workflows: Use new custom runners from github

[tstellar]

Summary

Github now provides custom runners for us to use with more CPU,
memory, and disk, so we are switching from the default runners
to 8 core runners for Linux and Windows. This change also allows
us to consolidate jobs now that we have runners with more disk space.

[alexbatashev]

Comments below are totally unrelated to this patch, I'm just curious.

.github/workflows/clang-tests.yml

12–17 ↗

(On Diff #432404)

What's the reason to include pull_request here if all pull requests are automatically closed?

.github/workflows/llvm-project-tests.yml

46–52

Did you consider using sccache here? I see that right now this job takes ~4 hours, and the majority of time is spent on builds, which correlates with my experience building llvm in github actions. Enabling sccache should bring that down to ~3-15 minutes (windows is still much slower than linux, but that's a big improvement over a few hours).

[tstellar]

.github/workflows/clang-tests.yml

12–17 ↗

(On Diff #432404)

This patch is for the release branch where we use pull requests for pre-commit testing. We put the pull requests in the llvm/llvm-project-release-prs repo to keep them separate, so they aren't automatically closed.

.github/workflows/llvm-project-tests.yml

46–52

Is sccache something that is integrated with GitHub or do we have to set up our own server somewhere?

With these new runners, most jobs take around 30 minutes which is a big improvement over the old runners. Here is an example pull requests that uses the new runners: https://github.com/llvm/llvm-project-release-prs/pull/1

[alexbatashev]

.github/workflows/clang-tests.yml

12–17 ↗

(On Diff #432404)

Oh, I see, makes sense then.

.github/workflows/llvm-project-tests.yml

46–52

That’s a fantastic improvement. But builds are still 39 minutes long.

For github-hosted machines I’ve had success with github.com/actions/cache (see for example https://github.com/intel/llvm/commit/95104c079bacd4dd2920c8439d97d3f93e601b18). In my config llvm produced ~1,5 gigs of binaries. 10 gigs cache was more than enough for that task.

On self-hosted machines that’s even easier. You can persist data on the machine itself: https://github.com/intel/llvm/blob/f6420c769d473c00be736b8df361e10dd98a6009/.github/workflows/sycl_windows_build_and_test.yml#L32

[tstellar]

.github/workflows/llvm-project-tests.yml

46–52

Thanks for this information. I will look into enabling this in a follow-up patch.

[asl]

Looks like there are some typos in runner names

.github/workflows/libclang-abi-tests.yml

85

Should be ubuntu-20.04-16x64, no?

.github/workflows/llvm-project-tests.yml

35–36

ubuntu-20.04-16x64 and windows-2022-16x64

.github/workflows/llvm-tests.yml

72

See above

[tstellar]

.github/workflows/libclang-abi-tests.yml

85

I created generic labels for the machines, so we can change the machine types without updating the workflow definition.

[tstellar]

.github/workflows/libclang-abi-tests.yml

85

I just checked and it looks like the labels are gone now. It says that unassigned labels will be removed periodically, so maybe they were removed, because we weren't using them? I can add them back after I commit the patch, or if you think it's better, we can just use the labels provided by github.

[asl]

.github/workflows/clang-tests.yml

3 ↗

(On Diff #432404)

I'd probably suggest to ignore forks here as well. And allow manual triggering

[tstellar]

.github/workflows/clang-tests.yml

3 ↗

(On Diff #432404)

Alright this patch is meant to be applied on top of D129526, so I will fix it in that patch.

[tstellar]

Rebase on top of the latest version of D129526.

[Trass3r]

.github/workflows/llvm-project-tests.yml

46–52

There is also hendrikmuhs/ccache-action which specializes the cache action for ccache/sccache. It's also of interest to cache the ThinLTO cache. The cache sizes need to be tuned a bit since GH limits the total size.
https://github.com/Trass3r/llvm-project/blob/main/.github/workflows/CI.yml

[kwk]

This is just a ping. About the status of this patch.

