# How compactly can your language perform accurate numerical integration?

WARNING: This challenge may need 128 bit floats.1

The task is to perform numerical integration. Consider the following three functions.

$$\ f(x) = cx^{c - 1}e^{-x^c} \$$

$$\ g_1(x) = 0.5e^{-x} \$$

$$\ g_2(x) = 5 e^{-10 x} \$$

We will have that $$\c \geq 0.2\$$. Your code should be correct for any value between 0.2 and 1.0.

The task is to perform the integral:

$$\ \int_{0}^{\infty} f(x) \ln\left(\frac{f(x)}{g_1(x)+g_2(x)}\right) \mathrm{d}x. \$$

Your output should be correct to within plus or minus $$\10^{-10}\$$ percent of the correct answer.

Input

A single real number $$\c\$$.

Output shown to 15 significant places

c = 0.2. Output: 119.2233798550179
c = 0.3. Outout: 8.077346771987397
c = 0.4. Output: 2.07833944013377
c = 0.5. Output: 0.8034827042913205
c = 0.6. Output: 0.3961739639940003
c = 0.7. Output: 0.2585689391629249
c = 0.8. Output: 0.2287758419066705
c = 0.9. Output: 0.2480070283065089
c = 1.0. Output: 0.2908566108890576


Timing

Your code must complete on TIO before it times out.

As is usual, this is a competition per language. In other words, Java coders should not be put off by the single character solution in a language they can't believe exists. You may use any language/library you choose.

Questions and notes

• Does this integral actually exist/converge?

Yes. It is the Kullback Leibler divergence between two probability distributions.

• Show me any code that can compute these integrals.

See this Wolfram Alpha example.

• Why such high precision that we need 128 bit floats?

It is quite easy to solve the integral to low precision using simple ideas. However, to get high precision seems to need careful thought both about the integral itself but also the quadrature method that will be used. I felt this made the question more challenging and interesting.

1. For lots of languages this is easy. E.g Fortran has real128, gcc/clang have _Float128 for C/C++, Python has Decimal (and many other options), Dyalog APL has 128 bit (decimal) floats, C# has the Decimal type, Haskell has rounded, PHP has decimal, MATLAB has the Multiprecision Computing Toolbox, Julia has BigFloat, Java has BigDecimal, JavaScript has big.js, Ruby has BigDecimal, R has Rmpfr, Perl has Math::BigFloat, SOGL has 200 digit decimal precision, etc. If your favourite golfing language doesn't, maybe consider adding it?

# Update

Alexey Burdin has shown that

$$\ \int_{0}^{\infty} f(x) \ln\left(\frac{f(x)}{g_1(x)+g_2(x)}\right) \mathrm{d}x = \int\limits_0^1 \left(\ln c+\left(\ln \left(-\ln t\right)\right)\cdot\frac{c-1}{c}+ \ln t- \ln\left(0.5+5e^{-9x(t)}\right) +(-\ln{t})^{1/c}\right)dt\$$

This potentially greatly simplifies the problem.

• Because of the very high precision requirements here, I think the answer to "How compactly can your language perform numerical integration?" is largely based on "Does your language have a high-precision real-number representation?"
– xnor
Nov 2, 2019 at 19:29
• @xnor It's a little more difficult than that :) But I think most (all?) main stream languages have at least libraries for 128 bit floats.
– user9207
Nov 2, 2019 at 19:40
• Many answers on this site are not written in mainstream languages. Nov 2, 2019 at 19:59
• @pppery that is true. But for every challenge there are a set of languages most suited and a set that are ill suited. Also the challenge is per language so even if you have to implement 128 bit multiplication/division yourself in your language, that's ok.
– user9207
Nov 2, 2019 at 20:08

## python3, 3982218318661821 1677, ~14s

from functools import*
#let's start with the comments. I'll exclude them from total count
#from functools I need only reduce
#I've been coding in python2 for 5 years so I'm used to built-in reduce much
#and it's very handy here
g=range
#just keep it in mind while reading) ranGe -> g
#and we're getting rid of classes in favor of functions, that gives
#us ~100 bytes more. It's how it was:
#class P:
#it's a Polynomial, self.c are coefficients and self.d is a divisor. So all polynomial computations are exact
#   c=[0]
#   d=1
#don't even know if these initial values are necessary
#   def __init__(self,c,d=1):
#       self.c=list(c)
#       self.d=d
#n'th derivative
#   def dn(self,n=1):
#       c=self.c[:]
#       for _ in g(n):
#           c=[c[i]*i for i in g(1,len(c))]
#       d=reduce(gcd,c,self.d)
#gcd is imported from math along with exp,log and gamma
#       return P([i//d for i in c],self.d//d)
#   def m(self,x):
#the value of poly at point x. __call__ may look better
#       return sum((x**i)*self.c[i] for i in range(len(self.c)-1,-1,-1))/self.d
#the reverse order was for floats and really needed there for precision,
#but python3 have exact Fractions, so maybe it's to be golfed too
#Now let's convert this class to functions
#derivative=lambda c:reduce(lambda c,_:[c[i]*i for i in g(1,len(c))],g(n),c)
#the_gcd=lambda c,d:reduce(gcd,c,d)
#the_result=lambda c,d,m:([i//m for i in c],d//m)
#where c is now the derivative
#and combined:
v=lambda c,d,n=1:(lambda c:
(lambda c,d,m:([i//m for i in c],d//m))(c,d,reduce(gcd,c,d))
)\
(reduce(lambda c,_:[c[i]*i for i in g(1,len(c))],g(n),c))
#the computing of a plynomial's value:
m=lambda c,d,x:sum((x**i)*c[i] for i in g(len(c)-1,-1,-1))/d
d={}
#https://en.wikipedia.org/wiki/Combination written recursively with Pascal's triangle, with the dict d for speed
def c(n,k):
if (n,k) not in d:
d[(n,k)]=1 if k==0 or k==n or n==0 else c(n-1,k)+c(n-1,k-1)
return d[(n,k)]
#p=30
#it's the floating point precision for Decimal
#it was needed for computing the polynomial roots, you can see the raw version
#maybe I'll get rid of it if it's used in the only place
from math import*
from fractions import Fraction as F
n=61
l=v(sum(([c(n,i)*(-1)**(n-i),0] for i in g(n+1)),[]),
2**n*reduce(lambda x,y:x*y,g(1,n+1),1),n)
#Legendre polynomial of order n, pretty straightforward, -3bytes for space
#you can't see these bytes in the tio version, there is 1 line
d=v(*l)
#and the derivative for weights
#see we can use c and d now 'cause we don't need them more
c=2**100
#maybe the hardest-to-read part, the roots of l
#taking too long to compute (2-3 mins) so hard-stored
#b(b'GKSmRyJoJz...') is the 390-bytes byte array,
#representing the <0 roots (the function is odd, so -x and x are both roots and we have to store only half)
#30 13-byte(100-bit, for 10^{-30} precision) integers are stored in lsbf order of bytes
#and c is common multiplier for all computed roots' denominators
from base64 import b64decode as b
#extract numbers from base-64,
r=(lambda b:[sum(2**(n*8)*i for n,i in enumerate(b[j:j+13])) for j in g(0,391,13)])(b(b'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'))
#and make fractions
#0 is in r for magically mistaken 391 for 390 )
#maybe +[0]+r[::-1] is 1 byte shorter (or longer?)
#but at least more readable )
#roots computation code is in the raw below
r=[F(i,c) for i in [-j for j in r]+r[-2::-1]]
#the weights
w=[2/((1-x**2)*(m(*d,x))**2) for x in r]
from decimal import*
D=Decimal
setcontext(Context(prec=30, rounding=ROUND_HALF_DOWN))
#that was float=
#it's only converting Fraction to Decimal. no built-in
f=lambda r:(lambda n,d:D(n)/D(d))(*str(r).split('/'))
#number of sub-intervals of (0,1)
d=2**8
s=input().strip()
c=D(s)
#the inverse of c, not the math.exp(1) ))
e=D(1)/c
#pretty straightforward, I is the integrator,
I=lambda h:sum(sum(f(v)*h(f(((x+1)/2+i)/d))/(2*d) for v,x in zip(w,r))
for i in range(d))
#the rest follows (at math:)
print(D(gamma(1/float(s)+1))+(c.ln())-I(lambda t:(D('0.5')+5*((-9*((-((t).ln()))**(e))).exp())).ln())-D('0.5772156649015328606065120900824')*(1-e)-1)


