1 files changed, 47 insertions, 0 deletions
diff --git a/gcc/testsuite/gcc.target/arm/fp16-rounding-ieee-1.c b/gcc/testsuite/gcc.target/arm/fp16-rounding-ieee-1.c
new file mode 100644
index 000000000..866d4d824
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/fp16-rounding-ieee-1.c
@@ -0,0 +1,47 @@
+/* Test intermediate rounding of double to float and then to __fp16, using
+   an example of a number that would round differently if it went directly
+   from double to __fp16.  */
+
+/* { dg-do run } */
+/* { dg-options "-mfp16-format=ieee" } */
+
+#include <stdlib.h>
+
+/* The original double value.  */
+#define ORIG 0x1.0020008p0
+
+/* The expected (double)((__fp16)((float)ORIG)) value.  */
+#define ROUNDED 0x1.0000000p0
+
+typedef union u {
+  __fp16 f;
+  unsigned short h;
+} ufh;
+
+ufh s = { ORIG };
+ufh r = { ROUNDED };
+
+double d = ORIG;
+
+int
+main (void)
+{
+  ufh x;
+
+  /* Test that the rounding is correct for static initializers.  */
+  if (s.h != r.h)
+    abort ();
+
+  /* Test that the rounding is correct for a casted constant expression
+     not in a static initializer.  */
+  x.f = (__fp16)ORIG;
+  if (x.h != r.h)
+    abort ();
+
+  /* Test that the rounding is correct for a runtime conversion.  */
+  x.f = (__fp16)d;
+  if (x.h != r.h)
+    abort ();
+
+  return 0;
+}