mirror of https://github.com/Michaelangel007/apple2_hgr_font_tutorial.git synced 2024-11-05 02:04:44 +00:00

Apple ][ //e HGR Font Tutorial

6502 6502-assembly apple2 font-drawing printing

Go to file

michaelangel007 b74f4988a0 emphasized visual mnemonic		2016-01-13 22:03:39 -08:00
pics	S:R in texture atlas cells	2016-01-12 18:06:59 -08:00
Apple2eFont7x8.png	Added Apple 2e Font 7x8	2016-01-09 11:48:01 -08:00
c_2_fontbin.c	Fixed Utilities bullet and broken filename	2016-01-10 10:06:04 -08:00
code_0300.bin	Added code_0300.bin	2016-01-10 18:26:07 -08:00
font_codepage_437_8x8.png	Added codepage 437 8x8	2016-01-11 16:13:01 -08:00
font_codepage_437_9x16.png	Rename codepage to 9x16	2016-01-11 16:15:26 -08:00
font.bin	Added utilities and font binary	2016-01-10 10:03:02 -08:00
font.h	Added utilities and font binary	2016-01-10 10:03:02 -08:00
hgr_scroll_up.bin	Added binary: hgr_scroll_up.bin	2016-01-10 18:36:14 -08:00
hgrtable.bin	Add missing hgrtable.bin	2016-01-10 15:57:42 -08:00
image_2_c.html	Added utilities and font binary	2016-01-10 10:03:02 -08:00
image_2_hex.html	Added image_2_hex.html	2016-01-09 19:21:26 -08:00
README.md	emphasized visual mnemonic	2016-01-13 22:03:39 -08:00
scroll_hgr_up_pixel.html	Added Javascript to unroll screen scroll up	2016-01-09 19:30:14 -08:00

README.md

#Apple ][ HGR Font Tutorial

Revision: 33, Jan 13, 2016.

Introduction
The Problem
The Solution
Hard-Coded: A
Functions we want & will write
Quirks of the Apple HGR screen
Non-Linear Memory
No FONT data in ROM
HGR bytes are reversed
Half-pixel shift
Font Data
Raw Font Data
Image to Font Data (JavaScript)
DrawChar()
Font -> Screen Memory Trace
DrawChar() version 1
X Cursor Position
CursorCol()
Introduction to Optimization
DrawChar() version 2
DrawChar() version 3
Character Inspector
Character Inspector version 2
Character Inspector version 3
Y Cursor Position
Natural Params SetCursorColRow()
DrawString()
Recap
Copy text screen to HGR
Exercise 1: ScrollHgrUpLine()
Exercise 2: ScrollHgrUpPixel()
Conclusion
Solution 1: ScrollHgrUpLine()
Solution 2: ScrollHgrUpPixel()
References
Misc. Utilities and Files
TODO:

Introduction

A lot of people in comp.sys.apple2.programmer and other places on the internet have wondered how to "print" text onto the Apple's High Resolution Graphics (HGR) screen. Here's a tutorial on "6502 Font Blitting."

Note: We will prefix hex numbers with $ (or C's notation of 0x). We will prefix binary numbers with %.

Fire up your favorite Apple emulator (cough AppleWin) or real hardware.

If you use:

AppleWin press F2 (to reboot), Ctrl-F2 to Ctrl-Reset, and then press F9 until you get a Monochrome screen.
Jace press Ctrl-Delete to reset.
You will also need the Java JRE.
On OSX, Jace has an copy-paste bug and won't paste in the first line of the clipboard. :-/
Virtual II press Ctrl-F12 to reset.

There are other emulators written in JavaScript but they are poor due to 2 reasons:

Don't support paste -- you'll be forced to manually enter in the hex code. :-/ (Yeah, right!
Don't emulate the half-pixel shift of real hardware at all -- not an issue, but you won't see the full effect for one section.

Some emulators that run in the browser:

Apple 2 js make sure you select:
Options, [x] Green Screen
David's Caldwall's Apple 2 JavaScript + WebGL

Note: If you are using an emulator -- I've added "comments" in the lines of machine code you would paste by having a semi-colon and a description at the end of the line.

You may want to mute your sound since the Apple will beep at the semi-colon "comments" as that part of the input is not tehcnically valid input. (The rest of the line WILL be processed, though.)

The Problem

When you are at the Applesoft ] prompt type in (or paste) the following:

CALL-151
FC58G
400:41

We used the the ASCII char A which has a hex value 0x41. Hmm, OK, so we see an A but why is it flashing?

**Note**: If you use AppleWin, select the lines, copy, switch back to the emulator, and press Shift-Insert to paste.

The 40x24 text screen of the Apple is "memory-mapped" -- that is by directly setting memory the Video Controller circuitry will read those contents and display that. In later computers this video ram or VRAM is not directly accessible by the CPU; all the CPU can do is set registers of the GPU.

One of the quirks of the Apple is that it has support for Inverse, Flashing, and Normal characters. It technically uses High-Bit ASCII to show normal characters. This doesn't really concern us but just so you understand:

400:01 41 C1

The control characters show up in inverse, ASCII characters show up flashing, and our normal character requires the high bit to be set. 0x41 + 080 = 0xC1.

A slight fun diversion: If you are on an enhanced Apple //e or //c we can activate a 2nd character set called `Mouse Text`. This replaces all the flashing text with special drawing charactes:

C00F:1

You should see the flashing A has been replaced with an open apple symbol.

To turn Mouse Text off:

C00E:1

Anyways, back to drawing text.

If we switch to the HGR screen and tried to enter in 0x41 what would happen? (Ignore the beeping the emulator will make.)

F399G    ; `TEXT`
F3E2G    ; `HGR`
2000:41  ; A

Hmm, that doesn't look like an A at all, only gibberish -- 2 dots. :-/ (If you see 2 magenta dots ignore the color for now.)

The Solution:

Hard-Coded: A

Enter in:

Voila!

You should see an uppercase A appear in the top left of the HGR screen.

Magic? :-)

Nah, just Computer Science. :-)

The first question you probably have is "How did I know what bytes to use?" We'll get to that in a second.

Functions we want & will write

When we are done we will have 6502 assembly code that implements the equivalent of these C functions names:

    void DrawChar();
    void DrawCharCol( char c, int col );
    void DrawCharColRow( char c, int col, int row );
    void SetCursorRow( int row );
    void SetCursorColRow3( int col, int row );
    void SetCursorCol( int col );
    void IncCursorCol();
    void DrawHexByte( char c );
    void DrawString( char *text );
    void CopyTextToHGR();
    void ScrollHgrUpPixel();

Quirks of the Apple HGR screen

There are couple of things we need to discuss first. The preceding example showed that the Apple's TEXT and HGR screen behaves a little "funky." The Apple's, shall we say, esoteric use of hardware, is one of the reasons us fans love (or hate) it. "There are 4 lights!" Er, There are 4 things that stand out:

Non-Linear Memory

First, we should notice that video memory is non-linear. :-( You'll want to get familiar with the HGR address for the various Y scanlines:

"Understanding the Apple II", page 5-14 has this table HGR Memory-mapped IO:

https://archive.org/stream/understanding_the_apple_ii#page/n105/mode/2up

With all the decimal cruft removed:

