[Tile] Add new tile test mode

[miscco]

The current testing is not sufficient. We need to test classical SIMT with --enable-tile and also --enable-tile within a tile program

For the latter there is now CCCL_FORCE_TILE_TESTS which changes lit to build a __tile_global__ kernel

Also universally add --enable-tile to the cccl compile options so that we can ensure that all projects are build with it

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

📝 Walkthrough

Summary by CodeRabbit

• New Features

  • Added an option to force tile-enabled test configurations.
  • Automatically enables tile support for builds and test runs when selected.
  • Tile-related tests now recognize and follow the forced-tile setting.

• Bug Fixes

  • Added validation for inconsistent tile test configurations.
  • Improved tile-specific compilation so attributes and macros are applied only when appropriate.
  • Ensured forced tile settings remain consistent across build, test configuration, and test execution.

Walkthrough

Changes

The build adds CCCL_FORCE_TILE_TESTS, propagates it through libcudacxx test configuration, applies tile compiler settings, validates macro consistency, and exposes a Lit feature for forced tile tests.

Tile test configuration

Layer / File(s)	Summary
Build tile enablement CMakeLists.txt, cmake/CCCLBuildCompilerTargets.cmake	Adds CCCL_FORCE_TILE_TESTS, uses it as the default for CCCL_ENABLE_TILE, and appends --enable-tile when tile support is enabled.
libcudacxx test propagation libcudacxx/cmake/LibcudacxxBuildCompilerTargets.cmake, libcudacxx/test/libcudacxx/CMakeLists.txt, libcudacxx/test/libcudacxx/lit.site.cfg.in, libcudacxx/test/utils/libcudacxx/test/config.py	Propagates the force option as a compile definition and Lit configuration value. Registers the force-tile feature.
Tile harness behavior libcudacxx/test/libcudacxx/force_include.h, libcudacxx/test/utils/libcudacxx/test/config.py	Rejects inconsistent tile macros and gates tile attributes on both macros. The configuration file also reformats existing setup and diagnostic code.

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (2)

CMakeLists.txt (1)

71-74: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

suggestion: Add an end-to-end forced-tile configuration test. This contract spans CMake, generated Lit configuration, compiler definitions, and __tile_global__ selection; test both forced and non-forced configurations. As per coding guidelines, changes should be validated with targeted builds and tests.

Source: Coding guidelines

libcudacxx/test/libcudacxx/force_include.h (1)

20-22: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

suggestion: Keep preprocessor annotations synchronized with their conditions. The comments omit defined(...) and/or the CCCL_FORCE_TILE_TESTS conjunct, so they describe different conditions than the code. Based on learnings, annotated #else/#endif branches must repeat the exact condition text from the corresponding #if.

Also applies to: 52-64, 138-141

Source: Learnings

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 58f934a5-5218-4773-8e1d-6814b3bf114d

📥 Commits

Reviewing files that changed from the base of the PR and between db07949 and 5c1f692.

📒 Files selected for processing (7)

• CMakeLists.txt
• cmake/CCCLBuildCompilerTargets.cmake
• libcudacxx/cmake/LibcudacxxBuildCompilerTargets.cmake
• libcudacxx/test/libcudacxx/CMakeLists.txt
• libcudacxx/test/libcudacxx/force_include.h
• libcudacxx/test/libcudacxx/lit.site.cfg.in
• libcudacxx/test/utils/libcudacxx/test/config.py

[coderabbitai]

🧹 Nitpick comments (1)

libcudacxx/test/libcudacxx/force_include.h (1)

20-22: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

suggestion: Keep preprocessor branch annotations exact.

Line 22 and Lines 142-144 omit parts of their corresponding #if conditions, including defined(...) and CCCL_FORCE_TILE_TESTS. Update each #else/#endif annotation to repeat the exact condition text.

Based on learnings, annotated preprocessor branch comments must repeat the corresponding condition verbatim.

Also applies to: 140-144

Source: Learnings

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 3e9ed72c-dd3c-4d38-8634-2b8a19fe684c

📥 Commits

Reviewing files that changed from the base of the PR and between 5c1f692 and b64513e.

📒 Files selected for processing (7)

• CMakeLists.txt
• cmake/CCCLBuildCompilerTargets.cmake
• libcudacxx/cmake/LibcudacxxBuildCompilerTargets.cmake
• libcudacxx/test/libcudacxx/CMakeLists.txt
• libcudacxx/test/libcudacxx/force_include.h
• libcudacxx/test/libcudacxx/lit.site.cfg.in
• libcudacxx/test/utils/libcudacxx/test/config.py

🚧 Files skipped from review as they are similar to previous changes (2)

• libcudacxx/test/libcudacxx/CMakeLists.txt
• libcudacxx/test/utils/libcudacxx/test/config.py

[miscco]

pre-commit.ci autofix

[jrhemstad]

@miscco please ensure that this PR is linked to an appropriate issue in #8086. There are already 2 relevant issues on this topic. Please either update one of those or create a new one.

[github-actions]

⏱️ CCCL compile-time benchmark comparison: Public headers compile-time bench

Result: 2 regression row(s), 4 improvement row(s) above threshold.

Run	Value
Config	public-headers-gcc13
Baseline	origin/main
Preset	all-dev
Targets	cub.headers.base, thrust.cpp.cuda.headers.base, libcudacxx.test.public_headers
GPU / launch args	rtx2080 / --cuda 13.3 --host gcc13

Artifacts: reports and traces

Direct file processing

-f file-processing exclusive --sort total

🔴 Direct file processing — Regressions

Rank	Regression impact	Selected Δ	Baseline	Current	Event	Matched traces
1	0.273872	0.273872	11.335505	11.609377	Processing Header File: libcudacxx/include/cuda/std/__type_traits/type_list.h	475
2	0.220949	0.220949	5.514983	5.735932	Processing Header File: libcudacxx/include/cuda/std/__iterator/iterator_traits.h	463

🟢 Direct file processing — Improvements

Rank	Improvement impact	Selected Δ	Baseline	Current	Event	Matched traces
1	1.234938	-1.234938	9.440576	8.205638	Processing Header File: libcudacxx/include/cuda/std/__cccl/prologue.h	548
2	0.558198	-0.558198	3.469345	2.911147	Processing Header File: libcudacxx/include/cuda/std/__cccl/epilogue.h	548
3	0.222234	-0.222234	20.175812	19.953578	Processing Header File: libcudacxx/include/cuda/__device/physical_device.h	94
4	0.213969	-0.213969	7.621734	7.407765	Processing Header File: cub/cub/util_type.cuh	278

[davebayer]

LGTM

[github-actions]

🥳 CI Workflow Results

🟩 Finished in 3h 33m: Pass: 100%/504 | Total: 5d 02h | Max: 2h 05m | Hits: 89%/1670405

See results here.