Try it online! -- takes input as a valid Decimal, like 0.2
Math:
0. We do google numerical integration.
1. Let's see provocative $$\f(x)=c\cdot x^{c-1}\cdot e^{-x^c}= -\left(e^{-x^c}\right)'\$$
2. So, we have $$\-\int\limits_0^\infty \ln\left(\frac{c\cdot x^{c-1}\cdot e^{-x^c}}{0.5e^{-x}+5e^{-10x}}\right)\left(e^{-x^c}\right)'dx\$$,
we change the variable to $$\t=e^{-x^c}\$$, $$\x^c=-\ln(t)\$$, $$\x=\left(-\ln t\right)^\frac1c\$$,
$$\t(0)=1\$$, $$\t(+\infty)=0\$$ and the integral is
$$\\int\limits_0^1 \ln\left(\frac{c\cdot x(t)^{c-1}\cdot t}{\left(0.5+5e^{-9x(t)}\right)e^{-x(t)}}\right)dt=\$$
$$\\int\limits_0^1 \left(\ln c+\left(\ln x(t)\right)\cdot(c-1)+ \ln t- \ln\left(0.5+5e^{-9x(t)}\right) +x(t)\right)dt=\$$
$$\\int\limits_0^1 \left(\ln c+\left(\ln \left(-\ln t\right)\right)\cdot\frac{c-1}{c}+ \ln t- \ln\left(0.5+5e^{-9x(t)}\right) +x(t)\right)dt\$$
3. Next we wolframalpha these things:
$$\\int\limits_0^1\ln(-\ln t)dt\$$, $$\\int\limits_0^1 (-\ln t)^a dt\$$, $$\\int\limits_0^1 \ln t\,dt\$$ and do the rest numerically (the latter thing maybe can be done with trapeziums or rectangles or Simpson -- the idea for short submissions).
The raw thing:

