RELEASE.md · chengbin/akg

# MindAKG Release Notes

[查看中文](./RELEASE.md)

## MindAKG 1.7.0 Release Notes

### Major Features and Improvements

* Support a new loop intrinsic for MindSpore HybridDSL. [!560](https://gitee.com/mindspore/akg/pulls/560) (ALL backends)
* Update the shared/local promotion strategy on GPU backend. [!556](https://gitee.com/mindspore/akg/pulls/556) (GPU)
* Use the new interface provided by isl to reconstruct the reschedule pass after the pluto scheduling algorithm on the Ascend backend.[!512](https://gitee.com/mindspore/akg/pulls/512) (Ascend)

### Bug fixes

* Fix the bug of Gemm/convolution operator due to repeated tiling. [!582](https://gitee.com/mindspore/akg/pulls/582) (GPU)
* Fixed the bug in the reduce operator when the Y direction is reduced and X is small. [!559](https://gitee.com/mindspore/akg/pulls/559) (CPU)

### Contributors

Thanks goes to these wonderful people:

yangsijia, polyhedral, zhangrenwei, yiyanzhi, hujiahui8, zhengzuohe, zhangzhaochuang, xuhui, liuchao, xiaruijie, DeshiChen, nelson.lossing, zhenzhang,

chenlei_autodiff, lingyunli63, wYann, hanhuifeng, gaoxiong, zichun_ye, chengbin, wangrao124, huangmengxi, harenome, lear, huoxinyou, yanglinfeng, Etienne, Trump

## Release 1.6.0
### Major Features and Improvements

* [STABLE] CPU backend support: AKG now supports automatic generation of high-performance code for CPUs with different architectures. [!413](https://gitee.com/mindspore/akg/pulls/413) (CPU)
* [STABLE] CSR operators support: Develops csr operators(csrmv/csr mul/csr reduce_sum)，and provides optimizing strategy to handle csr operators with dynamic upper bound. [!407](https://gitee.com/mindspore/akg/pulls/407) (GPU)
* [STABLE] Scheduler Optimizing: Outermost band node now will be treated as multiple filter nodes if there is no dependency relationship between multiple statements. [!460](https://gitee.com/mindspore/akg/pulls/460) (GPU)
* [STABLE] Replace gmp with imath on isl.[!455](https://gitee.com/mindspore/akg/pulls/455) (ALL)
* [STABLE] Apply Autodiff in Custom Op. [!464](https://gitee.com/mindspore/akg/pulls/464) (ALL)

### Bug fixes

* fix bugs in cpu tuning process:add "need_warm_up" attrs to avoid too many warm-up on cpu profiling.[!495](https://gitee.com/mindspore/akg/pulls/495) (CPU)
* fix auto-tiling output mem-flow bug:modify the GEMM's auto-tiling strategy to match the actual target constraint.[!504](https://gitee.com/mindspore/akg/pulls/504) (ASCEND)
* fix muilt output bug in emitter: when emit multi-output, should not convert real inputs.[!506](https://gitee.com/mindspore/akg/pulls/332) (ALL)

### Contributors

Thanks goes to these wonderful people:

yangsijia, polyhedral, zhangrenwei, yiyanzhi, hujiahui8, zhengzuohe, zhangzhaochuang, xuhui, liuchao, xiaruijie, DeshiChen, nelson.lossing, zhenzhang, chenlei_autodiff, lingyunli63, wYann, hanhuifeng, gaoxiong, zichun_ye, chengbin, wangrao124, huangmengxi, harenome, lear, huoxinyou, yanglinfeng, Etienne, Trump

Contributions of any kind are welcome!

