1 files changed, 46 insertions, 4 deletions

diff --git a/clang/lib/Headers/__clang_cuda_cmath.h b/clang/lib/Headers/__clang_cuda_cmath.h
index 8ba182689a4f..5bbb59a93b9e 100644
--- a/clang/lib/Headers/__clang_cuda_cmath.h
+++ b/clang/lib/Headers/__clang_cuda_cmath.h
@@ -66,10 +66,38 @@ __DEVICE__ float frexp(float __arg, int *__exp) {
 }
 
 // For inscrutable reasons, the CUDA headers define these functions for us on
-// Windows. For OpenMP we omit these as some old system headers have
-// non-conforming `isinf(float)` and `isnan(float)` implementations that return
-// an `int`. The system versions of these functions should be fine anyway.
-#if !defined(_MSC_VER) && !defined(__OPENMP_NVPTX__)
+// Windows.
+#if !defined(_MSC_VER) || defined(__OPENMP_NVPTX__)
+
+// For OpenMP we work around some old system headers that have non-conforming
+// `isinf(float)` and `isnan(float)` implementations that return an `int`. We do
+// this by providing two versions of these functions, differing only in the
+// return type. To avoid conflicting definitions we disable implicit base
+// function generation. That means we will end up with two specializations, one
+// per type, but only one has a base function defined by the system header.
+#if defined(__OPENMP_NVPTX__)
+#pragma omp begin declare variant match(                                       \
+    implementation = {extension(disable_implicit_base)})
+
+// FIXME: We lack an extension to customize the mangling of the variants, e.g.,
+//        add a suffix. This means we would clash with the names of the variants
+//        (note that we do not create implicit base functions here). To avoid
+//        this clash we add a new trait to some of them that is always true
+//        (this is LLVM after all ;)). It will only influence the mangled name
+//        of the variants inside the inner region and avoid the clash.
+#pragma omp begin declare variant match(implementation = {vendor(llvm)})
+
+__DEVICE__ int isinf(float __x) { return ::__isinff(__x); }
+__DEVICE__ int isinf(double __x) { return ::__isinf(__x); }
+__DEVICE__ int isfinite(float __x) { return ::__finitef(__x); }
+__DEVICE__ int isfinite(double __x) { return ::__isfinited(__x); }
+__DEVICE__ int isnan(float __x) { return ::__isnanf(__x); }
+__DEVICE__ int isnan(double __x) { return ::__isnan(__x); }
+
+#pragma omp end declare variant
+
+#endif
+
 __DEVICE__ bool isinf(float __x) { return ::__isinff(__x); }
 __DEVICE__ bool isinf(double __x) { return ::__isinf(__x); }
 __DEVICE__ bool isfinite(float __x) { return ::__finitef(__x); }
@@ -79,6 +107,11 @@ __DEVICE__ bool isfinite(float __x) { return ::__finitef(__x); }
 __DEVICE__ bool isfinite(double __x) { return ::__isfinited(__x); }
 __DEVICE__ bool isnan(float __x) { return ::__isnanf(__x); }
 __DEVICE__ bool isnan(double __x) { return ::__isnan(__x); }
+
+#if defined(__OPENMP_NVPTX__)
+#pragma omp end declare variant
+#endif
+
 #endif
 
 __DEVICE__ bool isgreater(float __x, float __y) {
@@ -142,6 +175,15 @@ __DEVICE__ float sqrt(float __x) { return ::sqrtf(__x); }
 __DEVICE__ float tan(float __x) { return ::tanf(__x); }
 __DEVICE__ float tanh(float __x) { return ::tanhf(__x); }
 
+// There was a redefinition error for this this overload in CUDA mode.
+// We restrict it to OpenMP mode for now, that is where it is actually needed
+// anyway.
+#ifdef __OPENMP_NVPTX__
+__DEVICE__ float remquo(float __n, float __d, int *__q) {
+  return ::remquof(__n, __d, __q);
+}
+#endif
+
 // Notably missing above is nexttoward.  We omit it because
 // libdevice doesn't provide an implementation, and we don't want to be in the
 // business of implementing tricky libm functions in this header.