Update CANN Open Software License to Version 2.0#5
Merged
Merged
Conversation
Update the license from CANN Open Software License Agreement Version 1.0 to Version 2.0. Key changes include expanding processor compatibility from Ascend processors to Huawei AI Processors (including Ascend, Kirin, and Yueying branded chipsets), clarifying usage restrictions and distribution terms, and strengthening the termination provisions with enhanced legal remedies.
Contributor
Summary of ChangesHello @Hzfengsy, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request updates the CANN Open Software License Agreement from Version 1.0 to Version 2.0. The primary goal is to broaden the scope of compatible hardware to include a wider range of Huawei AI Processors, beyond just Ascend, and to refine the legal terms regarding usage, distribution, and termination. This ensures the license accurately reflects the current product ecosystem and strengthens Huawei's intellectual property protections. Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request updates the CANN Open Software License from version 1.0 to 2.0. The changes expand the scope of compatible processors to include more Huawei AI Processors, clarify usage and distribution terms, and strengthen termination clauses. My review identifies a couple of minor opportunities to improve the clarity and formatting of the license text. Specifically, I've suggested a correction for inconsistent formatting in a list of processor brands and a grammatical improvement for better readability.
| 1.1 Software means the APIs, source code files, binaries, and related documents of Compute Architecture for Neural Networks("CANN") that are licensable by Huawei, and provided and licensed under this Agreement. | ||
|
|
||
| 1.2 Ascend processors means the chipsets branded with "Ascend" that are manufactured and supplied by Huawei. | ||
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. |
Contributor
There was a problem hiding this comment.
There appears to be a minor formatting inconsistency in the list of brands. The part "Kirin"," Yueying" has inconsistent spacing. For clarity and consistency, I suggest changing it to "Kirin", "Yueying".
1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin", "Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei.
|
|
||
| 3. Restrictions | ||
| 3.1 You are not authorized to, and shall not use, modify, or distribute this Software or its derivative works for any purpose except those expressly permitted by this Agreement. You shall not make any use of the Software or its derivative works to develop or distribute software for use in systems with processors other than Ascend processors. | ||
| 3.1 You are not authorized to, and shall not use, modify, or distribute this Software or its derivative works for any other purposes than those expressly permitted by this Agreement. You shall not make any use of the Software or its derivative works to develop or distribute any software for use in systems with processors other than Huawei AI Processors. All rights not expressly granted herein are expressly reserved by Huawei. |
Contributor
There was a problem hiding this comment.
The phrasing "for any other purposes than" is grammatically a bit awkward. For improved readability, consider rephrasing it to "for any purposes other than".
3.1 You are not authorized to, and shall not use, modify, or distribute this Software or its derivative works for any purposes other than those expressly permitted by this Agreement. You shall not make any use of the Software or its derivative works to develop or distribute any software for use in systems with processors other than Huawei AI Processors. All rights not expressly granted herein are expressly reserved by Huawei.
Contributor
There was a problem hiding this comment.
Pull request overview
This PR updates the CANN Open Software License from Version 1.0 to Version 2.0. The primary changes expand the license scope from Ascend processors to broader Huawei AI Processors (including Ascend, Kirin, and Yueying brands), refine distribution restrictions, and strengthen termination provisions with enhanced legal remedies.
Key Changes
- Expanded processor definition from "Ascend processors" to "Huawei AI Processors" including multiple Huawei-branded chipsets
- Added numbering to Section 2 (2.1) and clarified grant language to include systems with Software
- Enhanced termination provisions with "void ab initio" clause and survivability specifications
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| 1.1 Software means the APIs, source code files, binaries, and related documents of Compute Architecture for Neural Networks("CANN") that are licensable by Huawei, and provided and licensed under this Agreement. | ||
|
|
||
| 1.2 Ascend processors means the chipsets branded with "Ascend" that are manufactured and supplied by Huawei. | ||
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. |
There was a problem hiding this comment.
