Allow TMA inner reduction to test CI

[tbqh]

No description provided.

[tbqh]

!test

[github-actions]

Description

• Remove TMA reduction option check to enable testing in CI

• Allow TMA inner reduction to be used whenever mayUseTma() returns true

• Simplify use_tma boolean assignment by removing && condition

Changes walkthrough

	Relevant files
Tests	reduction.cpp Remove TMA option check for CI testing                                      csrc/scheduler/reduction.cpp Removed isOptionEnabled(EnableOption::TmaReduction) check from use_tma assignment Now use_tma depends only on mayUseTma(props) result Enables TMA reduction for CI testing by removing option guard +1/-2

PR Reviewer Guide

Here are some key observations to aid the review process:

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review

Removed Option Guard The PR removes the isOptionEnabled(EnableOption::TmaReduction) check, which means TMA reduction will now be enabled whenever mayUseTma(props) returns true, regardless of configuration. This removes important runtime control over TMA reduction behavior and could lead to unintended performance regressions or correctness issues in production environments. bool use_tma = mayUseTma(props);

Test failures

• (Medium, 4) nvFuser contiguity.size() mismatch in test_multidevice::test_welford across multiple runners

  Test Name	GB200	GB200 (dist.)	H100	H100 (dist.)	Source
  tests.python.multidevice.test_multidevice.test_welford	❌	❌	❌	❌

• (Medium, 2) NVFuser internal assert on cpAsync Bulk in tests.python.direct.test_repro::test_shared_memory_usage

  Test Name	GB200	Source
  tests.python.direct.test_repro.test_shared_memory_usage[nvfuser_direct_test=eager]	❌
  tests.python.direct.test_repro.test_shared_memory_usage[nvfuser_direct_test=lru_cache]	❌

• (Medium, 2) NVFuser internal assertion failure in NVFuserTest.InnerReductionUnrollVectorization across multiple runners

  Test Name	GB200	H100	Source
  NVFuserTest.InnerReductionUnrollVectorization	❌	❌	Link