Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.
部分用例性能劣于910B 典型失败case：x(2, 4738, 5201), scale(2, 5201), shift(2, 5201)，数据类型为float16
Ascend910B3、Ascend950PR CANN 9.0.0
用msprof op工具采集单算子性能，调用aclnnModulate执行即可。
预期Ascend950PR该算子性能不劣于Ascend910B3
采集性能数据如下：
/assign
Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.
Describe the current behavior / 问题描述 (Mandatory / 必填)
部分用例性能劣于910B
典型失败case：x(2, 4738, 5201), scale(2, 5201), shift(2, 5201)，数据类型为float16
Environment / 环境信息 (Mandatory / 必填)
Ascend910B3、Ascend950PR
CANN 9.0.0
Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)
用msprof op工具采集单算子性能，调用aclnnModulate执行即可。
Describe the expected behavior / 预期结果 (Mandatory / 必填)
预期Ascend950PR该算子性能不劣于Ascend910B3
Related log / screenshot / 日志 / 截图 (Mandatory / 必填)
采集性能数据如下：
Special notes for this issue/备注 (Optional / 选填)