On 05/23/2013 09:43 AM, Richard Earnshaw wrote: > So you're saying the documentation is wrong, or at the very least > misleading... What about dealing with address validation for soft accesses? > Surely we get better code if the SFP ~= HFP since we end up with fewer cases > where we fall back to sub-optimal sequences.
It's not wrong, exactly. It's exactly correct that the frame pointer in question is the soft frame pointer at the beginning of rtl generation. SFP != HFP can be tricky to do well, I agree. For Alpha I was able to get good results by allowing any signed 32-bit offset to SFP and then using LEGITIMIZE_RELOAD_ADDRESS to split the adjusted offset optimally. r~
