| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
VFFusion: drop tensor.insert_slice from fusion whitelist and update tests | 1 天前 | |
[Huawei][AscendNPU IR] sync ArithToAffine/AnnotationLowering/Decompose with A5 baseline | 30 天前 | |
[AscendNPU-IR] fix: replace llvm_unreachable with report_fatal_error | 19 天前 | |
[SIMT] Adds Triton and TritonExt dialects and related passes/utils | 25 天前 | |
!2134 merge fix-fold-cast-analysis into dev/regbase [Huawei][AscendNPUIR] Analyze marked fold-offset casts Created-by: zhonkebayev Commit-by: zhonkebayev Merged-by: gguo9 Description: --- title: "[AscendNPU IR]:" --- 请提供此 Pull Request 的高级别说明。 Please describe what this PR is about. - [ ] 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 - [ ] 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!2134 | 10 天前 | |
[Huawei][AscendNPU-IR] support custom AICore port | 15 天前 | |
!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 | 16 小时前 | |
!2697 merge bugfix_sync_block_lock into master [Huawei][AscendNPU-IR] fix: not add free_lock_var because version check on A5 Created-by: lrb Commit-by: lrb Merged-by: ascend-robot Description: --- title: "[AscendNPU IR]:" --- ## 描述 Description fix: not add free_lock_var because version check only check hivmc version on membase add todo: support check hivmc version on regbase ## 类型 Category - [x] 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 - [x] 我的代码遵循这个项目的代码风格 My code follows the project's code style - [x] 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 - [x] 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!2697 | 23 小时前 | |
!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 | 16 小时前 | |
[Huawei][AscendNPU-IR] fix unknown loc | 16 天前 | |
[Huawei][AscendNPU IR] add pass: (llvm) parameter-packing from A5 | 30 天前 | |
fix: unsigned cast sync solver fix | 8 个月前 | |
[AscendNPU-IR] sync MathExt Dialect definition from regbase | 1 个月前 | |
[AscendNPU IR][A5 Sync][PropagateReshape] Port A5 PropagateReshape codes | 18 天前 | |
[AscendNPU IR] Add HFusion Co-authored-by: chenxiangting<chenxiangting@huawei.com> | 10 个月前 | |
[one-shot-bufferize] fix conflict check in vf which have mulitple for loop | 4 天前 | |
!2396 merge mix2 into master [AscendNPU IR][SIMD&SIMT] fix: Support general tensor.extract and tensor.from_elements for simt scope Created-by: Maykin-PM Commit-by: Maykin-PM Merged-by: ascend-robot Description: --- [AscendNPU IR][SIMD&SIMT] fix: Support general tensor.extract and tensor.from_elements for simt scope --- ## 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 - [x] 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!2396 | 3 天前 | |
[Ascend NPUIR] fix : replace inappropriate llvm_unreachable with report_fatal_error [Ascend NPUIR] fix : replace inappropriate llvm_unreachable with report_fatal_error for Hfusion Dialect [Ascend NPUIR] fix : replace inappropriate llvm_unreachable with report_fatal_error for HIVM Dialect [Ascend NPUIR] fix : replace inappropriate llvm_unreachable with report_fatal_error | 3 个月前 | |
!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 | 16 小时前 | |
[Huawei][AscendNPU IR] adopt mlir:: qualifiers in LiteralDataTypeCast Add mlir:: to isa/dyn_cast calls. Port auto loc = op->getLoc() from A5. | 30 天前 | |
[Huawei][AscendNPU IR] [SIMT] Fix FMA dot operand linear layouts Motivation: Ensure dot-product operands use register ordinals compatible with the FMA ABI, especially for B operands whose K bits must remain in the register dimension. Design: Construct deterministic register-first basis orderings for A and B operands, then validate each register ordinal against the expected logical K and non-K coordinates. Add regression coverage for the corrected layouts. Risks: Layout ordering changes may affect generated LLVM IR and downstream shuffle behavior; existing layout and conversion tests should be run to verify compatibility. Assisted-by: Codex 5.3 | 4 天前 | |
[SIMT] Adds Triton and TritonExt dialects and related passes/utils | 25 天前 | |
[one-shot-bufferize] replace tensor.empty with alloc_tenosr in tensorCopyInsertion and add unit test | 4 天前 | |
[AscendNPU-IR][NormalizeVector] Use the inverse permutation map to compute gather indices permutation map maps the memref index space onto the result vector index space, so composing the result coordinates with it yields wrong memref indices for 3-D case, e.g. (d0, d1, d2) -> (d1, d2, d0). | 3 天前 | |
[SIMT] Adds Triton and TritonExt dialects and related passes/utils | 25 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 1 天前 | ||
| 30 天前 | ||
| 19 天前 | ||
| 25 天前 | ||
| 10 天前 | ||
| 15 天前 | ||
| 16 小时前 | ||
| 23 小时前 | ||
| 16 小时前 | ||
| 16 天前 | ||
| 30 天前 | ||
| 8 个月前 | ||
| 1 个月前 | ||
| 18 天前 | ||
| 10 个月前 | ||
| 4 天前 | ||
| 3 天前 | ||
| 3 个月前 | ||
| 16 小时前 | ||
| 30 天前 | ||
| 4 天前 | ||
| 25 天前 | ||
| 4 天前 | ||
| 3 天前 | ||
| 25 天前 |