International Symposia on Computer Architecture ISCA@50 Retrospective: 1996-2020

Amazing resource.

The retrospectives in particular are incredibly useful, taking this deep research work & connecting it & adding context.

The Corona paper was a great example. The team built incredibly advanced chip-scale 3d photonics. They assumed electrical connections would become more of a roadblock, but here we are sending 50+ gigatransfers/s across PAM4 copper in ethernet & soon PCIe. Epic. And they talk about the industry's proprietariness preventing things like effective adoption of Wavelength-Division-Multiplexing (wdm) which would help boost density/speed. Great paper to read really really good things getting built & they are happening, but much much slower, in lower-fi ways. A very just quote at the end for their paper, "Make no little plans, they have no magic in them to stir [the] blood."

On a damper note, it worries me a bit that the last decade feel like it has very few papers that will mean much to most of computing. Some of them might potentially see some break-out applications, that see truly great success. But very few seem likely to have broad applications & a significant chance at adoption. Where-as even 10 years ago it felt like most papers reflected the industry more broadly; even if it was talking about fancy data-center fabric, it was still fabric for workloads that were vaguely recognizable by most folk. It feels like we've descended into a much more speculative, much more niche mode of architecting.