Sarnold wrotes this package that is provided with tcllib.
Its goal is to handle floating-point numbers without accuracy limit - other than memory and CPU time, of course - and ensure all digits of computed numbers are true.
Example of use : the Runge-Kutta algorithm
package require math::bigfloat namespace import ::math::bigfloat::* # compute y'=f(t,y) with the initial value y(0) # here t0..tn and n are the steps and their number # f is the function (with two arguments : t and y) # y0 is y(0) proc RungeKutta4 {t0 tn f y0 n} { # f(x_0)=current value set t $t0 # y(initial_value) set y $y0 set step [div [sub $tn $t0] [fromstr $n]] set half_step [div $step [fromstr 2]] # from x_1 to x_N for {set i 0} {$i < $n} {incr i} { set k1 [$f $t $y] set k2 [$f [add $t $half_step] [add $y [mul $half_step $k1]]] set k3 [$f [add $t $half_step] [add $y [mul $half_step $k2]]] set k4 [$f [add $t $step] [add $y [mul $step $k3]]] set slope [div [add [div [add $k1 $k4] [fromstr 2]] [add $k2 $k3]] [fromstr 3]] set y [add $y [mul $slope $step]] set t [add $t $step] } # the return value is a bigfloat return $y } proc ident {t y} { set y } # y'(t)=y(t) and y(0)=1 set x0 [fromstr 0.0 10] set y0 [fromstr 1.0 10] set xEnd [fromstr 1.0 10] set steps 100 puts "With RK4 exp(1)=[tostr [RungeKutta4 $x0 $xEnd ident $y0 $steps]]" puts "Real value equals : [tostr [exp $xEnd]]"
See BigFloat for Tcl.