from copy import copy
from functools import reduce
class Poly:
c=[0]
d=1
def __init__(self,c,d=1):
self.c=list(c)
self.d=d
def dn(self,n=1):
c=copy(self.c)
for _ in range(n):
c=[c[i]*i for i in range(1,len(c))]
d=reduce(gcd,c,self.d)
return Poly([i//d for i in c],self.d//d)
def m(self,x):
return sum((x**i)*self.c[i] for i in range(len(self.c)-1,-1,-1))/self.d
def gcd(x,y):
return abs(gcd(y,x%y)) if y else x
cdict={}
def c(n,k):
if (n,k) not in cdict:
cdict[(n,k)]=1 if k==0 or k==n or n==0 else c(n-1,k)+c(n-1,k-1)
return cdict[(n,k)]
def root(x,f,f_):
x_old=0
while x!=x_old:# and abs(f(x))>1e-15:
print(x,f(x),f_(x))
x_old=x
x=x-f(x)/f_(x)
print('===')
return x
precision=30
def b(b,e,f,eps=1e-17):
mold,m=0,1
while e-b>eps and mold!=m:
mold=m
m=(e+b)/2
if (l.m(m)>0)!=(l.m(b)>0):
e=m
else:
b=m
#print(*map(float,[b,e,f(b),f(e)]))
print('.',sep='',end='')
return b
##class Frac:
##    n=0
##    d=1
##    def __init__(self,n,d):
##        self.n=n
##        self.d=d
##    def
lcm=lambda x,y:x*y//gcd(x,y)
import math
from fractions import Fraction
degree=61
n=degree
l=Poly(sum(([c(n,i)*(-1)**(n-i),0] for i in range(n+1)),[]),
2**n*reduce(lambda x,y:x*y,range(1,n+1),1)).dn(n)
ld=l.dn()
#print(l.coeffs,l.d)
#r=[root(math.cos(math.pi*(4*i-1)/(4*n+2)),l.m,ld.m) for i in range(1,n+1)]
##d=2**8#4096
##while True:
##    i_=[(b,e) for b,e in [(Fraction(i,d)-1,Fraction(i+1,d)-1)
##                          for i in range(2*d)] if (l.m(b)>0)!=(l.m(e)>0)]
##    print(d,len(i_))
##    if len(i_)==n:
##        break
##    d*=2
##r=[b(*i,l.m,Fraction(1,10**precision)) for i in i_]
##print('\n',sep='',end='')
cm=1267650600228229401496703205376
#cm=reduce(lambda x,i:lcm(int((str(i).split('/')+[1])[1]),x),r,1);cm
#[str(i*cm) for i in r]
r=[Fraction(int(i),cm) for i in ['-1266681605106811426160584860697', '-1262547799267707740863710314936', '-1255122086390021131342967954015', '-1244422184933863703004183427391', '-1230475806835976018691325584397', '-1213319288075032955237836344344', '-1192997371846808208190052844742', '-1169563069074221666262667138623', '-1143077514053466175428649175963', '-1113609802945881973204995949479', '-1081236812728978305124643427497', '-1046043000289730543348383120049', '-1008120181921745998647915597624', '-967567293702390287305984778942', '-924490133333634877286205966159', '-879001084101412705885157895540', '-831218821664589525108846925845', '-781268004433877062952267018393', '-729278948345971539276192261648', '-675387286880006687633990914140', '-619733617202494872284783888651', '-562463133363442075271048079222', '-503725247500289525435137450347', '-443673200037730123766618583031', '-382463659900222496924931293872', '-320256315780122313563780137228', '-257213459527711400732339664476', '-193499562749975057795207525453', '-129280847722706546595631169560', '-64724853735360852478581550597', '0', '64724853735360852478581550596', '129280847722706546595631169559', '193499562749975057795207525452', '257213459527711400732339664475', '320256315780122313563780137227', '382463659900222496924931293871', '443673200037730123766618583030', '503725247500289525435137450346', '562463133363442075271048079221', '619733617202494872284783888650', '675387286880006687633990914139', '729278948345971539276192261647', '781268004433877062952267018392', '831218821664589525108846925844', '879001084101412705885157895539', '924490133333634877286205966158', '967567293702390287305984778941', '1008120181921745998647915597623', '1046043000289730543348383120048', '1081236812728978305124643427496', '1113609802945881973204995949478', '1143077514053466175428649175962', '1169563069074221666262667138622', '1192997371846808208190052844741', '1213319288075032955237836344343', '1230475806835976018691325584396', '1244422184933863703004183427390', '1255122086390021131342967954014', '1262547799267707740863710314935', '1266681605106811426160584860696']]
#r=[Fraction(-72002513042367095, 72057594037927936), Fraction(-35883766692782835, 36028797018963968), Fraction(-35672715239059617, 36028797018963968), Fraction(-141474422995755015, 144115188075855872), Fraction(-69944451686068079, 72057594037927936), Fraction(-137938433007892963, 144115188075855872), Fraction(-135628098615462501, 144115188075855872), Fraction(-66481963419518119, 72057594037927936), Fraction(-32488216960856169, 36028797018963968), Fraction(-126602776952709289, 144115188075855872), Fraction(-61461197033678349, 72057594037927936), Fraction(-59460660411885473, 72057594037927936), Fraction(-7163124720376473, 9007199254740992), Fraction(-109999665903886841, 144115188075855872), Fraction(-105102359763536115, 144115188075855872), Fraction(-99930853605361871, 144115188075855872), Fraction(-23624659822433771, 36028797018963968), Fraction(-88819888837164975, 144115188075855872), Fraction(-41454708727199633, 72057594037927936), Fraction(-19195661220692463, 36028797018963968), Fraction(-70455555169530143, 144115188075855872), Fraction(-63944654967081315, 144115188075855872), Fraction(-28633465234423143, 72057594037927936), Fraction(-6304975386764939, 18014398509481984), Fraction(-10870271009372985, 36028797018963968), Fraction(-36408927801417385, 144115188075855872), Fraction(-29241784833142379, 144115188075855872), Fraction(-21998353389560073, 144115188075855872), Fraction(-14697530755564293, 144115188075855872), Fraction(-7358363943167303, 144115188075855872), Fraction(0, 1), Fraction(3679181971583651, 72057594037927936), Fraction(3674382688891073, 36028797018963968), Fraction(2749794173695009, 18014398509481984), Fraction(14620892416571189, 72057594037927936), Fraction(4551115975177173, 18014398509481984), Fraction(43481084037491939, 144115188075855872), Fraction(50439803094119511, 144115188075855872), Fraction(57266930468846285, 144115188075855872), Fraction(31972327483540657, 72057594037927936), Fraction(35227777584765071, 72057594037927936), Fraction(76782644882769851, 144115188075855872), Fraction(82909417454399265, 144115188075855872), Fraction(44409944418582487, 72057594037927936), Fraction(94498639289735083, 144115188075855872), Fraction(49965426802680935, 72057594037927936), Fraction(52551179881768057, 72057594037927936), Fraction(13749958237985855, 18014398509481984), Fraction(114609995526023567, 144115188075855872), Fraction(118921320823770945, 144115188075855872), Fraction(122922394067356697, 144115188075855872), Fraction(15825347119088661, 18014398509481984), Fraction(129952867843424675, 144115188075855872), Fraction(132963926839036237, 144115188075855872), Fraction(33907024653865625, 36028797018963968), Fraction(68969216503946481, 72057594037927936), Fraction(139888903372136157, 144115188075855872), Fraction(70737211497877507, 72057594037927936), Fraction(142690860956238467, 144115188075855872), Fraction(143535066771131339, 144115188075855872), Fraction(144005026084734189, 144115188075855872)]
#r=[Fraction(-144104924687379029, 144115188075855872), Fraction(-144061113633417193, 144115188075855872), Fraction(-71991152681636629, 72057594037927936), Fraction(-143868501524294767, 144115188075855872), Fraction(-35929931908622757, 36028797018963968), Fraction(-143536019344824829, 144115188075855872), Fraction(-71658710569760757, 72057594037927936), Fraction(-143063986084641675, 144115188075855872), Fraction(-142775775751233995, 144115188075855872), Fraction(-35613215044775229, 36028797018963968), Fraction(-142095317851501427, 144115188075855872), Fraction(-141703235672459401, 144115188075855872), Fraction(-141276708943911307, 144115188075855872), Fraction(-70407920670826367, 72057594037927936), Fraction(-140320744889653401, 144115188075855872), Fraction(-34947884983137447, 36028797018963968), Fraction(-139228355106226913, 144115188075855872), Fraction(-138631327306447751, 144115188075855872), Fraction(-69000300827755959, 72057594037927936), Fraction(-68668165733468573, 72057594037927936), Fraction(-136638678208163739, 144115188075855872), Fraction(-135907811461286037, 144115188075855872), Fraction(-67571954440908669, 72057594037927936), Fraction(-134347156155496359, 144115188075855872), Fraction(-16689718369143059, 18014398509481984), Fraction(-132655882883583863, 144115188075855872), Fraction(-131761773444627501, 144115188075855872), Fraction(-65417817986076227, 72057594037927936), Fraction(-64938847793634331, 72057594037927936), Fraction(-32222046285398959, 36028797018963968), Fraction(-127867345160661637, 144115188075855872), Fraction(-31703855946358703, 36028797018963968), Fraction(-62866338356003711, 72057594037927936), Fraction(-15577420891180097, 18014398509481984), Fraction(-123475765655790143, 144115188075855872), Fraction(-61151075136161875, 72057594037927936), Fraction(-121098806255952059, 144115188075855872), Fraction(-119866026109883705, 144115188075855872), Fraction(-118604109492524957, 144115188075855872), Fraction(-117313363144639965, 144115188075855872), Fraction(-115994100814789491, 144115188075855872), Fraction(-28661660795766563, 36028797018963968), Fraction(-28317829445786347, 36028797018963968), Fraction(-111868458922669023, 144115188075855872), Fraction(-110438407601984251, 144115188075855872), Fraction(-27245377857813575, 36028797018963968), Fraction(-26874531136490757, 36028797018963968), Fraction(-105988607520833283, 144115188075855872), Fraction(-26113331820545011, 36028797018963968), Fraction(-102892657018719659, 144115188075855872), Fraction(-50653488045428427, 72057594037927936), Fraction(-49848334969235775, 72057594037927936), Fraction(-98062129987227927, 144115188075855872), Fraction(-96403753553428489, 144115188075855872), Fraction(-94721943747436259, 144115188075855872), Fraction(-23254277343922147, 36028797018963968), Fraction(-45644832420663197, 72057594037927936), Fraction(-89540030043462969, 144115188075855872), Fraction(-10971078784389609, 