Learn Multiplatform ARM Assembly
Programming... For the Future!
Multiplatform Lessons
In this series we look at code which does not rely on the hardware of a
specific system, these are useful code fragments which will work on any ARM
based system

|
Lesson M1 - Random Numbers and Ranges
Let's write some simple routines we may need in a game!
We'll write a psuedo random number generator, which can create
repeatable 16 bit random numbers.
We'll also create a 'Range checker' which can be used as a
collision detection routine for unsigned 8 bit numbers |
 |

V1_Random.asm |
|
 |
 |
The
Random number generator shown today is far from the 'Best'...
but it does work!
It's used in these tutorials by all the games as it's been
tested without any problems, and has been ported to many
different systems and CPUs, and given the same random seed, it
gives the same results... meaning it can be used for
generating random levels in your game, and the level will be
the same on every system.
|
Random number generation
| To allow our random number
generator to create good varied results, we use a pair of 16
byte lookup tables |
 |
We have two random number generators,
The first returns a single byte in R0... using rotations and
EOR, using two 8 bit seeds R1/R4
|
|
We need the 'Low byte' to be very different, or the 16 bit
values won't be very random!
We use the lookup tables to get some new data, and combine these
to produce the second value |
 |
We have two routines we can use to get values
"DoRandomWord" will get a 16 bit value R3,R6 from seed R1, this
can be used to produce 'psuedorandom' data for times we want
repeatable results.
"DoRandom" is an easy function to return an 8 bit byte in R0...
it automatically updates its seed every time. |
 |
To
be usable, a random number generator needs to produce every
possible random number (0-255 or 0-65535)... otherwise you
could have serious problems with your program (if you're
waiting for a result that never happens).
The more 'random' the data the better... that is, if you
plotted the values on a graph there should be no patterns
present.
|
 |
Range checking
We need to do 'Collision detection'.
We specify a two targets and a 'range' ... if target
1 is within 'Range' of target
2, we return A=1... if it's out of range we return A=0...
This means we can do a BNE or BEQ on return. |
 |
The range check is simple... we test X and Y axis
We test the X axis first.
we subtract the 'RANGE' from each
axis of the object, and check if we went
below Zero (if so that direction is over the limit)... we then
compare to the position... if it's lower
we're out of range... if not we need to test more!
Next we add 'RANGE' twice... once to
move back to the center and once to move to the right... we then
compare to the position...
If it's higher we're out of range... if not we need
to test the Y axis!
We repeat the same tests for the Y axis. |
|

|
Lesson M2 - BCD, Binary Coded Decimal!
In games we may wish to use binary coded decimal to store scores in
a way that's easy to convert to Ascii
Here we'll look at some simple routines to support BCD on ARM. |
 |

BCD_Test.asm
|
|
 |
The Battle for BCD
Converting large numbers from Hex to Ascii is hard!
A common simple solution to this is Binary Coded Decimal... This
stores a single 0-9 digit in each nibble - Two per byte, This
is known as PACKED BCD!
This is easier to show to screen than normal hex numbers.
|

 |
BCD only uses a 0-9 value in each nibble... but of course each
nibble can really support 0-F
Suppose we repeatedly add 1 to the value 8... we'll need to correct
the value once the value goes over 9
values like $0A will need to be converted to the correct BCD value
of $10... and we'll need to do that ourselves! |
$08 +1
$09 +1
$0A -> $10 |
To ShowBCD, we need to step through each of
the 8 digits in our 32 bit value..... here we'll show R1 to the
screen
we rotate a digit (nibble) out of R1 into R0, add #48 to convert it
to Ascii, and show it to the screen with PrintChar
We repeat until all 8 characters are shown. |
 |
We'll need to AddBCD to add score... but this is
a little more tricky!
Here we're adding our values one digit at a time, but we need to
cope with any carry!...
If a digit goes over 9, we need to add 6, skipping over the A-F part
of the nibble, and storing any carry for the next digit in R6
we repeat for all 8 digits. |
 |
We can write SubBCD to subtract BCD in a
similar way
Once again we work through the digits keeping track of the carry. |
 |
 |
We'll also need to compare two BCD values
for our highscore routine, but we don't need any special code to
do this, we can just use the normal CMP to do the job!
|
Lesson
M3 - The GenericAnimator code - Part 1
GenericAnimator is based on the Animation Scripts from ChibiAkumas
episode 2.
They allow objects to follow simple movement sequences
'automatically' without extra program code or intervention. |
 |

