Skip to content

This repository contains files regarding my CPU designs

License

Notifications You must be signed in to change notification settings

martandrMC/cpu-design

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

26 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CPU Designs Library

This repository contains all relevant info about (most) processor architectures I designed. That includes schematics, assemblers, code examples, and detailed instruction set documentations.

Contents

A few words for each architecture

Puny Processing Unit

The PPU was concieved from a task I had laid out to create a Harvard memory layout architecture that would minimise the bits an instruction word would use. In the end that was 10 bits though with many compromises in the maximum sizes of the memory. (64 program locations, 16 data locations)

It of course had some other quirks like an unheard-of 2-operand + accumulator system in which an ADD would be done like this: [Dst = Acc + Src] whereas contemporary 2-operand architectures like x86 would do it like this: [Dst = Dst + Src]. There is no real reason for this decision other than I was naive and had no meaningful designing experience at that time.

Smolproc

The Smolproc was the first processor that I designed to feature a pipeline. It was a proof-of-concept that eventually even ended up on an FPGA. The architecture itself was based off of a predescessor of PHINIX which got scaled down from 16 to 8 bits. Due to the pipeline layout the system could be set up with a Harvard or a Von Neumann layout if a 2-port memory was available.

The instructions use the contemporary 2-operand system and are able to address 256 words (bytes) of memory and are variable in size. That is because it is the simplest way to get a full 8 bits immediate using word-based instructions. So in effect every instruction that needed an immediate would be 2 words in size. This would cause a NOP to be generated in the pipeline during that instruction's execution. Due to a happy accident, all memory operations are 2 words, which totally eliminated the possibility of a memory dependency requiring a stall. As a result, Smolproc never stalls and its forwarding system is complete. The branch delay slots were left in, however.

protoPHINIX

The (prototype) Pipelined INteger Instruction eXecutor is the processor that ended up emerging after the development of the architecture that Smolproc was based off of continued. It features an even deeper 6 stage pipeline with a dual write register file and a word size of 16 bits.

PHINIX

PHINIX is a successor to protoPHINIX for which are planned various tweaks to improve performace. Namely a Predicate instruction is planned to be included to reduce the slowdowns of full sized branches and also the dual-write register file is planned to be scaled back down to single-write due to the complexity of forwarding.

PHINIX+

PHINIX+ is a further enchancement on the lineage that is meant to be an approachable target for high level languages while still working with units of 16 bits.

Pipedream

Pipedream is the culmination of my current aspirations for a semi-modern processor featuring a modern modified Harvard memory layout, virtual memory, privalege levels, interrupts and many more. It rightfully deserves that name due to its purely hypotherical status.

About

This repository contains files regarding my CPU designs

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published