| Files | Last commit | Last update |
|---|---|---|
fix(cumsum): AddFactor循环边界显式转换为uint16_t,避免整型提升隐式转换 Co-authored-by: gcw_YBIAEfqJ<wangweidong15@huawei.com> # message auto-generated for no-merge-commit merge: !7108 merge fix/cumsum-sklansky-uint16-cast into master fix(cumsum): AddFactor循环边界显式转换为uint16_t,避免整型提升隐式转换 Created-by: gcw_YBIAEfqJ Commit-by: gcw_YBIAEfqJ Merged-by: cann-robot Description: ## 修改内容 op_kernel/arch35/cumsum_base/cumsum_core_sklansky.h 的 DoAddTailN 中,AddFactor 主循环边界 addFactorCircleLoop - 1 显式转换为 uint16_t: diff - for (uint16_t i = 0; i < addFactorCircleLoop - 1; i++) { + for (uint16_t i = 0; i < static_cast<uint16_t>(addFactorCircleLoop - 1); i++) { ## 原因 addFactorCircleLoop 为 uint16_t,- 1 运算经整型提升为 int,与 uint16_t 循环变量比较时存在隐式转换;显式 cast 消除隐式转换,意图更清晰。 ## 关联的Issue - [Bug-Report|缺陷反馈] cumsum Sklansky 尾块 AddFactor 循环边界存在整型提升隐式转换 #4459 https://gitcode.com/cann/ops-math/issues/4459 ## 影响范围 仅浮点 Sklansky 路径(tilingKey 1001/1002/1011/1012/1101/1102/1111/1112)的尾块 AddFactor 累加循环,纯类型修饰,无逻辑变化。 See merge request: cann/ops-math!7108 | 11 days ago | |
fix(cumsum): 修复整数空Tensor tiling除零 Co-authored-by: gcw_YBIAEfqJ<wangweidong15@huawei.com> # message auto-generated for no-merge-commit merge: !7177 merge fix/cumsum-int-empty-tensor into master fix(cumsum): 修复整数空Tensor tiling除零 Created-by: gcw_YBIAEfqJ Commit-by: gcw_YBIAEfqJ Merged-by: cann-robot Description: ## 问题描述 Cumsum 整数 tiling 未处理空 Tensor。当任一维度为 0 时,根据 0 维位于 axis 的前/中/后位置, leftAxisLen、midAxisLen 或 rightAxisLen 会变为 0,后续进入如下路径: - rightAxisLen=0:vlSize_ / (raLpUnit_ * dtypeSize) 触发除 0(SIGFPE); - midAxisLen=0:rLpUnit_=0,后续产生非法分核; - leftAxisLen=0:laLpCnt=0,CeilDiv(0, CeilDiv(0, coreNum)) 产生非法 tiling。 ACLNN 通路已有 self->IsEmpty() 早返回,但 TTK kernel/GE direct 会直接进入整数 tiling,因此仍需算子侧兜底。 ## 修改内容 1. 新增整数 empty tilingKey 11003; 2. GetInputDims() 后检测 leftAxisLen/midAxisLen/rightAxisLen == 0,下发 blockDim=1/key=11003/workspace=0; 3. kernel empty 分支通过 GET_TILING_DATA_WITH_STRUCT 注册 tiling struct size 后立即返回,不访问 GM; 4. 初始化整数 tiling 成员,保证 empty 分支序列化数据确定; 5. 增加 int32/int64/uint8 × axis 前/中/后三类空 Tensor UT。 ## 关联的Issue - [Bug-Report|缺陷反馈] Cumsum 整数 tiling 未处理空 Tensor——rightAxisLen=0 除零(SIGFPE)、midAxisLen=0 非法分核、leftAxisLen=0 非法 tiling #4634 https://gitcode.com/cann/ops-math/issues/4634 ## 验证结果 - Cumsum tiling UT:3/3 PASS; - empty UT:int32/int64/uint8 × axis 0/3/5,均 key=11003、blockDim=1、workspace=0; - Ascend950 kernel 二进制编译成功,opc 接受 11003; - TTK direct:int32/int64/uint8 空 Tensor 均 perf_status=PASS; - 非空回归 [105,2239]:axis0 key=11000/blockDim=35、axis1 key=11001/blockDim=53,precision 2/2 PASS; - git diff --check 通过。 See merge request: cann/ops-math!7177 | 9 days ago |
| Files | Last commit | Last update |
|---|---|---|
| 11 days ago | ||
| 9 days ago |