Table 1: HGR Y Address for every scanline

Y	Address	Y	Address	Y	Address	Screen Hole
0	$2000	64	$2028	128	$2050	$2078..$207F
1	$2400	65	$2428	129	$2450	$2478..$247F
2	$2800	66	$2828	130	$2850	$2878..$287F
3	$2C00	67	$2C28	131	$2C50	$2C78..$2C7F
4	$3000	68	$3028	132	$3050	$3078..$307F
5	$3400	69	$3428	133	$3450	$3478..$347F
6	$3800	70	$3828	134	$3850	$3878..$387F
7	$3C00	71	$3C28	135	$3C50	$3C78..$3C7F
8	$2080	72	$20A8	136	$20D0	$20F8..$20FF
9	$2480	73	$24A8	137	$24D0	$24F8..$24FF
10	$2880	74	$28A8	138	$28D0	$28F8..$28FF
11	$2C80	75	$2CA8	139	$2CD0	$2CF8..$2CFF
12	$3080	76	$30A8	140	$30D0	$30F8..$30FF
13	$3480	77	$34A8	141	$34D0	$34F8..$34FF
14	$3880	78	$38A8	142	$38D0	$38F8..$38FF
15	$3C80	79	$3CA8	143	$3CD0	$3CF8..$3CFF
16	$2100	80	$2128	144	$2150	$2178..$217F
17	$2500	81	$2528	145	$2550	$2578..$257F
18	$2900	82	$2928	146	$2950	$2978..$297F
19	$2D00	83	$2D28	147	$2D50	$2D78..$2D7F
20	$3100	84	$3128	148	$3150	$3178..$317F
21	$3500	85	$3528	149	$3550	$3578..$357F
22	$3900	86	$3928	150	$3950	$3978..$397F
23	$3D00	87	$3D28	151	$3D50	$3D78..$3D7F
24	$2180	88	$21A8	152	$21D0	$21F8..$21FF
25	$2580	89	$25A8	153	$25D0	$25F8..$25FF
26	$2980	90	$29A8	154	$29D0	$29F8..$29FF
27	$2D80	91	$2DA8	155	$2DD0	$2DF8..$2DFF
28	$3180	92	$31A8	156	$31D0	$31F8..$31FF
29	$3580	93	$35A8	157	$35D0	$35F8..$35FF
30	$3980	94	$39A8	158	$39D0	$39F8..$39FF
31	$3D80	95	$3DA8	159	$3DD0	$3DF8..$3DFF
32	$2200	96	$2228	160	$2250	$2278..$227F
33	$2600	97	$2628	161	$2650	$2678..$267F
34	$2A00	98	$2A28	162	$2A50	$2A78..$2A7F
35	$2E00	99	$2E28	163	$2E50	$2E78..$2E7F
36	$3200	100	$3228	164	$3250	$3278..$327F
37	$3600	101	$3628	165	$3650	$3678..$367F
38	$3A00	102	$3A28	166	$3A50	$3A78..$3A7F
39	$3E00	103	$3E28	167	$3E50	$3E78..$3E7F
40	$2280	104	$22A8	168	$22D0	$22F8..$22FF
41	$2680	105	$26A8	169	$26D0	$26F8..$26FF
42	$2A80	106	$2AA8	170	$2AD0	$2AF8..$2AFF
43	$2E80	107	$2EA8	171	$2ED0	$2EF8..$2EFF
44	$3280	108	$32A8	172	$32D0	$32F8..$32FF
45	$3680	109	$36A8	173	$36D0	$36F8..$36FF
46	$3A80	110	$3AA8	174	$3AD0	$3AF8..$3AFF
47	$3E80	111	$3EA8	175	$3ED0	$3EF8..$3EFF
48	$2300	112	$2328	176	$2350	$2378..$237F
49	$2700	113	$2728	177	$2750	$2778..$277F
50	$2B00	114	$2B28	178	$2B50	$2B78..$2B7F
51	$2F00	115	$2F28	179	$2F50	$2F78..$2F7F
52	$3300	116	$3328	180	$3350	$3378..$337F
53	$3700	117	$3728	181	$3750	$3778..$377F
54	$3B00	118	$3B28	182	$3B50	$3B78..$3B7F
55	$3F00	119	$3F28	183	$3F50	$3F78..$3F7F
56	$2380	120	$23A8	184	$23D0	$23F8..$23FF
57	$2780	121	$27A8	185	$27D0	$27F8..$27FF
58	$2B80	122	$2BA8	186	$2BD0	$2BF8..$2BFF
59	$2F80	123	$2FA8	187	$2FD0	$2FF8..$2FFF
60	$3380	124	$33A8	188	$33D0	$33F8..$33FF
61	$3780	125	$37A8	189	$37D0	$37F8..$37FF
62	$3B80	126	$3BA8	190	$3BD0	$3BF8..$3BFF
63	$3F80	127	$3FA8	191	$3FD0	$3FF8..$3FFF

Don't worry if the address pattern makes no sense right now -- we'll reveal that later -- but if you're curiuous it is this line using integer math:

        hgr[ y ] = 0x2000 + (y/64)*0x28 + (y%8)*0x400 + ((y/8)&7)*0x80;

Here's the JavaScript source code to generate a bare-bones table:

    function int2pad( n, pad )
    {
        return (Array( pad+1 ).join(' ') + n).slice(-pad);
    }

    function word2hex$( w )
    {
        return "$" + ("    " + w.toString(16).toUpperCase()).slice(-4);
    }

    var hgr = [];
    for( var y = 0; y < 193; ++y ) // Intentional 1 scanline too many!
        hgr[ y ] = 0x2000 + ((y/64)|0)*0x28 + ((y%8)|0)*0x400 + ((y/8)&7)*0x80;

    var text = "", s = " | ";
    for( y = 0; y < 64; ++y )
    {
        var a0 = hgr[ y +   0 ];
        var a1 = hgr[ y +  64 ];
        var a2 = hgr[ y + 128 ];
        text += "| "
        + int2pad( y +   0, 3 ) + s + word2hex$( a0 ) + s
        + int2pad( y +  64, 3 ) + s + word2hex$( a1 ) + s
        + int2pad( y + 128, 3 ) + s + word2hex$( a2 ) + s
        + word2hex$( a2 + 40 ) + ".."
        + word2hex$( a2 + 47 ) + " |\n";
    }
    console.log( text );

No FONT data in ROM

Second, each glyph in the Apple font is in a 7x8 cell -- the leading line on the bottom is usually blank but we'll store that too so that we have a true "underline" and bottom descender on 'j', 'y', etc. How do we know this?

The TEXT screen is 40x24 characters. The high resolution graphics HGR screen is 280x192.

