Thread: Here's a *if vs Switch* Benchmark program

Just thought you guys and gals would like to try a if vs Switch bench-mark code. It might not be the best but it seem to makes a decent point. On my machine Dell 3.00 Ghz -- 2 gigs of RAM "switch" was only one ms faster than "if" and when I ran it a second time, it was reverse so they are basically equal. It's only a matter of taste.

I found it here:
C++ Speed Test: If vs Switch Statement | Programming Designs
...

I actually found this interesting since it gave me a reason to look at some assembly code. It can be helpful to get familiar with what compilers are generating.

If the code is compiled as presented with optimizations, the compiler will see that val is constant and remove the if/switch statements, as expected. It didn't remove the loop, I guess since the compiler cannot know what side effects occur in the std::string assignment operator.

I changed it so val is assigned an random number modulus 7 to prevent this.
With the if statement, a cmp instruction is executed for each case. With the switch statement a single cmp instruction was executed to check for values over 5, otherwise it just used val as an index for an offset to select the string. My initial guess would have been that nearly identical code would be generated, so i learned something from this.

So, interesting journey into compiler code generation and optimization. It does look like the switch code is more optimal in this case by some small amount (not that I'd ever consider performance when selecting between using if/switch statements).

I used Visual Studio 2008 with full optimizations.

Code:

Your code also jumps in 99,99% of tests into the "default" case and the last if-else statement.

Who me

Code:

 If you hadn't use std::string the switch case would have got much better results. All cases should be used equally.

Lets not get personal


I know you're not talking about me. If I had wrote the code you best believe it be pack with in-line FASM some how. I just made it to charter-13 of Deitel C++. Now I can't wait to get out of school so I can explore C++. Just because the compile will do it all for me don't mean I don't want to peep under the s hood or do some things for myself. I don't have a true clue but interesting to figure that C++ can check SWITCH at bottom FIRST, but can't do the same for if ... if, if if, else, if, else etc. Kind of make since. It got to take it from the top to learn, but why loop EVERY statement, completely???

All though I saw no difference but the better programmers like you guy did, as a newbee this tells me to use switch when ever possible once I get started and let the standard continue to bicker about it. Optimization is optimization even if it save you an single clock and optimization turn on don't mean it's totally optimized. Humans can be smarter, even by accident. I can't wait to learn C++.

as a newbee this tells me to use switch when ever possible once I get started and let the standard continue to bicker about it. Optimization is optimization even if it save you an single clock and optimization turn on don't mean it's totally optimized.

Are you going to shoehorn all branching into a switch case in your programs now? You could, you know. But you will have to write additional functions to work out some Boolean logic in a switch case friendly way. The overhead from calling said functions will just spend any time that you save.

Are you going to shoehorn all branching into a switch case in your programs now?

Pay me no attention. I just been cocky because I'm writing programs beyond class assignments. I an expert at trial and error but can't explain what make it tick for days to weeks afterward so I might as well stay in the books. I use to always benchmark ASM code so I plan to find and use a lot of tools for bench-marking C++ code and to know what the compiler is doing. It be a while before I have any comments that makes any real since to others.


Elysia, I keep forgetting about , that will not happen again.