V1_GenericAnimator.asm |
|
 |
GenericAnimator Introduction
GenericAnimator performs all the movement animations for the
characters in the ChibiLife demo. Movement and sprite animation is
all done by these GenericAnimator scripts
The Collision detections are handled by special code, but the
GenericAnimator checks these collision flags are detected, and
characters 'Turn round' automatically
A pointer to the object to be animated is passed to the animator
code, and changes will be made to things like Xpos, Ypos and sprite
number etc to 'automatically' animate the object.
|
 |
Pointers for the animation scripts are held in a list.
The first entry is 'animator 1' as animator 0 is 'no animation'
|
|
Animation scripts start with a 1 byte 'Tick Mask'. This is ANDed
with the current game tick - if the result is zero, the next step of
the animation will occur.
Each line of an animation script is 4 bytes , the first byte is the
'command' the other three are parameters for that command.
Note that while a tick normally processes 1 line, it's possible to
process extra lines per tick with the 'aEXT' flag.
|
|
Code - Init
At the start of the script processor we will be
passed
An animator number in R1
An animator Frame/Tick in R4 (the line number we will process next
in the script)
We back these up, then check the animator number - 0 Means no
animation in which case we return.
We use GetAnimatorMemPos to calculate the address
of this script, then load in the first byte of the animation script
This first byte is the 'Tick Mask' - it's ANDed with the current
game tick, if the result is zero, the animation script processes a
line.
BUT we can override this, and use a custom tick, according to the
value in R3/R6
If the tick should occur this time, we run objectanimator_ProcessTick
Which will process a line of the script |

|
objectanimator_ProcessTick take's
the current script line number and multiplying it by four (as each
line is 4 bytes)
We load in the command byte number, the bottom 4 bits are the
command (16 commands total MAX) (The top bits are for conditionals
and multiline support)
we use this to load a pointer in the animator_vectorarray
which is the address of the command processor - we call it now
|

|
Code - Extended Commands
The first command is a
special one!
EXT allows for extended commands. 'Extended
commands' use the first parameter byte as a 'command number' - This
allows more commands (an extra 255!), but allows only 0-2 parameters
(rather than 0-3)
We load the command number and use it as an offset to the animator_vectorarrayext
list |


|
| LoopToStart resets the frame/tick number back to
0 |
 |
Halt stops the animation - by setting the
animator number to 0
|
|
SetAnim sets the Animator and Frame/line (0=first
frame).
|
|
vmexec Executes a call within the 'ChibiVM'
virtual machine.
The address to call is in 16 bit little endian.
|
|
Call runs a call natively on the ARM cpu.
As the destination address needs to be 32 bit, we load a 1 byte
number from the command parameter, and use it as an offset in a call
list
|
|
SwapIXIY allows a pointer to an alternate object
to be used (IX and IY were registers on the Z80)
This should only be used as part of a multiline sequence! (using
aEXT)
|
|
Lesson
M4 - The GenericAnimator code - Part 2
GenericAnimator is based on the Animation Scripts from ChibiAkumas
episode 2.
They allow objects to follow simple movement sequences
'automatically' without extra program code or intervention. |
 |