Char Width  (px/character) = Screen Width  (px) / Columns (characters) = 280/40 = 7

Char Height (px/character) = Screen Height (px) / Rows    (characters) = 192/24 = 8

Unfortunately, the data for the TEXT ROM 25123 hardware chip is not accessible from the 6502 unlike say the IBM PC BIOS Character Sets. :-/ T his means you will need to manually enter in the 8 bytes/character. :-( The good news is that I've already done this so you can copy / paste. :-)

You can find a picture of the Apple ][ ROM text font on Page 8-9, Figure 8.4 of "Understanding the Apple ][":

We're actually going to use the Apple //e ROM text font since it has lower case and the famous "Mouse Text" glyphs. See Page 8-25, Figure 8.8 of "Understanding the Apple //e":

HGR bytes are reversed

Third, the video scanner for HGR mode scans bits in reverse. :-/ This means that we need to "flip" the bits in a byte if we want it to appear properly. Not hard, just inconvenient. We'll store the pre-flipped bits so we don't have to do this at run-time. :-)

For example, If we want these 4 scan-lines of \:

X____
_X___
__X__
___X_

You would normally encode the pixels in binary as:

%1000_0000 = $80
%0100_0000 = $40
%0010_0000 = $20
%0001_0000 = $10

And if we tried entering in:

We would only get:

3 scanlines instead of the expected 4 (see the next point), and
the image would be flipped along the left-right (X axis) like this: /

On the Apple we need to flip each byte:

%0000_0001 = $01
%0000_0010 = $02
%0000_0100 = $04
%0000_1000 = $08

Enter in:

2200:1
2600:2
2A00:4
2E00:8

And we see the correct: \

Half-pixel shift

Fourth, we mentioned above that when we entered in $80 that the Apple didn't display any pixels for this byte. This is because the Apple uses the high-bit as a flag to shift that group of 7 pixels over HALF a pixel. (Yes, half a pixel.) This means the monochrome effective resolution is a pseudo 560x192. We can't individually access every 560 pixels, only part of them so it is not a "true" 560 resolution. :-( What this means in practice is that we can use this half-pixel shift / byte to get very smooth slopes for Y, etc. :-)

For example this will give us a "sharp" Y:

If we change the 2nd and 4th scan line to use this half-pixel shift we can't just set the high bit as we won't get quite the correct image:

Enter in: (Again, ignore the beeping.)

2302:22
2702:A2 ;
2B02:14
2F02:88 ;
3302:8
3702:8
3B02:8

We actually also need to move the right-edge pixel of these 2 scanlines over left by 1 pixel so it appears in the correct location when shifted.

Enter in:

2304:22
2704:92 ;
2B04:14
2F04:8C ;
3304:8
3704:8
3B04:8

Ah-ha! We've got a "smooth" Y.

Note: The emulators Virtual ][ and Apple2js are broken emulators. They do not emulate the half-pixel shift of real hardware at all. This is another reason we won't worry about it for now.

We're going to ignore the half-pixel shift since it is easy to touch up the font data later if we wish.

At the beginning we said to view the HGR screen in monochrome. Notice how the extra colors make the Hi-Res text much harder to read. If you are running on real hardware the Apple Color Composite Monitor had a push-button on the front to toggle the screen between color and monochrome. Now we know why!

Font Data

Alrighty then, let's get the font data!

Here is a picture of the Apple //e character set:

If we wanted only uppercase ASCII we could get away with 4 rows of 16 characters (symbols, numbers, letters) = 64 glyphs:

64 glyphs * 8 bytes/glyph = 512 bytes.

Since the font data chews up memory anyways we'll "splurge" and use the full 128 ASCII glyphs:

128 glyphs * 8 bytes/glyph = 1024 bytes = 1K of data.

Ouch! We're using 1K of our precious 64K. Now we know why all those font glyphs was in a ROM chip.

Raw Font Data

I've saved you the trouble of converting all the pixels to hex. Remember, you may want to mute your sound since the Apple will beep at the semi-colon "comments".

Enter in (or download the raw binary font.bin and with AppleWin press F7, type bload font.bin,6000, press F7):

6000:10 08 36 7F 3F 3F 7E 36  ; ^@
6008:10 08 36 41 21 21 4A 36  ; ^A
6010:00 00 02 06 0E 1E 36 42  ; ^B
6018:7F 22 14 08 08 14 2A 7F  ; ^C
6020:00 40 20 11 0A 04 04 00  ; ^D
6028:7F 3F 5F 6C 75 7B 7B 7F  ; ^E
6030:70 60 7E 31 79 30 3F 02  ; ^F
6038:00 18 07 00 07 0C 08 70  ; ^G
6040:08 04 02 7F 02 04 08 00  ; ^H
6048:00 00 00 00 00 00 00 2A  ; ^I
6050:08 08 08 08 49 2A 1C 08  ; ^J
6058:08 1C 2A 49 08 08 08 08  ; ^K
6060:7F 00 00 00 00 00 00 00  ; ^L
6068:40 40 40 44 46 7F 06 04  ; ^M
6070:3F 3F 3F 3F 3F 3F 3F 3F  ; ^N
6078:13 18 1C 7E 1C 18 10 6F  ; ^O
6080:64 0C 1C 3F 1C 0C 04 7B  ; ^P
6088:40 48 08 7F 3E 1C 48 40  ; ^Q
6090:40 48 1C 3E 7E 08 48 40  ; ^R
6098:00 00 00 7F 00 00 00 00  ; ^S
60A0:01 01 01 01 01 01 01 7F  ; ^T
60A8:08 10 20 7F 20 10 08 00  ; ^U
60B0:2A 55 2A 55 2A 55 2A 55  ; ^V
60B8:55 2A 55 2A 55 2A 55 2A  ; ^W
60C0:00 3E 41 01 01 01 7F 00  ; ^X
60C8:00 00 3F 40 40 40 7F 00  ; ^Y
60D0:40 40 40 40 40 40 40 40  ; ^Z
60D8:08 1C 3E 7F 3E 1C 08 00  ; ^[
60E0:7F 00 00 00 00 00 00 7F  ; ^\
60E8:14 14 77 00 77 14 14 00  ; ^]
60F0:7F 40 40 4C 4C 40 40 7F  ; ^^
60F8:01 01 01 01 01 01 01 01  ; ^_
6100:00 00 00 00 00 00 00 00  ;  
6108:08 08 08 08 08 00 08 00  ; !
6110:14 14 14 00 00 00 00 00  ; "
6118:14 14 3E 14 3E 14 14 00  ; #
6120:08 3C 0A 1C 28 1E 08 00  ; $
6128:06 26 10 08 04 32 30 00  ; %
6130:04 0A 0A 04 2A 12 2C 00  ; &
6138:08 08 08 00 00 00 00 00  ; '
6140:08 04 02 02 02 04 08 00  ; (
6148:08 10 20 20 20 10 08 00  ; )
6150:08 2A 1C 08 1C 2A 08 00  ; *
6158:00 08 08 3E 08 08 00 00  ; +
6160:00 00 00 00 08 08 04 00  ; ,
6168:00 00 00 3E 00 00 00 00  ; -
6170:00 00 00 00 00 00 08 00  ; .
6178:00 20 10 08 04 02 00 00  ; /
6180:1C 22 32 2A 26 22 1C 00  ; 0
6188:08 0C 08 08 08 08 1C 00  ; 1
6190:1C 22 20 18 04 02 3E 00  ; 2
6198:3E 20 10 18 20 22 1C 00  ; 3
61A0:10 18 14 12 3E 10 10 00  ; 4
61A8:3E 02 1E 20 20 22 1C 00  ; 5
61B0:38 04 02 1E 22 22 1C 00  ; 6
61B8:3E 20 10 08 04 04 04 00  ; 7
61C0:1C 22 22 1C 22 22 1C 00  ; 8
61C8:1C 22 22 3C 20 10 0E 00  ; 9
61D0:00 00 08 00 08 00 00 00  ; :
61D8:00 00 08 00 08 08 04 00  ; ;
61E0:10 08 04 02 04 08 10 00  ; <
61E8:00 00 3E 00 3E 00 00 00  ; =
61F0:04 08 10 20 10 08 04 00  ; >
61F8:1C 22 10 08 08 00 08 00  ; ?
6200:1C 22 2A 3A 1A 02 3C 00  ; @
6208:08 14 22 22 3E 22 22 00  ; A
6210:1E 22 22 1E 22 22 1E 00  ; B
6218:1C 22 02 02 02 22 1C 00  ; C
6220:1E 22 22 22 22 22 1E 00  ; D
6228:3E 02 02 1E 02 02 3E 00  ; E
6230:3E 02 02 1E 02 02 02 00  ; F
6238:3C 02 02 02 32 22 3C 00  ; G
6240:22 22 22 3E 22 22 22 00  ; H
6248:1C 08 08 08 08 08 1C 00  ; I
6250:20 20 20 20 20 22 1C 00  ; J
6258:22 12 0A 06 0A 12 22 00  ; K
6260:02 02 02 02 02 02 3E 00  ; L
6268:22 36 2A 2A 22 22 22 00  ; M
6270:22 22 26 2A 32 22 22 00  ; N
6278:1C 22 22 22 22 22 1C 00  ; O
6280:1E 22 22 1E 02 02 02 00  ; P
6288:1C 22 22 22 2A 12 2C 00  ; Q
6290:1E 22 22 1E 0A 12 22 00  ; R
6298:1C 22 02 1C 20 22 1C 00  ; S
62A0:3E 08 08 08 08 08 08 00  ; T
62A8:22 22 22 22 22 22 1C 00  ; U
62B0:22 22 22 22 22 14 08 00  ; V
62B8:22 22 22 2A 2A 36 22 00  ; W
62C0:22 22 14 08 14 22 22 00  ; X
62C8:22 22 14 08 08 08 08 00  ; Y
62D0:3E 20 10 08 04 02 3E 00  ; Z
62D8:3E 06 06 06 06 06 3E 00  ; [
62E0:00 02 04 08 10 20 00 00  ; \
62E8:3E 30 30 30 30 30 3E 00  ; ]
62F0:00 00 08 14 22 00 00 00  ; ^
62F8:00 00 00 00 00 00 00 7F  ; _
6300:04 08 10 00 00 00 00 00  ; `
6308:00 00 1C 20 3C 22 3C 00  ; a
6310:02 02 1E 22 22 22 1E 00  ; b
6318:00 00 3C 02 02 02 3C 00  ; c
6320:20 20 3C 22 22 22 3C 00  ; d
6328:00 00 1C 22 3E 02 3C 00  ; e
6330:18 24 04 1E 04 04 04 00  ; f
6338:00 00 1C 22 22 3C 20 1C  ; g
6340:02 02 1E 22 22 22 22 00  ; h
6348:08 00 0C 08 08 08 1C 00  ; i
6350:10 00 18 10 10 10 12 0C  ; j
6358:02 02 22 12 0E 12 22 00  ; k
6360:0C 08 08 08 08 08 1C 00  ; l
6368:00 00 36 2A 2A 2A 22 00  ; m
6370:00 00 1E 22 22 22 22 00  ; n
6378:00 00 1C 22 22 22 1C 00  ; o
6380:00 00 1E 22 22 1E 02 02  ; p
6388:00 00 3C 22 22 3C 20 20  ; q
6390:00 00 3A 06 02 02 02 00  ; r
6398:00 00 3C 02 1C 20 1E 00  ; s
63A0:04 04 1E 04 04 24 18 00  ; t
63A8:00 00 22 22 22 32 2C 00  ; u
63B0:00 00 22 22 22 14 08 00  ; v
63B8:00 00 22 22 2A 2A 36 00  ; w
63C0:00 00 22 14 08 14 22 00  ; x
63C8:00 00 22 22 22 3C 20 1C  ; y
63D0:00 00 3E 10 08 04 3E 00  ; z
63D8:38 0C 0C 06 0C 0C 38 00  ; {
63E0:08 08 08 08 08 08 08 08  ; |
63E8:0E 18 18 30 18 18 0E 00  ; }
63F0:2C 1A 00 00 00 00 00 00  ; ~
63F8:00 2A 14 2A 14 2A 00 00  ;