18014398509481984), Fraction(-85975896119833679, 144115188075855872), Fraction(-42081131673510875, 72057594037927936), Fraction(-20582043201506859, 36028797018963968), Fraction(-40237035159487741, 72057594037927936), Fraction(-39300203286200335, 72057594037927936), Fraction(-19176909251924261, 36028797018963968), Fraction(-74796221710411347, 144115188075855872), Fraction(-72866625298407583, 144115188075855872), Fraction(-35459658404464939, 72057594037927936), Fraction(-68954769584591107, 144115188075855872), Fraction(-33486730579157493, 72057594037927936), Fraction(-32487936568629803, 72057594037927936), Fraction(-31481245542875313, 72057594037927936), Fraction(-60933804407252565, 144115188075855872), Fraction(-58890306225406907, 144115188075855872), Fraction(-56832493264165915, 144115188075855872), Fraction(-54760865727051299, 144115188075855872), Fraction(-26337963587783547, 72057594037927936), Fraction(-25289092203398143, 72057594037927936), Fraction(-48468147330210951, 144115188075855872), Fraction(-46346328843725859, 144115188075855872), Fraction(-11053311177256411, 36028797018963968), Fraction(-21034706713097939, 72057594037927936), Fraction(-19957678053844251, 72057594037927936), Fraction(-4718949543956531, 18014398509481984), Fraction(-35578660114658865, 144115188075855872), Fraction(-33397075583856055, 144115188075855872), Fraction(-15603686524289111, 72057594037927936), Fraction(-29010084771446223, 144115188075855872), Fraction(-1675359053686717, 9007199254740992), Fraction(-24594889131807827, 144115188075855872), Fraction(-22378054994346827, 144115188075855872), Fraction(-20155781304247947, 144115188075855872), Fraction(-4482152060343901, 36028797018963968), Fraction(-15697077176510773, 144115188075855872), Fraction(-6730865269878557, 72057594037927936), Fraction(-5611555844344803, 72057594037927936), Fraction(-4490882388138869, 72057594037927936), Fraction(-6738234618615355, 144115188075855872), Fraction(-2246533281244659, 72057594037927936), Fraction(-2246806352819167, 144115188075855872), Fraction(0, 1), Fraction(1123403176409583, 72057594037927936), Fraction(4493066562489317, 144115188075855872), Fraction(3369117309307677, 72057594037927936), Fraction(8981764776277737, 144115188075855872), Fraction(11223111688689605, 144115188075855872), Fraction(13461730539757113, 144115188075855872), Fraction(3924269294127693, 36028797018963968), Fraction(17928608241375603, 144115188075855872), Fraction(10077890652123973, 72057594037927936), Fraction(11189027497173413, 72057594037927936), Fraction(12297444565903913, 72057594037927936), Fraction(26805744858987471, 144115188075855872), Fraction(14505042385723111, 72057594037927936), Fraction(31207373048578221, 144115188075855872), Fraction(16698537791928027, 72057594037927936), Fraction(2223666257166179, 9007199254740992), Fraction(37751596351652247, 144115188075855872), Fraction(39915356107688501, 144115188075855872), Fraction(42069413426195877, 144115188075855872), Fraction(44213244709025643, 144115188075855872), Fraction(23173164421862929, 72057594037927936), Fraction(24234073665105475, 72057594037927936), Fraction(50578184406796285, 144115188075855872), Fraction(52675927175567093, 144115188075855872), Fraction(27380432863525649, 72057594037927936), Fraction(28416246632082957, 72057594037927936), Fraction(29445153112703453, 72057594037927936), Fraction(15233451101813141, 36028797018963968), Fraction(62962491085750625, 144115188075855872), Fraction(64975873137259605, 144115188075855872), Fraction(66973461158314985, 144115188075855872), Fraction(34477384792295553, 72057594037927936), Fraction(70919316808929877, 144115188075855872), Fraction(36433312649203791, 72057594037927936), Fraction(37398110855205673, 72057594037927936), Fraction(76707637007697043, 144115188075855872), Fraction(78600406572400669, 144115188075855872), Fraction(80474070318975481, 144115188075855872), Fraction(82328172806027435, 144115188075855872), Fraction(84162263347021749, 144115188075855872), Fraction(42987948059916839, 72057594037927936), Fraction(87768630275116871, 144115188075855872), Fraction(11192503755432871, 18014398509481984), Fraction(91289664841326393, 144115188075855872), Fraction(93017109375688587, 144115188075855872), Fraction(47360971873718129, 72057594037927936), Fraction(12050469194178561, 18014398509481984), Fraction(49031064993613963, 72057594037927936), Fraction(99696669938471549, 144115188075855872), Fraction(101306976090856853, 144115188075855872), Fraction(51446328509359829, 72057594037927936), Fraction(104453327282180043, 144115188075855872), Fraction(52994303760416641, 72057594037927936), Fraction(107498124545963027, 144115188075855872), Fraction(108981511431254299, 144115188075855872), Fraction(55219203800992125, 72057594037927936), Fraction(55934229461334511, 72057594037927936), Fraction(113271317783145387, 144115188075855872), Fraction(114646643183066251, 144115188075855872), Fraction(57997050407394745, 72057594037927936), Fraction(29328340786159991, 36028797018963968), Fraction(29651027373131239, 36028797018963968), Fraction(14983253263735463, 18014398509481984), Fraction(60549403127976029, 72057594037927936), Fraction(122302150272323749, 144115188075855872), Fraction(61737882827895071, 72057594037927936), Fraction(124619367129440775, 144115188075855872), Fraction(125732676712007421, 144115188075855872), Fraction(126815423785434811, 144115188075855872), Fraction(31966836290165409, 36028797018963968), Fraction(128888185141595835, 144115188075855872), Fraction(129877695587268661, 144115188075855872), Fraction(130835635972152453, 144115188075855872), Fraction(32940443361156875, 36028797018963968), Fraction(66327941441791931, 72057594037927936), Fraction(133517746953144471, 144115188075855872), Fraction(67173578077748179, 72057594037927936), Fraction(135143908881817337, 144115188075855872), Fraction(33976952865321509, 36028797018963968), Fraction(68319339104081869, 72057594037927936), Fraction(137336331466937145, 144115188075855872), Fraction(138000601655511917, 144115188075855872), Fraction(69315663653223875, 72057594037927936), Fraction(4350886097069591, 4503599627370496), Fraction(139791539932549787, 144115188075855872), Fraction(17540093111206675, 18014398509481984), Fraction(140815841341652733, 144115188075855872), Fraction(70638354471955653, 72057594037927936), Fraction(17712904459057425, 18014398509481984), Fraction(71047658925750713, 72057594037927936), Fraction(142452860179100915, 144115188075855872), Fraction(71387887875616997, 72057594037927936), Fraction(71531993042320837, 72057594037927936), Fraction(143317421139521513, 144115188075855872), Fraction(35884004836206207, 36028797018963968), Fraction(143719727634491027, 144115188075855872), Fraction(71934250762147383, 72057594037927936), Fraction(143982305363273257, 144115188075855872), Fraction(18007639204177149, 18014398509481984), Fraction(36026231171844757, 36028797018963968)]
w=[2/((1-x**2)*(ld.m(x))**2) for x in r]
from decimal import Decimal,Context,setcontext,ROUND_HALF_DOWN
setcontext(Context(prec=30, rounding=ROUND_HALF_DOWN))
float=lambda r:(lambda x:Decimal(x[0])/Decimal(x[1]))(str(r).split('/'))
d=2**8
s='{120.000000000000000000000000000,9.26052826812554737559992621571,3.32335097044784255118406403126,2.00000000000000000000000000000,1.50457548825155601882809780906,1.26582350605728327736200759570,1.13300309631934634747833911121,1.05218372089129332722399903882,1.00000000000000000000000000000}'
l=list(map(Decimal,s[1:-1].split(',')))
gamma=Decimal('0.5772156649015328606065120900824024310421593359399235988057672348848677267776646709369470632917467495')
for (c,c_1,real),l_ in zip([
(Decimal('0.1')*i,Decimal('10')/i,10/i) for i in range(2,11)],l):
#let t=e^{-x^c},\ln t=-x^c,x=(-\ln t)^{\frac1c}
#t(0)=1,t(\infty)=0
#dt=-e^{-x^c}\cdot x^{c-1}\cdot c
#f(x)=e^{-x^c}\cdot x^{c-1}\cdot c=t\cdot (-\ln t)^{\frac{c-1}c}\cdot c
xt=lambda t:(-((t).ln()))**(c_1)
##    f=lambda x:c*(x**(c-1))#*(math.exp(-(x**c)))
fln=lambda t:((-((t).ln())).ln())#+(t.ln())
##    g=lambda x:Decimal('0.5')+5*((-9*x).exp())#*math.exp(-x)
g=lambda t:Decimal('0.5')+5*((-9*xt(t)).exp())#*math.exp(-x)
##    h=lambda x:f(x)*((-(x**c)).exp())*\
##       ((f(x)).ln()-(x**c)-((g(x)).ln()-x))
##    #0,inf->0,1: x=t/(t-1), xt-x=t, (x-1)t=x, t=x/(x-1)
##    h1=lambda t:h(t/(1-t))/((1-t)**2) # 0,1
##    h2=lambda v:h1((v+1)/2)/2
h1=lambda t:g(t).ln()#f(t).ln()-((g(t).ln())-xt(t))
int_=lambda h1:sum(
sum(float(w_)*h1(float(((x+1)/2+i)/d))/(2*d) for w_,x in zip(w,r))
for i in range(d))
print(c,Decimal(math.gamma(real+1))+(c.ln())-int_(h1)-gamma*(1-c_1)-1)
h1_=r'''0.2 0.476044892154085268532810576285
0.3 0.326045243249124607083205564208
0.4 0.194544295792217104304557844994
0.5 0.0805857800502670858642542661562
0.6 -0.0176136562407463311404194282348
0.7 -0.102042234943715186336125809639
0.8 -0.174612380676135222634071174702
0.9 -0.237048749195042896134197564780
1.0 -0.290856610889057599214369459143'''
x=[i.split(': ') for i in r'''c = 0.2. Output: 119.2233798550179
c = 0.3. Outout: 8.077346771987397
c = 0.4. Output: 2.07833944013377
c = 0.5. Output: 0.8034827042913205
c = 0.6. Output: 0.3961739639940003
c = 0.7. Output: 0.2585689391629249
c = 0.8. Output: 0.2287758419066705
c = 0.9. Output: 0.2480070283065089
c = 1.0. Output: 0.2908566108890576'''.split('\n')]
y=[i.split(' ') for i in r'''0.2 119.223379855017945799292478451
0.3 8.07734677198740392689540304712
0.4 2.07833944013376923348715605315
0.5 0.80348270429132046532502570246
0.6 0.39617396399400051391328713871
0.7 0.25856893916292307961043114604
0.8 0.22877584190665494099938058972
0.9 0.24800702830645810748432393277
1.0 0.29085661088905759921436945914'''.split('\n')]
#[(lambda l:(l and min(l)) or 0)([n for n,(d1,d2) in enumerate(zip(i[1],j[1])) if d1!=d2]) for i,j in zip(x,y)]
#[0, 14, 15, 17, 17, 16, 15, 14, 17]


