acme/docs/Floats.txt

65 lines
2.6 KiB
Plaintext
Raw Normal View History

ACME
...the ACME Crossassembler for Multiple Environments
--- floating-point support ---
In release 0.92, ACME gained some experimental support for floating-
point maths, so you can finally build sin/cos tables directly in ACME.
But the expression parser still uses integer calculations per default.
Here are the rules:
a) if a maths operation is useless when done with integers, it is done
with floats and returns a float. Applies to sin(), cos(), tan(),
arcsin(), arccos(), arctan() and float(): These are always computed in
float mode and always return floats.
b) if a maths operation is useles when done with floats, it is done
with integers and returns an integer. Applies to NOT, AND, OR, XOR,
MOD, DIV, LSR, lowbyteof, highbyteof, bankbyteof and int(). These are
always computed in integer mode and always return integers.
c) All other mathematical operations are done in float mode if and
only if at least one of the operands is a float. So "1/2*2" will give
zero because it is done in integer mode, but "1.0/2*2" will give 1
because it is done in float mode.
To force a numerical value to be flagged as being a float, just add
a decimal point and a zero. If a decimal value ends with a
dot character, ACME switches to using the C type "double" and keeps
reading digits. The value is then flagged internally as being float.
CAUTION: Float usage is only activated when a decimal point has been
found, so don't expect "100000000000000000000" to work.
"100000000000000000000.0" won't work either, because when the decimal
point gets parsed, the integer value will have overflown already. As
there is no support yet for the "e" format (1e20 for 1*10^20) either,
the only way to enter this value is by writing "1.0 * 10.0 ^ 20.0".
Examples:
!byte 1 / 2 * 2 ; gives 0 (integer maths)
!byte 1 / 2 * 2.0 ; gives 0 (1/2 => 0 in integer maths,
; float usage comes too late)
!byte 1 / 2.0 * 2 ; gives 1 (FP in effect)
!byte 1 / 2.0 * 2.0 ; gives 1 (FP in effect)
!byte 1.0 / 2 * 2 ; gives 1 (FP in effect)
!byte 1.0 / 2 * 2.0 ; gives 1 (FP in effect)
!byte 1.0 / 2.0 * 2 ; gives 1 (FP in effect)
!byte 1.0 / 2.0 * 2.0 ; gives 1 (FP in effect)
You can use the new float() and int() functions to ensure the type of
maths:
!byte a / b * c ; depends on a/b/c's internal flags
!byte float(a) / b * c ; calculation is done as float
!byte int(a) / int(b) * int(c) ; calculation is done as int
As you will have guessed, the trigonometric functions assume radians
for measuring angles (90 degrees equals PI/2).
Have a look at the example source code "trigono.a", it builds some
sin/cos tables.