about summary refs log tree commit diff
path: root/src/math/aarch64/fmin.c
diff options
context:
space:
mode:
authorSzabolcs Nagy <nsz@port70.net>2017-03-19 05:26:45 +0100
committerRich Felker <dalias@aerifal.cx>2017-03-21 12:39:18 -0400
commit54807d47acecab778498ced88ce8f62bfa16e379 (patch)
tree2203cc44d6061f092dd104ef6f89d4f2ca9fa1d7 /src/math/aarch64/fmin.c
parentb6e1fe0d5e78dac647e85d49c2d537bb071ba49e (diff)
downloadmusl-54807d47acecab778498ced88ce8f62bfa16e379.tar.gz
musl-54807d47acecab778498ced88ce8f62bfa16e379.tar.xz
musl-54807d47acecab778498ced88ce8f62bfa16e379.zip
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.
Diffstat (limited to 'src/math/aarch64/fmin.c')
-rw-r--r--src/math/aarch64/fmin.c7
1 files changed, 7 insertions, 0 deletions
diff --git a/src/math/aarch64/fmin.c b/src/math/aarch64/fmin.c
new file mode 100644
index 00000000..f1e99808
--- /dev/null
+++ b/src/math/aarch64/fmin.c
@@ -0,0 +1,7 @@
+#include <math.h>
+
+double fmin(double x, double y)
+{
+	__asm__ ("fminnm %d0, %d1, %d2" : "=w"(x) : "w"(x), "w"(y));
+	return x;
+}