mirror of
https://github.com/michaelcmartin/Ophis.git
synced 2024-12-27 06:29:16 +00:00
142 lines
3.7 KiB
Plaintext
142 lines
3.7 KiB
Plaintext
<chapter>
|
|
<title>Advanced Memory Segments</title>
|
|
|
|
<para>
|
|
By now we've covered the basics of every command in the assembler;
|
|
in this final installment we show the full capabilities of
|
|
the <literal>.text</literal> and <literal>.data</literal> commands
|
|
as we produce a more sophisticated set of Commodore 64 header
|
|
files.
|
|
</para>
|
|
|
|
<section>
|
|
<title>The Problem</title>
|
|
|
|
<para>
|
|
Our <literal>print'str</literal> routine
|
|
in <xref linkend="tutor6-src" endterm="tutor6-fname"> accesses
|
|
memory locations $10 and $11 directly. We'd prefer to have
|
|
symbolic names for them. This reprises our concerns back in
|
|
<xref linkend="ch5-link"> when we concluded that we wanted two
|
|
separate program counters. Now we realize that we really need
|
|
three; one for the text, one for the data, and one for the zero
|
|
page data. And if we're going to allow three, we really should
|
|
allow any number.
|
|
</para>
|
|
</section>
|
|
<section>
|
|
<title>The Solution</title>
|
|
|
|
<para>
|
|
The <literal>.data</literal> and <literal>.text</literal>
|
|
commands can take a label name after them—this names a new
|
|
segment. We'll define a new segment
|
|
called <literal>zp</literal> (for <quote>zero page</quote>) and
|
|
have our zero-page variables be placed there. We can't actually
|
|
use the default origin of $0000 here either, though, because the
|
|
Commodore 64 reserves memory locations 0 and 1 to control its
|
|
memory mappers:
|
|
</para>
|
|
|
|
<programlisting>
|
|
.data zp
|
|
.org $0002
|
|
</programlisting>
|
|
|
|
<para>
|
|
Now, actually, the rest of the zero page is reserved too:
|
|
locations $02-$8F are used by the BASIC interpreter, and
|
|
locations $90-$FF are used by the KERNAL. We don't need the
|
|
BASIC interpreter, though, so we can back up all of $02-$8F at
|
|
the start of our program and restore it all when we're done.
|
|
</para>
|
|
|
|
<para>
|
|
In fact, since we're disablng BASIC, we can actually also swap
|
|
out its ROM entirely and get a contiguous block of RAM from
|
|
$0002 to $CFFF:
|
|
</para>
|
|
|
|
<programlisting>
|
|
.scope
|
|
; Cache BASIC zero page at top of available RAM
|
|
ldx #$8e
|
|
* lda $01, x
|
|
sta $cf81, x
|
|
dex
|
|
bne -
|
|
|
|
; Swap out the BASIC ROM for RAM
|
|
lda $01
|
|
and #$fe
|
|
ora #$06
|
|
sta $01
|
|
|
|
; Run the real program
|
|
jsr _main
|
|
|
|
; Restore BASIC ROM
|
|
lda $01
|
|
ora #$07
|
|
sta $01
|
|
|
|
; Restore BASIC zero page
|
|
ldx #$8e
|
|
* lda $cf81, x
|
|
sta $01, x
|
|
dex
|
|
bne -
|
|
|
|
; Back to BASIC
|
|
rts
|
|
|
|
_main:
|
|
; _main points at the start of the real program,
|
|
; which is actually outside of this scope
|
|
.scend
|
|
</programlisting>
|
|
|
|
<para>
|
|
Our <literal>print'str</literal> routine is then rewritten to
|
|
declare and use a zero-page variable, like so:
|
|
</para>
|
|
|
|
<programlisting>
|
|
; PRINTSTR routine. Accumulator stores the low byte of the address,
|
|
; X register stores the high byte. Destroys the values of $10 and
|
|
; $11.
|
|
|
|
.scope
|
|
.data zp
|
|
.space _ptr 2
|
|
.text
|
|
printstr:
|
|
sta _ptr
|
|
stx _ptr+1
|
|
ldy #$00
|
|
_lp: lda (_ptr),y
|
|
beq _done
|
|
jsr chrout
|
|
iny
|
|
bne _lp
|
|
_done: rts
|
|
.scend
|
|
</programlisting>
|
|
|
|
<para>
|
|
Also, we ought to put in an extra check to make sure our
|
|
zero-page allocations don't overflow, either:
|
|
</para>
|
|
|
|
<programlisting>
|
|
.data zp
|
|
.checkpc $80
|
|
</programlisting>
|
|
|
|
<para>
|
|
The final source file is <xref linkend="tutor7-src"
|
|
endterm="tutor7-fname">.
|
|
</para>
|
|
</section>
|
|
</chapter>
|