There is an inconsistent spacing issue with the quotation marks around brand names. The term "Kirin"," has a comma inside the closing quote followed by another quote, when it should be "Kirin", with the comma outside. This creates incorrect punctuation.
Suggested change
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. | |
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin", "Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. |
| 1.1 Software means the APIs, source code files, binaries, and related documents of Compute Architecture for Neural Networks("CANN") that are licensable by Huawei, and provided and licensed under this Agreement. | ||
|
|
||
| 1.2 Ascend processors means the chipsets branded with "Ascend" that are manufactured and supplied by Huawei. | ||
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. |
There was a problem hiding this comment.
There's a missing space between "Kirin"," and "Yueying". After fixing the quote issue, there should be a space before "Yueying" to properly separate the listed brand names.
Suggested change
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. | |
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin", "Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. |
| 1.1 Software means the APIs, source code files, binaries, and related documents of Compute Architecture for Neural Networks("CANN") that are licensable by Huawei, and provided and licensed under this Agreement. | ||
|
|
||
| 1.2 Ascend processors means the chipsets branded with "Ascend" that are manufactured and supplied by Huawei. | ||
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. |
There was a problem hiding this comment.
The phrase "including have manufactured" appears grammatically awkward. The correct phrasing should be "including having manufactured" to maintain parallel structure with the gerund form.
Suggested change
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including have manufactured), supplied (including have supplied) or designed (including have designed) by Huawei. | |
| 1.2 Huawei AI Processors mean AI chipsets (i) branded with "Ascend", "Kirin"," Yueying" or other brands owned or controlled by Huawei; or (ii) manufactured (including having manufactured), supplied (including having supplied) or designed (including having designed) by Huawei. |
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Feb 22, 2026
- Fix conftest docstring to say BEFORE_AND_AFTER (#1, hw-native-sys#5, hw-native-sys#8) - Align InitMemRef pass property table and docs with code (hw-native-sys#2) - Add INTERNAL_CHECK for null pass result in Pass::operator() (hw-native-sys#3) - Switch SSA assignment tracking to pointer identity (hw-native-sys#4) - Add null instrument checks in RunBeforePass/RunAfterPass (hw-native-sys#6) - Add strict=True to xfail with descriptive reason (hw-native-sys#7) - Fix pass_manager.md verification mode to BEFORE_AND_AFTER (hw-native-sys#9) - Add ExitContext stack invariant check (hw-native-sys#10)
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Feb 22, 2026
- Use memory_order_relaxed for atomic ID counter (comment hw-native-sys#3) - Fix UniqueId doc: "stable for the lifetime of the process" (comment hw-native-sys#2) - Clarify ir.pyi docstring: variable identity is part of hash (comment #1) - Fix Type overload docstring: note enable_auto_mapping scope (comment hw-native-sys#4) - Remove redundant static_cast in hash_var_identity (comment hw-native-sys#5) - Optimize triple As<T> to single GetKind() check (comment hw-native-sys#6)
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Feb 26, 2026
- Add tuple arity validation in Function binding (comments hw-native-sys#2/hw-native-sys#4) - Add strict validation in param_directions deserialization (comments hw-native-sys#3/hw-native-sys#5)
syfeng-bot
pushed a commit
that referenced
this pull request
Mar 1, 2026
- Fix closure variable merging order to follow lexical scoping (comment #1) - Fix error.span isinstance check - span is ir.Span not dict (comment #6) - Use forward-compatible AST field copying with ast.copy_location (comment #3) - Add return handling in inline mode with early termination after return (comment #4) - Prevent leaking inline function variables to caller scope (comments #2, #5, #7)
4 tasks
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Mar 2, 2026
- Reorder __all__ to match import order (comment #1) - Document all 5 system ops in en/zh-cn syntax docs (comment hw-native-sys#3) - Fix frame_offset=2 in helper functions for accurate span capture (comment hw-native-sys#4) - Exclude 'system' from unified dispatch to prevent misleading errors (comment hw-native-sys#5)
