mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-05-29 16:52:28 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Speed up fmaf() on x86

Browse source
This commit is contained in:
Justine Tunney 2024-03-22 19:26:56 -07:00
parent 640668931d
commit 40b7da8422
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
1 changed files with 10 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

10
libc/tinymath/fmaf.c
View file

@ -100,6 +100,16 @@ float fmaf(float x, float y, float z)
#else #else
#ifdef __x86_64__
if (X86_HAVE(FMA)) {
asm("vfmadd132ss\t%1,%2,%0" : "+x"(x) : "x"(y), "x"(z));
return x;
} else if (X86_HAVE(FMA4)) {
asm("vfmaddss\t%3,%2,%1,%0" : "=x"(x) : "x"(x), "x"(y), "x"(z));
return x;
}
#endif
/* A double has more than twice as much precision than a float, /* A double has more than twice as much precision than a float,
so direct double-precision arithmetic suffices, except where so direct double-precision arithmetic suffices, except where
double rounding occurs. */ double rounding occurs. */