# Pari/GP, 88 87 bytes, < 1 sec

Great work from @AlexeyBurdin!

I also saw the "provocative" way in which $$\f\$$ was a derivative of an easier function, but only had the idea to use this for partial integration, without success.

For golfing reasons I recombined all the logarithms, except the one giving $$\-1\$$ because it makes no difference for the code length and subtracting one seemed slightly more efficient than multiplying the integrand by $$\t\$$. I also took the logarithms of $$\1/t\$$ instead of $$\t\$$ to be able to get rid of the otherwise needed parentheses for the minus sign. The 2 in the last argument to intnum reduces the integration step size. That is needed to achieve the required precision.

l=log;i(c)=intnum(t=0,1,l(c*l(1/t)^(1-1/c)/(.5+5*exp(-9*l(1/t)^c^-1)))+l(1/t)^c^-1,2)-1


Try it online!

• This is pretty awesome. Thank you.
– user9207
Nov 6, 2019 at 16:48

# Python 3, 150133 122 bytes

(thanks to Delta and Alexey Burdin for further golfing this)

SymPy's mpmath module is a great way to handle this problem. All math functions (exp, log, euler, gamma and quad for integration) are run in high precision.

lambda c:gamma(1/c+1)+l(c)-euler*(1-1/c)-1-quad(lambda t:l(.5+5*exp(-9*((-l(t))**(1/c)))),[0,1])
from mpmath import*
l=log


Try it online!

