Albin Ahlbäck

Currently on my `arm_assembly` branch: ``` mpn_mul vs flint_mpn_mul m = 1: 4.67 m = 2: 4.68 3.61 m = 3: 4.01 3.30 3.04 m = 4: 2.89 2.39 2.27...

> I would really prefer to just insert more forward declarations instead of littering the codebase with pragmas except in very special circumstances. How about `*-impl.h` headers? That would allow...

> > How about *-impl.h headers? That would allow us to keep track of global private functions and avoid us from using pragmas. > > I don't see the need....

> Disadvantages: > > * There would be a performance hit working with $R^{**}$ or $R^*$ instead of $R$ (due to needing to wrap and inspect elements). But surely this...

It stems from GMP's definition of `__mpz_struct`, where the allocation and size fields are `int`s. Now, `int` doesn't have to be four bytes, but for most systems they are. Now,...

There is nothing stopping you from using bigger numbers than $2^{32} - 1$, but then you'd have to use the low-level interfaces instead (see [GMP's low-level interface](https://gmplib.org/manual/Low_002dlevel-Functions)). If you ever...

I mean, for specialized purposes (such as your case), one can just compile GMP and FLINT with the fields to long int, no? Or are there hardcoded things in the...

I suppose that it could speed some things up to avoid zero extensions (i.e. promotion from 32-bit unsigned to 64-bit unsigned). Still don't like having such this struct not being...

`flint_mpn_mullow_basecase` is now implemented, so only higher cases left to do here.

Yeah, there are also problems with non-x86 systems using `profiler.h` (I suspect you are affected by this as well). If you find a solution, please let us know!