(For the advanced user, you can save this: BSAVE FONT.BIN,A$6000,L$400)

**AppleWin** users: We can save the state of the virutal machine via `F11`.

Image to Font Data (JavaScript)

If you were wondering how this data was generated, you see the great thing about computers is that they can automate all the tedious and boring crap, er, calculations for us. Here's a HTML + JavaScript program I wrote to convert the image to HEX:

     <!DOCTYPE HTML>
    <html>
    <head>
        <script>
            function byte2hex$( byte )
            {
                return ("0" + byte.toString(16)).toUpperCase().substr(-2)
            }

            function OnLoad()
            {
                var image   = document.getElementById( "Apple2eFont7x8" );
                var canvas  = document.createElement( "Canvas" );
                var context = canvas.getContext( "2d" );

                canvas.width  = image.width;
                canvas.height = image.height;
                context.drawImage( image, 0, 0 );

                var CW = 7, CH = 8; // Cell Width Height
                var address = 0x6000, pixel, rgba, lines = "";
                for( var ty = 0; ty < image.height/CH; ++ty )
                {
                    for( var tx = 0; tx < image.width/CW; ++tx )
                    {
                        var text = "";
                        for( var y = 0; y < CH; ++y )
                        {
                            var hex = 0, mask = 0x1;
                            for( var x = 0; x < CW; ++x, mask <<= 1 )
                            {
                                pixel = context.getImageData( tx*CW+x, ty*CH+y, 1, 1 );
                                rgba  = pixel.data;
                                hex  += rgba[0] ? mask : 0; // assume R=G=B
                            }
                            text += byte2hex$( hex ) + " ";
                        }
                        var c = (16*ty)+tx, d =       String.fromCharCode( c );
                        if (c < 32)         d = "^" + String.fromCharCode( c + 0x40 );
                        text += "; " + d + "\n";
                        lines += "" + address.toString(16).toUpperCase() + ":" + text;
                        address += 8;
                    }
                }
                console.log( lines );
                var pre = document.getElementById( "hexdump" );
                pre.innerHTML = lines;
            }
        </script>
    </head>
    <body onload="OnLoad()">
        <img id="Apple2eFont7x8" src="Apple2eFont7x8.png">
        <hr>
        <pre id="hexdump"></pre>IncCursorCol
    </body>
    </html>

Note: If you get a retarded Uncaught SecurityError: Failed to execute 'getImageData' on 'CanvasRenderingContext2D': The canvas has been tainted by cross-origin data. with Chrome you need to start it with the command line:

--allow-file-access-from-files

Another solution is to use a web browser that isn't "broken" such as Firefox, etc. when trying to read local files.

DrawChar()

Font -> Screen Memory Trace

OK, so now that we have the font data how the the heck do we actually draw a character "on screen" ?

We need to transfer 8 consecutive bytes (1 byte / scanline) to 8 different scanlines scattered all over memory.

Assuming we want to draw the A glyph at the top-left of the screen we would need to transfer bytes from the (source) font glyph memory locations to these (destination) HGR screen memory locations:

($6208) -> $2000
($6209) -> $2400
($620A) -> $2800
($620B) -> $2C00
($620C) -> $3000
($620D) -> $3400
($620E) -> $3800
($620F) -> $3C00

For simplicity, we're going to "quantize" our destination Y so that we render font glyphs only on the start of every 8 rows and every 7 pixel columns. (See Table 2 down below in section Y Cursor Position). If we then had the starting address we simply could move to the next scan line by successively adding $0400 to our destination screen pointer.

