AscendNPU-IR/bishengir/lib · Ascend/AscendNPU-IR

Llrb[Huawei][AscendNPU-IR] fix: not add free_lock_var because version check

文件	最后提交记录	最后更新时间
Analysis	[SIMT] Adds Triton related analysis utils	24 天前
Bindings	[AscendNPU IR] Add HFusion Co-authored-by: chenxiangting<chenxiangting@huawei.com>	10 个月前
CAPI	fix: add triton generated header deps	24 天前
Conversion	!2568 merge fix/fp-bitwise-vv-broadcast into master [AscendNPU-IR][VV] Support bitwise VV broadcast fusion Created-by: EricHaHaHaHa Commit-by: Eric Merged-by: ascend-robot Description: title: "[AscendNPU IR]: VF融合增强 --- 支持vbrc VF融合" 描述 Description 原始问题可以概括为：上游 HIVM 到 Upstream / HFusion 的 lowering 中，有些 op 会过早或不一致地落到 linalg.map，导致后续 pass 难以统一处理。典型问题包括： linalg.map 对 HFusion named op 生态不友好，hfusion-inline-brc 无法直接把 fill/broadcast mask 内联成 scalar operand。 FP bitwise 需要表达 bit-pattern 语义，不能当作普通 FP 数值运算处理。 broadcast / transpose 需要准确表达 indexing map，简单 named op 不足以表达全部 shape transform。 convert-to-named-op=true 和 convert-to-named-op=false 两种模式原来的 CHECK 和 lowering 预期不够清晰。 integer、unsigned integer、FP 的 bitwise 语义边界需要明确，否则容易把 unsigned 类型错误改写成 signless 路径。高精度除法、mod、shift 等相邻 op 也需要从 map-based 或 arith 直降路径调整到更适合 HFusion 的形态。因此本次修改的本质是：减少不必要的 linalg.map -> 保留更多结构化语义 -> 按类型和 shape transform 选择正确 lowering 形态 -> 让 InlineBrc / Generalize / AutoVectorize / VF outline 能继续工作 ## 类型 Category - [ ] Bug修复 Bugfix - [ ] 新功能 New feature - [ ] 代码风格更新（格式化，局部变量）Style changes - [ ] 重构（即不是新增功能，也不是修改bug的代码变动）Refactor - [ ] 构建过程或辅助工具的变动 Build-related - [ ] 文档内容更新 Docs ## Checklist - [ ] 我的代码遵循这个项目的代码风格 My code follows the project's code style - [ ] 我已经自己测试过我的代码 I have self-tested my code - [ ] 我已经更新了相应的文档 I have updated the relevant documentation - [ ] 我已经根据需要更新了对应的变更日志 I have updated the changelog as needed - [ ] 我已经在标题中正确使用了类型标签（例如：`feat:`, `fix:`）I have correctly used type labels in the title (e.g., feat:, fix:) See merge request: Ascend/AscendNPU-IR!2568	1 天前
Dialect	!2583 merge flatten-perf into master [AscendNPU-IR] Reapply "enable flatten by default" (!2326) & fix performance degradation caused by this patch Created-by: kiansitov Commit-by: kiansitov;Roman Rusyaev Merged-by: ascend-robot Description: --- title: "[AscendNPU IR]: Reapply "enable flatten by default" (!2326) and fix performance degradation caused this patch" --- ## 描述 Description 1. Revert "Revert "!2326 merge migration into master"". Reapply previously reverted commit "[AscendNPU-IR] enable flatten by default" (af7a256c8d3e7e3cdbe4153b18a9fc747a07b2a8) 2. Fix performance degradation. `hfusion\FlattenOp` pass is applied only to MIX kernels, while the pre-flattening `PropagateReshape` pass is also applied to AIV kernels. `PropagateReshape` and `hfusion\FlattenOp` should be applied consistently: `PropagateReshape` should be applied only when `hfusion\FlattenOp` is also applicable. Applying `PropagateReshape` without `hfusion\FlattenOp` leads to performance degradation because `PropagateReshape` pass can affect operation fusion. Skip `PropagateReshape` pass for AIV kernels (when flattening is not applicable) ## 类型 Category - [ ] Bug修复 Bugfix - [ ] 新功能 New feature - [ ] 代码风格更新（格式化，局部变量）Style changes - [ ] 重构（即不是新增功能，也不是修改bug的代码变动）Refactor - [ ] 构建过程或辅助工具的变动 Build-related - [ ] 文档内容更新 Docs ## Checklist - [ ] 我的代码遵循这个项目的代码风格 My code follows the project's code style - [ ] 我已经自己测试过我的代码 I have self-tested my code - [ ] 我已经更新了相应的文档 I have updated the relevant documentation - [ ] 我已经根据需要更新了对应的变更日志 I have updated the changelog as needed - [ ] 我已经在标题中正确使用了类型标签（例如：`feat:`, `fix:`）I have correctly used type labels in the title (e.g., feat:, fix:) See merge request: Ascend/AscendNPU-IR!2583	1 小时前
