forked from M-Labs/libfringe
document why we're not using a straigh `ret`
This commit is contained in:
parent
d06e7c7cdd
commit
ae00fa2346
|
@ -31,6 +31,8 @@ jmp 2f
|
||||||
movq %rax, %rsp
|
movq %rax, %rsp
|
||||||
|
|
||||||
// jump into the new context (return to the call point)
|
// jump into the new context (return to the call point)
|
||||||
|
// doing this instead of a straight `ret` is 8ns slower,
|
||||||
|
// presumably because the branch predictor tries to be clever about it
|
||||||
popq %rax
|
popq %rax
|
||||||
jmpq *%rax
|
jmpq *%rax
|
||||||
|
|
||||||
|
|
Loading…
Reference in New Issue