Further upgrading put pixel.

[NaturalDemon]

A test will only make the game slow down. I want a performance boost to make the game run faster.

wel the basic putpixel ... does already 4 tests out of the box.

Code: Select all

   assert( x >= 0 );
   assert( y >= 0 );
   assert( x < SCREENWIDTH );
   assert( y < SCREENHEIGHT );

this is much more complex than

Code: Select all

if (x >= 0 && x <= screenwidth && y >= 0 && y <= screenheight)
{
PutPixel (.........);
}

you should use7create a frametimer.

Code: Select all

pSysBuffer[ x + SCREENWIDTH * y ]

is any array of the type D3DCOLOR ... you acces the induvidual index directly ... i don´t see how you can speed that up .... unles the cpu it self has some special tools i don´t know off.

read ... http://www.planetchili.net/forum/viewto ... a+blitting

[Musi]

you're wrong ... sorry to say

Code: Select all

( x < 0 || x + 20 > SCREENWIDTH ||
            y < 0 || y + 20 > SCREENHEIGHT )

(see image) http://www.planetchili.net/forum/viewto ... e&start=10

is much slower than

Code: Select all

if (x >= 0 && x <= screenwidth && y >= 0 && y <= screenheight)
{
PutPixel (.........);
}

Is it realy slower? im not calling it for every pixel though. Surely one if statment is faster than 400. What about a 100x100 sprite, is one if statement slower than 10,000?

Also i think asserts get removed on the release build.

[NaturalDemon]

Is it realy slower? im not calling it for every pixel though. Surely one if statment is faster than 400. What about a 100x100 sprite, is one if statement slower than 10,000?

you are doing ... atleast 6 operation vs atleast 4 operations .. so it's 33,333333% slower than 4
on a 10.000 pixel image .. you would do this test also 10.000x and all of them 33,3333% slower than the most simple solution

x < 0 (operation)
|| (operation)
x + 20 (operation)
(sum) > SCREENWIDTH (operation)
||(operation)
y < 0 (operation)
|| (operation)
y + 20(operation)
(sum) > SCREENHEIGHT (operation)

Also i think asserts get removed on the release build.

yeah, chili mentioned some ... this means you need to add test on place to prevent any outside of the client coordinates in the currect state (800x600) of the framework

wiki: stack

[Musi]

you are doing ... 6 operation vs 4 operations .. so it's 33,333333% faster
on a 10.000 pixel image .. you would do this test also 10.000x and all of them 33,3333% slower than the most simple solution

Ugh, but like i said im not testing every pixel, im testing the sprites coordinates, which is one test to find out if the sprite is all on the screen or not. THEN if the sprite is partly off the screen you test every pixel.

[NaturalDemon]

Ugh, but like i said im not testing every pixel, im testing the sprites coordinates, which is one test to find out if the sprite is all on the screen or not. THEN if the sprite is partly off the screen you test every pixel.

oh, yeah now i see ... sorry

but there are various methods(adaptive loops instead of omitting pixels) ... and i think ... all this stuff aint realy needed ....
because i have been reading all this time and i still don`t have a clear picture ....
but the 2d stuff seams to be harder than the 3d stuff.

[cameron]

The asserts only run in debug mode. So it dosnt slow the game down its just to check if something goes wrong so the tests don't help. The asserts are the tests. So more tests just slows it even more. I need a speed boost. If a speed boost is even possible.

[NaturalDemon]

what are you doing?

are you using images?
mutile of the same?
are you loading them on the fly or preload the images?
1920 x 1080?

[cameron]

Im pre loading them.

[Zionich]

@Cameron

I see what your saying as far as the initial test, before the individual test. I like it.

[Musi]

I don't think put pixel can get any simpler really, all it does now is change a value in an array. The only thing i can think of that might improve drawing speed is to use something other than memcpy(); to copy it to the back buffer. But again i don't even know if there's a faster way of doing that either.

Are your programs running slow?