OT: Leaving Rust gamedev after 3 years

[Bruce Carneal]

On Monday, 29 April 2024 at 01:44:20 UTC, monkyyy wrote:
> On Monday, 29 April 2024 at 00:55:54 UTC, Walter Bright wrote:
>> On 4/26/2024 4:14 PM, Matheus Catarino wrote:
>>> I'm not a game developer, but when it comes to graphics 
>>> programming dealing with floating point, NaN will be a 
>>> problem that not everyone has experience or patience in 
>>> dealing with.
>>
>> When a NaN shows up unexpectedly in the results, it's 
>> indicative of a bug in the program.
>
> nans dont show up when programming video games they end in 
> silently failed draw calls

If they're silent how do you know they failed?  Also, if you *do* 
know they failed why would NaN be unhelpful in localizing the 
problem?  It could be initialization, or it could be something 
else along the way.

Zero initialization has a few things to recommend it over NaN 
(consistency with our non-FP types, zero is often (not always) 
the correct initialization value anyway, aligns with zero fill 
new-page machinery, simpler if you write bug free code :-) ...).

OTOH, NaNs can be very helpful when trying to track down FP 
problems, with improper (unthinking) initialization probably near 
the top of the list.