ExecutionEngine	[AscendNPU-IR] Support bitwise VV broadcast fusion	2 天前
Pass	[AscendNPU IR] feature-inject-sync-sp-1114-squashed Co-authored-by: h00930278<haj.ahmad.ammar@huawei.com> # message auto-generated for no-merge-commit merge: !142 merge feature-inject-sync-sp-1114-squashed into master [AscendNPU IR] feature-inject-sync-sp-1114-squashed Created-by: AmmarDab3an Commit-by: h00930278 Merged-by: ascend-robot Description: ## 描述 Description 请提供此 Pull Request 的高级别说明。 Please describe what this PR is about. ## 类型 Category - [ ] Bug修复 Bugfix - [ ] 新功能 New feature - [ ] 代码风格更新（格式化，局部变量）Style changes - [ ] 重构（即不是新增功能，也不是修改bug的代码变动）Refactor - [ ] 构建过程或辅助工具的变动 Build-related - [ ] 文档内容更新 Docs ## Checklist - [ ] 我的代码遵循这个项目的代码风格 My code follows the project's code style - [ ] 我已经自己测试过我的代码 I have self-tested my code - [ ] 我已经更新了相应的文档 I have updated the relevant documentation - [ ] 我已经根据需要更新了对应的变更日志 I have updated the changelog as needed - [ ] 我已经在标题中正确使用了类型标签（例如：`feat:`, `fix:`）I have correctly used type labels in the title (e.g., feat:, fix:) See merge request: Ascend/AscendNPU-IR!142	8 个月前
Template	fix apply padding b64	2 天前
Tools	!2583 merge flatten-perf into master [AscendNPU-IR] Reapply "enable flatten by default" (!2326) & fix performance degradation caused by this patch Created-by: kiansitov Commit-by: kiansitov;Roman Rusyaev Merged-by: ascend-robot Description: --- title: "[AscendNPU IR]: Reapply "enable flatten by default" (!2326) and fix performance degradation caused this patch" --- ## 描述 Description 1. Revert "Revert "!2326 merge migration into master"". Reapply previously reverted commit "[AscendNPU-IR] enable flatten by default" (af7a256c8d3e7e3cdbe4153b18a9fc747a07b2a8) 2. Fix performance degradation. `hfusion\FlattenOp` pass is applied only to MIX kernels, while the pre-flattening `PropagateReshape` pass is also applied to AIV kernels. `PropagateReshape` and `hfusion\FlattenOp` should be applied consistently: `PropagateReshape` should be applied only when `hfusion\FlattenOp` is also applicable. Applying `PropagateReshape` without `hfusion\FlattenOp` leads to performance degradation because `PropagateReshape` pass can affect operation fusion. Skip `PropagateReshape` pass for AIV kernels (when flattening is not applicable) ## 类型 Category - [ ] Bug修复 Bugfix - [ ] 新功能 New feature - [ ] 代码风格更新（格式化，局部变量）Style changes - [ ] 重构（即不是新增功能，也不是修改bug的代码变动）Refactor - [ ] 构建过程或辅助工具的变动 Build-related - [ ] 文档内容更新 Docs ## Checklist - [ ] 我的代码遵循这个项目的代码风格 My code follows the project's code style - [ ] 我已经自己测试过我的代码 I have self-tested my code - [ ] 我已经更新了相应的文档 I have updated the relevant documentation - [ ] 我已经根据需要更新了对应的变更日志 I have updated the changelog as needed - [ ] 我已经在标题中正确使用了类型标签（例如：`feat:`, `fix:`）I have correctly used type labels in the title (e.g., feat:, fix:) See merge request: Ascend/AscendNPU-IR!2583	1 小时前
Transforms	[Huawei][AscendNPU IR] Debugging Capabilities: UB Probe: human readable	12 天前
CMakeLists.txt	[SIMT] Adds Triton related analysis utils	24 天前