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Mar 5, 2026
- Add dtype guard to is_const_int for round-trip fidelity (hw-native-sys#3) - Document 1-3 arg concise forms in syntax blocks (hw-native-sys#5-7) - Note concise forms are sugar; IR stores start/stop/step (#1-2) - Fix range(0,10,1) comment to pl.range(10) (hw-native-sys#8-9) - Fix inaccurate test comment about Var start printing (hw-native-sys#10)
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Mar 12, 2026
- Derive repo owner/name dynamically via gh repo view (hw-native-sys#2) - Use comments(last: 1) to fetch latest comment per thread (hw-native-sys#4) - Add pagination with pageInfo/endCursor for >100 threads (hw-native-sys#6) - Extract run ID from check link field for gh run view (hw-native-sys#5, hw-native-sys#7) - Document external check fallback (open link URL directly)
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Mar 13, 2026
- Expanded GraphQL query for readability (gemini hw-native-sys#2) - Added gh pr checks command to define $LINK variable (gemini #1, copilot hw-native-sys#7) - Replaced grep -oP with portable sed for run ID extraction (gemini #1, copilot hw-native-sys#7) - Fixed grep pattern to handle whitespace in JSON (copilot hw-native-sys#6) - Replaced non-standard BRE with grep -E (copilot hw-native-sys#5) - Added local reproduction fallback for CI logs (gemini hw-native-sys#3) - Removed ellipsis placeholder, save output inline (copilot hw-native-sys#4)
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
Mar 13, 2026
…e-sys#501 - Fix bugprone-unchecked-optional-access in op_registry.h fluent methods by extracting local ref after EnsureMemorySpec() with NOLINT - Add missing #include <cstddef> in op_registry.h - Add missing #include "pypto/ir/memory_space.h" in 7 tile op files - Fix bugprone-parent-virtual-call in infer_tile_memory_space_pass.cpp: use IRMutator::VisitExpr/VisitStmt instead of ExprFunctor/StmtFunctor - Remove unused #include "pypto/core/any_cast.h" from init_memref.cpp - Return empty list instead of None for unconstrained input constraints in get_op_memory_spec binding (review comment hw-native-sys#5)
luohuan19
added a commit
to luohuan19/pypto
that referenced
this pull request
Mar 29, 2026
- Replace global _simpler_stamp_value with @functools.lru_cache(maxsize=1) - Replace bare bool globals with list[bool] containers to avoid PLW0603 - Move all inline imports (ctypes, subprocess, torch, importlib.util, os, sys, logging) to top-level to fix PLC0415 violations - Use importlib.import_module() for optional Simpler deps (code_runner, kernel_compiler, runtime_builder) to fix PLC0415 and pyright errors - Write cached binaries atomically via temp file + os.replace() (hw-native-sys#5) - Add _init_simpler_root_if_needed() to fix SIMPLER_ROOT timing in pytest_collection_finish (runs before session fixtures) (hw-native-sys#7) - Log pre-build task failures instead of silently swallowing them (hw-native-sys#2) - Extract _collect_test_case_from_item() helper to fix PLR0912 in pytest_collection_finish - Fix E501 long print lines in conftest.py - Move module-level logger _log to module scope in test_runner.py
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
May 6, 2026
- inline pass: extend DefVarCollector + VarSubstituteMutator to cover WhileStmt return_vars and ForStmt return_vars (Copilot hw-native-sys#5). - inline pass: reject inline bodies with non-trailing ReturnStmts via NestedReturnCounter — silent miscompile if hand-built IR ever does early-return inside an If/For branch (Gemini #1). - inline pass: drop the stale "Inline function as program entry: detected before splicing; raises" doc-comment line; the actual behaviour is silent removal in the cleanup phase (Copilot hw-native-sys#6). - verifier: also flag *dangling* Calls — Calls whose callee GlobalVar isn't in program->functions_ — with error_code 2. The previous code short-circuited when no Inline functions survived, missing the case where the InlineFunctions pass dropped the function but left a Call behind (Copilot hw-native-sys#4). - JIT _resolve_int: handle ast.Pow (only with non-negative int exponents) and ast.UAdd (Gemini hw-native-sys#2). - docs/zh-cn/dev/passes/22-lower_pipeline_loops.md: narrow CanonicalizeIOOrder scope wording from "全程序每一个 SeqStmts" to "ForKind::Pipeline 作用域内的 SeqStmts" (CodeRabbit hw-native-sys#9). - examples/models/qwen3_jit/kernels/rmsnorm.py: cast resid chunks to FP32 in post_rmsnorm to mirror input_rmsnorm and avoid BF16 accumulation precision loss (CodeRabbit hw-native-sys#11).