How did I know to use $0400 when going to the next line? One quirk of the HGR screen is that every 8 successive scan lines start this many bytes away. Refer back to the HGR Memory-mapped IO table listed above.

DrawChar() version 1

Before we can start a simple DrawChar(char c) function, we also first need to assign some zero page memory locations for our static and temporary variables, our 16-bit address of where want to draw to. Since we also have our font data, we need a symbol for that too.

                  HgrLo   EQU $E5   ; Low  byte Pointer to screen destination
                  HgrHi   EQU $E6   ; High byte Pointer to screen destination
                  TmpLo   EQU $F5   ; Low  byte Working pointer to screen byte
                  TmpHi   EQU $F6   ; High byte Working pointer to screen byte

                  Font    EQU $6000

Here's the disassembly of our (hard-coded) DrawChar() program:

    ; FUNC: PrintChar()
    ; NOTES: A, X, Y is destroyed
                     ORG $0900
    0900:         PrintChar
    0900:20 0A 09    JSR HgrToTmpPtr
    0903:A9 00       LDA #00        ; glyph 'c' to draw (not used yet)
    0905:A0 00       LDY #00        ; Y = column to draw at (hard-coded)
    0907:4C 10 03    JMP DrawChar

    ; FUNC: HgrToTmpPtr()
    090A:         HgrToTmpPtr
    090A:A5 E5       LDA HgrLo      ; Copy initial screen
    090C:85 F5       STA TmpLo      ; destination pointer
    090E:A5 E6       LDA HgrHi      ; to working pointer
    0910:85 F6       STA TmpHi
    0912:60          RTS

    ; FUNC: DrawChar()
    ; PARAM: A = glyph to draw
    ; PARAM: Y = column to draw at; $0 .. $27 (Columns 0 .. 39) (not modified)
    ; INPUT : $F5,$F6 pointer to the destination screen scanline
    ;         Must start at every 8 scanlines.
    ; OUTPUT: The Y-Register (cursor column) is automatically incremented.
                     ORG $0310
    0310:         DrawChar
    0310:4C 50 03    JMP _DrawChar

                     ORG $0350
    0350:         _DrawChar
    0350:A2 00       LDX #0
    0352:         _LoadFont         ; A = font[ offset ]
    0352:BD 00 62    LDA Font+#$200,X
    0355:91 F5       STA (TmpLo),Y  ; screen[col] = A
    0357:18          CLC
    0358:A5 F6       LDA TmpHi
    035A:69 04       ADC #4         ; screen += 0x400
    035C:85 F6       STA TmpHi
    035E:E8          INX
    035F:E0 08       CPX #8
    0361:D0 EF       BNE _LoadFont
    0363:60          RTS

Enter in:

900:20 0A 09 A9 00 A0 00 4C 10 03
90A:A5 E5 85 F5 A5 E6 85 F6 60
310:4C 50 03
350:A2 00 BD 00 62 91 F5 18
358:A5 F6 69 04 85 F6 E8 E0
360:08 D0 EF 60

We're almost ready to run this! We just need to initialize one variable -- where to draw the glyph at:

E5:00 20
900G

And with any luck you should see the at sign @ in the top-left.

**AppleWin** users: You can enter these symbols into the debugger to make the disassembly more readable. Press `F7`, then type in (paste with `Ctrl-V`):

sym HgrLo       = E5
sym HgrHi       = E6
sym TmpLo       = F5
sym TmpHi       = F6
sym PrintChar   = 0900
sym HgrToTmpPtr = 090A
sym DrawChar    = 0310
sym _DrawChar   = 0350
sym _LoadFont   = 0352
sym Font        = 6000
350L

When you are done with the debugger, press F7 to return to the emulator.

(Screenshot of debugger forthcoming)

X Cursor Position

If we wanted to draw in columns 1 and 2 instead of column 0 then we need to set the Y register which controls which "column" we'll draw at.

Enter in:

By changing the Y register value we can control the column of where to draw the cursor. This works because we are using the 6502 Indirect Zero-Page Y addressing mode to store the destination pixels with the STA instruction. Since the Y-register must always be used in this addressing mode -- we (effectively) get a column offset "for free." :-)

    0355:91 F5       STA (TmpLo),Y  ; screen[col] = A

Here's the C pseudo-code of the assembly code:

    char  c      = '@'; // 0x40;
    int   col    = 0;
    char  FONT[] = { ... };      // our font data glyphs starting at 0x6000
    char *screen = 0x2000 + col; // destination
    char *font   = 0x6200;       // eventually want: &FONT[ c*8 ]
    for( y = 0; y < 8; y++, screen += 0x400 )
       *screen = *font++;

CursorCol( col )

Since the Y-register controls the column we can inline this function and have the caller take care of setting the Y-Register before calling DrawChar().

                     LDY #column

After drawing a character with DrawChar() it is handy if we can advance both:

the column of the cursor
the pointer to the screen where the next glyph will be drawn

Notice how after 8 scan lines we end up with and Tmp address of $4xxx (or $6xxx if we were drawing to HGR page 2.) This means we need to subtract off $20 from the top byte of the 16-bit address to the temp destination screen pointer.

    ; FUNC: IncCursorCol1()
    ; OUTPUT: Y-Register (column) is incremented
    ; Increment the cursor column and move the destination screen pointer back
    ; up 8 scan lines previously to what it was when DrawChar() was called.
    ; Version 1
                     ORG $0364
    0364:         IncCursorCol1
    0364:C8          INY
    0365:18          CLC            ; Note:
    0366:A5 F6       LDA TmpHi      ;     (To the astute reader)
    0368:E9 1F       SBC #$1F       ; <-- ??? Shouldn't this be #20 ?!
    036A:85 F6       STA TmpHi
    036C:60          RTS

Introduction to Optimization

One tip for beginner 6502 assembly programmers. It is tempting just to always clear the carry flag CLC before doing any addition or subtraction. Unfortunately, for subtraction we'll have an off-by-one bug (fence-post error) so we need to subtract ONE less then the value. This makes reading the code a little unintuitive. Is there a way to remedy this? Yes.

Op	Carry	Opcode
+	Clear	CLC
-	Set	SEC

We change the carry flag state before we do the operation depending on whether we are adding or subtracting. The Rule of Thumb is CLC before ADD and SEC before SUB. A mnemonic to help you remember is that both SEC and SUB start with S.

    ; FUNC: IncCursorCol2()
    ; OUTPUT: Y-Register (column) is incremented
    ; Increment the cursor column and move the destination screen pointer back
    ; up 8 scan lines previously to what it was when DrawChar() was called.
    ; Version 2
                     ORG $0364
    0364:         IncCursorCol2
    0364:C8          INY
    0365:38          SEC            ; CLC SBC #1F
    0366:A5 F6       LDA TmpHi      ; was not obvious that we really
    0368:E9 20       SBC #$20       ; meant A - #$20 !!
    036A:85 F6       STA TmpHi
    036C:60          RTS

One thing when writing 6502 assembly is to pay attention to all optimization opportunities due to the slow ~1 MHz of the 6502. Since we only need to modify the upper few bits instead of doing a bulky subtraction SEC SBC we might be tempted to see if there is a faster and/or smaller alternative. We just need to be careful that our optimization is "shuffling" the bits behaves around in the exact same way at the end of the day. i.e. The Right Place at the Right Time.

