Name: Vladyslav Shtabovenko (email_not_shown)
Date: 10/17/16-02:10:20 AM Z


Dear Feng,

I believe that with the latest commit to the development version the problem should be resolved. For

exp = (mc + GSD[k1]).GAD[c[li3]].(mc + GSD[k1 + k2 + p - q]).GAD[
    c[ga]].(-mc + GSD[p] - GSD[q]).GA[5].(mc + GSD[p] + GSD[q]).GAD[
    c[li2]].(-mc + GSD[k2]).GAD[li2].(mc + GSD[p] + GSD[q]).GA[
    5].(-mc + GSD[p] - GSD[q]).GAD[ga].(mc - GSD[k2 + 2 p]).GAD[li3] //FCI

res = AbsoluteTiming[DiracTrace[exp, DiracTraceEvaluate -> True]];
res // First

I obtain 33.27 seconds on my Thinkpad X230 and Mathematica 11.0

Running the same code with FeynCalc 4.2.0 on Mathematica 5.2 the timing is

118.77 seconds.

So FeynCalc 9.1 is now factor 3.5 faster.

If one simplifies the calculation by removing the c-Head:

exp2 = (mc + GSD[k1]).GAD[li3].(mc + GSD[k1 + k2 + p - q]).GAD[
    ga].(-mc + GSD[p] - GSD[q]).GA[5].(mc + GSD[p] + GSD[q]).GAD[
    li2].(-mc + GSD[k2]).GAD[li2].(mc + GSD[p] + GSD[q]).GA[
    5].(-mc + GSD[p] - GSD[q]).GAD[ga].(mc - GSD[k2 + 2 p]).GAD[li3] //
   FCI

res = AbsoluteTiming[DiracTrace[exp2, DiracTraceEvaluate -> True]];
res // First

then FeynCalc 9.1 needs only 2.08 seconds, while FeynCalc 4.2.0 requires 5.96.

Cheers
Vladyslav

>
>
> Dear All,
>
> I have found that The Tr operation is terriblely slow in FeynCalc 6.0, while the same code run very faster with fc5beta3.
>
> Here is my testing code, and has anyone else experienced the same problem?
>
> ===========================
> << HighEnergyPhysics`fc`
>
> exp = (mc + GSD[k1]).GAD[c[li3]].(mc + GSD[k1 + k2 + p - q]).GAD[
> c[\\\\\\\\\\\\\\\\[Gamma]]].(-mc + GSD[p] - GSD[q]).GA[
> 5].(mc + GSD[p] + GSD[q]).GAD[c[li2]].(-mc + GSD[k2]).GAD[
> li2].(mc + GSD[p] + GSD[q]).GA[
> 5].(-mc + GSD[p] - GSD[q]).GAD[\\\\\\\\\\\\\\\\[Gamma]].(mc - GSD[k2 + 2 p])
> .GAD[li3] // FCI
>
> (*$VeryVerbose=5;*)
>
> DateString[]
> Tr[exp]
> DateString[]



This archive was generated by hypermail 2b29 : 08/19/17-02:00:02 AM Z CEST