拟合可以调用matlab的命令ployfit来做，例如：
x=[start:
:
stop];
y=atan（x）;
pa=polyfit（x,y,2）;
上式中的运算都是简单的乘法运算，较为简单。
开方运算
浮点开方运算描述为：
floatx,y;
y=sqrt（x）;
定点求开方有多种方法，各种方法在收敛速度上不尽相同，下面介绍几种常用的迭代算法。
1．Newton-Raphson-Babylonian算法：
给定整数N,求sqrt（N）。
首先确定初值x[0],然后利用一个简单的迭代公式：
x[n+1]=（x[n]+N/x[n]）/2
迭代次数的选择：
迭代次数与初值x[0]的选取很有关系，x[0]越接近sqrt（N）,收敛越快。
但总的来说，该方法收敛较快。
缺点是收敛时间不确定。
2．确定收敛速度的算法：
该方法描述如下：
intsqrt（intx）
{inttest,step;
if（x<0）return（-1）;if（x==0）return（0）;
step=1<<15;
test=0;
while（step!
=0）
{
registerinth;
h=（test+step）*（test+step）;
if（h<=x）{test+=step;}
if（h==x）break;
step>>=1;
}
return（test）;}
以上例子是32位开放运算，32位以上的开方运算可参考附录1voidfixsqrt（UINT4*a,UINT4*b,intdigits），方法同上。
求开方还可以运用线性拟合的方法，由于曲线变化较快，必须根据自变量的范围分段拟合才能达到理想的精度。
附录
附录1：
定点函数库
/*___________________________________________________________________________
||
|FunctionName:
L_add|
||
|Purpose:
|
||
|32bitsadditionofthetwo32bitsvariables（L_var1+L_var2）with|
|overflowcontrolandsaturation;theresultissetat+47when|
|overflowoccursorat-48whenunderflowoccurs.|
||
|Complexityweight:
2|
||
|Inputs:
|
||
|L_var132bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var3<=0x7fffffff.|
||
|L_var232bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var3<=0x7fffffff.|
||
|Outputs:
|
||
|none|
||
|ReturnValue:
|
||
|L_var_out|
|32bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var_out<=0x7fffffff.|
|___________________________________________________________________________|
*/
Word32L_add（Word32L_var1,Word32L_var2）
/*___________________________________________________________________________
||
|FunctionName:
L_sub|
||
|Purpose:
|
||
|32bitssubtractionofthetwo32bitsvariables（L_var1-L_var2）with|
|overflowcontrolandsaturation;theresultissetat+7when|
|overflowoccursorat-8whenunderflowoccurs.|
||
|Complexityweight:
2|
||
|Inputs:
|
||
|L_var132bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var3<=0x7fffffff.|
||
|L_var232bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var3<=0x7fffffff.|
||
|Outputs:
|
||
|none|
||
|ReturnValue:
|
||
|L_var_out|
|32bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var_out<=0x7fffffff.|
|___________________________________________________________________________|
*/
Word32L_sub（Word32L_var1,Word32L_var2）
/*___________________________________________________________________________
||
|FunctionName:
add|
||
|Purpose:
|
||
|Performstheaddition（var1+var2）withoverflowcontrolandsaturation;|
|the16bitresultissetat+32767whenoverflowoccursorat-32768|
|whenunderflowoccurs.|
||
|Complexityweight:
1|
||
|Inputs:
|
||
|var1|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var1<=0x00007fff.|
||
|var2|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var1<=0x00007fff.|
||
|Outputs:
|
||
|none|
||
|ReturnValue:
|
||
|var_out|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var_out<=0x00007fff.|
|___________________________________________________________________________|
*/
Word16add（Word16var1,Word16var2）
/*___________________________________________________________________________
||
|FunctionName:
sature|
||
|Purpose:
|
||
|Limitthe32bitinputtotherangeofa16bitword.|
||
|Inputs:
|
||
|L_var1|
|32bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var1<=0x7fffffff.|
||
|Outputs:
|
||
|none|
||
|ReturnValue:
|
||
|var_out|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var_out<=0x00007fff.|
|___________________________________________________________________________|
*/
Word16sature（Word32L_var1）
/*___________________________________________________________________________
||
|FunctionName:
sub|
||
|Purpose:
|
||
|Performsthesubtraction（var1+var2）withoverflowcontrolandsatu-|
|ration;the16bitresultissetat+32767whenoverflowoccursorat|
|-32768whenunderflowoccurs.|
||
|Complexityweight:
1|
||
|Inputs:
|
||
|var1|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var1<=0x00007fff.|
||
|var2|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var1<=0x00007fff.|
||
|Outputs:
|
||
|none|
||
|ReturnValue:
|
||
|var_out|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var_out<=0x00007fff.|
|___________________________________________________________________________|
*/
Word16sub（Word16var1,Word16var2）
/*___________________________________________________________________________
||
|FunctionName:
L_mult|
||
|Purpose:
|
||
|L_multisthe32bitresultofthemultiplicationofvar1timesvar2|
|withoneshiftleft.:
|
|L_mult（var1,var2）=shl（（var1timesvar2）,1）and|
|L_mult（-32768,-32768）=47.|
||
|Complexityweight:
1|
||
|Inputs:
|
||
|var1|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var1<=0x00007fff.|
||
|var2|
|16bitshortsignedinteger（Word16）whosevaluefallsinthe|
|range:
0xffff8000<=var1<=0x00007fff.|
||
|Outputs:
|
||
|none|
||
|ReturnValue:
|
||
|L_var_out|
|32bitlongsignedinteger（Word32）whosevaluefallsinthe|
|range:
0x80000000<=L_var_out<=0x7fffffff.|
|___________________________________________________________________________|
*/
Word32L_mult（Word16var1,Word16var2）
/*Computesthesquarerootofafixpointnumbera=square（b）.*/
/*length:
a[digits],b[2*digits]*/
voidfixsqrt（UINT4*a,UINT4*b,intdigits）
{
Returnscarry.
Lengths:
a[digits],b[digits],c[digits].
*/
NN_DIGITNN_Add（a,b,c,digits）
NN_DIGIT*a,*b,*c;
unsignedintdigits;
/*Computesa=b-c.Returnsborrow.
Lengths:
a[digits],b[digits],c[digits].
*/
NN_DIGITNN_Sub（a,b,c,digits）
NN_DIGIT*a,*b,*c;
unsignedintdigits;
/*Computesa=b*c.
Lengths:
a[2*digits],b[digits],c[digits].
Assumesdigits*/
voidNN_Mult（a,b,c,digits）
NN_DIGIT*a,*b,*c;
unsignedintdigits;
/*Returnssignofa-b.*/
intNN_Cmp（a,b,digits）
NN_DIGIT*a,*b;
unsignedintdigits;
/*Computesa=b*2^c.,shiftsleftcbits）,returningcarry.
RequirescNN_DIGITNN_LShift（a,b,c,digits）
NN_DIGIT*a,*b;
unsignedintc,digits;
/*Computesa=bdiv2^c.,shiftsrightcbits）,returningcarry.
Requ
升级会员 