build: Update C++ standard from c++17 to c++20

[mc-nv]

What does the PR do?

• Upstream PyTorch raised ATen's minimum required C++ standard from 17 to 20 ([12/12] Enforce C++20 minimum in header guards (#178150) pytorch/pytorch#178150), breaking the pytorch_backend build.
• Bumps this repo's C++ standard default from 17 to 20 for consistency across the Triton stack.
• Also fixes a genuine C++20 compile break surfaced by the bump: explicit this capture in a lambda in src/backend_model.cc (implicit capture via [=] is deprecated in C++20, and this repo builds with -Werror=deprecated).

Checklist

• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Related issues are referenced.
• Populated github labels field
• Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• Verified copyright is correct on all changed files.
• Added succinct git squash message before merging ref.
• All template sections are filled out.
• Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type box here and add the label to the github PR.

• build
• ci
• docs
• feat
• fix
• perf
• refactor
• revert
• style
• test

Related PRs:

• build: Update C++ standard from c++17 to c++20 pytorch_backend#210
• build: Update C++ standard from c++17 to c++20 backend#129
• build: Update C++ standard from c++17 to c++20 checksum_repository_agent#16
• build: Update C++ standard from c++17 to c++20 identity_backend#38
• build: Update C++ standard from c++17 to c++20 local_cache#18
• build: Update C++ standard from c++17 to c++20 openvino_backend#123
• build: Update C++ standard from c++17 to c++20 redis_cache#26
• build: Update C++ standard from c++17 to c++20 repeat_backend#21
• build: Update C++ standard from c++17 to c++20 square_backend#24
• build: Update C++ standard from c++17 to c++20 tensorrt_backend#128
• build: Update C++ standard from c++17 to c++20 third_party#82
• build: Update C++ standard from c++17 to c++20 common#166
• build: Update C++ standard from c++17 to c++20 client#922
• build: Update C++ standard from c++17 to c++20 onnxruntime_backend#359
• build: Update C++ standard from c++17 to c++20 server#8962
• build: Update C++ standard from c++17 to c++20 python_backend#455
• build: Update C++ standard from c++17 to c++20 fil_backend#461
• build: Update C++ standard from c++17 to c++20 dali_backend#311

Where should the reviewer start?

• CMakeLists.txt - src/backend_model.cc

Test plan:

• CI Pipeline ID: 67062220

Caveats:

This repo's own build broke under C++20 (see the fix commit) — the deprecated-lambda-capture error was found and fixed as part of this same change, not a separate follow-up.

Background

Part of a coordinated C++17->C++20 bump across ~18 Triton component repos (TRI-1877), triggered by upstream PyTorch's ATen.h now requiring C++20.

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• Resolves: TRI-1877

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with the previous Python-binding inconsistency fully resolved and no new actionable failures identified.

Summary

• Propagates TRITON_MIN_CXX_STANDARD to the Python binding target.
• Explicitly captures this in the asynchronous model-instance creation lambda for C++20 compatibility.
• Renames the internal memory.h and memory.cc files to avoid a C++20-era header-name conflict and updates their references.
• The previous Python-binding standard inconsistency is fully fixed.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Standard[TRITON_MIN_CXX_STANDARD = 20] --> Stub[Server stub]
  Standard --> Core[Core library]
  Standard --> Python[Python bindings]
  Core --> Compatibility[Explicit this capture]
  Core --> Memory[triton_memory header and source]

Loading

_{Reviews (2) · Last reviewed commit: "fix: use TRITON_MIN_CXX_STANDARD for pyt..."}

[mattwittwer]

LGTM