The problem is we want to see if we can simply the transform of TmpHi after 8 scanlines (basically reset the cursor back to the original scanline before we drew all 8 scanlines of the glyph):

Tmp   = Hgr + (8 * $0400)

Since we only care about the high byte:

TmpHi = HgrHi + (8 * $04)
      = HgrHi + $20

Legend:
    Initial = destination HGR address before we draw the glyph
    TmpHi   = destination HGR address after drawing all 8 lines
    Final   = destination HGR address set back to initial value

Y	Initial	TmpHi	Final	(T and $1F)	(T and $1F) or $20
0	$2000	$40	$20	$00	$20
8	$2080	$40	$20	$00	$20
16	$2100	$41	$21	$01	$21
24	$2180	$41	$21	$01	$21
32	$2200	$42	$22	$02	$22
40	$2280	$42	$22	$02	$22
48	$2300	$43	$23	$03	$23
56	$2380	$43	$23	$03	$23
64	$2028	$40	$20	$00	$20
72	$20A8	$40	$20	$00	$20
80	$2128	$41	$21	$01	$21
88	$21A8	$41	$21	$01	$21
96	$2228	$42	$22	$02	$22
104	$22A8	$42	$22	$02	$22
112	$2328	$43	$23	$03	$23
120	$23A8	$43	$23	$03	$23
128	$2050	$40	$20	$00	$20
136	$20D0	$40	$20	$00	$20
144	$2150	$41	$21	$01	$21
152	$21D0	$41	$21	$01	$21
160	$2250	$42	$22	$02	$22
168	$22D0	$42	$22	$02	$22
176	$2350	$43	$23	$03	$23
184	$23D0	$43	$23	$03	$23

Hmm, we would need to replace SEC SBC with AND OR which we might think would be a littler faster and takes less code to boot but let's verify our assumption:

    ; FUNC: IncCursorCol3()
    ; OUTPUT: Y-Register (column) is incremented
    ; Increment the cursor column and move the destination screen pointer back
    ; up 8 scan lines previously to what it was when DrawChar() was called.
    ; Version 3
                     ORG $0364
    0364:         IncCursorCol3
    0364:C8          INY
    0365:A5 F6       LDA TmpHi
    0367:29 1F       AND #%00011111 ; Requires an extra OR
    0369:09 20       ORA #20        ; Hard-code to HGR page 1
    036B:85 F6       STA TmpHi
    036D:60          RTS

Hmm, so the code isn't any smaller on a 6502 CPU. It might be on other CPUs.

Second, is it any faster?

    SEC    ; 2 cycles
    SBC #n ; 2 cycles

    AND #n ; 2 cycles
    ORA #m ; 2 cycles

Nope. Bummer. :-(

The lessons?

Verify our assumptions and Profile!
Even though we "failed" this time, we shouldn't be afraid to experiment with "out-the-box" thinking using the 6502 instructions; sometimes there are clear "wins" but you won't know unless you try! Also, it isn't always obvious if we should optimize to minimize space (with the potential to run slower) or to optimize for higher performance (at the cost of more code.) The "proper" solution depends on the context of your needs. With 8-bit CPU's we tend to focus on code density -- cram as much code in as little space as possible. Graphics / Rendering unfortunately "needs" to run as fast as possible so this means unrolling loops, etc., to run "flat out" even though we lose valuable memory.

We'll briefly touch upon this topic of optimization again with bit-shifts and memcpy().

Wait, you say! There IS a way to solve this problem -- and it doesn't take lateral thinking. What we really are doing is just restoring TmpHi back to its previous value! We need to save TmpHi when we set the rows to draw to 0, and restore it after drawing 8 rows.

                     ORG $034C
    034C:          _DrawChar1
    034C:A6 F6       LDX TmpHi
    034E:86 FD       STX TopHi

    ;                === _DrawChar begin ===
    ;                 ORG $0350
    ;0350:         _DrawChar
    ;                 ...
    ;035F:E0 08       CPX #8
    ;0361:D0 EF       BNE _LoadFont
    ;0363:60          RTS
    ;                === DrawChar end ===

    ; FUNC: IncCursorCol()
                     ORG $0363      ; intentional extend _DrawChar
    0363:C8          INY
    0364:A6  FD      LDX TopHi      ; Move cursor back to top of scanline
    0366:86  F6      STX TmpHi
    0368:60          RTS

We just need to touch up our entry point PrintChar at $0310 instead of calling _DrawChar ($0350) we need to call our new _DrawChar1 ($034C):

                     ORG $0310
    0310:         PrintChar
    0310:4C 4C 03    JMP _DrawChar1 ; NEW entry point

                     ORG $0A00
    0A00:         DemoDraw3Char
    0A00:20 00 09    JSR DrawChar
    0A03:20 10 03    JSR PrintChar
    0A06:20 10 03    JMP PrintChar
    0A09:60          RTS

Enter in:

310:4C 4C 03
34C:A6 F6 86 FD
363:C8 A6 FD 86 F6 60

Let's try it out:

A00: 20 00 09 20 10 03 20 10 03 60
A00G

We are one step closer to printing a string. We have a total of 5 @ because we didn't change our initial column from above. We are only printing 3 chars, the previous 2 are "left over" from the previous demo.

(Screenshot showing 5 @ forthcoming)

**AppleWin** users: Press `F7`, copy & paste the below, press `F7` when done.

SYM TopHgr       = FD
SYM _DrawChar1   = 34C
SYM IncCursorCol = 363
34CL

DrawChar() version 2

The glyph to draw is currently hard-coded to $40 (@). The pointer to the start of this glyph is located at:

source = $6000 + ($40*8) = $6000 + $200 = $6200

If we wanted to draw a different glyph, say D we would need to modify the source pointer of the font glyph data.

Recall that our font has this memory layout:

Char	Index	Address
^@	$00	$6000
^A	$01	$6008
^B	$02	$6010
^C	$03	$6018
	:	:
Spc	$20	$6100
!	$21	$6108
	:	:
0	$30	$6180
1	$31	$6188
2	$32	$6190
3	$33	$6198
	:	:
?	$3F	$61F8
@	$40	$6200
A	$41	$6208
B	$42	$6210
C	$43	$6218
D	$44	$6220
:	:	:
_	$5F	$62F8

The 6502 stores and loads 16-bit addresses in Little-Endian format so for glyph D we need to store the bytes of the address $6220 in reverse order.

Enter in:

353:20 62

And to draw the new glyph, enter in:

900G

We should see the third character of @ change to D.

DrawChar() version 3

Let's remove the hard-coded printing of the glyph and use the character data we really want to draw. This means we need to "fix-up" the temporary source pointer to the font glyph data. Since we have 8 bytes/glyph we need to manually calculate the array offset.

