• Create Account

### #5Splat  Members   -  Reputation: 122

Like
Likes
Like

Posted 26 November 1999 - 07:10 AM

LaMothe is not god! Shifts are faster (methinks) than integer mods.

#define _RGB16BIT555(r,g,b) (b >> 3) + ((g >> 3) << 5) + ((r >> 3) << 10))

#define _RGB16BIT565(r,g,b) (b >> 3) + ((g >> 2) << 5) + ((r >> 3) << 11))

- Splat

### #6 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

Likes

Posted 26 November 1999 - 07:54 AM

You can just use fast modulus. Replace %32 with &31.
It only works with powers of 2. Formula is %n = &(n - 1)
or %32 = &(32 - 1) = &31.

### #7Splat  Members   -  Reputation: 122

Like
Likes
Like

Posted 26 November 1999 - 09:07 AM

Shoot, I should've seen the error before...

Problem with "modulus"ing the color values is that we want the MOST significant bits of the source color, not the least. & 31 and % 32 will cut off the top three bits that have the _most_ relevant data for a 5-bit / 6-bit color version.

With shifting - the correct way -, 10111011 (187) will become 10111 (23) in 5-bit, while using modulus will get you 11011 (27). If you were to convert these back to 8-bit, the shifted 10111 would become 10111000 (184) and the modulus one would become 11011000 (216). Which is closer to the original 187?

An even more ridiculous example is 00011111 (31) in 8-bit, is converted to 5-bit as 11111 (31), which is the brightest color in 5-bit, while in 8-bit it is quite dark. I can't believe I read this in LaMothe and didn't realize. I guess that book has more errors than I saw at first...

- Splat

[This message has been edited by Splat (edited November 26, 1999).]

### #8CoolMike  Members   -  Reputation: 122

Like
Likes
Like

Posted 26 November 1999 - 09:08 AM

Ok, I tried those different macros but they still are not working for me. When I change the RGB values, they don't work correctly. All I can get are greens and blues. Any other ideas?

### #9 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

Likes

Posted 26 November 1999 - 06:40 PM

What I should of explained was that the modulus is not a substitute for shifting.
I was just optimizing the use of %32.

To use the Lamothe 16-bit macro correctly, you have to use 16-bit values, e.g. you
can't stick in 24-bit values directly like (255,255,255) even though this particular
example will come out correctly as (1F,1F,1F). You have to shift FIRST, then use
the 16-bit macro.

### #10Jim Adams  Members   -  Reputation: 440

Like
Likes
Like

Posted 27 November 1999 - 04:27 AM

It may be your bitmap_buffer pointer is not a char pointer. Since you are using x*2 to get word alignment and are storing only one byte...

Jim

### #11CoolMike  Members   -  Reputation: 122

Like
Likes
Like

Posted 27 November 1999 - 06:08 AM

I removed the *2 in the x*2 part, and that seemed to give me a LITTLE control over my colors, although it put my pixel farther to the left on my screen. Here is some code I used to make my RGB macro, define my buffer, and draw my pixel ):

#define _RGB16BIT(r,g,b) ((b%32) + ((g%32) << 5) + ((r%32) << 10))

typedef unsigned char UCHAR;
UCHAR *bitmap_buffer = NULL;
bitmap_buffer = (UCHAR *)ddsd.lpSurface;

bitmap_buffer[x+ddsd.lPitch*y]=_RGB16BIT(255,255,255);

Now the colors are better, but not perfect. (0,0,0) makes black, (255,255,255) makes white, but then (0,0,255) makes a really bright blue (it should be dark blue), and other combinations don't work either.

Now about that shifting my color values: I am fairly new to programming, and I am only beginning to grasp the whole shifting thing. Now Vader said that (255,255,255) is 24-bit values, so how EXACTLY do I convert those to 16-bit values?

[btw, I tried those other macros and they all worked the same for me, however I didn't time them so their speeds may vary)

[This message has been edited by CoolMike (edited November 27, 1999).]

### #12Jim Adams  Members   -  Reputation: 440

Like
Likes
Like

Posted 27 November 1999 - 09:03 AM

I prefer to shift the values:

#define _RGB16BIT(r,g,b) ((r>>3)<<10) | ((g>>3)<<5) | (b>>3)

But, you should replace the pointer as such:

typedef unsigned short USHORT
USHORT *bitmap_buffer = NULL;
bitmap_buffer = (*USHORT)ddsd.lpSurface;
bitmap_buffer[x+ddsd.lpPitch*y] = _RGB16BIT(255,255,255);

I don't remember offhand if you have to divide lpPitch by two beforehand. Try both ways.

BTW - RGB(0,0,255) should make a bright blue (unless you mean it's adding some green for a teal color).

Remember - when writing 16 bit colors, you should use a 16-bit pointer (USHORT), not a byte pointer (CHAR).

Jim

### #13Tuan  Members   -  Reputation: 122

Like
Likes
Like

Posted 30 November 1999 - 04:27 PM

1/ Your explaination is very clear, Splat! I have ever written on 16 bit colors. The problem is: we must get the 5 most significant bits of each component, ignore the 3 low bits. So, we must use the shift bit macro.
2/ The pixel's format of each Video card is different. Some card are 555, and anothers are 565 (S3 Trio 64V+ is an ex.). So, you must get the pixel's format to get the correct macro.
3/ I have a problem on ATI card too! Sorry, I don't remember the version of my card. But I want to tell you know that: the GDI functions will destroy the DirectSurface interface! It means: in my program, I use the LoadBitmap function to load the external bitmap file, get Device context of surface GetDC(), BitBlt the memory device context to Surface. And I had error when blitting with src color key. So, I must load bitmap into surface manually, not using the GDI functions! And it work well.
Goodluck.

------------------

### #14CoolMike  Members   -  Reputation: 122

Like
Likes
Like

Posted 07 December 1999 - 06:13 PM

I am getting nowhere with this. I failed to get it to work when I changed my UCHAR's to USHORT's and I know my ATI card is 555 because I can already load sprites from bitmaps and display them just fine on the screen. It is only when I blit pixels and use my _RGB16BIT macro that it doesn't work right.
BTW, would anyone be willing to be emailed my project? I would really like it if someone could actually look at my entire project (as small as it is) and tell me what's wrong. It is about 70kb and is for Visual C++ 5.0. Thanks in advance for any gracious soul willing to undertake this task!

-Mike

### #15Splat  Members   -  Reputation: 122

Like
Likes
Like

Posted 08 December 1999 - 12:29 PM

Send away! My email should be listed in my profile...

- Splat

Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

PARTNERS