summaryrefslogtreecommitdiff
path: root/src/math/erff.c
diff options
context:
space:
mode:
authorSzabolcs Nagy <nsz@port70.net>2013-05-15 23:08:52 +0000
committerSzabolcs Nagy <nsz@port70.net>2013-05-15 23:08:52 +0000
commite216951f509b71da193da2fc63e25b998740d58b (patch)
tree391381eafd8b0b7eb4fdbb693608072df5c181d5 /src/math/erff.c
parent2c184264eae3797de028403ed1e86c1f7ae5b813 (diff)
downloadmusl-e216951f509b71da193da2fc63e25b998740d58b.tar.gz
math: use double_t for temporaries to avoid stores on i386
When FLT_EVAL_METHOD!=0 (only i386 with x87 fp) the excess precision of an expression must be removed in an assignment. (gcc needs -fexcess-precision=standard or -std=c99 for this) This is done by extra load/store instructions which adds code bloat when lot of temporaries are used and it makes the result less precise in many cases. Using double_t and float_t avoids these issues on i386 and it makes no difference on other archs. For now only a few functions are modified where the excess precision is clearly beneficial (mostly polynomial evaluations with temporaries). object size differences on i386, gcc-4.8: old new __cosdf.o 123 95 __cos.o 199 169 __sindf.o 131 95 __sin.o 225 203 __tandf.o 207 151 __tan.o 605 499 erff.o 1470 1416 erf.o 1703 1649 j0f.o 1779 1745 j0.o 2308 2274 j1f.o 1602 1568 j1.o 2286 2252 tgamma.o 1431 1424 math/*.o 64164 63635
Diffstat (limited to 'src/math/erff.c')
-rw-r--r--src/math/erff.c5
1 files changed, 3 insertions, 2 deletions
diff --git a/src/math/erff.c b/src/math/erff.c
index e2cfc984..ed5f3975 100644
--- a/src/math/erff.c
+++ b/src/math/erff.c
@@ -86,7 +86,7 @@ sb7 = -2.2440952301e+01; /* 0xc1b38712 */
static float erfc1(float x)
{
- float s,P,Q;
+ float_t s,P,Q;
s = fabsf(x) - 1;
P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
@@ -96,7 +96,8 @@ static float erfc1(float x)
static float erfc2(uint32_t ix, float x)
{
- float s,z,R,S;
+ float_t s,R,S;
+ float z;
if (ix < 0x3fa00000) /* |x| < 1.25 */
return erfc1(x);