If statements inside switch statements?
- From: Chris Hofstader <cdh@xxxxxxxxxxxxx>
- To: programmingblind@xxxxxxxxxxxxx
- Date: Mon, 26 Jul 2010 12:21:53 -0400
Hi all,
Is it efficient on the computer's resources if I place an if statement
inside a switch statement?
The reason I ask is because for whatever reason, I can't get my else
clause to work. Switch, on the other hand, works just fine.
Jes
__________
View the list's information and change your settings at
//www.freelists.org/list/programmingblind
Most i86 based C compilers generate code that converts all of the case
statements inside a switch to a set of integers that it sorts from top
to bottom and looks like:
mov esi, numCases
mySwitch:
move doSomething, doSomethingElse[esi]
sub myValue, internalValue
jz dosomething
dec internalValue
jz mySwitch
jmp endSwitch
doSomething0 proc
; do SomethingUseful
jmp endSwitch
doSomething[1] proc
;Do something else useful
jmp endSwitch
[[repeat for all cases]
endSwitch:
[doTheRestOfYourProgram
This can be made more efficient by using an xlat instruction but, as of
the last time I looked, few compilers had that operation.
So, adding an "if" statement to a switch would simply put some sort of
compare instruction into one of the cases which, if an integer for
instance, would just looklike:
sub myInteger, theItemToCompareWith
jnz notEqual
[Do something that you want your if to do.]
jmp endSwitch
Historically, compare operations (cmp) were considered slow (I think
they took 3 or 4 cycles on a box with an 80286) so we tried to avoid
them in tight loops. Now that a cycle takes roughly one third of a
billionth of a second, even if it is still a slow instruction,it doesn't
really matter much where you use it. Do not mistake a switch with a
tight loop. A switch is really efficient for a compiler to optimize but
sometimes loops that try to do too many things too tightly are the cause
of what will get your profiler screaming bloody murder.
Happy Hacking,
cdh
__________
View the list's information and change your settings at
//www.freelists.org/list/programmingblind
Other related posts:
