I feel like doubling the workload is better than quadrupling the size of the project inheriting a bevy of features and tools you likely won’t touch at all. Sure it’s stripped out later (ideally), but I like less bloat and that includes during dev when I might have to dig through 3rd party code with its own conventions and standards packed into a ‘source available’ library with potentially dogshit or absent documentation.
Unused code is stripped out by the compiler, but will your homemade library properly use all the fancy instruction set extensions for matrices? IIRC it’s not as simple as just compiling for the correct microarch. But I could be wrong.
I feel like doubling the workload is better than quadrupling the size of the project inheriting a bevy of features and tools you likely won’t touch at all. Sure it’s stripped out later (ideally), but I like less bloat and that includes during dev when I might have to dig through 3rd party code with its own conventions and standards packed into a ‘source available’ library with potentially dogshit or absent documentation.
Also yes, it’s good practice
Unused code is stripped out by the compiler, but will your homemade library properly use all the fancy instruction set extensions for matrices? IIRC it’s not as simple as just compiling for the correct microarch. But I could be wrong.