• This is great. Thank you.
– user9207
Nov 7, 2019 at 7:21
• Try it online! you can save some bytes if you use a lamda instead of the input() Nov 7, 2019 at 13:48
• @Delta thanks! I thought we needed to output a fixed number of digits, thus I hang on to the nprint, but indeed that is not required. That saves 17 bytes. Nice one. Nov 7, 2019 at 21:00
• 128 bytes Nov 8, 2019 at 23:51
• @AlexeyBurdin thanks! I further golfed it to 122 bytes. Nov 9, 2019 at 12:04

# Wolfram Language (Mathematica), 105 bytes ~7.5 seconds

Mathematica is good at automatically determining the best method needed to perform the integration. Setting the working precision high gets us to results that match the stated answer. The time stated is to perform all the example integrals.

NIntegrate[((f=#*x^(#-1))*Log[f/(E^x^#*(5/E^(10*x)+0.5/E^x))])/E^x^#,{x,0,-Log@0},WorkingPrecision->128]&


Try it online!

• At least**!** ))) Nov 5, 2019 at 15:27
• At least someone posted wolfram mathematica solution. The language is perfect for numerical integration) or for all math) Nov 5, 2019 at 15:51
• 62 bytes. In Mathematica, precision is specified in decimal digits. MachinePrecision offers 15.9 decimal digits (53 bits) of precision by default on TIO, which is sufficient for the stated requirement.
– att
Nov 6, 2019 at 8:12
• @Anush You only required accuracy to within +/-10^-10. 119.22337985501592 is sufficiently close.
– att
Nov 6, 2019 at 22:43
• @attinat, I think you've got some good edits, but I'm not convinced that dropping the WorkingPrecision option works when c=0.8. As the problem states, 10^-10% implies a relative error of 10^-12 or less, but the relative difference between your answer and the stated correct answer is 7.5*10^-11 which is just a tad bit too much error. I figured that was a corner case left in by the questioner to force the use of extra precision. Nov 7, 2019 at 22:48

# Desmos, 62 59 bytes

a=ln(1/t)^{1/c}
f(c)=∫_0^1(ln(a^c/a/e(.5+5e^{-9a})c)+a)dt


This is a port of Christian Siever's solution with some modifications for numeric precision.

Try it on Desmos!

-3 bytes thanks to Aiden Chow.

The final column in the table shows the relative error, which reaches 5.71×10^-13 as the worst case at c=0.2. This barely slips under the 10^-12 relative error requirement.

As for the speed requirement, this run in under 1500ms on my computer.

## Precision notes

Desmos uses 64-bit floats for everything, so it's tough to get the desired precision, considering the machine precision is on the order of 10^-16. However, Desmos's quadrature routine is pretty good for general integrals because it works by recursively splitting the integration interval when it needs more precision.

ln(1/t) can be changed to (-lnt) for -1 byte, but this costs about two orders of magnitude of relative precision.

• I try absorbing the -1 into the ln and it seem to work for 59 bytes. One of the relative errors is negative, but its absolute value is under $10^{-12}$ so I think it's fine. Try It On Desmos!. This runs anywhere from 1500ms to 2000ms on my computer. May 7, 2022 at 22:33
• @AidenChow Nice observation! Putting the -1 into the integral on its own doesn't work, but putting it inside the ln as ln(1/e) (as you did) does work May 8, 2022 at 4:09

# APL(NARS), 929 chars, time 1' 7''

⎕fpc←128⋄grv←gerr←0⋄ge←q←0v⋄funOne←{f←q×(⍵*q-1v)×*-⍵*q⋄f×⍟f÷(0.5v×*-⍵)+5×*-10v×⍵}

r←(f S)w
gerr←1⋄r←¯1J1⋄→0×⍳3≠≢w⋄→0×⍳w[3]>⎕fpc÷3.322⋄gerr←grv←0⋄ge←÷10v*w[3]⋄r←f It w[1]w[2]
→0×⍳∼gerr≠0⋄r←¯1J1

r←(f I)w;a;b;n;nn;h;s1;s2;h1;h2;i;t;p;m
r←⍬⋄→0×⍳gerr≠0⋄h2←h1←s2←s1←0v⋄(a b n)←w⋄→2×⍳n≤0⋄h←(2×nn←2v×n)÷⍨b-a⋄i←0v⋄→3
r←⍬⋄gerr←1⋄→0
t←f a+h×2×i+←1⋄s1+←t⋄→4×⍳∼0=2∣i⋄h1+←t⋄→5
h2+←t
→3×⍳i<nn-1⋄i←0v⋄→2×⍳∞=∣s1
s2+←f a+h×¯1+2×i+←1⋄→6×⍳i<nn⋄→2×⍳∞=∣s2⋄p←f¨a b⋄→2×⍳∞=+/∣p⋄t←+/p
p←(3÷⍨h×t+(4×s2)+2×s1)(3÷⍨h×2×t+(4×h2)+2×h1)
m←(h×s2+s1+t÷2)(h× 2×s1+t÷2)⋄→2×⍳(∞=∣m[1])∨∞=∣p[1]
s2←∣-/p⋄s1←∣-/m⋄→10×⍳∼s2>s1⋄r←m,s1⋄→0
r←p,s2

r←(f It)w;t;y;m
r←0v⋄→2×⍳0≠gerr⋄→3×⍳∼grv>250⋄gerr←1
r←0v⋄→0
grv+←1⋄t←f I w,1⋄→2×⍳0≠gerr⋄→6×⍳t[3]≥1⌊ge×1E5⋄→5×⍳ge>t[3]⋄t←f I w,19⋄→2×⍳0≠gerr
→6×⍳∼ge>t[3]
y←f I w,23⋄→2×⍳0≠gerr⋄→6×⍳ge≤∣y[1]-t[1]⋄grv-←1⋄r←↑y⋄→0
m←2v÷⍨-/⌽w⋄r←(f It w[1],w[1]+m)+(f It(w[1]+m),w[2])⋄grv-←1

:for q :in 0.1v+10÷⍨⍳9⋄q,(11⍕ funOne S 1E¯64v (7E8v) 11)⋄:endfor


82+ 8+82+18+2 +39+74+13+40+5+25+63+44+50+37+6+10+ 15+35+7+79+12+54+58+6 +65=929

"grv" is one global variable used for make not too deep the recursive function

"gerr" is one global variable used for find some possible error

"ge" is one global variable used for store epsilon of precision

"funOne" would be the function use in exercise where the "c" is here "q".

"S" is the function to call for integration of one continue function in one interval, it has arguments on the left the function, in the right 3 numbers, 2 are a b for the integration interval [a, b], and one for the precision of calculation. It is possible that S function fail (return ¯1J1 in that case) or not show the error in some integral even if the number result is wrong (example some integral of periodic functions). It is possible S give the wrong value to the function that stop the computation with "DOMAIN ERROR".

