[fpc-devel] Sorry for poor testing

[Martok]

Am 14.01.2019 um 15:01 schrieb J. Gareth Moreton:
> Martok mentioned doing some checks differently in the bug report in question,
> such as 6 comparisons being faster than a jump table.  Are there any others
> worth mentioning?
Not neccessarily faster, but in that code definitely smaller. Is there a way to
directly estimate how large the generated code might be, maybe something like
"numlabels*constant"?

For the "faster" question, the cache line occupation matters more than
instruction throughput. If your jumptable gets large enough to force a full data
cache line flush, you could have done a lot of instructions in the time the CPU
waits for the data. And I haven't found reliable information how the
Spectre-related microcode updates changed that, given that both cmp/jmp and
jmp[indirect] used to be heavily optimized, but were both affected.


-- 
Regards,
Martok