[tstellar]

This is on hold, until we figure out if GitHub will charge us from the custom runners.

.github/workflows/libclang-abi-tests.yml

steps:
             echo ::set-output name=ABI_HEADERS::.
             echo ::set-output name=ABI_LIBS::libclang.so libclang-cpp.so
           fi
 
 
   abi-dump:
     if: github.repository_owner == 'llvm'
     needs: abi-dump-setup
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04-llvm
     strategy:
       matrix:
         name:
           - build-baseline
           - build-latest
         include:
           - name: build-baseline
             llvm_version_major: ${{ needs.abi-dump-setup.outputs.BASELINE_VERSION_MAJOR }}

.github/workflows/llvm-project-tests.yml

 
 jobs:
   lit-tests:
     name: Lit Tests
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         os:
-          - ubuntu-latest
-          # Use windows-2019 due to:
-          # https://developercommunity.visualstudio.com/t/Prev-Issue---with-__assume-isnan-/1597317
-          - windows-2019
+          - ubuntu-20.04-llvm
+          - windows-2022-llvm
           # We're using a specific version of macOS due to:
           # https://github.com/actions/virtual-environments/issues/5900
           # We need addtional testing to see if our workaround works for
           # new macOS builds.
           - macOS-10.15
     steps:
       - name: Setup Windows
         if: startsWith(matrix.os, 'windows')
         uses: llvm/actions/setup-windows@main
         with:
           arch: amd64
       - name: Install Ninja
         uses: llvm/actions/install-ninja@main
       # actions/checkout deletes any existing files in the new git directory,
       # so this needs to either run before ccache-action or it has to use
       # clean: false.
       - uses: actions/checkout@v3
         with:
           fetch-depth: 250
       - name: Setup ccache
         uses: hendrikmuhs/ccache-action@v1.2

.github/workflows/llvm-tests.yml

concurrency:
   # Cancel intermediate builds: only if it is a pull request build.
   group: ${{ github.workflow }}-${{ github.ref }}
   cancel-in-progress: ${{ startsWith(github.ref, 'refs/pull/') }}
 
 
 jobs:
   check_all:
     if: github.repository_owner == 'llvm'
-    name: Test llvm,clang,libclc
+    name: Test llvm,clang,lld,lldb,libclc
     uses: ./.github/workflows/llvm-project-tests.yml
     with:
       build_target: check-all
-      projects: clang;libclc
-
-  # These need to be separate from the check_all job, becuase there is not enough disk
-  # space to build all these projects on Windows.
-  build_lldb:
-    if: github.repository_owner == 'llvm'
-    name: Build lldb
-    uses: ./.github/workflows/llvm-project-tests.yml
-    with:
-      build_target: ''
-      projects: clang;lldb
-
-  check_lld:
-    if: github.repository_owner == 'llvm'
-    name: Test lld
-    uses: ./.github/workflows/llvm-project-tests.yml
-    with:
-      build_target: check-lld
-      projects: lld
+      projects: clang;lld;lldb;libclc
 
   abi-dump-setup:
     if: github.repository_owner == 'llvm'
     runs-on: ubuntu-latest
     outputs:
       BASELINE_REF: ${{ steps.vars.outputs.BASELINE_REF }}
       ABI_HEADERS: ${{ steps.vars.outputs.ABI_HEADERS }}
       BASELINE_VERSION_MAJOR: ${{ steps.vars.outputs.BASELINE_VERSION_MAJOR }}
steps:
           else
             echo ::set-output name=BASELINE_VERSION_MAJOR::${{ steps.version.outputs.LLVM_VERSION_MAJOR }}
             echo ::set-output name=ABI_HEADERS::.
           fi
 
   abi-dump:
     if: github.repository_owner == 'llvm'
     needs: abi-dump-setup
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04-llvm
     strategy:
       matrix:
         name:
           - build-baseline
           - build-latest
         include:
           - name: build-baseline
             llvm_version_major: ${{ needs.abi-dump-setup.outputs.BASELINE_VERSION_MAJOR }}