diff options
| author | bde <bde@FreeBSD.org> | 2008-03-30 18:07:12 +0000 |
|---|---|---|
| committer | bde <bde@FreeBSD.org> | 2008-03-30 18:07:12 +0000 |
| commit | 2916ad3e28c0f7fea67ffe9b7af7f51b6e3aaaf0 (patch) | |
| tree | 94cad26c2877839b64fe3a8e8a6b961683a86d57 /lib/msun/src | |
| parent | b06e3a074e790b19403a03d0155466a343805194 (diff) | |
| download | FreeBSD-src-2916ad3e28c0f7fea67ffe9b7af7f51b6e3aaaf0.zip FreeBSD-src-2916ad3e28c0f7fea67ffe9b7af7f51b6e3aaaf0.tar.gz | |
Use fabs[f]() instead of bit fiddling for setting absolute values.
This makes little difference in float precision, but in double
precision gives a speedup of about 30% on amd64 (A64 CPU) and i386
(A64). This depends on fabs[f]() being inline and efficient. The
bit fiddling (or any use of SET_HIGH_WORD(), which libm does too
much because it was best on old 32-bit machines) always causes
packing overheads and sometimes causes stalls in the packing, since
it operates on only part of a variable in the double precision case.
It apparently did cause stalls in a critical path here.
Diffstat (limited to 'lib/msun/src')
| -rw-r--r-- | lib/msun/src/e_hypot.c | 4 | ||||
| -rw-r--r-- | lib/msun/src/e_hypotf.c | 4 |
2 files changed, 4 insertions, 4 deletions
diff --git a/lib/msun/src/e_hypot.c b/lib/msun/src/e_hypot.c index 8f1cd10..30177fb 100644 --- a/lib/msun/src/e_hypot.c +++ b/lib/msun/src/e_hypot.c @@ -60,8 +60,8 @@ __ieee754_hypot(double x, double y) GET_HIGH_WORD(hb,y); hb &= 0x7fffffff; if(hb > ha) {a=y;b=x;j=ha; ha=hb;hb=j;} else {a=x;b=y;} - SET_HIGH_WORD(a,ha); /* a <- |a| */ - SET_HIGH_WORD(b,hb); /* b <- |b| */ + a = fabs(a); + b = fabs(b); if((ha-hb)>0x3c00000) {return a+b;} /* x/y > 2**60 */ k=0; if(ha > 0x5f300000) { /* a>2**500 */ diff --git a/lib/msun/src/e_hypotf.c b/lib/msun/src/e_hypotf.c index c70668e..c82c6e7 100644 --- a/lib/msun/src/e_hypotf.c +++ b/lib/msun/src/e_hypotf.c @@ -30,8 +30,8 @@ __ieee754_hypotf(float x, float y) GET_FLOAT_WORD(hb,y); hb &= 0x7fffffff; if(hb > ha) {a=y;b=x;j=ha; ha=hb;hb=j;} else {a=x;b=y;} - SET_FLOAT_WORD(a,ha); /* a <- |a| */ - SET_FLOAT_WORD(b,hb); /* b <- |b| */ + a = fabsf(a); + b = fabsf(b); if((ha-hb)>0xf000000) {return a+b;} /* x/y > 2**30 */ k=0; if(ha > 0x58800000) { /* a>2**50 */ |
