You're using a 16.16 format? That seems a bit overkill, and it even makes adding/subtracting hard, right?

Edit: OHWAIT this is for the Nspire contest. durrrr

Edit2:
I think something like this is how it's done:
int quotient = ((long long)dividend << 16)/divisor;

Thanks!
Also, just want to make sure, am I doing multiplication right?
product = (a >> * (b >>

If you want a fully accurate answer, I think you'd need ((long long)a * b) >> 16

Hmm, maybe I should write some ARM ASM routines for fixed-point multiplication/division. Doing it normally through GCC is going to be a bit inefficient.

Edit:
Just to confirm, are you using signed values?

But the way I did it would prevent overflows, because I am only using 32 bit integers the whole time.
Yes, I'm using signed numbers. Will that introduce problems?
Also, am I understanding the division correctly? a / b in fixed point is [integer part of a]/b?

You can have overflow when multiplying 32-bit integers, if the absolute value of the result is more than 2^31. If you want, I could implement saturation in my routines (so anything outside of the range will be turned into the maximum or minimum value).

And the division is (a * 2^16)/b (this is to nullify the factor of 2^16 present in b)

Apparently ARM7 does not natively support long long integers. So that means I would get odd results if I tried division with any number with an absolute value of over 1.
Would (a <<  8 ) / (b >>  8 ) work? I would still get an overflow if I used anything over 65536 in fixed point, but I don't think I will get to numbers that high.
Also, is there a tag to disable smileys?

No ARM processor natively supports 64-bit, but GCC should be able to generate code anyway.
A big problem with (a << 8) / (b >> 8) is that the top 8 bits of a are lost before the division. This means that numbers greater than 256 wouldn't work.

By the way, I've finished with a fixed-point multiplication routine now and I'm working on division. You should be able to put this in a .S file and include it in your makefile, and you can use it if you define the function prototype in your C code.
@ int fixed_mul(int, int);
.global fixed_mul
fixed_mul:
@ Multiply 32x32->64 and shift right by 16
smull r0,r2,r1,r0
mov r0,r0,lsr #16
orr r0,r0,r2,lsl #16
@ Overflow check start
teq r2,r0,asr #16
movne r0,#0x7FFFFFFF
eorne r0,r0,r2,asr #31
@ Overflow check end
bx lr
Edit: Here's the fixed-point division routine (note: neither of these routines have been tested yet)
@ int fixed_div(int, int);
.global fixed_div
fixed_div:
@ r2 = abs(r0)
eors r2,r0,r0,asr #32
adc r2,r2,#0
@ r3 = abs(r1)
eors r3,r1,r1,asr #32
adc r3,r3,#0
@ Get sign of result
eor r1,r1,r0
@ Initialize result
mov r0,#0x7FFFFFFF
@ Check for overflow
cmp r3,r2,lsr #15
bls fixed_div_end

.irp shift,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
cmp r3,r2,lsr #\shift
subls r2,r2,r3,lsl #\shift
bichi r0,r0,#0x10000 << \shift
.endr
.irp shift,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
rsbs r2,r3,r2,lsl #1
addlo r2,r2,r3
biclo r0,r0,#1 << \shift
.endr

fixed_div_end:
@ Adjust sign
eors r0,r0,r1,asr #32
adc r0,r0,#0
bx lr
Edit2: It's looking likely that the division routine doesn't work :(

Edit3: Replaced the division routine with one that might work (at least, the equivalent C code seems to work)

I was getting major errors so I replaced my multiply and divides with the ones in C you mentioned above, using long longs, and now I am getting errors when I compile:
$ make
nspire-gcc -Os -Wall -W -c main.c
main.c: In function 'main':
main.c:52:7: warning: unused variable 'oldtime'
main.c:51:7: warning: unused variable 'time'
main.c:46:8: warning: unused variable 'timer'
nspire-gcc -Os -Wall -W -c utils.c
utils.c: In function 'filesize':
utils.c:129:3: warning: passing argument 2 of 'stat' from incompatible pointer type
c:/ndless-v2.0/sdk/include/os.h:288:1: note: expected 'struct stat *' but argument is of type 'char *'
nspire-gcc -Os -Wall -W -c graphics.c
nspire-gcc -Os -Wall -W -c math.c
nspire-ld  main.o utils.o graphics.o math.o -o raycaster.elf
c:/program files/yagarto/bin/../lib/gcc/arm-none-eabi/4.5.1\libgcc.a(unwind-arm.o): In function `get_eit_entry':
C:\msys\1.0\home\yagarto\gcc-build\arm-none-eabi\libgcc/../../../gcc-4.5.1/libgcc/../gcc/config/arm/unwind-arm.c:614: undefined reference to `__exidx_start'
C:\msys\1.0\home\yagarto\gcc-build\arm-none-eabi\libgcc/../../../gcc-4.5.1/libgcc/../gcc/config/arm/unwind-arm.c:614: undefined reference to `__exidx_end'
c:/program files/yagarto/bin/../lib/gcc/arm-none-eabi/4.5.1/../../../../arm-none-eabi/lib\libc.a(lib_a-abort.o): In function `abort':
C:\msys\1.0\home\yagarto\newlib-build\arm-none-eabi\newlib\libc\stdlib/../../../../../newlib-1.18.0/newlib/libc/stdlib/abort.c:63: undefined reference to `_exit'
c:/program files/yagarto/bin/../lib/gcc/arm-none-eabi/4.5.1/../../../../arm-none-eabi/lib\libc.a(lib_a-signalr.o): In function `_kill_r':
C:\msys\1.0\home\yagarto\newlib-build\arm-none-eabi\newlib\libc\reent/../../../../../newlib-1.18.0/newlib/libc/reent/signalr.c:61: undefined reference to `_kill'
c:/program files/yagarto/bin/../lib/gcc/arm-none-eabi/4.5.1/../../../../arm-none-eabi/lib\libc.a(lib_a-signalr.o): In function `_getpid_r':
C:\msys\1.0\home\yagarto\newlib-build\arm-none-eabi\newlib\libc\reent/../../../../../newlib-1.18.0/newlib/libc/reent/signalr.c:96: undefined reference to `_getpid'
collect2: ld returned 1 exit status
make: *** [raycaster.tns] Error 1


Okay, I just removed the function from utils.c that gave the warning since I'm not using it anyways. I also tried using -nostdlib, and the compiler output looks normal, but I don't get a tns document. Also, not having 32 bit division will be an issue for me if I get -nostdlib working.

I have a multiplication routine which almost works. Basically, I isolate the upper and lower halves of a and b each to the bottom half of a new variable, use the FOIL method and add them together.
inline long mul(long a, long b) {
  long a1 = a >> 16;
  long a2 = a & 0x0000FFFF;
  long b1 = b >> 16;
  long b2 = b & 0x0000FFFF;
  return ((a1 * b1) << 16) + (a1 * b2) + (a2 * b1) + ((a2 + b2) << 16);
}I tried to multiply 0x00028000 and 0x0002000 (2 * 2.5), but I got 0x80050000, it seems that the sign bit gets messed up. What am I doing wrong?