Support pp on xpu

[sljlp]

softmax_mask_fuse_upper_triangle is not supported if paddle is not compiled with cuda/rocm

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}