[ssbuffer](feat) Revert Non-Interruptive block_id Allocation

[cxtverygood]

This reverts commit e9a0f9c.

New contributor declaration

• I am not making a trivial change, such as fixing a typo in a comment.

• I have written a PR description following these
  rules.

• I have run pre-commit run --from-ref origin/main --to-ref HEAD.

• Select one of the following.

  • I have added tests.
    • /test for lit tests
    • /unittest for C++ tests
    • /python/test for end-to-end tests
  • This PR does not need a test because FILL THIS IN.

• Select one of the following.

  • I have not added any lit tests.
  • The lit tests I have added follow these best practices,
    including the "tests should be minimal" section. (Usually running Python code
    and using the instructions it generates is not minimal.)

[github-actions]

🔍 OpenCodeReview found 17 issue(s) in this PR.

• ✅ Successfully posted inline: 14 comment(s)
• 📝 In summary (no line info): 2 comment(s)
• ❌ Failed to post inline: 1 comment(s)

---

[bug · high]

📄 third_party/ascend/include/DynamicCVPipeline/PlanComputeBlock/ComputeBlockIdManager.h

⚠️ GitHub could not post this as an inline comment: No line information provided

The constructor previously skipped block IDs where blockId <= 0, filtering out both zero and negative values. The new code uses if (auto blockId = blockIdAttr.getInt()) which only skips zero (falsy) but now accepts negative block IDs. This is problematic because -1 is used throughout the codebase as a special sentinel value for 'no block ID' (getBlockIdByOp returns -1 as 'not found', updateBlockId treats -1 as 'remove attribute'). Accepting negative block IDs into the manager could lead to incorrect lookups or collisions with the sentinel value.

---

[bug · medium]

📄 third_party/ascend/lib/DynamicCVPipeline/PlanComputeBlock/PlanCubeBlock.cpp

⚠️ GitHub could not post this as an inline comment: No line information provided

The condition for identifying cube operations has changed from isCubeSimpleOpOrCf (which walked into region-branch ops to check child ops' core types) to isCubeOp (which rejects all SCF ops entirely). The old isCubeSimpleOpOrCf was specifically designed to handle RegionBranchOpInterface ops by inspecting their inner ops. The new isCubeOp returns false for any SCF op, including those that may contain only CUBE_ONLY operations. This changes the set of ops recognized as "cube", potentially causing scheduling failures for cube code wrapped in SCF control flow.

---

[documentation · medium]

📄 third_party/ascend/unittest/DynamicCVPipeline_ut/test_sdf22_mlir_four_layer_v2c_inner_dep_outer.py (L11-L11)

⚠️ GitHub could not post this as an inline comment: Line 11 could not be resolved (outside PR diff hunks)

The module-level docstring (line 11) says 'ensure the MLIR code contains the "scope" keyword', which contradicts the new assertion that expects 'scope' NOT to be present. This should be updated to describe the new fallback-scenario validation (scope should NOT be present).

💡 Suggested Change

Before:

  3. Add MLIR content validation in test functions to ensure the MLIR code contains the "scope" keyword

After:

  3. Add MLIR content validation in test functions to ensure the MLIR code does NOT contain the "scope" keyword in fallback scenarios

[cxtverygood]

/retry