Hzfengsy
added a commit
to Hzfengsy/pypto
that referenced
this pull request
May 6, 2026
- inline pass: extend DefVarCollector + VarSubstituteMutator to cover WhileStmt return_vars and ForStmt return_vars (Copilot hw-native-sys#5). - inline pass: reject inline bodies with non-trailing ReturnStmts via NestedReturnCounter — silent miscompile if hand-built IR ever does early-return inside an If/For branch (Gemini #1). - inline pass: drop the stale "Inline function as program entry: detected before splicing; raises" doc-comment line; the actual behaviour is silent removal in the cleanup phase (Copilot hw-native-sys#6). - verifier: also flag *dangling* Calls — Calls whose callee GlobalVar isn't in program->functions_ — with error_code 2. The previous code short-circuited when no Inline functions survived, missing the case where the InlineFunctions pass dropped the function but left a Call behind (Copilot hw-native-sys#4). - JIT _resolve_int: handle ast.Pow (only with non-negative int exponents) and ast.UAdd (Gemini hw-native-sys#2). - docs/zh-cn/dev/passes/22-lower_pipeline_loops.md: narrow CanonicalizeIOOrder scope wording from "全程序每一个 SeqStmts" to "ForKind::Pipeline 作用域内的 SeqStmts" (CodeRabbit hw-native-sys#9). - examples/models/qwen3_jit/kernels/rmsnorm.py: cast resid chunks to FP32 in post_rmsnorm to mirror input_rmsnorm and avoid BF16 accumulation precision loss (CodeRabbit hw-native-sys#11).
lyfne123
added a commit
to lyfne123/pypto
that referenced
this pull request
May 11, 2026
CI fix (root cause of pypto-lib-model + system-tests failures): - P6's call-site injector previously inlined `tensor.as_layout(arg, DN)` into the kernel-call args directly, which the orchestration codegen rejects with `Call to '<callee>' arg N is neither a variable nor a recognized constant literal`. Refactor `CallSiteAsLayoutInjector` to operate at the statement level: for every AssignStmt / EvalStmt / ReturnStmt whose RHS targets a promoted callee, emit one `bridged_<param> = tensor.as_layout(arg, DN)` AssignStmt immediately before the call statement and replace the inline Call arg with the bound Var. Net IR is SSA-form and matches what orchestration codegen consumes per arg slot (Var | const-literal). Review comments addressed: - gemini #1: codegen `tensor.as_layout` now special-cases the identity flip (target layout == source layout) and emits a plain `Tensor result = input;` alias instead of a spurious `.transpose()`. Simplify still folds these before codegen in the default pipeline, but the codegen is now robust against ad-hoc compile paths that skip Simplify. - coderabbitai hw-native-sys#2 / hw-native-sys#3: drop the "next default pass" wording in en/zh-cn doc 17 — `MaterializeTensorStrides` runs later in the pipeline (after `CanonicalizeIOOrder`), not immediately after. The zh-cn doc's "17th pass" text is also clarified — the 17 is the docs/passes/ slot, not a literal pipeline call-count. - coderabbitai hw-native-sys#4: `DeduceTensorAsLayoutType` now preserves the source TensorView's `valid_shape` (with trailing-pair swap on cross-layout flips) and `pad` through `tensor.as_layout`. Previously these fields were dropped, making the reinterpret silently lossy for sliced or fill-padded inputs. - coderabbitai hw-native-sys#6: `MaterializeTensorStrides` direct-ctor rebuild path now forwards `op->attrs_`. The previous version preserved type and kwargs but dropped attrs, which would have silently discarded call metadata (arg_directions, manual_dep_edges) attached by earlier passes. - coderabbitai hw-native-sys#7: update the stale comment block above `VisitExpr_` in `simplify_pass.cpp` — it still described the dropped shape-bearing `as_layout(x, shape, layout)` form and the never-implemented chain folding. New comment accurately describes the single identity-elimination rule and explains why chain folding is deferred. - coderabbitai hw-native-sys#8 / hw-native-sys#9: the unit-test pattern that