On Monday, 8 June 2015 at 11:44:47 UTC, Daniel Kozák wrote:
> No difference even with GC.disable() results are same.

Profile! Callgrind is your friend~

On Mon, 08 Jun 2015 18:16:57 +0000
Anonymouse via Digitalmars-d-learn <digitalmars-d-learn@puremagic.com> wrote:

> On Monday, 8 June 2015 at 11:44:47 UTC, Daniel Kozák wrote:
> > No difference even with GC.disable() results are same.
> 
> Profile! Callgrind is your friend~
Yep, but I dont care, I am the one who makes transcode faster, so I am happy with results :P.

P.S. I care and probably when I have some spare time I will improve to!dstring too

On Monday, 8 June 2015 at 18:48:17 UTC, Daniel Kozak wrote:
> Yep, but I dont care, I am the one who makes transcode faster, so I am happy
> with results :P.
>
> P.S. I care and probably when I have some spare time I will
> improve to!dstring too

Ah, so you are. I confused you with Kadir Erdem Demir.

Am Mon, 08 Jun 2015 11:13:25 +0000
schrieb "Daniel Kozak" <kozzi11@gmail.com>:

> BTW on ldc(ldc -O3 -singleobj -release -boundscheck=off) transcode is the fastest:
> 
> f0 time: 1 sec, 115 ms, 48 μs, and 7 hnsecs // to!dstring
> f1 time: 449 ms and 329 μs // toUTF32
> f2 time: 272 ms, 969 μs, and 1 hnsec // transcode

Three functions, each twice as fast and twice as hidden as the one before. :)

-- 
Marco

Am Mon, 8 Jun 2015 12:59:31 +0200
schrieb Daniel Kozák via Digitalmars-d-learn
<digitalmars-d-learn@puremagic.com>:

> 
> On Mon, 08 Jun 2015 10:41:59 +0000
> Kadir Erdem Demir via Digitalmars-d-learn
> <digitalmars-d-learn@puremagic.com> wrote:
> 
> > I want to use my char array with awesome, cool std.algorithm functions. Since many of this algorithms requires like slicing etc.. I prefer to create my string with Utf32 chars. But by default all strings literals are Utf8 for performance.
> > 
> > With my current knowledge I use to!dhar to convert Utf8[](or char[]) to Utf32[](or dchar[])
> > 
> > dchar[] range = to!dchar("erdem".dup)
> > 
> > How costly is this?
> 
> import std.conv;
> import std.utf;
> import std.datetime;
> import std.stdio;
> 
> void f0() {
>     string somestr = "some not so long utf8 string forbenchmarking";
>     dstring str = to!dstring(somestr);
> }
> 
> 
> void f1() {
>     string somestr = "some not so long utf8 string forbenchmarking";
>     dstring str = toUTF32(somestr);
> }
> 
> void main() {
>     auto r = benchmark!(f0,f1)(1_000_000);
>     auto f0Result = to!Duration(r[0]);
>     auto f1Result = to!Duration(r[1]);
>     writeln("f0 time: ",f0Result);
>     writeln("f1 time: ",f1Result);
> }
> 
> 
> /// output ///
> f0 time: 2 secs, 281 ms, 933 μs, and 8 hnsecs
> f1 time: 600 ms, 979 μs, and 8 hnsecs
> 

Please have the result of the transcode influence the program
output. E.g. Add the first character of the UTF32 string to
some global variable and print it out. At the moment - at
least in theory - you allow the compiler to deduce f0/f1 as
pure, return-nothing functions and you will benchmark anything
from your written code to an empty loop. I'm talking out of
experience here:
https://github.com/mleise/fast/blob/master/source/fast/internal.d#L99

-- 
Marco