Differences in results when using the same function in CTFE and Runtime

[Abdulhaq]

On Thursday, 15 August 2024 at 09:13:31 UTC, Carsten Schlote 
wrote:
> On Tuesday, 13 August 2024 at 07:56:34 UTC, Abdulhaq wrote:
>> One little piece of anecdata, when working with python I saw 
>> differing results in the 17th d.p. when running the exact same 
>> calculation in the same python session. The reason turned out 
>> to be which registers were being used each time, which could 
>> vary. This is not considered to be a bug.
>
> I would consider it a bug.
>

I'd say that whether it's a bug or not comes down to the 
definition of "bug", and I don't think a discussion of that is 
useful, and also it's subjective and reasonable people can differ 
on it.

But I would say that, in my python example, both answers were 
wrong, in the sense that they were not exactly right (and can't 
be), but they were both wrong in a different way.

In your example I'd ask you to consider this: all the images you 
created, except the trivial ones, were wrong, in the sense that 
some of the calculations were inexact and some of them had 
sufficient error to change the color of the pixel. It's just that 
the CTFE image was wrong in a different way to the non-CTFE 
image. Why do you want to reproduce the errors?

In general floating point calculations are inexact, and we 
shouldn't (generalising here) expect to get the same error across 
different platforms.

In this specific case you could argue that it's the "same 
platform", I get that.