previously inspected an inline `tensor.as_layout` Call as a kernel-call arg no longer applies after the SSA refactor above. Tests now look up the bridge via `_find_assign_rhs(orch, var)` and guard `op is not None` before reading `op.name` (matching the defensive pattern already used in the B^T test). Skipped (with reason): - coderabbitai hw-native-sys#5: "Handle 3-arg `tile.load`". `tile.load` registers four mandatory args (tensor, offsets, shapes, valid_shapes) and the Python builder always materializes `valid_shapes` (defaults to `shapes` when the caller omits it). Once IR is constructed, every `tile.load` is 4-arg — the 3-arg form only exists at the DSL surface. The internal check stays as-is.
lyfne123
added a commit
to lyfne123/pypto
that referenced
this pull request
May 11, 2026
CI fix (root cause of pypto-lib-model + system-tests failures): - P6's call-site injector previously inlined `tensor.as_layout(arg, DN)` into the kernel-call args directly, which the orchestration codegen rejects with `Call to '<callee>' arg N is neither a variable nor a recognized constant literal`. Refactor `CallSiteAsLayoutInjector` to operate at the statement level: for every AssignStmt / EvalStmt / ReturnStmt whose RHS targets a promoted callee, emit one `bridged_<param> = tensor.as_layout(arg, DN)` AssignStmt immediately before the call statement and replace the inline Call arg with the bound Var. Net IR is SSA-form and matches what orchestration codegen consumes per arg slot (Var | const-literal). Review comments addressed: - gemini #1: codegen `tensor.as_layout` now special-cases the identity flip (target layout == source layout) and emits a plain `Tensor result = input;` alias instead of a spurious `.transpose()`. Simplify still folds these before codegen in the default pipeline, but the codegen is now robust against ad-hoc compile paths that skip Simplify. - coderabbitai hw-native-sys#2 / hw-native-sys#3: drop the "next default pass" wording in en/zh-cn doc 17 — `MaterializeTensorStrides` runs later in the pipeline (after `CanonicalizeIOOrder`), not immediately after. The zh-cn doc's "17th pass" text is also clarified — the 17 is the docs/passes/ slot, not a literal pipeline call-count. - coderabbitai hw-native-sys#4: `DeduceTensorAsLayoutType` now preserves the source TensorView's `valid_shape` (with trailing-pair swap on cross-layout flips) and `pad` through `tensor.as_layout`. Previously these fields were dropped, making the reinterpret silently lossy for sliced or fill-padded inputs. - coderabbitai hw-native-sys#6: `MaterializeTensorStrides` direct-ctor rebuild path now forwards `op->attrs_`. The previous version preserved type and kwargs but dropped attrs, which would have silently discarded call metadata (arg_directions, manual_dep_edges) attached by earlier passes. - coderabbitai hw-native-sys#7: update the stale comment block above `VisitExpr_` in `simplify_pass.cpp` — it still described the dropped shape-bearing `as_layout(x, shape, layout)` form and the never-implemented chain folding. New comment accurately describes the single identity-elimination rule and explains why chain folding is deferred. - coderabbitai hw-native-sys#8 / hw-native-sys#9: the unit-test pattern that previously inspected an inline `tensor.as_layout` Call as a kernel-call arg no longer applies after the SSA refactor above. Tests now look up the bridge via `_find_assign_rhs(orch, var)` and guard `op is not None` before reading `op.name` (matching the defensive pattern already used in the B^T test). Skipped (with reason): - coderabbitai hw-native-sys#5: "Handle 3-arg `tile.load`". `tile.load` registers four mandatory args (tensor, offsets, shapes, valid_shapes) and the Python builder always materializes `valid_shapes` (defaults to `shapes` when the caller omits it). Once IR is constructed, every `tile.load` is 4-arg — the 3-arg form only exists at the DSL surface. The internal check stays as-is.
