U++ forum: Welcome to the forum

Search on this site

Search in forums

Home » U++ Library support » U++ Core » Small optimization

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Small optimization [message #57056]

Tue, 18 May 2021 15:36

Didier
Messages: 726
Registered: November 2008
Location: France

Contributor

Hello Mirek,

I have a small optimisation proposition for cases where fundamental data types are used

in Vcont.h

Replace:

template <class T>
inline void Construct(T *t, const T *lim) {
	while(t < lim)
		new(t++) T;
}

with:

template <class T>
inline void Construct(T *t, const T *lim) {
	if ( ! std::is_fundamental<T>::value ) {
		while(t < lim)
			new(t++) T;
	}
}

Other methods near this one can benefit from this small optimisation (Destroy for example)

Note: Tested on my app : everything seems to work fine

[Updated on: Tue, 18 May 2021 16:58]

Report message to a moderator

Re: Small optimization [message #57119 is a reply to message #57056]

Thu, 27 May 2021 09:07

mirek
Messages: 14257
Registered: November 2005

Ultimate Member

Didier wrote on Tue, 18 May 2021 15:36

Hello Mirek,

I have a small optimisation proposition for cases where fundamental data types are used

in Vcont.h

Replace:
template <class T>
inline void Construct(T *t, const T *lim) {
	while(t < lim)
		new(t++) T;
}
with:
template <class T>
inline void Construct(T *t, const T *lim) {
	if ( ! std::is_fundamental<T>::value ) {
		while(t < lim)
			new(t++) T;
	}
}
Other methods near this one can benefit from this small optimisation (Destroy for example)

Note: Tested on my app : everything seems to work fine

I am not fundamentally against it, but I am pretty sure that compiler already optimizes that out as dead code - which means it is just added noise.

never_inline
void Foo() {
	RLOG("---");
}

CONSOLE_APP_MAIN
{
	int x[4];
	Foo();
	Construct(x, x + 4);
	Foo();
}

400018B0  sub rsp,byte +0x28 
400018B4  call dword 0x1400017e0 // call to Foo
400018B9  nop // this to align
400018BA  add rsp,byte +0x28 
400018BE  jmp dword 0x1400017e0 // call to Foo

Mirek

Report message to a moderator

Re: Small optimization [message #57125 is a reply to message #57119]

Thu, 27 May 2021 20:01

Didier
Messages: 726
Registered: November 2008
Location: France

Contributor

You are perfectly right then: clearner code is better Thumbs Up

Report message to a moderator

Previous Topic:	DarkThemeCached
Next Topic:	SIGPIPE problem

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sat May 10 00:44:09 CEST 2025

Total time taken to generate the page: 0.03554 seconds