Our array offset for the source glyph data is:

address = $6000 + (glyph * 8)

Technically the C pseudo-code would be the more elegant:

    char* GetGlyphAddress( char c )
    {
        static char FONT[]  = { ... };      // our font data glyphs starting at 0x6000
        return &FONT[ c * 8 ];
    }

A more equivalent 6502 version would be:

    int GetAddress( char c )
    {
               int  offset  = c * 8;
        static char FONT[]  = { ... };      // our font data glyphs starting at 0x6000
               int  address = 0x6000 + offset;

        return address;
    }

Since we are dealing with a 16-bit address offset it is simpler to break this down into a low-byte and high-byte calculation for the 6502 since it can't natively do 16-bit offsets. Every 32 characters we need to offset 256 bytes.

    int AddressHi = FontAddressHi + (c / 32)

But since the 6502 doesn't have a division instruction we need to use bit-shifts instead. The calculation c / 32 is the same as c >> 5.

    char c        = 'D'; // 0x44

    int Font      = 0x6000;
    int FontHi    = (Font >> 8) & 0xFF;
    int FontLo    = (Font >> 0) & 0xFF;
    int AddressHi = FontHi + ((c >> 5) & 0x07);
    int AddressLo = FontLo + ((c << 3) & 0xF8);

We'll assign unique letters to each bit of c:

+-------------------------------+
| P | Q | R | s | t | u | v | w | glyph to draw
+-------------------------------+
  7   6   5   4   3   2   1   0   bit position

You see what I did there? :-) I put the low bits in lowercase and the High bits in Uppercase as a visual mnemonic to help remember which bits belong to which part of the address.

IF this is confusing, remember, we are calculating a 16-bit offset:

Offset =       %PQRstuvw * 8
Offset =       %PQRstuvw * 2^3;
Offset =       %PRQstuvw << 3
Offset = |%00000PQR|stuvw000|
         |High Byte|Low Byte|

A naive 6502 glyph/32 calculation would be to use 5 shift right bit-shifts:

    48       PHA      ; save c
    --       --       ; calc low byte offset
    68       PLA      ; pop c  = %PQRstuvw to draw
    4A       LSR      ; c /  2 = %0PQRstuv C=w
    4A       LSR      ; c /  4 = %w0PQRstu C=V
    4A       LSR      ; c /  8 = %wv0PQRst C=u
    4A       LSR      ; c / 16 = %uvw0PQRs C=t
    4A       LSR      ; c / 32 = %tUvw0PQR C=s
    29 07    AND #07  ;        = %00000PQR
    18       CLC      ;

We can optimize the CLC out by clearing the bottom bits and then doing the shift:

    48       PHA      ; save c
    --       --       ; calc low byte offset
    68       PLA      ; pop c  = %PQRstuvw to draw
    29 E0    AND #E0  ;        = %PQR00000 s=0, Optimization: implicit CLC
    4A       LSR      ; c /  2 = %0PQR0000
    4A       LSR      ; c /  4 = %00PQR000
    4A       LSR      ; c /  8 = %000PQR00
    4A       LSR      ; c / 16 = %0000PQR0
    4A       LSR      ; c / 32 = %00000PQR

However we can save one instruction (and 2 cycles) if we optimize c/32 to use the counter-intuitive 6502's ROL instruction -- which only requires 4 instructions instead:

    48       PHA      ; save c
    --       --       ; calc low byte offset
    68       PLA      ; pop c  = %PQRstuvw to draw
    29 E0    AND #E0  ;        = %PQR00000 s=0, Optimization: implicit CLC
    2A       ROL      ;        = %QR000000 C=P
    2A       ROL      ;        = %R000000P C=Q
    2A       ROL      ;        = %000000PQ C=R
    2A       ROL      ; c / 32 = %00000PQR C=0

Our prefix code to setup the source address becomes:

    ; FUNC: _DrawChar2a( c, col )
    ; PARAM: A = glyph to draw
    ; PARAM: Y = column to draw at; $0 .. $27 (Columns 0 .. 39) (not modified)
    ; NOTES: X is destroyed
                     ORG $0335
    0335:         _DrawChar2a
    0335:48          PHA            ; push c = %PQRstuvw to draw
    0336:29 1F       AND #1F        ;        = %000stuvw R=0, implicit CLC
    0338:0A          ASL            ; c * 2    %00stuvw0
    0339:0A          ASL            ; c * 4    %0stuvw00
    033A:0A          ASL            ; c * 8    %stuvw000
    033B:69 00       ADC #<Font     ; += FontLo; Carry = 0 since R=0 from above
    033D:8D 53 03    STA _LoadFont+1; AddressLo = FontLo + (c*8)
    0340:68          PLA            ; pop c  = %PQRstuvw to draw
    0341:29 E0       AND #E0        ;        = %PQR00000 s=0, implicit CLC
    0343:2A          ROL            ;        = %QR000000 C=P
    0344:2A          ROL            ;        = %R000000P C=Q
    0345:2A          ROL            ;        = %000000PQ C=R need one more
    0346:2A          ROL            ; c / 32 = %00000PQR C=0 shift to get R
    0347:69 60       ADC #>Font     ; += FontHi; Carry = 0 since S=0 from above
    0349:8D 54 03    STA _LoadFont+2; AddressHi = FontHi + (c/32)
                     ; intentional fall into _DrawChar1
    034C:          _DrawChar1
    034C:A6 F6       LDX TmpHi
    034E:86 FD       STX TopHi
    0350:         _DrawChar
    0350:A2 00       LDX #0         ; Note: next instruction is self-modified !
    0352:         _LoadFont         ; A = font[ offset ]
    0352:BD 00 00    LDA Font+#$200,X

Did you catch that note ? One popular trick on the 6502 was self-modifying code. Instead of wasting memory with yet-another-variable we directly change the load/store instructions themselves! This actually has 2 advantages:

It lets us avoid an expensive indirection pointer access, and
It is the fastest way to load/store/copy an array. The 6502 addressing mode is LDA address,X or LDA address,Y.

However, there are still 2 more optimizations we can make:

If we assume our font is "page aligned", that is, starts at a multiple of 256 bytes -- we could remove the redundant AddressLo += (FontLo + (c*8)) and replace with the direct AddressLo = (c*8). Technically, in C you would keep the bottom 8-bits with & 0xFF but since the 6502 registers are only 8-bit and we're storing a byte the & 0xFF is not needed.

    033B:69 00       ADC #<Font     ; += FontLo; Carry = 0 since R=0 from above

It is "funny" how we end up shifting and rotating in the same direction: Left! :-) It would nice to leverage this work for both the high and low byte address calculation. That is the technical term for "Don't do dumb (redundant) work" or in the immortal words of Back to the Future: "Think, McFly!" :-)

Given: c

Start	AddressHi	AddressLo
PQRstuvw	00000PQR	stuvw000

Our original glyph fits in one byte. Our 16-bit address offset technically could also fit in one byte -- we just have shifted it over 3 bits with zeroes.

If we had the byte: stuvwPQR that would be extremely convenient as it would be trivial to calculate the offset:

    PHA             ; push A = %stuvwPQR% initial glyph
    AND #$F8        ;      A = %stuvw000
    STA AddressLo
    PLA             ; pop  A = %stuvwPQR
    AND #$07        ;      A = %00000PQR
    CLC
    ADC FontHi
    STA AddressHi

Let us trace the rotate left ROL instruction 4 times paying attention to the Carry register C and the Accumulator A:

ROL   C   A         Comment
0     ?   PQRstuvw  initial glyph to draw
1     P   QRstuvw?
2     Q   Rstuvw?P
3     R   stuvw?PQ

Hmm, that is suspiciously close to what we want! Is there any way we can end up with?

