[CI] Disable auto CUDA arch injection to avoid duplicate gencode flags

[EmmonsCurse]

Motivation

Recent changes in Paddle #78704 modified the behavior of CUDAExtension, introduced automatic CUDA architecture flag injection via PADDLE_CUDA_ARCH_LIST, even when custom -gencode flags are already specified.

This results in duplicated CUDA arch flags during compilation, increasing binary size and potentially causing linker errors such as:

• relocation truncated to fit

To maintain stable builds and avoid unnecessary code generation, a workaround is required.

Modifications

• Patched extension_utils._get_cuda_arch_flags to return an empty list when user-defined -gencode flags are detected, preventing Paddle from auto-injecting CUDA arch flags.
• Added a secondary safeguard by overriding CUDAExtension._add_cuda_arch_flags to ensure no additional arch flags are appended internally.
• Explicitly controlled CUDA architecture flags via get_gencode_flags, avoiding reliance on PADDLE_CUDA_ARCH_LIST.
• Effectively disabled Paddle’s automatic CUDA arch injection mechanism to prevent duplicated -gencode entries.
• Ensured correct generation of arch=compute_xxa,code=sm_xxa pairs (e.g., 90a, 100a) and avoided incomplete flags like arch=compute_90a.
• Reduced the risk of compilation and linking issues (e.g., relocation overflow) caused by conflicting or duplicated CUDA arch flags.

Usage or Command

N/A

Accuracy Tests

N/A

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[EmmonsCurse]

/skip-ci ci_iluvatar
/skip-ci ci_hpu
/skip-ci stable_test
/skip-ci pre_ce_test

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (develop@a0c39cc). Learn more about missing BASE report.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #7513   +/-   ##
==========================================
  Coverage           ?   73.58%           
==========================================
  Files              ?      398           
  Lines              ?    54988           
  Branches           ?     8616           
==========================================
  Hits               ?    40462           
  Misses             ?    11817           
  Partials           ?     2709           

Flag	Coverage Δ
GPU	73.58% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[EmmonsCurse]

/skip-ci all

[PaddlePaddle-bot]

🤖 AI Code Review | 2026-04-23 22:11:56

📋 Review 摘要

PR 概述：禁用 Paddle 自动 CUDA arch 注入，避免重复 gencode flags 导致编译/链接错误。
变更范围：custom_ops/setup_ops.py（构建脚本 patch）、scripts/check_approval.sh（CI 审批检查）
影响面 Tag：CI OP

问题

级别	文件	概述
🟡 建议	custom_ops/setup_ops.py:47	flag 匹配条件过宽，"compute_" in flag 可能误匹配非架构 flag
❓ 疑问	custom_ops/setup_ops.py:56	cflags=None 时 patch 不生效，是否覆盖所有触发场景需确认

总体评价

check_approval.sh 中统一使用 --merge-base 替换直接 diff，修复了 CI 检查在存在合并提交时误判的问题，改动正确。setup_ops.py 中的 Workaround 思路合理，但 flag 匹配逻辑建议收窄，降低误匹配风险；同时建议补充说明 cflags=None 场景是否已由第二道防线（_add_cuda_arch_flags override）兜底。

[PaddlePaddle-bot]

🟡 建议 flag 匹配条件过于宽松，存在误判风险

当前条件 "compute_" in flag or "sm_" in flag 会对任意包含这些子串的 flag 触发（例如编译器优化参数 --use_compute_mode=sm_xx、日志参数等非架构类 flag），可能导致在不应屏蔽时意外返回空列表。

建议收窄匹配范围，使用更精确的前缀或正则：

import re

if isinstance(flag, str) and (
    flag.startswith("-gencode")
    or re.match(r'^-arch=compute_\d+', flag)
    or re.match(r'^-code=sm_\d+', flag)
):
    return []

或至少改为更严格的前缀检查（如 "-arch=compute_" / "-code=sm_"），避免子串误匹配。

[PaddlePaddle-bot]

❓ 疑问 cflags=None 时 patch 逻辑不生效，是否已有兜底？

当前实现在 cflags 为空/None 时直接透传给原始函数，若 Paddle 内部调用 _get_cuda_arch_flags() 时不传 cflags（例如在 _add_cuda_arch_flags 内部独立调用），第一道 patch 将无效，自动 gencode 注入仍会触发。

第二道防线（_add_cuda_arch_flags = lambda self, flags: flags）依赖 extension_utils.CUDAExtension 存在该方法，但如果 Paddle 版本不满足 hasattr 条件，两道防线均失效。

建议在注释中明确说明：

1. cflags=None 场景是否在实际调用链中不会发生；
2. 或者两道防线如何互补，覆盖不同的 Paddle 版本。