search

intel / torch-xpu-ops

Apache License 2.0
30 stars 21 forks source link

Update default_sorter to work with 2025 compiler #1032

Closed ZzEeKkAa closed 3 weeks ago

ZzEeKkAa commented 3 weeks ago

Fixes: #1027

ZzEeKkAa commented 3 weeks ago

Part of the code is compiled with clang++, so __INTEL_LLVM_COMPILER can't be used there, so I don't know how to address it

fengyuan14 commented 3 weeks ago

Hi, @ZzEeKkAa. We have got a pr about it https://github.com/intel/torch-xpu-ops/pull/1017. We build torch-xpu-ops with separate compilation solution, compile host with gcc and compile device code with dpc++. And the macro is not a source-code-defined macro in SYCL headers but a dpc++ builtin. So the macro is not defined in host part as your changes. Please refer the pr pasted, we have a pre-process in cmake process to expose the macro.

ZzEeKkAa commented 3 weeks ago

PR #1017 does exactly what we need. Thank you!