Skip to content

ROADMAP.md

Latest commit

 

History

History
114 lines (94 loc) · 3.98 KB

ROADMAP.md

File metadata and controls

114 lines (94 loc) · 3.98 KB
Raw

Roadmap

The ultimate goal is to make Go, the branded modern C, become the actual modern C, featuring great usability and platform native support.

We are well aware it is an ambitious project regarding to our small scale, and it takes time and effort in order to make it actually happen, currently we are working on following tasks:

Long-term

  • Encourage Go developers to build more allocation-free modules.

Phase 1: Make Go Suitable for System Programming (ongoing)

Stage 1: Up & Running - Single-Thread Application (current)

NOTE At this stage, we will prioritize the js/wasm support.

  • Be compatible with go1.21 toolchain.
  • Better structuring of the std.
  • Expose all low-level & reusable bits of the runtime.
  • Provide apis for custom memory allocator.
  • Stacktrace.
  • Stack growth.
    • Copy stack.
  • Debugging with delve.
  • Program startup
    • js/wasm
    • wasip1/wasm
    • linux/{arm64, amd64}
    • android/{arm64, amd64}
    • darwin/{arm64, amd64}
    • ios/arm64
    • windows/{arm64, amd64}
  • FFI support (call and callback)
    • js
    • wasm/wasi
    • syscall/linux
    • vdso
    • C
      • aapcs64
      • sysv{32, 64}
      • fastcall
      • stdcall
      • win64
    • objc (based on C)
    • jni (based on C)
  • Platform SDK (pcz codegen)
    • DOM apis (code generation with WebIDL)
      • Support converting WebIDL to YAML spec in h2y.
      • Support Go package generation for WebIDL in ffigen.
      • Use browser-compat-data for comments.
    • WebExtension apis (Manifest Version 3).
      • Normalize chromium web extension idls & json.
      • Resolve package conflicts among namespaces.
    • nodejs apis.
    • android NDK
      • Add C header file support in h2y
      • Support Go package generation for C in ffigen.
    • darwin XCode (*.sdk/*.framework)
      • Add Objective-C header file support in h2y
      • Support Go package generation for Objective-C in ffigen.
    • win32 Windows SDK (see tasks for android NDK)

NOTE A lot of design decisions to be made in the stage 1 tasks require taking stage 2 tasks into consideration, we won't create any barrier for any disscussions under any circumstances, stay ontopic and you are always welcome.

Stage 2: The go keyword reMastered - Multi-Thread Application

  • Define apis for building custom goroutine support.
  • Port Sched program from the official Go std for goroutine.
    • Generalize and expose internal/poll.
  • Port GC program from the official Go std to work with custom allocators.
  • Runtime feature support.
    • race
    • asan
    • msan
    • pprof
    • fuzz

Stage 3: Dream Come True? - Cross-platform Native GUI

  • GPU support
    • vulkan
    • directx
    • metal
  • Wrap native GUI widgets exposed by platform apis.
  • Constraint layout for native GUI widgets.
  • opentype font support.

TBD.

Phase 2: Compile C in Go

  • Support static binary build when working with C (go build -buildmode=static)
    • Implement a pcz subcommand to support go external linking mode.
  • Standardize linker scripts in YAML (in our opinion, they resembles a lot)
    • Implement tool (l2y) to convert existing linker scripts to YAML.
    • Implement support for linking custom sections in pcz.
  • Support unsafe.SelectABI for foreign ABIs in go compiler (inspired by golang/go#44065).
  • A libc implemented in Go.
  • A configurable run-time linker (ldso).
    • Define platform agnostic standard for thread-local storage relocation.

Phase 3: Build kernels in Go

  • Add missing privileged instructions to $GOROOT/src/cmd/asm.
  • UEFI libraries in Go
    • OS Loader
  • Drivers
  • Unikernel library (Inspired by eliasnaur/unik).

Phase 4: Make Go Toolchain the Modern LLVM?

TBD.