aarch64: add single instruction math functions

this should increase performance and reduce code size on aarch64.

the compiled code was checked against using __builtin_* instead
of inline asm with gcc-6.2.0.

lrint is two instructions.

c with inline asm is used because it is safer than a pure asm
implementation, this prevents ll{rint,round} to be an alias
of l{rint,round} (because the types don't match) and depends
on gcc style inline asm support.

ceil, floor, round, trunc can either raise inexact on finite
non-integer inputs or not raise any exceptions. the new
implementation does not raise exceptions while the generic
c code does.

on aarch64, the underflow exception is signaled before rounding
(ieee 754 allows both before and after rounding, but it must be
consistent), the generic fma c code signals it after rounding
so using single instruction fixes a slight conformance issue too.

1 files changed, 7 insertions, 0 deletions

diff --git a/src/math/aarch64/fmaxf.c b/src/math/aarch64/fmaxf.c
new file mode 100644
index 00000000..ee5eac2d
--- /dev/null
+++ b/src/math/aarch64/fmaxf.c
@@ -0,0 +1,7 @@
+#include <math.h>
+
+float fmaxf(float x, float y)
+{
+	__asm__ ("fmaxnm %s0, %s1, %s2" : "=w"(x) : "w"(x), "w"(y));
+	return x;
+}