| HN Mirror

I don't have a full list, but generally the Toronto office has the following types of software roles: (1) Quartus compiler - this is what takes a digital circuit netlist and tries to map it in an optimal way to the FPGA hardware - doing things like clustering, placement, routing, etc. Lots of very interesting optimization problems. Hardware knowledge is helpful but not required, it's much more important to be able to reason about algorithms and write efficient C++ code to implement them. (The team I'm hiring for is a part of this group). (2) High-level design - basically, how can we make FPGA programming accessible for software engineers who know nothing about hardware? This involves a bunch of compiler work to translate C-like languages (or even higher-level constructs, like machine learning models) automatically to a hardware circuit description, at which point the Quartus compiler from #1 above takes over. Again, hardware knowledge is optional, ability to reason about compilers / algorithms is key. (3) Device modeling - how can we model all of the enormous complexity of a real physical device (with all of its complex timing and power characteristics) into a simplified software model that the Quartus compiler can then use efficiently? Other than these 3 main groups, there are others (e.g. working on specific hardware pre-built blocks that customers can use right away, or working on combination of hardware / software / embedded firmware to make it easier to interface between FPGAs and modern memory microchips).