A Stall-Free Real-Time Garbage Collector for Reconfigurable Hardware (2012) [pdf] | Hacker News Reader

A Stall-Free Real-Time Garbage Collector for Reconfigurable Hardware (2012) [pdf]

researcher.watson.ibm.com

63 pointsmpweiher6 comments

Here's an IPFS backup if anyone finds it handy: https://ipfs.io/ipfs/QmPGJSqSgvFBkXryxeHofPLs11cpLURevDFu9rn...

A prior work [1] on implementing Scheme in hardware mentioned doing garbage collection in hardware, too. Such work might be combined with hardware GC’s like this for something like a modern, LISP machine or even imperative like Wirth’s (or Go language).

[1] https://www.cs.indiana.edu/ftp/techreports/TR413.pdf

It is good that people are trying to get computation closer to the memory again.

Yes. I think in the long run we'll look back at the death of Moore's law scaling as the beginning of a new era in innovation. Since it will no longer be possible to defeat clever chip designs through "brute force" process shrinking we'll see an explosion of new (and old) designs.

It is my hope that more fabrication for existing node sizes will be built such that prices will drop precipitously, enabling even small teams such as university engineering schools to design and build radically different computer architectures and actually have them fabricated.

The one that was most interesting to me on bridging the CPU and memory gap was Venray's TOMI:

http://venraytechnology.com/Strategy_Papers.htm

https://www.edn.com/design/systems-design/4429530/Two-Views-...

Their site is in a state of rot, though. That might mean they didn't make it anywhere. They did build something brilliant, though. Hopefully, someone licenses and/or acquires it that will crank out more at lower cost like Adapteva was doing. The second article has a nice summary of what it's doing along with another interesting piece of hardware.

maybe we will see people create chips with thousands of simplified cores on them, which will unleash a new era in graphics and machine learning...

Reply on news.ycombinator.com