## Release 1.5.0
### Major Features and Improvements
  * [STABLE] New operators developing: Tensor of Tensor operators(Gather/GatherNd/TensorScatterAdd/UnsortedSegmentSum), which can be used to support GNN networks.[!323](https://gitee.com/mindspore/akg/pulls/323)(GPU)
  * [STABLE] Add a topi op of UserDefine Op in AKG, which can be compiled from func_source_string or op_imply_path.[!319](https://gitee.com/mindspore/akg/pulls/319)(GPU)
  * [STABLE] Lower interface Refactor: add StageLower for stage lower case.[!310](https://gitee.com/mindspore/akg/pulls/310)(GPU)
  * [STABLE] The profiling suit for new runtime process.[!306](https://gitee.com/mindspore/akg/pulls/306)(ASCEND)

### Bug fixes
  * Fixed memory promotion bug: sort the clusters before merging.[!338](https://gitee.com/mindspore/akg/pulls/338) (ASCEND)
  * Fixed irregular-reduce bug: replace shfl.down with shared memory reducetion.[!332](https://gitee.com/mindspore/akg/pulls/332) (GPU)
  * Fixed foldDimension bug: build wrong axis relation of relation.[!302](https://gitee.com/mindspore/akg/pulls/302) (GPU)

### Contributors
Thanks goes to these wonderful people:

yangsijia, xxxxxxw, polyhedral, zhangrenwei, yiyanzhi, xixixian, hujiahui8, zhengzuohe, lishanni, zhangzhaochuang, xuhui, liu
chao, gengzhen, xiaruijie,chenlei_autodiff, lingyunli63, wYann, lvwenyuan, peiwenfang, hanhuifeng, gaoxiong, chengyun
Contributions of any kind are welcome!

## Release 1.3.0
### Major Features and Improvements
  * [STABLE] Support optimizing GEMM && Conv by using polyhedral + Tensorcore, as well as providing an akg::fragment_add/sub/mul/div library for GEMM op fusions. [!156](https://gitee.com/mindspore/akg/pulls/156) (GPU)
  * [STABLE] Optimize layout related operators(Transpose && pad/unpad) by adjusting the autotiling strategy and solving bank conflict for these ops. [!152](https://gitee.com/mindspore/akg/pulls/152/) (GPU)
  * [STABLE] Add Kahan algorithm for reducetion operators. [!107](https://gitee.com/mindspore/akg/pulls/107) (GPU)
  * [STABLE] Support transdata + matmul prefusion pattern in akg. [!103](https://gitee.com/mindspore/akg/pulls/103) (ASCEND)

### Bug fixes
  * Fixed stitch_fusion bug when store is shared but load is not shared. [!109](https://gitee.com/mindspore/akg/pulls/109) (GPU)
  * Fixed reshape bug: when add reshape, should update tensor shape. [!140](https://gitee.com/mindspore/akg/pulls/140) (GPU)
  * Fixed autofuse bug: when set autofuse config, should record broadcast. [!111](https://gitee.com/mindspore/akg/pulls/111) (GPU)

### Contributors
Thanks goes to these wonderful people:

## Release 1.2.0
## Major Features and Improvements
  * [STABLE] Rebuild the AKG repository for providing a new way to support ascend backend by linking a static library contained all the ascend passes. (Ascend)
  * [STABLE] Optimize the reduction add operation in ascend backend. (Ascend)
  * [STABLE] Add support for tuning elemwise&&reduction operators. (GPU)

### Bug fixes
  * Fixed a problem that data prefetch cannot be enabled by attributes in DSL.
  * Fixed bugs of autotiling algorithms (tiling too small, cannot adapted matmul+bias, etc.) in Ascend platform.
  * Fixed local memory promotion for large thread (2980!)
  * Fixed reduce binding dimension issue on gpu platform (ff38!)

### Contributors
Thanks goes to these wonderful people:

yangsijia, xxxxxxw, polyhedral, zhangrenwei, yiyanzhi, xixixian, hujiahui8, zhengzuohe, lishanni, zhangzhaochuang, xuhui, liuchao, gengzhen, xiaruijie, 
chenlei_autodiff, lingyunli63, wYann, lvwenyuan, peiwenfang, hanhuifeng, gaoxiong, chengyun
Contributions of any kind are welcome!

## Initial Version
* Upload the initial framework
* Basic support for Ascend910 platform and gpu v100
* Integration with GraphKernel fusion of MindSpore.

chengbin/akg

简介

发行版

贡献者

近期动态

chengbin/akg .gitee-modal { width: 500px !important; }

简介

发行版

贡献者

近期动态

搜索帮助

chengbin/akg