ci.yml

name: CI

on:
  pull_request:
    branches:
      - main
      - gh/**
  push:
    branches:
      - main
    tags:
      - ciflow/rocm/*
  workflow_dispatch: # Allows manual triggering

concurrency:
  group: ${{ github.workflow }}-${{ github.ref == 'refs/heads/main' && github.run_number || github.ref }}
  cancel-in-progress: true

permissions:
  id-token: write
  contents: read

jobs:
  build-gpu:
    name: Build GPU
    uses: ./.github/workflows/build-gpu.yml

  # CPU jobs skip on ciflow/rocm tag push since they already run for the pull_request trigger
  build-cpu:
    name: Build CPU
    if: ${{ !startsWith(github.ref, 'refs/tags/ciflow/rocm/') }}
    uses: ./.github/workflows/build-cpu.yml
    with:
      artifact-name: monarch-cpu-${{ github.sha }}-py3.10

  test-cpu-python:
    name: Test CPU Python
    if: ${{ !startsWith(github.ref, 'refs/tags/ciflow/rocm/') }}
    needs: build-cpu
    uses: ./.github/workflows/test-cpu-python.yml
    with:
      artifact-name: monarch-cpu-${{ github.sha }}-py3.10

  # Runs pyright as its own check so type errors surface as a separate
  # signal. This job is intentionally omitted from status-check.needs so type
  # failures never gate test results or merging.
  type-check-python:
    name: Type Check Python
    if: ${{ !startsWith(github.ref, 'refs/tags/ciflow/rocm/') }}
    needs: build-cpu
    uses: ./.github/workflows/type-check-python.yml
    with:
      artifact-name: monarch-cpu-${{ github.sha }}-py3.10

  test-gpu-python:
    name: Test GPU Python
    needs: build-gpu
    uses: ./.github/workflows/test-gpu-python.yml

  test-cpu-rust:
    name: Test CPU Rust
    if: ${{ !startsWith(github.ref, 'refs/tags/ciflow/rocm/') }}
    uses: ./.github/workflows/test-cpu-rust.yml

  test-gpu-rust:
    name: Test GPU Rust
    needs: build-gpu
    uses: ./.github/workflows/test-gpu-rust.yml

  build-docker:
    name: Build Docker image
    needs: build-gpu
    if: ${{ !startsWith(github.ref, 'refs/tags/ciflow/rocm/') }}
    uses: ./.github/workflows/build-docker.yml
    with:
      # Docker image requires python 3.12
      artifact-name: monarch-cuda-${{ github.sha }}-py3.12

  status-check:
    name: Status Check
    runs-on: ubuntu-latest
    needs: [test-cpu-python, test-gpu-python, test-cpu-rust, test-gpu-rust]
    if: always()
    steps:
      # Fail if any job failed or was cancelled; skipped jobs are OK
      - name: Check all jobs status
        if: ${{ contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled') }}
        run: exit 1