"I" is a function for calculate the integral in 2 differents way the trapezes and Simpson formula, return the two sums have less difference.

"It" is a recursive function from the model of qsort, that use function "I" and change partition for recalculate the result each little interval it find ok for to be a little more sure of result even if that means 2x slow (so it is possible one silent fail).

  :for q :in 0.1v+10÷⍨⍳9⋄q,(11⍕ funOne S 1E¯64v (7E8v) 11)⋄:endfor
0.2  119.22337985502
0.3  8.07734677200
0.4  2.07833944015
0.5  0.80348270430
0.6  0.39617396400
0.7  0.25856893917
0.8  0.22877584191
0.9  0.24800702831
1  0.29085661089


1E¯64v would be one value not 0 sufficient little, the value for +oo that not make wrong results in function funOne, it seems to be 7E8v.It gets 1' 7'' for all value of c (here it is q variable) from 0.1 to 1. The function integrate of sys (∫) is more fast, but the right result it seems to be only the first

     q←0.2v⋄30⍕  1E¯70v funOne ∫7E8v
119.223379855016584445335168065350
q←0.3v⋄30⍕  1E¯70v funOne ∫7E8v
0.253368882611387716254867856476
q←0.4v⋄30⍕  1E¯70v funOne ∫7E8v
0.016195181203412657409325335505
q←1v⋄30⍕  1E¯70v funOne ∫7E8v
0.001677768739402358961260938889


For some function f, +/f¨a b is not the same +/f a b. One has to eliminate all the places where is used "=" or use ⎕ct←0 in each function use "=".

In the function "I" these are the 4 sums, the first (p[1]) the more precise Simpson sum, than the second (p[2]) less precise Simpson the 3rd the more precise use trapezes (m[1]), the less precise use trapezes (m[2]).

p←(3÷⍨h×t+(4×s2)+2×s1)(3÷⍨h×2×t+(4×h2)+2×h1)
m←(h×s2+s1+t÷2)(h× 2×s1+t÷2)


# C (gcc), 1933 bytes, 3''

    #include <quadmath.h>
#include <stdio.h>
#define C signed char
#define U unsigned
#define I int
#define LD __float128
#define G goto
#define P printf
#define R return
#define FM FLT128_MAX
I PP(C*a,C*fmt,LD v){I r;C arr[1010];arr[0]=0;r=quadmath_snprintf(arr,1000,fmt,v);if(r<0)R r;R P("%s%s",a,arr);}
I grv, gerr; LD ge,t[3],y[3],gq=0.2q;I Er(LD x){if(x<0)R-x>=FM;R x>=FM;}
LD*In(LD*rf,LD(*f)(LD),LD a,LD b,U n){LD  h,h1,h2,s1,s2,t,x,p[2],m[2],r;U nn,i;if(gerr)R rf;h2=h1=s2=s1=0;if(n==0||n>1e6){L0:gerr=1;R rf;}nn=2*n;h=(b-a)/(2*nn);i=0;L1:++i;s1+=t=f(a+h*2*i);if(Er(t))G L0;if(i%2==0)h1+=t;else h2+=t;if(i<nn-1)G L1;i=0;if(Er(s1))G L0;L2:++i;s2+=t=f(a+h*(2*i-1));if(Er(t))G L0;if(i<nn)G L2;if(Er(s2))G L0;t=f(a);if(Er(t))G L0;x=f(b);if(Er(x))G L0;x=f(b);t+=x;p[0]=h*(t+4*s2+2*s1)/3;p[1]=2*h*(t+4*h2+2*h1)/3;m[0]=h*(s2+s1+t/2);m[1]=2*h*(s1+t/2);if(Er(p[0])||Er(m[0]))G L0;s1=fabsq(p[0]-p[1]);s2=fabsq(m[0]-m[1]);if(s1<s2){rf[0]=p[0];rf[1]=p[1];rf[2]=s1;}else{rf[0]=m[0];rf[1]=m[1];rf[2]=s2;}R rf;}
LD It(LD(*f)(LD),LD a,LD b){LD r,m;if(gerr){L3: r=0;R r;}if(grv>250){gerr=1;G L3;}++grv;In(t,f,a,b,1);if(gerr)G L3;if(t[2]>=1||t[2]>=ge*1e5)G L6;if(ge>t[2])G L4;In(t,f,a,b,19);if(gerr)G L3;if(ge>t[2]){L4:In(y,f,a,b,23);if(gerr)G L3;if(ge<=fabsq(y[0]-t[0]))G L6;--grv;R y[0];}L6:m=(b-a)/2;r=It(f,a,a+m)+It(f,a+m,b);--grv;R r;}
LD Si(LD(*f)(LD),LD a,LD b,U p){LD r; gerr=1;if(p>38)R FM;gerr=grv=0;ge=powq(10,-(LD)p);r=It(f,a,b);R gerr?FM:r;}
LD fN(LD x){LD f,h,z;f=logq(gq)+(gq-1.0q)*logq(x)-powq(x,gq)+x-logq(0.5q+expq(-9.0q*x));return gq*powq(x,gq-1.0q)*expq(-powq(x,gq))*f;}
LD fO(LD w){LD f=gq*powq(w,gq-1.0q)*expq(-powq(w,gq)),z;z=0.5q*expq(-w)+5.0q*expq(-10.0q*w);R f*logq(f/z);}
LD funOne(LD x){R x<1000?fO(x):fN(x);}
I main(){LD r[3],q,i,f,a,b;for(gq=0.2q;gq<=1;gq+=0.1q){q=Si(funOne,1e-64,7e8,11);PP("","%.2Qf",gq);PP(", ","%.30Qf",q);P("\n");}R 0;}



Try it online!

ungolfed:

// > gcc this.c -lquadmath
#include <stdio.h>

#define C  signed char
#define U  unsigned
#define I  int
#define LD __float128
#define G   goto
#define P   printf
#define R   return
#define FM  FLT128_MAX

I PP(C*a,C*fmt,LD v)
{I r;C arr[1010];
R P("%s%s",a,arr);
}

I grv, gerr; LD ge,t[3],y[3],gq=0.2q;

