(no British English spelling checker active - yet) On one Q&A platform
one question sparked my interest in coding multiplication for AVR, and
some of the results (especially retiring two multiplier bits instead of
one at very little additional cost) have looked like GCC for AVR should
profit from it(in part for Atmel Studio still including a libgcc
predating gjl's 2012/08/24
rewrites)(<http://stackoverflow.com/questions/29812009/faster-16bit-multiplication-algorithm-for-8-bit-mcu>).
With my mind still not made up about how to contribute (dumping it to
bug tracking being the timid&lazy option), let me start with a meta²
question: Do you prefer each of a few subjectsbrought up by the same
occasion(like discriminate "optimise time" by expected vs. worst case;
implementation of symmetrical operation with decided asymmetry in cycle
count between operands; code generation for multiplication by constants
for 2 vs. 3 address architectures (and weird intermediates like the
AVRs); choosing between not totally ordered implementations, testing
multiplication; avr.md vs. -fixed; multilibs) in a separate posting, or
should they be kept together, if at the risks of postings as long-winded
as this question?
regards,
HoW
--
Wolfgang Hospital
_______________________________________________
AVR-GCC-list mailing list
AVR-GCC-list@nongnu.org
https://lists.nongnu.org/mailman/listinfo/avr-gcc-list