Is there a way I can improve performance by using special prefetch instructions and if so where/how to deploy them?

55 Views Asked by At

I have written a recursive descent parser, so one that is not table-driven, and it typically only takes a very small amount of input text which might only be 30-100 code units (optionally either 16-bit or 32-bits each) with the total depending greatly on the presence of comments within the text that it is parsing. I am wondering if I can improvement performance by using prefetch instructions on x86-64 and AArch64 - prfm instruction iirc ? If so how and where should I deploy them ?

0

There are 0 best solutions below