Re: using built in compiler
- To: mathgroup at smc.vnet.net
- Subject: [mg128339] Re: using built in compiler
- From: Chris <kristoph.steikert at gmail.com>
- Date: Mon, 8 Oct 2012 02:31:04 -0400 (EDT)
- Delivered-to: l-mathgroup@mail-archive0.wolfram.com
- Delivered-to: l-mathgroup@wolfram.com
- Delivered-to: mathgroup-newout@smc.vnet.net
- Delivered-to: mathgroup-newsend@smc.vnet.net
- References: <k3jo30$2cj$1@smc.vnet.net>
Am Samstag, 22. September 2012 09:08:33 UTC+2 schrieb (unbekannt):
> dear esteemed mathematica experts,
>
> hello there & greetings from malaysia.
>
> i have been trying to use the compiler on a module to see whether it speeds up the execution or not. kindly see below.
>
> unfortunately, mathematica returns a couple of error messages.
>
> can anybody help ? will appreciate any assistance from this forum.
>
>
>
> cheers
>
> rajesh
>
>
>
> fC := Compile[{{l, _Real}, {nsteps, _Integer}, {En, _Real} },
>
> Module[{meV, \[Angstrom], \[Sigma], \[Mu], \[Lambda], k, s,
>
> scat, \[Delta]},
>
> meV = 1./(27.3*1000.);
>
> en = En*meV; \[Angstrom] = 1./0.52; \[Sigma] =
>
> 3.57 \[Angstrom]; \[Epsilon] = 5.9 meV;
>
> V[r_] = \[Epsilon] ((\[Sigma]/r)^12 - 2 (\[Sigma]/r)^6);
>
>
>
> \[Mu] = (mH mKr)/(mH + mKr) /. {mH -> 1836.0, mKr -> 1836.0*136.0};
>
> (*Print[\[Mu]];*)
>
> \[Mu] // N;
>
> \[Lambda] = 2 \[Pi]/Sqrt[2.0 \[Mu] en] // N; k = 2 \[Pi]/\[Lambda];
>
> (*Print["\[Lambda]= ",\[Lambda]]*);
>
> s = 5. \[Lambda]/nsteps;
>
> x = {0.5 \[Sigma], 0.5 \[Sigma] + s}; \[Psi] = {0.0, 0.0001};
>
> F[r_] = 2. \[Mu]1 (V[r] + 1/(2. \[Mu]1 r^2) l (l + 1) - en);
>
> Do[\[Psi]m = \[Psi][[n - 1]]; \[Psi]n = \[Psi][[n]];
>
> xm = x[[n - 1]]; xn = x[[n]];
>
> xp = xn + s; x = Append[x, xp]; Fn = F[xn]; Fp = F[xp]; Fm = F[xm];
>
> \[Psi]p = (2. \[Psi]n - \[Psi]m +
>
> s^2/12. (10. Fn \[Psi]n + Fm \[Psi]m))/(1. - s^2/12. Fp);
>
> \[Psi] = Append[\[Psi], \[Psi]p], {n, 2, nsteps}];
>
> \[Psi] = \[Psi]/Max[\[Psi]]; scat = Transpose[{x, \[Psi]}];
>
> nmax = Length[scat]; {r1, u1} = scat[[nmax]]; {r2, u2} =
>
> scat[[nmax - 1]];
>
> K = (r1 u2)/(r2 u1); j1 = j[l, k r1]; j2 = j[l, k r2];
>
> n1 = n[l, k r1]; n2 = n[l, k r2];
>
> \[Delta] = ArcTan[(K j1 - j2)/(K n1 - n2)]]]
>
>
>
> fC[2.0, 1000, 2.0] // AbsoluteTiming
>
> CompiledFunction::cfse: Compiled expression 1822.5985401459855` should be a machine-size integer. >>
>
>
>
> CompiledFunction::cfex: Could not complete external evaluation at instruction 7; proceeding with uncompiled evaluation. >>
Hi
Here is some code that works. There are still some lines that are unnecessary but I think it will do. Note, that the code is not complete because it is not clear what you mean after your function K. It seems that you are new
to the topic, thus I advise to debug your code one line at a tim. Cheers
fC:=Compile[{{l,_Real},{nsteps,_Integer},{En,_Real}},Block[{meV=1./(27. 3*1000.),en,\[Angstrom]=1./0.52,\[Sigma],\[Epsilon],mH=1836.,mKr=249696.,\[Mu],\[Lambda],k,s,x,\[Psi]={0.0,0.0001},scat,vec1,vec2,K},
en=En*meV;
\[Sigma]=3.57 \[Angstrom];
\[Epsilon]=5.9 meV;
\[Mu]=(mH mKr)/(mH+mKr);
\[Lambda]=2 \[Pi]/Sqrt[2.* \[Mu]* en];
k=2 \[Pi]/\[Lambda];
s=5.*\[Lambda]/nsteps;
x={0.5 \[Sigma],0.5 \[Sigma]+s};
Block[{n=2,xn,xm,xp,\[Psi]n,\[Psi]m,\[Psi]p,Fn,Fp,Fm},
Do[
xn=x[[n]];
xp=xn+s;
AppendTo[x,xp];
\[Psi]n=\[Psi][[n]];
\[Psi]m=\[Psi][[n-1]];
xm=x[[n-1]];
Fn=2.*\[Mu]*(\[Epsilon] *((\[Sigma]/xn)^12-2.* (\[Sigma]/xn)^6)+1./(2. \[Mu]* xn^2) l (l+1.)-en);
Fp=2.*\[Mu]*(\[Epsilon] *((\[Sigma]/xp)^12-2.* (\[Sigma]/xp)^6)+1./(2. \[Mu]* xp^2) l (l+1.)-en);
Fm=2.*\[Mu]*(\[Epsilon] *((\[Sigma]/xm)^12-2.* (\[Sigma]/xm)^6)+1./(2. \[Mu]* xm^2) l (l+1.)-en);
\[Psi]p=(2.*\[Psi]n-\[Psi]m+s^2/12. *(10. Fn* \[Psi]n+Fm* \[Psi]m))/(1.-s^2/12.*Fp);
AppendTo[\[Psi],\[Psi]p];
n++
,{nsteps-2}]
];
\[Psi]=\[Psi]/Max[\[Psi]];
scat=Transpose[{x,\[Psi]}];
vec1=scat[[-1]];
vec2=scat[[-2]];
K=(vec1[[1]]* vec2[[2]])/(vec2[[1]]* vec1[[2]])
]
]