diff options
author | Szabolcs Nagy <nsz@port70.net> | 2017-03-19 05:26:45 +0100 |
---|---|---|
committer | Rich Felker <dalias@aerifal.cx> | 2017-03-21 12:39:18 -0400 |
commit | 54807d47acecab778498ced88ce8f62bfa16e379 (patch) | |
tree | 2203cc44d6061f092dd104ef6f89d4f2ca9fa1d7 /src/math/aarch64/sqrt.c | |
parent | b6e1fe0d5e78dac647e85d49c2d537bb071ba49e (diff) | |
download | musl-54807d47acecab778498ced88ce8f62bfa16e379.tar.gz musl-54807d47acecab778498ced88ce8f62bfa16e379.tar.bz2 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/sqrt.c')
-rw-r--r-- | src/math/aarch64/sqrt.c | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/src/math/aarch64/sqrt.c b/src/math/aarch64/sqrt.c new file mode 100644 index 00000000..fe93c3e6 --- /dev/null +++ b/src/math/aarch64/sqrt.c @@ -0,0 +1,7 @@ +#include <math.h> + +double sqrt(double x) +{ + __asm__ ("fsqrt %d0, %d1" : "=w"(x) : "w"(x)); + return x; +} |