MorphZone

The irony... Rosetta not translating...

ausPPC

# 1
Priest of the Order of the Butterfly

Posts: 543 from 2007/8/6

From: Pending...

Is this a doomed example or is there a simple fix to produce the desired output?

https://rosettacode.org/wiki/Elliptic_curve_arithmetic

I can't even compile it under MorphOS and on Linux all but the fixed co-ordinates are Not-A-Numbers.
PPC assembly ain't so bad... ;)
»26.05.15 - 11:07

igracki

# 2

Order of the Butterfly

Posts: 421 from 2003/2/24

From: Berlin

The python version works here!
Code:


0:00.60 [ 1] MOSBOOT:> c:python3  rosetta.py
a = (-1.817, 1.000)
b = (-1.442, 2.000)
c = a + b = (10.375, -33.525)
d = -c = (10.375, 33.525)
c + d = Zero
a + b + d = Zero
a * 12345 = (10.759, 35.387)

»26.05.15 - 11:40

bigfoot

# 3
MorphOS Developer

Posts: 510 from 2003/4/11
There's a bug in the C code, or at least a difference from how the Python version works.

In the from_y() function, replace
Code:
```
        r.x = pow(y * y - C, 1.0/3);
```
with
Code:
```
        double tmp;
        tmp = y * y - C;
        if (tmp >= 0)
                r.x = pow(tmp, 1.0/3);
        else
                r.x = -pow(-tmp, 1.0/3);
```
That'll make it behave identically to the Python version, and also give an identical result.

The code as it is compiles fine for MorphOS, at least if you use GCC 4.
I rarely log in to MorphZone which means that I often miss private messages sent on here. If you wish to contact me, please email me at [username]@asgaard.morphos-team.net, where [username] is my username here on MorphZone.
»26.05.15 - 18:44

ausPPC

# 4
Priest of the Order of the Butterfly

Posts: 543 from 2007/8/6

From: Pending...

Thank you. Now I get the expected output when compiled on Linux (gcc 4.8.2) but I might be experiencing a corrupted SDK. I'm downloading the sdk again and will do a checksum against the file I installed from...
PPC assembly ain't so bad... ;)
»27.05.15 - 01:00

ausPPC

# 5
Priest of the Order of the Butterfly

Posts: 543 from 2007/8/6

From: Pending...

The sdk archive I installed from seems to be the same as the newly downloaded one but gccselect isn't behaving as expected. I only managed to compile the fixed source code by entering ppc-morphos-gcc-4

Are gccselect and gccselect4 supposed to require a full path to work? After trying:

> sdk:gccselect4 4.4.5

and:

> gcc --version

it still reported 2.95.3
PPC assembly ain't so bad... ;)
»27.05.15 - 01:59

jacadcaps

# 6
MorphOS Developer

Posts: 3146 from 2003/3/5

From: Canada

gccselect4 selects the version to use when invoking ppc-morphos-gcc-4. To change the default gcc link, use gccselect.
»27.05.15 - 06:13