Most of us are familiar with the concept of speedrunning, which is trying to complete a game (or game levels) in the shortest possible time. However, this is the first time I’ve seen the same CPU treatment as Hackaday writer Julian Scheffers managed to achieve. simulate a functional processor from scratch in Logisim in just six hours.
The project is called Kitchen pipedue to the fact that he needed a name quickly and not a reference to the popular one 19th century style top hat. Scheffers says Stovepipe’s hardware was built in less than four hours, with an additional two hours spent building the assembler afterwards.
ISA (Instruction Set Architecture) was created by the process of removing things that were not absolutely necessary, resulting in the eight main opcodes representing over 512 bits, significantly smaller than one of the previous 8192-bit Sceffers processors, i.e. GR8CPU.
Described as an “exercise in minimalism”, the end result includes just 256 bytes of RAM, zero I/O ports, and a battery acting as the only register accessible to the user. The instructions are described as lasting one fetch cycle and one to three execution cycles.
Compared to Boa³²the latest previous version of Scheffer, the simplified chip is much slower thanks to the 32 registers of this chip. However, it outperformed GR8CPU due to the fact that it only needs one cycle to load an instruction compared to GR8CPU’s three cycles. However, Boa³² looks like a Ryzen 7 9800X3D in comparison thanks to its pipelined architecture and separate address and data buses.
However, the time factor must be taken into account. Scheffers estimates that Boa³² was completed in two months, while this compilation appears to have been put together in a single six-hour session.
While it’s not a record-breaker (to the creator’s knowledge) in terms of performance relative to size, it’s still an extremely impressive achievement – especially for someone like me whose eyes start to water just by explaining the build rather than building it for themselves.
Scheffer claims that if they try to make Stovepipe 2, they will record the time it takes to actually measure speed, but even if you take them at their word, this is a great example of pushing the CPU to its limits. The practical application of designing and simulating a working processor in such a tiny time is questionable at best, but can it be done at all? Pretty amazing if you ask me.
