on the speed of std::string::find

[Paolo Carlini]

Paolo Carlini wrote:

>> Dennis, thanks for your help, but something is certainly wrong in the 
>> benchmarking, we can't trust it. Likely, it's because of the 
>> optimization level, or something else. As I said already, you must 
>> use -O2, otherwise we are measuring something which nobody will see 
>> in practice.
>
> Just to say one, without inlining at -O0, going through type_traits 
> (the tiny eq, in particular) means function calls which are certainly 
> heavy to do for each char. This thing alone makes the "new" 
> implementation appear much slower than memmem where the algorithm is 
> in fact the same.

Anyway, something else must be also wrong in your testing, sorry. These 
are the numbers I'm seeing on a P4-2400 from your fullwatch.cpp, -O0:

std::string::find : 1.53305
GNU memmem        : 1.85151
Factor            : 0.827999

Paolo.