This is sort of intriguing for me. I couldn't find any saturation instructions using general purpose register in ARMv8. However, there are saturation instructions for Neon registers I couldn't find the same involving general purpose register. Can anyone suggest the alternative of implementing such instructions such as QSUB in ARMv8 architecture?
Those operations were mostly used for DSP type work like mixing audio, and the SIMD is good for that as it can do a number of the operations in parallel. Do you have some idea in mind where it would only deal with one value at at time and the job is big?