lyfne123
added a commit
to lyfne123/pypto
that referenced
this pull request
May 12, 2026
CI fix (root cause of pypto-lib-model + system-tests failures): - P6's call-site injector previously inlined `tensor.as_layout(arg, DN)` into the kernel-call args directly, which the orchestration codegen rejects with `Call to '<callee>' arg N is neither a variable nor a recognized constant literal`. Refactor `CallSiteAsLayoutInjector` to operate at the statement level: for every AssignStmt / EvalStmt / ReturnStmt whose RHS targets a promoted callee, emit one `bridged_<param> = tensor.as_layout(arg, DN)` AssignStmt immediately before the call statement and replace the inline Call arg with the bound Var. Net IR is SSA-form and matches what orchestration codegen consumes per arg slot (Var | const-literal). Review comments addressed: - gemini #1: codegen `tensor.as_layout` now special-cases the identity flip (target layout == source layout) and emits a plain `Tensor result = input;` alias instead of a spurious `.transpose()`. Simplify still folds these before codegen in the default pipeline, but the codegen is now robust against ad-hoc compile paths that skip Simplify. - coderabbitai hw-native-sys#2 / hw-native-sys#3: drop the "next default pass" wording in en/zh-cn doc 17 — `MaterializeTensorStrides` runs later in the pipeline (after `CanonicalizeIOOrder`), not immediately after. The zh-cn doc's "17th pass" text is also clarified — the 17 is the docs/passes/ slot, not a literal pipeline call-count. - coderabbitai hw-native-sys#4: `DeduceTensorAsLayoutType` now preserves the source TensorView's `valid_shape` (with trailing-pair swap on cross-layout flips) and `pad` through `tensor.as_layout`. Previously these fields were dropped, making the reinterpret silently lossy for sliced or fill-padded inputs. - coderabbitai hw-native-sys#6: `MaterializeTensorStrides` direct-ctor rebuild path now forwards `op->attrs_`. The previous version preserved type and kwargs but dropped attrs, which would have silently discarded call metadata (arg_directions, manual_dep_edges) attached by earlier passes. - coderabbitai hw-native-sys#7: update the stale comment block above `VisitExpr_` in `simplify_pass.cpp` — it still described the dropped shape-bearing `as_layout(x, shape, layout)` form and the never-implemented chain folding. New comment accurately describes the single identity-elimination rule and explains why chain folding is deferred. - coderabbitai hw-native-sys#8 / hw-native-sys#9: the unit-test pattern that previously inspected an inline `tensor.as_layout` Call as a kernel-call arg no longer applies after the SSA refactor above. Tests now look up the bridge via `_find_assign_rhs(orch, var)` and guard `op is not None` before reading `op.name` (matching the defensive pattern already used in the B^T test). Skipped (with reason): - coderabbitai hw-native-sys#5: "Handle 3-arg `tile.load`". `tile.load` registers four mandatory args (tensor, offsets, shapes, valid_shapes) and the Python builder always materializes `valid_shapes` (defaults to `shapes` when the caller omits it). Once IR is constructed, every `tile.load` is 4-arg — the 3-arg form only exists at the DSL surface. The internal check stays as-is.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update the license from CANN Open Software License Agreement Version 1.0 to Version 2.0. Key changes include expanding processor compatibility from Ascend processors to Huawei AI Processors (including Ascend, Kirin, and Yueying branded chipsets), clarifying usage restrictions and distribution terms, and strengthening the termination provisions with enhanced legal remedies.