puzzles: faster, smaller sqrt()
Change-Id: I18e170ee49bff131fe76fa4bb1b9e0f120818b82
This commit is contained in:
parent
f2f1889b10
commit
ec1a74a37c
1 changed files with 7 additions and 61 deletions
|
@ -984,70 +984,16 @@ float atan2_wrapper(float y, float x)
|
|||
}
|
||||
}
|
||||
|
||||
/* Square root function, original. */
|
||||
float sqrt_wrapper(float x)
|
||||
{
|
||||
float z;
|
||||
int32_t sign = (int)0x80000000;
|
||||
int32_t ix,s,q,m,t,i;
|
||||
uint32_t r;
|
||||
/* find inverse, Quake-style */
|
||||
float xhalf = .5f * x;
|
||||
int i = *(int*)&x;
|
||||
i = 0x5f3759df - (i >> 1);
|
||||
x = *(float*)&i;
|
||||
x = x * (1.5f - (xhalf * x * x));
|
||||
|
||||
GET_FLOAT_WORD(ix,x);
|
||||
|
||||
/* take care of Inf and NaN */
|
||||
if((ix&0x7f800000)==0x7f800000) {
|
||||
return x*x+x; /* sqrt(NaN)=NaN, sqrt(+inf)=+inf
|
||||
sqrt(-inf)=sNaN */
|
||||
}
|
||||
/* take care of zero */
|
||||
if(ix<=0) {
|
||||
if((ix&(~sign))==0) return x;/* sqrt(+-0) = +-0 */
|
||||
else if(ix<0)
|
||||
return (x-x)/(x-x); /* sqrt(-ve) = sNaN */
|
||||
}
|
||||
/* normalize x */
|
||||
m = (ix>>23);
|
||||
if(m==0) { /* subnormal x */
|
||||
for(i=0;(ix&0x00800000)==0;i++) ix<<=1;
|
||||
m -= i-1;
|
||||
}
|
||||
m -= 127; /* unbias exponent */
|
||||
ix = (ix&0x007fffff)|0x00800000;
|
||||
if(m&1) /* odd m, double x to make it even */
|
||||
ix += ix;
|
||||
m >>= 1; /* m = [m/2] */
|
||||
|
||||
/* generate sqrt(x) bit by bit */
|
||||
ix += ix;
|
||||
q = s = 0; /* q = sqrt(x) */
|
||||
r = 0x01000000; /* r = moving bit from right to left */
|
||||
|
||||
while(r!=0) {
|
||||
t = s+r;
|
||||
if(t<=ix) {
|
||||
s = t+r;
|
||||
ix -= t;
|
||||
q += r;
|
||||
}
|
||||
ix += ix;
|
||||
r>>=1;
|
||||
}
|
||||
|
||||
/* use floating add to find out rounding direction */
|
||||
if(ix!=0) {
|
||||
z = one-tiny; /* trigger inexact flag */
|
||||
if (z>=one) {
|
||||
z = one+tiny;
|
||||
if (z>one)
|
||||
q += 2;
|
||||
else
|
||||
q += (q&1);
|
||||
}
|
||||
}
|
||||
ix = (q>>1)+0x3f000000;
|
||||
ix += (m <<23);
|
||||
SET_FLOAT_WORD(z,ix);
|
||||
return z;
|
||||
return 1.0f / x;
|
||||
}
|
||||
|
||||
/* hack, simple trig */
|
||||
|
|
Loading…
Reference in a new issue