[clang] [clang][CUDA] Disable float128 diagnostics for device compilation (PR #83918)

[via cfe-commits]

================
@@ -0,0 +1,18 @@
+// CPU-side compilation on x86 (no errors expected).
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -aux-triple nvptx64 -x cuda -fsyntax-only -verify=cpu %s
+
+// GPU-side compilation on x86 (no errors expected)
+// RUN: %clang_cc1 -triple nvptx64 -aux-triple x86_64-unknown-linux-gnu -fcuda-is-device -x cuda -fsyntax-only -verify=gpu %s
+
+// cpu-no-diagnostics
+typedef _Complex float __cfloat128 __attribute__ ((__mode__ (__TC__)));
+typedef __float128 _Float128;
----------------
piotrrak wrote:

@pranavk 

Hi, Do you happen to remember if `_Float128` name is of importance for this test?

Since libstdc++ in its `<stdfloat>` even for C++ header uses C23 _FloatN types etc. I'd be nice to make it _Float128 KEYALL keyword.

Such typedef would be incorrect if `_Float128` is a keyword.


https://github.com/llvm/llvm-project/pull/83918