Post Reply 
[BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
12-19-2016, 04:15 AM
Post: #1
[BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
I was reading this post in another forum http://forum.hp-prime.de/showthread.php?tid=183

and can confirm that all of the following result in the wrong root 0 on the Prime:
zeros(x^4 - 3x^3 - 2.75x^2 + 12x - 5)
solve(x^4 - 3x^3 - 2.75x^2 + 12x - 5 = 0)
fsolve(x^4 - 3x^3 - 2.75x^2 + 12x - 5 = 0)
-> [2.5, 2., -2., 0]

The correct result would be:
-> [2.5, 2., -2., 0.5]

Firmware version: 2016 08 29 (10637)

The same commands work correctly with giac 1.2.2.

Also, the exact version works on the Prime:
zeros(x^4 - 3x^3 - (275/100)*x^2 + 12x - 5)
-> [2, -2, 5/2, 1/2]


This looks more significant than a small rounding error.
Find all posts by this user
Quote this message in a reply
12-19-2016, 07:53 AM
Post: #2
RE: [BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
Another case that's not affected is when I give a parameter range to solve:
solve(x^4 - 3x^3 - 2.75x^2 + 12x - 5 = 0, x=(-3)..3)
Find all posts by this user
Quote this message in a reply
12-19-2016, 04:37 PM
Post: #3
RE: [BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
This has been fixed in the source code, I forgot to report it here.
Find all posts by this user
Quote this message in a reply
12-19-2016, 06:47 PM
Post: #4
RE: [BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
I have the Android app and noticed the same problem in version 8151.

I had reasons to believe that the bug had been solved in the most recent update, but now I have updated my app to version 2016 12 08 (11226), CAS Version: 1.1.2 – 11, and I still have the same problems.

See this thread:
http://h30434.www3.hp.com/t5/Calculators...-p/5269068
Find all posts by this user
Quote this message in a reply
12-20-2016, 03:18 AM
Post: #5
RE: [BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
Jan, the calculator firmware is lagging a bit behind the latest CAS source code. I think we have to wait for a future update until the fix actually ships.
Find all posts by this user
Quote this message in a reply
12-20-2016, 01:14 PM
Post: #6
RE: [BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
(12-20-2016 03:18 AM)danielmewes Wrote:  Jan, the calculator firmware is lagging a bit behind the latest CAS source code. I think we have to wait for a future update until the fix actually ships.
Hello Danielmewes,

Can you, or anyone else, explain why Maké in message 12 of this thread:

http://h30434.www3.hp.com/t5/Calculators...068/page/2

can get the correct factorisation of the polynomial with version 10077 on his calculator, whereas I, on my Android app, and the most recent update,
version 2016 12 08 (11226), CAS version 1.1.2 – 11,
get the wrong factorisation, namely a factor x instead of (x-1.0)?
Find all posts by this user
Quote this message in a reply
12-20-2016, 01:54 PM (This post was last modified: 12-20-2016 01:55 PM by toml_12953.)
Post: #7
RE: [BUG?] Wrong numeric result with CAS `fsolve`/`zeros`/`solve`
(12-19-2016 04:15 AM)danielmewes Wrote:  I was reading this post in another forum http://forum.hp-prime.de/showthread.php?tid=183

and can confirm that all of the following result in the wrong root 0 on the Prime:
zeros(x^4 - 3x^3 - 2.75x^2 + 12x - 5)
solve(x^4 - 3x^3 - 2.75x^2 + 12x - 5 = 0)
fsolve(x^4 - 3x^3 - 2.75x^2 + 12x - 5 = 0)
-> [2.5, 2., -2., 0]

The correct result would be:
-> [2.5, 2., -2., 0.5]

Firmware version: 2016 08 29 (10637)

The same commands work correctly with giac 1.2.2.

Also, the exact version works on the Prime:
zeros(x^4 - 3x^3 - (275/100)*x^2 + 12x - 5)
-> [2, -2, 5/2, 1/2]


This looks more significant than a small rounding error.

On the Windows Virtual Calculator (build 11226), it's weird but the second function (solve) gets it right but the first and third get it wrong, as you have noted.

Tom L

Tom L
Cui bono?
Find all posts by this user
Quote this message in a reply
Post Reply 




User(s) browsing this thread: 1 Guest(s)