v0.0.0

Repository reference

Project root

5 files▾

Overview

Start with the product brief and repository contract.

1▾

README.md
project root

Build and packaging

Native build graph and Python packaging entry points.

2▾

Repository policy

Ignore rules and licensing metadata for the whole project.

2▾

Python wrapper

6 files▾

Entry points

Package marker and the Python CLI command surface.

2▾

Configuration

Default values and runtime settings for the Python wrapper.

2▾

Output

Terminal printing and report shaping for Python-facing flows.

2▾

Native commands

8 files▾

Entry point

Top-level native executable dispatch.

1▾

main.cpp
src

Check command

Interface and implementation for inspection-only checks.

2▾

Doctor command

Interface and implementation for diagnosis and repair flows.

2▾

Planned commands

Reserved files for build, setup, and validate behaviors.

3▾

Native core

16 files▾

Reporting and process

Core report types and subprocess execution primitives.

3▾

Platform and hardware detection

Host, driver, GPU, and CUDA environment discovery.

8▾

Repair and build helpers

Configuration, repo patching, and builder orchestration.

5▾

Kernels and UI

9 files▾

Elementwise kernel

Simple elementwise CUDA sample and shared declarations.

2▾

Memory benchmark

Memory transfer benchmarking kernels and declarations.

2▾

Reduction kernel

Reduction sample kernels and matching headers.

2▾

UI layer

Terminal placeholders and native report rendering.

3▾

Automation and tests

11 files▾

Automation

Bootstrap helpers and Docker environments for local reproduction.

3▾

C++ tests

Native coverage for CUDA environment, driver, and GPU probes.

3▾

Python tests

CLI, configuration, and report-output verification in Python.

3▾

Vendored placeholders

Reserved directories for future third-party dependencies.

2▾

Auto-configuration contract

Implementedv0.0.0

include/core/configure.hpp

This header exposes one job: write `.cuda-doctor.env` when a usable toolkit root can be found and the platform allows it.

configure.cpp doctor.cpp

Reference

Public surfacecpp

struct AutoConfigureResult {
  Probe probe;
  std::filesystem::path env_file;
  bool changed;
};

AutoConfigureResult apply_cuda_env(const std::string& os, const std::filesystem::path& cwd);

src/core/cuda_env.cpp

src/core/configure.cpp

Related files

Core implementation

src/core/configure.cpp

Writes `.cuda-doctor.env` with `CUDA_HOME`, `PATH`, and `LD_LIBRARY_PATH` for the discovered toolkit root.

Command implementation

src/commands/doctor.cpp

Extends `check` with repo scanning, optional auto-configuration, auto-fix behavior, and next-step generation.