Help with GBA ARM assembly program

I tried to modify this ARM assembly program (see bottom of the page) to use subroutines. It should display a red screen on the GBA (and compiling the exampl开发者_如何学编程e program it actually does, so it's not a toolchain use problem), but, when I run it, the screen goes black instead. I'm using devkitARM 30

.arm
.text
.global main
main:
    mov r0, #0x4000000
    mov r1, #0x400
    add r1, r1, #3
    str r1, [r0]
    mov r0, #0x6000000
    stmdb sp!, {r0, r2, lr} @push registers before calling
    mov r0, #0xff
    mov r1, #0
    mov r2, #0
    bl set_rgb
    mov r1, r0
    ldmdb sp!, {r0, r2, lr} @restore registers
    mov r2, #0x9600
loop1:
    strh r1, [r0], #2
    subs r2, r2, #1
    bne loop1
infin:
    b infin

set_rgb:
    @r0: R; r1: G; r2: B; r0: RGB

    @truncate everything but the least significant five bits
    and r0, r0, #0x1f
    and r1, r1, #0x1f
    and r2, r2, #0x1f
    @form the color
    orr r0, r0, r1, lsl #5
    orr r0, r0, r2, lsl #10
    mov pc, lr @return

The question is, what's wrong in this program?

I solved it on my own.

The problem was the way I used the stack. Instead of stmdb and ldmdb I needed to use stmfd and ldmfd.

stmdb means decrement before and then use that address to start writing to the stack, this is proper.

ldmia means increment after so start with the current stack pointer to read the values back to their registers then increment the stack pointer.

The fd nomenclature never made sense to me. Like jump if equal and jump if zero are the same instruction for all processors and some asms offer both, there are only two flavors of ldm and stm the db,ia,fd all map into those two flavors.

I find it easier to remember increment after (ldmia) and decrement before (ldmdb). Or if you turn the load/store direction bit around for some reason then you still choose the right increment before or after depending on what you are trying to do.

In C it is like *(ptr++) vs *(++ptr)