C faults, etc

[Lutger]

bearophile wrote:

> Daniel Keep:
>>Why on earth should the D compiler special-case a function that's not even
>>supposed to be used?!<
> 
> What's the purpose of being able to use the C std lib in D programs if
> then such functions aren't supposed to be used?
> 
> 
>> Actually, let's check... (Goes off to test.)<
> 
> This is a first little test, with DMD V.2.031, on Windows, Phobos2:
> 
> import std.stdio: printf;
> void main() {
>     const int N = 2_000_000;
>     for (int i; i < N; i++)
>         printf("test %d\n", i);
> }
> 
> 
> A second version that uses writefln:
> 
> import std.stdio: writefln;
> void main() {
>     const int N = 2_000_000;
>     for (int i; i < N; i++)
>         writefln("test %d", i);
> }
> 
> I have redirected their output to a file (26.2 MB file).
> The timings are: 1.16 seconds for the version with printf, and 14.93
> seconds for the version with writefln.
> 

writefln does a flush where writef doesn't. On my system (fedora 11) all 
timings are nearly the same however, while writef is actually a bit faster 
than printf. What OS are you on?