[Driver][NFC] Don't specify offloading model in help text for -Xarch_device/host

[jinge90]

Current implementation for -Xarch_device/host can work in any offloading model besides CUDA/HIP, so remove the specific offloading model in help text to align with implementation.

[llvmbot]

@llvm/pr-subscribers-clang

Author: None (jinge90)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/165503.diff

1 Files Affected:

• (modified) clang/include/clang/Driver/Options.td (+2-2)

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 8784c9d7d206d..77a5ed67beb29 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -950,9 +950,9 @@ def Xarch__
   the host system, which can be used to suppress incompatible GPU arguments.}]>,
       MetaVarName<"<arch> <arg>">;
 def Xarch_host : Separate<["-"], "Xarch_host">, Flags<[NoXarchOption]>,
-  HelpText<"Pass <arg> to the CUDA/HIP host compilation">, MetaVarName<"<arg>">;
+  HelpText<"Pass <arg> to the CUDA/HIP/SYCL host compilation">, MetaVarName<"<arg>">;
 def Xarch_device : Separate<["-"], "Xarch_device">, Flags<[NoXarchOption]>,
-  HelpText<"Pass <arg> to the CUDA/HIP device compilation">, MetaVarName<"<arg>">;
+  HelpText<"Pass <arg> to the CUDA/HIP/SYCL device compilation">, MetaVarName<"<arg>">;
 def Xassembler : Separate<["-"], "Xassembler">,
   HelpText<"Pass <arg> to the assembler">, MetaVarName<"<arg>">,
   Group<CompileOnly_Group>;

[jinge90]

Hi, @bader
-Xarch_device/host can work for SYCL and we have used them in intel/llvm, this PR shows SYCL is supported in -Xarch_device/host help text.
Thanks very much.

[bader]

we have used them in intel/llvm

Do they do anything useful for SYCL in https://github.com/llvm/llvm-project/?

[jinge90]

we have used them in intel/llvm

Do they do anything useful for SYCL in https://github.com/llvm/llvm-project/?

Hi, @bader
The driver code handling -Xarch_device/host doesn't do specific things to SYCL or CUDA, HIP : https://github.com/llvm/llvm-project/blob/main/clang/lib/Driver/Compilation.cpp#L58
If you think we can update the help text, I will add a lit test for -Xarch in SYCL.
Thank very much.

[bader]

The driver code handling -Xarch_device/host doesn't do specific things to SYCL or CUDA, HIP : https://github.com/llvm/llvm-project/blob/main/clang/lib/Driver/Compilation.cpp#L58

I see that @jhuber6 enabled these options for OpenMP offload in eec516a. The patch enabled them not only for OpenMP offload, but for any offloading programming model.

Considering that, I suggest we don't specify offloading models at all. Let's align the help text with the implementation by removing mentions of CUDA/HIP/SYCL.

[bader]

LGTM.
@shiltian, @jhuber6, FYI.