News for hackers. Stuff that matters.
Re:Resurrecting the SuperH architecture by nickpsecurity on Wednesday July 01, 2015 @03:15PM (#9814427)
Good to see them doing it. I included SuperH in my list [1] of non-Intel architectures and old hardware to use post-Snowden. Additionally, I proposed that it falling out of favor despite Japanese chip-makers backing it might make it a nice candidate for trying to get them to open the design but still sell it. Concept is a proven design which can be verified by third parties, masked by whoever, and taped out at fab of their choice. Although, there's work in moving to new nodes and that cost would be on whoever did it. The precedent is Gaisler's SPARC-based processors and I.P. [2] that are dual-licensed as commercial and GPL with tools for easy customization.

Alternatively, I proposed the security enhancements for processors showing up in academia be applied to this or another processor with low market share as a differentiator. Some of these enhancements take almost no chip real-estate, esp simple tags & tag-checks. The chip designer could also make money for the semi-custom work. Time has passed, that didn't happen for low market chips, and did happen for AMD+Intel for non-security applications [that I know of]. Matter of fact, even though my scheme didn't happen, AMD is making so much money off the other half of my proposal that they could be cited when trying to convince chip-makers to do it for security enhancements with mass-market availability. So long as they don't bear the cost of failure (huge in ASIC's) they might go for it.

Finally, anyone wanting to deploy this or other things, remember the Structured ASIC's with FPGA conversions. eASIC [3] has a long track record in this with offers down to 28nm. Gigoptix [4] does S-ASIC's down to 28nm. Tekmos [5] offers a similar product at 350nm (good for budget masks). Just make sure you design in FPGA's with ASIC transition in mind from the start & follow published advice on that (available with Google or consultation). The result is you prove it in FPGA's, even use it on FGPA boards, and then move it to S-ASIC later for reduced costs/power + maybe speed increase. Authors are right that 180nm is a sweet spot although proving it at 350nm or higher first might be smarter given costs.

[1] https://www.schneier.com/blog/archives/2013/09/surreptitious...

[2] http://www.gaisler.com/

[3] http://www.easic.com/products/28-nm-easic-nextreme-3/

[4] http://www.gigoptix.com/products/asics/asic-type/structured-...

[5] http://www.tekmos.com/products/asics/process-technologies

Re:Resurrecting the SuperH architecture by nickpsecurity on Wednesday July 01, 2015 @05:31PM (#9815328)
EDIT to add: Triad Semiconductor [1] has a mixed-signal ASIC take on S-ASIC's. Interesting stuff. Just found it.

[1] http://www.triadsemi.com/vca-technology/