V1_GenericAnimator.asm |
|
 |
Commands
condjmp will jump to an alternative
frame/tick based on a condition
condprocess sets the T flag according to a
condition. objectanimator_condprocess will perform the comparison -
The condition is selected by the top 3 bits of the command byte.
Parameter 1 is the offset in the object data to be tested/
Parameter 2 is the 'comparison value'
Parameter 3 is the new tick/frame for the animation script
The top 3 bits of the command byte define the condition. |
 |
condbra will branch to an alternative frame/tick
based on a condition, This is a relaitve offset to the tick number,
so allows a number of ticks to be 'skipped' and may be more useful
than jump for skipping a few commands.
Parameter 1 is the offset in the object data to be tested/
Parameter 2 is the 'comparison value'
Parameter 3 is the new offset
The top 3 bits of the command byte define the condition. |
|
load16 will load a 16 bit value into an offset of
the objects variables. The data is in LITTLE ENDIAN format.
This is useful for changing a 16 bit sprite pointer.
|
|
load8 will load an 8 bit value into
an offset of the objects variables.
This is useful for changing a Xpos or Ypos |
|
load8dual Sets two separate 8 bit values.
The first byte is a pair of offsets (in the object data) - one is
defined by the low nibble, the other by the high byte of the nibble.
This means only the first 16 bytes of the object data can be changed
in this way. |
|
add8dual Adds two separate 8 bit values.
The first byte is a pair of offsets (in the object data) - one is
defined by the low nibble, the other by the high byte of the nibble.
This means only the first 16 bytes of the object data can be changed
in this way. |
 |
addmasked performs an add and mask operation
The first parameter is added to the destination
The second parameter is ANDed with the result.
By setting the AND mask to a value of #3 it would be possible to
cycle between values 0,1,2,3 (maybe for toggling sprite frames)
By setting the AND mask to #255, this acts as as simple ADD |
 |
loadadd8dual Loads one 8 bit value to one object
value, and adds another 8 bit value to another object value
The first byte is a pair of offsets (in the object data) - one is
defined by the low nibble, the other by the high byte of the nibble. |
 |
condld8 will load a new 8 bit value to the object
value if the condition is true.
Parameter 1 is the offset in the object data to be tested/changed
Parameter 2 is the 'comparison value'
Parameter 3 is the new value to set if the condition is true.
The top 3 bits of the command byte define the condition.
|
|
condadd8 will conditionally add an 8
bit value to the object value if the condition is true.
Parameter 1 is the offset in the object data to be tested/changed
Parameter 2 is the 'comparison value'
Parameter 3 is the new value to set if the condition is true.
The top 3 bits of the command byte define the condition. |
|
condanim will conditionally switch to a new
animator script (starts from frame 0)
Parameter 1 is the offset in the object data to be tested/changed
Parameter 2 is the 'comparison value'
Parameter 3 is the new animator to set if the condition is true.
The top 3 bits of the command byte define the condition.
|
|
Helper routines
| objectanimator_update is executed after the
command, Bit 5 of the original command number is tested, this is the
'EXTra line' flag, which allows multiple lines to be processed
in a single tick. |
|
offsetdestnibble routines will set
R5 to the object offsets according to two 4 bit offsets.
The top nibble can optionally switch to R8/IY as the source object,
based on bit 7 of the command byte (used for conditions on some
commands)
|
|
offsetdestfromhl will set R5 to an offset defined
by a single byte.
There are two versions, one for conditions (which cannot swap IY) ,
and one for non-conditions (which can swap IY)
|
|
Condition processing
The conditions
take a byte from the object data, and compare it to an immediate
value parameter.
One of 8 possible conditions can be tested, decided by the top 3
bits of the command byte.
If the condition is true an action will be performed.
Here we load the parameters, and branch to the condition processing
routines. |
 |
Condition type 7 is a special case, it's ANDed with the world
tick. and is true if the result is zero.
This allows events to occur only rarely, for example anding with a
value %00000010 would make something happen for two ticks, every two
ticks.
|
|
We perform the required comparison and set the carry flag
according to the result.
Note we clear R4 once we're done - it contained the command before,
but the top bit is used by some 'non conditional' commands for R7-R8
/ IX-IY flip
|
|
| |
Buy my Assembly programming book on Amazon in Print or Kindle!



Available worldwide! Search 'ChibiAkumas' on your local Amazon website!
Click here for more info!
Buy my Assembly programming book on Amazon in Print or Kindle!



Available worldwide! Search 'ChibiAkumas' on your local Amazon website!
Click here for more info!
Buy my Assembly programming book on Amazon in Print or Kindle!



Available worldwide! Search 'ChibiAkumas' on your local Amazon website!
Click here for more info!
|