Well, it's not a "tool" exactly, but a kind of simulator. The purpouse of this is to give a opportunity to freely experiment with user selectable AA patterns. The tool consist of two parts - the AA render and the AA sampler:
The AA samples coordiantes are stored in .txt file (samples.txt) and the count is limited to 100 of them. The sample coord's within a pixel ranges from 0.000000 to 1.000000 (note the precision), but the kernel size is not limited to one pixel, so you can make some offpixel sampling. There is no fixed grid however - if you want some ordered position for the samples, do some math and edit the coord's in the .txt file.
To determine the size of the window, you can start the program with following switches: "--xres= --yres=", ex.: fsaatest.exe --xres=128 --yres=128 for a smaller and faster window (note, that the performance decrease is linear to the screen pixels and number of samples).
The program is in a veeery alpha stage, implicting limited functionality (for now) and it's simple rendering engine is 100% software based - from the transformations to the final sampling and blending. The internal precision is 32-bit FP, however the rendertarget is 8-bit (so the autor is claiming that). Any feedback and reccomendations are welcome, and I hope the autor may join here, for a conversation.
Grab the current build right here. <- updated!
The AA samples coordiantes are stored in .txt file (samples.txt) and the count is limited to 100 of them. The sample coord's within a pixel ranges from 0.000000 to 1.000000 (note the precision), but the kernel size is not limited to one pixel, so you can make some offpixel sampling. There is no fixed grid however - if you want some ordered position for the samples, do some math and edit the coord's in the .txt file.
To determine the size of the window, you can start the program with following switches: "--xres= --yres=", ex.: fsaatest.exe --xres=128 --yres=128 for a smaller and faster window (note, that the performance decrease is linear to the screen pixels and number of samples).
The program is in a veeery alpha stage, implicting limited functionality (for now) and it's simple rendering engine is 100% software based - from the transformations to the final sampling and blending. The internal precision is 32-bit FP, however the rendertarget is 8-bit (so the autor is claiming that). Any feedback and reccomendations are welcome, and I hope the autor may join here, for a conversation.
Grab the current build right here. <- updated!
Last edited by a moderator: