Gregory Maxwell

> size of size_t, see for example Might be better to make that config macro get size_t's bits and base the limit on that. Otherwise its reducing portability for unexposed...

https://github.com/bitcoin-core/secp256k1/blob/770b3dcd6f1edb0a6355a55d92b7bea1e9524042/src/ecmult_impl.h#L58 the window_size absurd maximum of 24 is driven by size_t overflow.

> I would not be surprised if we assume sizeof(int) >= 4 somewhere Nope, test_fixed_wnaf_small has some bugs where it uses 32-bit values as int arguments, but other than that...

Woops! Will do. I'm excited about this work-- but I've been redoing all the power at home and have had systems in various states of off and disconnected and couldn't...

@peterdettman Oh crap. I am terribly sorry for completely forgetting your request. I've tested on two different 32-bit arm systems: A faster superscalar cortex-a9, and a slower in order cortex-a8...

Given that it was so compiler sensitive I thought it would be useful to upgrade to a newer GCC and try clang too: --------- Cortex-a8 gcc version 10.2.0 (GCC) Master:...

I don't think removing it would be doing anything to benefit the world. Instead, it would make the library less useful to users who need a quasi-DER decoder. Creating a...

If you asked me if I thought #701 should apply to contrib I would have said probably not (among other things "DER encoded private key" is actually what it's called,...

It's fine by me. I hope it comes back later in some documentation or example or something like that.

Good news: tests pass, including valgrind ct tests. Bad news: First is before the patch. x86_64 64-bit GCC 10.2.1 (asm disabled, obviously) scalar_sqr: min 0.0421us / avg 0.0422us / max...