MathGroup Archive 2007

[Date Index] [Thread Index] [Author Index]

Search the Archive

5.2 vs 4.0

I know that definite integration convergence tests in versions 5.X try
"harder" and this fact may cause some definite integrals to be
evaluated in more time than in previous version but even in view of
this fact I can't understand the following difference in the time-
performances of 5.2 and 4.0 in the following definite integral

(*version 5.2*)
Timing[Integrate[x^2*Sin[x]^4*Cos[x]^3*Log[x], {x, 0, Pi}]]
    Second, -((152*Pi*(1 + 2*Log[Pi]))/3675) + (3*SinIntegral[Pi])/
      32 - (1/288)*SinIntegral[3*Pi] - SinIntegral[5*Pi]/4000 +


(*version 4.0*)
Timing[Integrate[x^2*Sin[x]^4*Cos[x]^3*Log[x], {x, 0, Pi}]]
{2.141*Second, (-510720*Pi - 1021440*Pi*Log[Pi] +
1157625*SinIntegral[Pi] -
        42875*SinIntegral[3*Pi] - 3087*SinIntegral[5*Pi] +


(*numerical check, both 5.2 and 4.0*)
NIntegrate[x^2*Sin[x]^4*Cos[x]^3*Log[x], {x, 0, Pi}]

Any ideas/comments?


  • Prev by Date: Re: Passing a list as seperate parameters?
  • Next by Date: Would someone confirm they also see this problem?
  • Previous by thread: Sparse Arrays question
  • Next by thread: Would someone confirm they also see this problem?