I Er(LD x){if(x<0)R-x>=FM;R x>=FM;}

LD*In(LD*rf,LD(*f)(LD),LD a,LD b,U n)
{LD  h,h1,h2,s1,s2,t,x,p[2],m[2],r;U nn,i;
if(gerr)R rf;
h2=h1=s2=s1=0;if(n==0||n>1e6){L0:gerr=1;R rf;}
nn=2*n;h=(b-a)/(2*nn);i=0;
L1: ++i;s1+=t=f(a+h*2*i);if(Er(t))G L0;if(i%2==0)h1+=t;else h2+=t;if(i<nn-1)G L1;i=0;if(Er(s1))G L0;
L2: ++i;s2+=t=f(a+h*(2*i-1));if(Er(t))G L0;if(i<nn)G L2;if(Er(s2))G L0;t=f(a);if(Er(t))G L0;x=f(b);if(Er(x))G L0;x=f(b);t+=x;
p[0]=h*(t+4*s2+2*s1)/3;p[1]=2*h*(t+4*h2+2*h1)/3;
m[0]=h*(s2+s1+t/2);m[1]=2*h*(s1+t/2);
if(Er(p[0])||Er(m[0]))G L0;
s1=fabsq(p[0]-p[1]);s2=fabsq(m[0]-m[1]);
if(s1<s2){rf[0]=p[0];rf[1]=p[1];rf[2]=s1;}
else     {rf[0]=m[0];rf[1]=m[1];rf[2]=s2;}
R rf;
}

LD It(LD(*f)(LD),LD a,LD b)
{LD r,m;
if(gerr){L3:  r=0;  R r;}
if(grv>250){gerr=1;G L3;}
++grv;In(t,f,a,b,1);if(gerr)G L3;if(t[2]>=1||t[2]>=ge*1e5)G L6;if(ge>t[2])G L4;
In(t,f,a,b,19);if(gerr)G L3;
if(ge>t[2])
{L4:In(y,f,a,b,23);if(gerr)G L3;if(ge<=fabsq(y[0]-t[0]))G L6;--grv;R y[0];}
L6: m=(b-a)/2;r=It(f,a,a+m)+It(f,a+m,b);--grv;
R r;
}

LD Si(LD(*f)(LD),LD a,LD b,U p){LD r; gerr=1;if(p>38)R FM;gerr=grv=0;ge=powq(10,-(LD)p);r=It(f,a,b);R gerr?FM:r;}

LD fN(LD x)
{LD f,h,z;
f=logq(gq)+(gq-1.0q)*logq(x)-powq(x,gq)+x-logq(0.5q+expq(-9.0q*x));
return gq*powq(x,gq-1.0q)*expq(-powq(x,gq))*f;
}
LD fO(LD w)
{LD f=gq*powq(w,gq-1.0q)*expq(-powq(w,gq)),z;
z=0.5q*expq(-w)+5.0q*expq(-10.0q*w);
R f*logq(f/z);
}
LD funOne(LD x){R x<1000?fO(x):fN(x);}

I main()
{LD r[3],q,i,f,a,b;
for(gq=0.2q;gq<=1;gq+=0.1q)
{q=Si(funOne,1e-64,7e8,11);
PP("","%.2Qf",gq);PP(", ","%.30Qf",q);P("\n");
}
R 0;
}


This would be one port of APL NARS answer to C....

"grv" is one global variable used for make not too deep the recursive function

"gerr" is one global variable used for find some possible error

"ge" is one global variable used for store epsilon of precision

"funOne" would be the function use in exercise where the "c" is here "q".

"Si" is the function to call for integration of one continue function in one interval, it has arguments on the left the function, in the right 3 numbers, 2 are a b for the integration interval [a, b], and one for the precision of calculation. It is possible that Si function fail (return FLT128_MAX in that case) or not show the error in some integral even if the number result is wrong (example some integral of periodic functions). It is possible S give the wrong value to the function that stop the computation.

"In" is a function for calculate the integral in 2 differents way the trapezes and Simpson formula, return the two sums have less difference.

"It" is a recursive function from the model of qsort, that use function "In" and change partition for recalculate the result each little interval it find ok for to be a little more sure of result even if that means 2x slow (so it is possible one silent fail).

0.20, 119.223379855017788373126945485560
0.30, 8.077346772003659028828166522766
0.40, 2.078339440146599411449845152080
0.50, 0.803482704300521900602181696767
0.60, 0.396173964001614436175399637559
0.70, 0.258568939171353519263491445983
0.80, 0.228775841912709730320218436446
0.90, 0.248007028314673262106569598459
1.00, 0.290856610889722950747717956860


1E¯64 would be one value not 0 sufficient little, the value for +oo that not make wrong results in function funOne, it seems to be 7E8.It gets 3'' for all value of c (here it is q variable) from 0.1 to 1.

There was some problem but one see the different behaviors of functions one know what to do: Cut and paste 2 function in 1...

fO                fN
args: 1..10*114000
inf, 7.146180987711778877830383715838e-05
inf, 1.783936268160517786280124237027e-05
inf, 7.145191389859739510802225203435e-06
inf, 3.548880901445954254961267159123e-06
inf, 1.999831511094571465956022639528e-06
inf, 1.225727400360527391484147042523e-06
inf, 7.980629270833437955446020180540e-07
inf, 5.439501158831436480969842592496e-07
inf, 3.843424608223386417573364622201e-07

fO                                             fN
args: 1..10*10
3.675246391455555827226018131679e-02, 3.675246391455555827226018131679e-02
4.382838598960554328826955735253e-02, 4.382838598960554328826955735253e-02
4.457247966758306754921782013739e-02, 4.457247966758306754921782013739e-02
4.397010507150066869496325771966e-02, 4.397010507150066869496325771966e-02
4.298551944407423573690051874755e-02, 4.298551944407423573690051874755e-02
4.190240448327716144966191450276e-02, 4.190240448327716144966191450276e-02
4.082071320292452338988660547692e-02, 4.082071320292452338988660547692e-02
3.977794012002491268811702806721e-02, 3.977794012002491268811702806721e-02
3.878734785898533961995102431946e-02, 3.878734785898533961995102431946e-02

fO                                           fN
args: 1/1..10
-8.243106934757116432656018641562e-01, -1.344494569228624722492452318334e-01
-1.290429743446295599947997195038e+00, 2.015992534869343736972957346392e-01
-1.460826278910182993294146550044e+00, 7.738815018537929006207034760483e-01
-1.455905510003998243340646950725e+00, 1.487514582923200556163817499455e+00
-1.330519622929357039421233297650e+00, 2.297630406711373536179161098540e+00
-1.115252743630621385523306314288e+00, 3.179047478956116831216159688529e+00
-8.293358937016342749872510061477e-01, 4.115949553797663332308038109261e+00
-4.858491347654902257043425759574e-01, 5.097618989802348559994390744177e+00
-9.419553179254290987076726693566e-02, 6.116387165599517173497262275172e+00

• An heroic effort!
– Simd
Feb 18 at 11:48
• @Simd if one has 2 or more functions suppose be equal or two implementation same algo, is not so difficult to debug for find the differences. i found that the type and the value of constants used, are fundamental if one want more than 8 digits of precision Feb 18 at 14:56
• 1123 bytes Mar 3 at 6:40