diff options
author | Szabolcs Nagy <nsz@port70.net> | 2018-09-22 21:43:42 +0000 |
---|---|---|
committer | Rich Felker <dalias@aerifal.cx> | 2018-10-15 14:45:28 -0400 |
commit | e9016138886527a739804634048aeac16092dc1e (patch) | |
tree | b7581245a4ff1da3f27bfdc5826ff362a8ab8430 | |
parent | 7396ef0a05b834bf92c4f268a3336c0bc10c3593 (diff) | |
download | musl-e9016138886527a739804634048aeac16092dc1e.zip musl-e9016138886527a739804634048aeac16092dc1e.tar.gz musl-e9016138886527a739804634048aeac16092dc1e.tar.bz2 |
x86_64: add single instruction fma
fma is only available on recent x86_64 cpus and it is much faster than
a software fma, so this should be done with a runtime check, however
that requires more changes, this patch just adds the code so it can be
tested when musl is compiled with -mfma or -mfma4.
-rw-r--r-- | src/math/x32/fma.c | 23 | ||||
-rw-r--r-- | src/math/x32/fmaf.c | 23 | ||||
-rw-r--r-- | src/math/x86_64/fma.c | 23 | ||||
-rw-r--r-- | src/math/x86_64/fmaf.c | 23 |
4 files changed, 92 insertions, 0 deletions
diff --git a/src/math/x32/fma.c b/src/math/x32/fma.c new file mode 100644 index 0000000..4dd53f2 --- /dev/null +++ b/src/math/x32/fma.c @@ -0,0 +1,23 @@ +#include <math.h> + +#if __FMA__ + +double fma(double x, double y, double z) +{ + __asm__ ("vfmadd132sd %1, %2, %0" : "+x" (x) : "x" (y), "x" (z)); + return x; +} + +#elif __FMA4__ + +double fma(double x, double y, double z) +{ + __asm__ ("vfmaddsd %3, %2, %1, %0" : "=x" (x) : "x" (x), "x" (y), "x" (z)); + return x; +} + +#else + +#include "../fma.c" + +#endif diff --git a/src/math/x32/fmaf.c b/src/math/x32/fmaf.c new file mode 100644 index 0000000..30b971f --- /dev/null +++ b/src/math/x32/fmaf.c @@ -0,0 +1,23 @@ +#include <math.h> + +#if __FMA__ + +float fmaf(float x, float y, float z) +{ + __asm__ ("vfmadd132ss %1, %2, %0" : "+x" (x) : "x" (y), "x" (z)); + return x; +} + +#elif __FMA4__ + +float fmaf(float x, float y, float z) +{ + __asm__ ("vfmaddss %3, %2, %1, %0" : "=x" (x) : "x" (x), "x" (y), "x" (z)); + return x; +} + +#else + +#include "../fmaf.c" + +#endif diff --git a/src/math/x86_64/fma.c b/src/math/x86_64/fma.c new file mode 100644 index 0000000..4dd53f2 --- /dev/null +++ b/src/math/x86_64/fma.c @@ -0,0 +1,23 @@ +#include <math.h> + +#if __FMA__ + +double fma(double x, double y, double z) +{ + __asm__ ("vfmadd132sd %1, %2, %0" : "+x" (x) : "x" (y), "x" (z)); + return x; +} + +#elif __FMA4__ + +double fma(double x, double y, double z) +{ + __asm__ ("vfmaddsd %3, %2, %1, %0" : "=x" (x) : "x" (x), "x" (y), "x" (z)); + return x; +} + +#else + +#include "../fma.c" + +#endif diff --git a/src/math/x86_64/fmaf.c b/src/math/x86_64/fmaf.c new file mode 100644 index 0000000..30b971f --- /dev/null +++ b/src/math/x86_64/fmaf.c @@ -0,0 +1,23 @@ +#include <math.h> + +#if __FMA__ + +float fmaf(float x, float y, float z) +{ + __asm__ ("vfmadd132ss %1, %2, %0" : "+x" (x) : "x" (y), "x" (z)); + return x; +} + +#elif __FMA4__ + +float fmaf(float x, float y, float z) +{ + __asm__ ("vfmaddss %3, %2, %1, %0" : "=x" (x) : "x" (x), "x" (y), "x" (z)); + return x; +} + +#else + +#include "../fmaf.c" + +#endif |