Maamoun TK maamoun.tk@googlemail.com writes:
Sure. I updated the pull request.
Thanks. Merged (first time I try the merge button on gitlab).
Yes, it makes sense to avoid unaligned loads in the main loop by checking low-order bits of address, but still I can't imagine it would be more simple in this case. Allocating stack buffers used very often along the lifespan of process and I think it's ok to be used for this purpose.
It's no big problem, it just seems slightly wasteful with an extra round of load and store to memory.
We could revisit it if we ever get to rearranging the loads for the main loop.
Regards, /Niels