Instructor for Research Computing, etc. at the University of Wisconsin -- Milwaukee
User Details
- User Since
- Nov 27 2017, 4:37 PM (372 w, 4 d)
Yesterday
Wed, Jan 15
Tue, Jan 14
Mon, Jan 13
Sun, Dec 29
Dec 12 2024
Dec 6 2024
Dec 3 2024
Nov 28 2024
Nov 27 2024
Nov 25 2024
Nov 24 2024
Nov 22 2024
Nov 21 2024
Nov 19 2024
Nov 17 2024
Nov 16 2024
Nov 15 2024
Nov 8 2024
Nov 7 2024
Nov 3 2024
Nov 2 2024
Oct 29 2024
Oct 27 2024
Oct 25 2024
Oct 24 2024
Oct 21 2024
Oct 18 2024
Oct 16 2024
I tested on aarch64 using SIMDE. See https://github.com/outpaddling/freebsd-ports-wip/tree/master/trimadap, and let me know if it works on arm7.
I don't think it's a good idea to add a new port that's basically a small fraction of an existing, well-supported port. Redoing the patches may be a drag, but it would benefit a lot of users by supporting other platforms, and avoid leading future developers away from a better solution by example. Using SIMDE also allows ONLY_FOR_ARCHS to be removed entirely, simplifying the port. The changes needed for SIMDE are similar to what you did. Just replace emmintrin.h with something like simde/x86/sse2.h. The conversion from sse2neon to SIMDE should be easy.
Oct 15 2024
I think SIMDE will do the same thing for neon, while at the same time supporting a variety of other architectures with no additional patching. I've used it successfully with aarch64, powerpc64, and riscv. Probably worth a look. Check out biology/bwa, biology/bowtie2, etc.
Did you consider using devel/simde? That's usually how I get around this sort of issue.
Did you consider using devel/simde? That's usually how I get around this sort of issue.