glTexSubImage2d vs WEBGL's GL.texSubImage2d

Too lazy to dig through all this? Visit me on twitch and I'll help you, since I obviously know!

www.twitch.com/KanjiCoder

www.twitch.com/KanjiCoder

 I could get GL.texSubImage2D working... but when I tried to use the OpenGL version... I noticed an initial offset into the cpu-side bitmap (rgba byte array) was NOT a parameter to the OpenGL version of the function.

Here are my fix notes for OpenGL call:

QUESTION( via facebook):

I can get WebGL's "texSubImage2D" to work... But no luck in OpenGL. The signature for WEBGL includes a "stride". But the OpenGL signature does not. Think that is what is messing me up? ANSWER( answered my own question later on facebook)

Answer:
Because OpenGL's version does NOT include an initial offset into the data like WebGL's version, you need to do some pointer math on the uint8_t byte array on CPU side that houses your image.
I have a 512x512 bitmap on CPU side.
I have a 512x512 texture on GPU side.
To update a 32x32 section of 512x512 texture using a selection rectangle
from the 512x512 bitmap, you must specify:
STATE:
1. Unpack alignment( 4 because I use RGBA, 4 bytes per pixel)
2. Row alignment (stride, in pixels for me)
3. First byte of data strip in glTexSubImag2D call.
To get item 3, your last param to glTexSubImage needs to be something
along the lines of this:
&( my_cpu_side_bitmap[
( top_left_pixel_of_selection_rect_x + (top_left_pixel_of_selection_rect_y * 512 ) ) * 4
]
4 is because RGBA, 4 bytes per pixel.
512 is the pixel stride of an RGBA bitmap byte array.
In general, we are using an XY-to-INDEX formula and then
multiplying by the components per pixel size.

If you need more info on how his works.... You can download my code base from github.

https://github.com/KanjiCoder/AAC2020

https://github.com/KanjiCoder/AAC2020

The relevant function you want is:

aac2020_tausync_glTexSubImage2D

Within this file:

https://github.com/KanjiCoder/AAC2020/blob/master/LIB/TAUSYNC.F._

-John Mark Isaac Madison



Comments

Post a Comment

Popular posts from this blog

How to compile C without an IDE

  How to compile C without an IDE. Specifically Notepad++... Come visit my stream on twitch and I will walk you through it. www.twitch.com/kanjicoder If you can, come prepared with: 1. Notepad++ installed 2. Git for windows installed 3. GCC Downloaded But if that's too much work, just drop on by and I'll walk you through all of that setup too. -John Mark Isaac Madison
Read more

Clean Code Sucks , Make Up New Words!

Lego Naming Prevents Higher Level Thinking : GunPowderedPoweredLargeIronBallLauncher <<-- Clean Code Cannon <<-- Terse Invented Concept In Your Code If you agree. Check out my coding stream. I do Data Oriented Game Engine Design. (DOGED) www.twitch.com/kanjicoder www.twitch.com/kanjicoder //:2021_02_23:===============================================:// WindowExternalVisibilityController ==> BLIND ExtraFlufflyCushionedChair ==> LazyBoy MultiBarrelRotatingMiniCanno ==> GatlingGun LargeIronBallLauncher ==> Trebuchet GunPowderedPoweredLargeIronBallLauncher ==> Cannon You cannot think of "GunPowderedPoweredLargeIronBallLauncher" at the next level of abstraction until you condense the word into a TERSE/SHORT word. Example: You won't invent a "Gauss Cannon" until you STOP CALLING YOU CANNON A: "GunPowderedPoweredLargeIronBallLaunched" Same applies to your code. ...
Read more

0x400000

 0x400000 : The nearest power of 2 (rounded down) when taking the max value of a signed 32 bit integer and dividing it by 256.      ROUND_DOWN_TO_NEAREST_POWER_OF_2(  I32_MAX_VALUE / 256 )  Why? Working on rendering algorithm where I want to do floating point math using discrete units. In order for this to work, I need to upscale all of my 1x1 pixel tiles into larger grids. Since I have a grid of 256x256 "pixel sized tiles" on my "canvas user view" (cuvipix), I can get away with scaling up each pixel by a factor of 0x400000 before I start doing my math. I choose to use the max value of SIGNED integer rather than unsigned because everything is eventually getting converted to 32 bit float. And since floats can go negative, trying to partition the max value of a SIGNED int may result in TOO MUCH addressing space. We don't want to lose precision when we convert of discrete integer fractions into floating point if we can help it. Hmm... Now that I think of it......
Read more