi-architecture like POWER machines, we also are very active in open source communities for software and hardware.
Tell me about your project?
With prior kind funding from NLnet and NGI POINTER we have established and implemented the principles of the SVP64 Extension to the Power ISA. This Extension is optimised to greatly simplify mathematical calculations, providing benefits in the fields of Cryptography, Machine Learning, Autonomous systems, Audio/Video, DSP and many other general-purpose areas.
The building-blocks of SVP64 are designed to provide by combination these solutions and we are now working, with the support of NGI Search, to establish new building-blocks to include Advanced Search capabilities, which are also applicable to all of the above.
What milestones have you set and how are you going about achieving them
The three key milestones are: analysis, review and enhancement, and we are already well into the analysis process. The review process will complete Stage 2, and implementation will be Stage 3.
Across the three participating organisations we have a multi-skilled team and there will be cross-organisational cooperation on all Milestones. For those interested, anyone will be able to follow our work in real-time at https://libre-soc.org
What are your goals for the middle/long-time future?
The ultimate goal is to create a progressive family of micro-processors and this grant allows us to validate the critical elements of this project, ensuring that the processor family is efficient at string, data and machine learning and provides support for standard open source software libraries.
How is the NGI Search money helping you?
At a fundamental level it is buying us "thinking time". A considerable amount of time is spent to analyse and understand, to a purpose. This purpose is the consequences for Search as opposed to previously funded developments such as NGI POINTER, which addressed different goals.
How has being part of NGI Search helped you?
We have been part of the NGI Ecosystem now for five years: NLnet (NGI Trust and Ensure), NGI POINTER and now NGI Search. Whilst NGI Search is new and evolving we expect it to be as useful and productive as our previous interactions.
In particular we really appreciate the additional support that comes with the NGI Family: the "added-value" activities, such as training, sharing of ideas, and Mentoring on strategies for Business Development.
Any piece of advice for those who are looking for public funding?
It's actually quite simple: read the "call", make sure your application addresses it directly, and have a coherent vision (a "why" - best explained by Simon Sinek's TED Talk). Grant Applications come primarily in two phases: a "points" system followed by a "human review". It is relatively easy to answer the first phase by ensuring that you have answered the "call", but the second phase is where you need to make people "feel" like they get what you are doing, and that's where Simon Sinek's TED Talk comes into play.