I believe GMP is just a bignum library and it is present and working well on arm.
As well as 64 bit values can be (and probably are) emulated by the compiler with 32 bit instructions.
Let me grab a cup of coffee and give me 2 hours
Perhaps I didn't state it clearly. The code *in the miner* makes some assumptions about the internal GMP data structures. Some of it is code inherited from the initial xptMiner that I haven't looked at very deeply because it was just working, such as parts of the code relating to share submission:
(this is in an ifdef win64, but it gives you a taste of what's there:
*(uint64*)(nOffset+d* = z_temp2->_mp_d[d];
)
I haven't tested any of that outside of x86 and have no idea how robust it is.
I also wasn't as careful as I should have been about making sure I didn't have any weird size assumptions in my parts of the code. The one Supercomputing noted is a quite possible bug - while I tried to use (u)int64_t in any place I touched, there's still room for cleanup.
And, finally, there's that windows bug that I haven't tracked down that's *probably* just a generic buffer overwrite somewhere that isn't killing linux due to differences in memory allocation. That might be biting you on arm.
More time to work on Riecoin again starting next week.