There's cost to transfer it through memory and cache hierarchies and there's cost when it causes the CPU to drop some other L1C cache line.
I think I'll test whether use of exceptions can affect optimizations concerning register allocation and spilling, like function prologues and epilogues. Unwinding will also need to call destructors, and for that you'll need to have the object pointer somewhere in (stack frame) memory. But it might be in just a register. Unless, of course, the unwinding engine can look into registers as well.
I guess it's not a huge deal either way, especially in 32-bit x86 code. For large register file systems it might be a different story, forcing otherwise unnecessary spills. Regardless, I'll find out when I got some time.
Anyways, my point was more generic. So-called zero-cost abstractions are my pet peeve, because they're often not zero cost precisely because of larger generated code size. Typical microbenchmarks don't capture this, because their hot path can usually fit comfortably in L1 code and data caches.
Regarding spilling and register allocation, IIRC[1] the unwind tables contain a simple bytecode that is interpreted by the undwinder to fixup stack and restore registers (indexed by the current ip address), so no actual registration of object to destroy need to be done, nor does the frame pointer need to exist at all: the unwinder interprets the bytecode to recursively adjust each stack frame and registers to the expected values.
[1] At some point I had actually learned to write unwind tables by hand, of course I completely forgot everything about that: look at trampoline_thunk at https://github.com/gpderetta/delimited/blob/master/delimited... even with extensive comments is still write-only code. It doesn't help that there aren't a lot of DWARF CFI tutorials around.
You only use up "cache" if you actually hit the cache. If you're on a different L1 cache line (and never execute), you'll never be in L1 instruction cache, maybe never even in L2 or even L3 cache lines.
Not that I'm an expert on compilers... but whenever I look at assembly, there are lots of "NOPs" across the code. I've always assumed it was for some kind of boundary alignment, probably cache.
I'd sure hope they're on an entirely different page, not just cache line!
Concerning compiler code alignment: they seem to place my code on 16-byte boundaries - for a good reason too, the smallest code size while keeping optimal 16-byte aligned branch targets.
In past I actually had to drop to assembler to get 64-byte (cache line) alignment for functions (MSVC).
GCC does have optimize pragma, like:
#pragma GCC optimize ("align-functions=64")
Although I want to point out one shouldn't usually worry about low level details like function alignment. Use profile guided optimization, etc. instead. Also remember due to L1 set associativity limitations excessive alignment can even cause pathological L1 churn in the worst case.Regarding execution times, I don't know from first-hand experience, I did some research and found this[1]. The third sentence in that reply ("The error code is not sensitive to the percentage of occurrence") is AFAIK wrong (because of speculative execution, I believe?), but the rest seems quite solid.