Don't try to enable use of ARM NEON intrinsics when compiling in CUDA device mode. They are not available in that configuration, even if the host supports them.

PiperOrigin-RevId: 468232254 Change-Id: I1cd26eda2962778b62135a31f477928c3bde1124
2 years ago · e81f6ba042
parent fcfc7a6d15
commit e81f6ba042
1 changed files with 6 additions and 1 deletions
--- a/absl/base/config.h
+++ b/absl/base/config.h
@ -915,9 +915,14 @@ static_assert(ABSL_INTERNAL_INLINE_NAMESPACE_STR[0] != 'h' ||

 // ABSL_INTERNAL_HAVE_ARM_NEON is used for compile-time detection of NEON (ARM
 // SIMD).
+//
+// If __CUDA_ARCH__ is defined, then we are compiling CUDA code in device mode.
+// In device mode, NEON intrinsics are not available, regardless of host
+// platform.
+// https://llvm.org/docs/CompileCudaWithLLVM.html#detecting-clang-vs-nvcc-from-code
 #ifdef ABSL_INTERNAL_HAVE_ARM_NEON
 #error ABSL_INTERNAL_HAVE_ARM_NEON cannot be directly set
-#elif defined(__ARM_NEON)
+#elif defined(__ARM_NEON) && !defined(__CUDA_ARCH__)
 #define ABSL_INTERNAL_HAVE_ARM_NEON 1
 #endif