These registers are responsible for controlling how framebuffer data can be DMA'd from the DS GPU, and also for configuring the upscaling matrix.
The physical address can be calculated by subtracting 0xEB00000 from the virtual address.
There are two matrix units, one at +0x200 for vertical (Y) scaling, and the other one at +0x300 for horizontal (X) scaling.
|0x1EC1xn00||KRN_WIDTH||4||Kernel width - 1 is written here, 1 <= width <= 8
This decides how many pixels are written each batch.
|0x1EC1xn04||KRN_PATTERN_BITS||4||If the corresponding bit for the current batch iteration index is set then a new pixel is read.
The amount of set bits determine how many pixels are read each batch. Any bit indexes past KRN_WIDTH are ignored.
This value is 8 bits, but it has to be written with a 32bit write.
|0x1EC1xn40||KRN_MTX||0xC0||int kerneldata; - matrix data is written here, height is always 6|
|0||Enable bit (?)|
|1||Enable vertical matrix|
|2||Enable horizontal matrix|
|8-9||Input pixel mode? 0 = 4byte color, 1 = 3byte color, 2 = 2byte color, 3 = 2byte color|
|10-11||Output framebuffer rotation: 0 = normal, 1 = 90° CW (right), 2 = 180° CW (upside down, not mirrored), 3 = 270° CW (left)|
|12||Output tiling for use with the GPU. When set, the output width and height must be a multiple of 8.|
|15||Interrupt enable (?)|
|16||Data still available flag (?)|
|0-8||Output framebuffer width - 1 is written here, 1 <= width <= 512|
|16-25||Output framebuffer height - 1 is written here, 1 <= height <= 512|
Reading this register will return pending interrupts. Writing this register will acknowledge the pending interrupt.