[FFmpeg-devel] [PATCH] RoQ encoder

[Guillaume Poirier]

Hi,

On May 5, 2007, at 10:35 , Michael Niedermayer wrote:

> On Sat, May 05, 2007 at 10:08:25AM +0200, Guillaume Poirier wrote:
>> On May 5, 2007, at 3:00 , Rich Felker wrote:
>>> On Sat, May 05, 2007 at 12:57:35AM +0100, Robert Swain wrote:
>>>> Out of personal interest and being a lesser experienced programmer,
>>>> I've always been lead to believe that 'goto' shouldn't be used  
>>>> unless
>>>> absolutely necessary. Grepping the FFmpeg source throws up far more
>>>> results than I expected so now I'm wondering - why have you  
>>>> suggested
>>>> the use of goto over something like a 'while' loop? If it's  
>>>> better to
>>>> use a goto in some cases, I'd like to know when and why. :)
>>>
>>> This is nonsensical dogma. Continue and break statements are exactly
>>> the same thing as goto. Use whatever control construct makes  
>>> sense to
>>> the code you're writing. Inserting dummy loops so you can break just
>>> to "hide" your goto in a break statement is ridiculous. If you mean
>>> goto, say goto.
>>>
>>> FWIW, goto is extremely useful for error case handling can cleanup
>>> before error returns.
>>
>> It's maybe useful for the programmer, but it confuses compilers so
>> that some optimization passes can't be performed on a code with  
>> goto's.
>> Well, maybe not all compilers are like this, but i bet it confuses
>> GCC, who's not the smartest compiler around...
>
> ffmpeg contains a few gotos, can you point to one where the code  
> would be
> faster without the goto?

I'll try my best to gather this information for you. I was merely  
repeating what one of our corporate in-house genius has told me to  
explain why GOTO are evil. I'm now convinced that they should really  
be avoided in all cases.

I'll try to go through FFmpeg's GOTOs and see if they can be  
problematic.

Guillaume