LaPlace: HP's Erable 3.2 beats by far the best TI method I know

[Prev][Next][Index][Thread]

LaPlace: HP's Erable 3.2 beats by far the best TI method I know

[ LaPlace: HP's Erable 3.2 beats by far the best TI method I know ]

After I knew that Bernard Parisse had released a new version of Erable (3.2),
I wanted to compare again the speed of HP's best tool I knew with TI's best
tool I knew for LaPlace, to find out how the new Erable changed (or not) the
results I had achieved. And I found that with the new Erable 3.2, HP is
faster and TI for LaPlace transforms. Let's see...

000HP's corner000
HP48GX's boxer for LaPlace transforms is the latest version of the superb
program by Bernard Parisse, Erable 3.2. We made a small program that performs
the operation and displays time: << TIME SWAP LAP EXPAND TIME ROT - 10000 * >>
Erable 3.2 takes 84KB of memory and is much faster than the previous version,
3.1, which we used yesterday to compare speeds. So, we have decided to compare
TI and HP again today, for ver 3.2 does changes the rules of the game.

000TI's corner000
TI-89's boxer for LaPlace transforms is the "First LaPlace proposal by Perez-
Franco", which demonstrated to be, as we showed in yesterday's article, the
fastest and smallest formula you can type in TI to obtain LaPlace transforms.
It is described as:
     -'(e^(-s*t)*z,t)|t=0
It uses 20 bytes of memory. It can be defined as a userd defined function. If
you know of any other tool for TI that solves LaPlace *faster* than this ultra
fast instruction, let me know.

000How was time measured?000 Since TI does not have an inner clock as HP
does, times were measured in different ways. In HP, the program we made gave
us the time. In TI, we used a Casio stop-watch and our eye. :o) We decided to
use huge functions in the transformations, in order to get big times which
can be compared without taking care my eye-hand's reaction time. Anyway, time
was rounded to just one decimal after the point.

000How was this comparation made?000 1) HP thing was made in a HP48GX R
runing Java (this is necessary for pretty print, because TI gives answer in
pretty print and conditions must be equal for both), with Erable installed in
Port:1: and nothing else in user memory but a CST and the small program we
mentioned (125KB free) and new batteries. 2) TI thing was made in a TI89 with
nothing else in user memory but the lap() user defined function and new
batteries.



000Results000

Function: 1 + t + 2*t^2 + 3*t^3 + 4*t^4 + 5*t^5 + 6*t^6
HP: 8.9 sec
TI: 10.6 sec
Rate: 0.84 (HP wins!)
Both gived identical answer display than MapleV.

Function: 1 + t*sin(t) + 2*t^2*sin(2*t) + 3*t^3*sin(3*t)
HP: 10.5 sec !!!
TI: 137.3 sec
Rate: 0.077 (HP wins!)
Both gived different answer display than MapleV.

Function: exp(5*t-3)+4*t^2*sin(7*t-Pi/3)+2*t^3/5
HP: 74.6 sec
TI: 79.1 sec
Rate: 0.94 (HP wins!)
Both gived different answer display than MapleV.



000Conclussion000

Although Erable 3.1 is much slower than my TI's formula, the new Erable 3.2 is
clearly faster than both.

If you have $300 to spend and know how to install in your calculator the
Erable 3.2 library of the genial Bernard Parisse, then the best hand held
tool you'll find to solve LaPlace transforms is HP48GX with 128KB RAMCard and
Erable 3.2 and Java (for pretty print) in port 1. This will allow you to make
LaPlace and InvLaPlace in the fastest way a human hand held device have ever
known. It is the best choice, but the most expensive.

For half the money, you can get a TI-89 and type in the entry line my LaPlace
proposal (-'(e^(-s*t)*z,t)|t=0) and solve LaPlace transforms in the second
fastest way a human hand held device have ever known. No need for RAMCards. No
need for additional software. It is still a very good choice and it is worth
every cent of the price.

- Roberto Perez-Franco

HP48 and TI89 user
http://surf.to/Perez-Franco

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
Follow-Ups: