This commit is contained in:
Virgil Dupras 2020-05-03 12:54:22 -04:00
parent 47a9c89f19
commit d3dbeb450f
2 changed files with 0 additions and 52 deletions

1
.gitignore vendored
View File

@ -1,2 +1 @@
*.o *.o
/kernel/user.h

View File

@ -1,51 +0,0 @@
This file describe tricks and conventions that are used throughout the code and
might need explanation.
*** Quickies
or a: Equivalent to "cp 0", but results in a shorter opcode.
xor a: sets A to 0 more efficiently than ld a, 0
and 0xbf: Given a letter in the a-z range, changes it to its uppercase value
if it's already uppercased, then it stays that way.
*** Z flag for results
Z if almost always used as a success indicator for routines. Set for success,
Reset for failure. "xor a" (destroys A) and "cp a" (preserves A) are used to
ensure Z is set. To ensure that it is reset, it's a bit more complicated and
"unsetZ" routine exists for that, although that in certain circumstances,
"inc a \ dec a" or "or a" can work.
*** Little endian
z80 is little endian in its 16-bit loading operations. For example, "ld hl, (0)"
will load the contents of memory address 0 in L and memory address 1 in H. This
little-endianess is followed by Collapse OS in most situations. When it's not,
it's specified in comments.
This get a bit awkward with regards to 32-bit. There are no "native" z80 32-bit
operations, so z80 doesn't mandate an alignment. In Collapse OS, 32-bit numbers
are stored as "big endian pair of little endian 16-bit numbers". For example,
if "ld dehl, (0)" existed and if the first 4 bytes of memory were 0x01, 0x02,
0x03 and 0x04, then DE (being the "high" word) would be 0x0201 and HL would be
0x0403.
*** DAA
When it comes to dealing with decimals, the DAA instruction, which look a bit
obscur, can be very useful. It transforms the result of a previous arithmetic
operation involving two BCD (binary coded decimal, one digit in high nibble,
the other digit in low nibble. For example, 0x99 represents 99) into a valid
BCD. For example, 0x12+0x19=0x2b, but after calling DAA, it will be 0x31.
To clear misunderstanding: this does **not** transform an arbitrary value into
BCD. For example, "ld a, 0xff \ daa" isn't going to magically give you a binary
coded 255 (how could it?). This is designed to be ran after an arithmetic
operation.
A common trick to transform an arbitrary number to BCD is to loop 8 times over
your bitstream, SLA your bits out of your binary value and then run
"adc a, a \ daa" over it (with provisions for carries if you expect numbers
over 99).