Skip to content

unlimitedsoftwareworks/type-v

Repository files navigation

Type-V is a virtual machine and runtime environment for type-c programs. While the VM can be used for generic purposes, its instructions are heavily optimized for type-c source code, such as offset-based data types (structs, arrays, classes, interfaces, etc),

About

Type-V is a virtual machine and runtime environment for type-c programs. The Type-C compiler can be found at https://github.com/unlimitedsoftwareworks/type-c.

Type-V is written in C11 and has minimal dependencies.

The VM tries to store minimal RTTI and is currently very unoptimized. Type-V development goes hand in hand with the Type-C compiler.

Here are the current planne and implemented features:

  • All instructions for all data types
  • Closures
  • Coroutines
  • Source Mapping
  • GC (So far only scavenger is implemented, can allocate up to 1Mb)
  • JIT (Just in time compilation in 2050)
  • Disassembler
  • Lots of optimizations

Type-V performance is very comparable to node in interpreter mode (JIT off, since Type-V doesn't have JIT yet). Type-V even beats node on some benchmarks (unpublished).

Why Type-V

Type-V is a runtime environment for type-c. All instructions are optimized for type-c source code, they are all typed and the VM tries to store minimal RTTI (yet somehow i feel like i'm not succeeding).

Licenses

Name License Usage Link
libtable MIT Table printing https://github.com/marchelzo/libtable

Need help

Head over to the discussion section, and start a discussion. Please do not open an issue, start with a discussion and if the issue is verified (or a feature request is accepted), then we will create an issue. Thank you!

Notes

  • Type-V is not production ready yet.
  • Type-V is not optimized yet.
  • Type-V is not secure yet.
  • Type-V is not documented yet.
  • Type-V is not fully tested yet.
  • Type-V is not JIT'ed yet.

Benchmarks

These benchmarks are compiled with the Type-C compiler and ran with Type-V. Currently Type-C doesn't perform any optimizations at all. So there is a lot of room for improvement.

Hardware & OS:

The performance evaluation was conducted on a system powered by AMD EPYC Processor (Family 25, Model 1, Stepping 1).

  • Cores and Threads: The CPU consists of 2 logical processors with 1 physical core (2 threads per core).
  • Clock Speed: The base frequency is 2.4 GHz (2396.398 MHz).
  • Cache:
    • L2 Cache Size: 512 KB per core.
  • Microarchitecture: AMD EPYC series
  • Memory and Cache Alignment: The cache line size is 64 bytes, with TLB (Translation Lookaside Buffer) size supporting 1024 4K pages.
  • Address Sizes: The processor supports 40-bit physical and 48-bit virtual address spaces.
  • RAM: 2Gb
  • OS: Ubuntu 24.04.1 LTS codename noble

  • Python: Python 3.12.3 (main, Sep 11 2024, 14:17:37) [GCC 13.2.0] on linux
  • Node: Node.js v18.19.1.

Fibonacci(40), recursive

Non-optimized versions, just to test context switching.


Large array addition (100M elements)

Elements are preallocated for two arrays.


Nested member access within large loop