Skip to content

Commit 3e7f03b

Browse files
zhzhcookiezhzhcookie
authored
[DOC] Update v0.2.0 release reports (#35)
1 parent 248073a commit 3e7f03b

3 files changed

Lines changed: 46 additions & 6 deletions

File tree

‎reports/v0.2.0/release_notes_v0.2.0.md‎

Lines changed: 38 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,38 @@
1+
[中文版](./release_notes_v0.2.0_cn.md)
2+
3+
## FlagTree 0.1.0 Release
4+
5+
### Highlights
6+
7+
FlagTree inherits capabilities from the previous version, continuously integrates new backends, expands support for Triton versions, and provides hardware-aware optimization capabilities. The project is currently in its early stages, aiming to be compatible with existing adaptation solutions for various AI chip backends, unify the code repository, build a code co-construction platform, and quickly implement multi-backend support in a single repository.
8+
9+
### New features
10+
11+
* Added multi-Backend Support
12+
13+
Currently supported backends include triton_shared cpu, iluvatar, xpu (klx), mthreads, __metax__, __aipu__(arm npu), __ascend__ npu & cpu, __tsingmicro__, cambricon, with __bold__ indicating newly added ones. <br>
14+
Each new backend maintains the capabilities of the previous version: cross-platform compilation and rapid verification, plugin-based high-differentiation modules, CI/CD, and quality management capabilities. <br>
15+
Jointly developing common extensions for the middleware layer with backend vendors, and open-sourcing standardized PyTorch backend extensions to support Triton / FlagTree practices. <br>
16+
17+
* Dual Compilation Path Support
18+
19+
Supports TritonGPU and Linalg compilation paths. Provides multiple integration paradigms for non-GPGPU backends, adds FLIR repository support for Linalg Dialect extensions and MLIR extensions for backend compilation.
20+
21+
* Added support for Triton versions
22+
23+
Currently supported Triton versions include 3.0.x, 3.1.x, __3.2.x__, __3.3.x__, with __bold__ indicating newly added ones.
24+
25+
* Hardware-aware optimization support
26+
27+
Supports providing guided programming interfaces for backend-common or specific hardware features. Through compatible extensions, adding guidance information at the frontend to provide flexibility in operator writing and performance tuning.
28+
29+
* Joint construction with FlagGems operator library
30+
31+
Collaborating with the [FlagGems](https://github.com/FlagOpen/FlagGems) operator library to support related features in version adaptation, backend interfaces, registration mechanisms, and test modifications.
32+
33+
### Looking ahead
34+
35+
GPGPU backend code will be integrated, decoupling backend differentiation changes from TritonGPU; non-GPGPU backends will be horizontally integrated on the FLIR foundation, with unified design for common passes. <br>
36+
Providing Triton adaptation version upgrade guides for backend vendors: 3.0 -> 3.1 -> 3.2 -> 3.3. <br>
37+
CI/CD will add FlagGems operator library functional testing. <br>
38+
Integrating C++ Runtime functionality to reduce runtime overhead outside of kernels to be on par with CUDA. <br>

‎reports/v0.2.0/release_notes_v0.2.0_cn.md‎

Lines changed: 7 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -10,27 +10,29 @@ FlagTree 继承前一版本的能力，持续集成新的后端，拓展对 Trit
1010

1111
* 新增多后端支持
1212

13-
目前支持的后端包括 iluvatar、xpu (klx)、mthreads、__metax__、__aipu__(arm npu)、__ascend__ npu & cpu、__tsingmicro__、cambricon，其中 __加粗__ 为本次新增。 <br>
13+
目前支持的后端包括 triton_shared cpu、iluvatar、xpu (klx)、mthreads、__metax__、__aipu__(arm npu)、__ascend__ npu & cpu、__tsingmicro__、cambricon，其中 __加粗__ 为本次新增。 <br>
1414
各新增后端保持前一版本的能力：跨平台编译与快速验证、高差异度模块插件化、CI/CD、质量管理能力。 <br>
15+
与后端厂商联合研发中间层通用扩展，并开源 pytorch 标准化扩展后端以支持 Triton / FlagTree 的实践。 <br>
1516

1617
* 两种编译路径支持
1718

18-
支持 TritonGPU、Linalg 两种编译路径。对非 GPGPU 后端提供多种接入范式，新增 FLIR 仓库支持基于 Linalg Dialect 扩展的后端编译。
19+
支持 TritonGPU、Linalg 两种编译路径。对非 GPGPU 后端提供多种接入范式，新增 FLIR 仓库支持基于 Linalg Dialect 扩展、、MLIR 扩展的后端编译。
1920

2021
* 新增 Triton 版本支持
2122

2223
目前支持的 Triton 版本包括 3.0.x、3.1.x、__3.2.x__、__3.3.x__，其中 __加粗__ 为本次新增。
2324

2425
* 硬件感知优化支持
2526

26-
支持为后端通用或特有的硬件特性提供指导编程接口。通过注解等兼容式地扩展，在语言层添加指导信息以提高具体后端的性能。当前已实现对 Async DMA 的硬件感知优化支持。
27+
支持为后端通用或特有的硬件特性提供指导编程接口。通过注解等兼容式地扩展，在语言层添加指导信息，提供算子编写与性能调优的灵活性。
2728

2829
* 与 FlagGems 算子库联合建设
2930

30-
在版本适配、后端适配、推理芯片特性适配等方面，与 [FlagGems](https://github.com/FlagOpen/FlagGems) 算子库联合支持相关特性。
31+
在版本适配、后端接口、注册机制、测试修改等方面，与 [FlagGems](https://github.com/FlagOpen/FlagGems) 算子库联合支持相关特性。
3132

3233
### Looking ahead
3334

34-
GPGPU 后端代码将作整合，将后端差异化改动与 TritonGPU 解耦；非 GPGPU 后端将在 FLIR 基础上横向整合，对通用 Pass 进行统一设计。 <br>
35+
GPGPU 后端代码将进行整合，将后端差异化改动与 TritonGPU 解耦；非 GPGPU 后端将在 FLIR 基础上横向整合，对通用 Pass 进行统一设计。 <br>
3536
为后端厂商提供 Triton 适配版本升级指南：3.0 -> 3.1 -> 3.2 -> 3.3。 <br>
3637
CI/CD 将添加 FlagGems 算子库功能测试。 <br>
38+
集成 C++ Runtime 功能，将 kernel 之外的运行时开销降低到与 CUDA 持平。 <br>

‎reports/v0.2.0/report_tests.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@ FlagTree tests are validated on different backends, but currently the tests cons
88

99
| 　　　　　　　　　　　　 | default | xpu (klx) | iluvatar | mthreads | metax |
1010
|----------------------|---------------------------|-------------------------------------------|------------------------------------------------|------------------------------------------------|---------------------------------------------|
11-
| Number of unit tests | 11353 items | 12623 items | 14808 items | 10392 items | 13728 items |
11+
| Number of unit tests | 18518 items | 12623 items | 14808 items | 10392 items | 13728 items |
1212
| Script location | flagtree/python/test/unit | flagtree/third_party/xpu/python/test/unit | flagtree/third_party/iluvatar/python/test/unit | flagtree/third_party/mthreads/python/test/unit | flagtree/third_party/metax/python/test/unit |
1313
| Test command | python3 -m pytest -s | python3 -m pytest -s | python3 -m pytest -s | python3 -m pytest -s | python3 -m pytest -s |
1414
| Passing rate | 100% | 100% | 100% | 100% | 100% |

0 commit comments

Comments
 (0)