ROL   C   A
?     ?   stuvwPQR

We would need to start with the carry pre-loaded with P, and the QR already shifted over one to the left:

ROL   C   A
n-3   P   QR?stuvw
n-2   Q   R?stuvwP
n-1   R   ?stuvwPQ
n     ?   stuvwPQR

That might look like something like this:

    PHA        ; push c
    AND #$E0
    TAX        ; save %PQR00000
    PLA        ; pop c
    AND #$1F
    STA Temp   ; save %000stuvw
    TXA        ;     A=PQR00000
    ASL        ; C=P A=QR000000  <- stuvw not taking advantage of shift :-(
    OR  Temp   ; C=P A=QR0stuvw
    ROL        ; C=Q A=R0stuvwP
    ROL        ; C=R A=0stuvwPQ
    ROL        ; C=0 A=stuvwPQR
    PHA
    AND #1F        ;        = %000stuvw
    STA _LoadFont+1; AddressLo = FontLo + (c*8)
    PLA
    AND #E0
    CLC
    ADC AddressHi
    STA _LoadFont+2

Hmm, that seems like an awful lot of work just for some bit-shuffling!! For one thing we're doing a shift and stuvw is not taking advantage of it. Can we not we make use of the fact that we will eventually be doing AND #1F and AND #E0 ?

The lateral thinking is to use partial results.

ROL   C   A = glyph to draw
0     ?   PQRstuvw
1     P   QRstuvw?
2     Q   Rstuvw?P
3     R   stuvw?PQ
          --push A--
          --A & #F8
          --store low byte offset--
          --pop A--
4     s   tuvw?PQR
          --CLC--
          --add FontHi to A
          --store high byte offset--

Let's code this up:

    ; FUNC: DrawCharCol( c, col ) alias _DrawChar2
    ; PARAM: A = glyph to draw
    ; PARAM: Y = column to draw at; $0 .. $27 (Columns 0 .. 39) (not modified)
    ; NOTES: X is destroyed
                     ORG $033A
    033A:          DrawCharCol      ;     A=%PQRstuvw
    033A:2A          ROL            ; C=P A=%QRstuvw?
    033B:2A          ROL            ; C=Q A=%Rstuvw?P
    033C:2A          ROL            ; C=R A=%stuvw?PQ
    033D:AA          TAX            ;     X=%stuvw?PQ push glyph
    033E:29 F8       AND #F8        ;     A=%stuvw000
    0340:8D 53 03    STA _LoadFont+1; AddressLo = (c*8)
    0343:8A          TXA            ;     A=%stuvw?PQ pop glyph
    0344:29 03       AND #3         ; Optimization: s=0 implicit CLC !
    0346:2A          ROL            ; C=s A=%00000PQR and 1 last ROL to get R
    0347:69 60       ADC #>Font     ; += FontHi; Carry=0 since s=0 from above
    0349:8D 54 03    STA _LoadFont+2; AddressHi = FontHi + (c/32)
                                    ; intentional fall into _DrawChar1 @ $034C

Since we'll re-use our existing font drawing code _DrawChar1 at $034C it is always a good idea to document why there is no RTS at the end.

Here is a comparison between the original and final version (clock cycle timings are the #'s):

ORG $0335 (old)   ORG $033A (new)
3 PHA             2 ROL              
2 AND #1F         2 ROL              
2 ASL             2 ROL              
2 ASL             2 TAX              
2 ASL             2 AND #F8          
2 ADC #<Font      4 STA _LoadFont+1  
4 STA _LoadFont+1 2 TXA              
4 PLA             2 AND #3           
2 AND #E0         2 ROL              
2 ROL             3 ADC #>Font       
2 ROL             4 STA _LoadFont+2  
2 ROL             -
2 ROL             -
2 ADC #>Font      -
4 STA _LoadFont+2 -

       Original   Final
Bytes  23         18
Cycles 37         27

Much better!!!

We need to (again) touch up our PrintCharentry point at $0310 calling _DrawChar2 ($034C) to call DrawCharCol ($033A):

    310:4C 3A 03         JMP DrawCharCol

Enter in:

B00:20 0A 09 A9 00 A0 00 4C 10 03
310:4C 3A 03
33A:2A 2A 2A AA 29 F8
340:8D 53 03 8A 29 03 2A 69
348:60 8D 54 03
B00G

We should now see an closed apple glyph!

To change which glyph is printed:

B04:41
B00G

And we should see an A printed.

We now have the ability to print any of the 128 ASCII characters!

**AppleWin** users: Press `F7`, copy & paste the below, press `F7` when done.

SYM DrawCharCol = 33A
33AL

(Debugger screenshot forthcoming.)

Character Inspector

Let's verify this by writing a character inspector. We'll use the arrow keys to select the glyph and ESC to exit.

    ; FUNC: DemoCharInspect()
                  KEYBOARD    EQU $C000
                  KEYSTROBE   EQU $C010
                  glyph       EQU $FE

                     ORG $1000
    1000:         DemoCharInspect
    1000:A9 00       LDA #0         ; glyph=0
    1002:85 FE       STA glyph      ; save which glyph to draw
    1004:A9 00    .1 LDA #0         ; screen = 0x2000
    1006:85 F5       STA HgrLo      ;
    1008:A9 20       LDA #20        ;
    100A:85 F6       STA HgrHi      ;
    100C:A5 FE       LDA glyph      ; A = glyph
    100E:A0 00       LDY #00        ; Y = col
    1010:20 10 03    JSR PrintChar
    1013:AD 00 C0 .2 LDA KEYBOARD   ; read A=key
    1016:10 FB       BMI .2         ; no key?
    1018:8D 10 C0    STA KEYSTROBE  ; debounce key
    101B:C9 88       CMP #88        ; key == <-- ?
    101D:D0 0A       BNE .4         ;
    101F:C6 FE       DEC glyph      ; yes, --glyph
    1021:A5 FE    .3 LDA glyph      ; glyph &= 0x7F
    1023:29 7F       AND #7F        ;
    1025:85 FE       STA glyph      ;
    1027:10 DB       BPL .1         ; always branch, draw prev char
    1029:C9 95    .4 CMP #95        ; key == --> ?
    102B:D0 05       BNE .5         ;
    102D:E6 FE       INC glyph      ; yes, ++glyph
    102F:18          CLC            ; always branch
    1030:90 EF       BCC .3         ;   draw prev char
    1032:C9 9B    .5 CMP #9B        ; key == ESC ?
    1034:D0 DD       BNE .2         ;
    1036:60          RTS            ; yes, exit

Enter in this code:

1000:A9 00 85 FE A9 00 85 F5
1008:A9 20 85 F6 A5 FE A0 00
1010:20 10 03 AD 00 C0 10 FB
1018:8D 10 C0 C9 88 D0 0A C6
1020:FE A5 FE 29 7F 85 FE 10
1028:DB C9 95 D0 05 E6 FE 18
1030:90 EF C9 9B D0 DD 60
1000G

We now have an ASCII char inspector!

**AppleWin** users: Press `F7`, copy & paste the below, press `F7` when done.

SYM glyph = FE
1000L

Character Inspector version 2

Let's fix it up to print the hex value of the current character we are inspecting:

                     ORG $1010
    1010:20 3C 10    JSR Patch1

                     ORG $1037
    103C:         Patch1
    103C:48          PHA            ; save c
    103D:20 10 03    JSR DrawChar
    1040:68          PLA            ; restore c so we can print it in hex
    1041:4C 01 03    JMP DrawHexByte

                     ORG $0303
    ; FUNC: DrawHexByte( c ) = $0301
    ; PARAM: A = byte to print in hex
    0301:         DrawHexByte
    0301:48          PHA            ; save low nibble
    0302:6A          ROR            ; shift high nibble
    0303:6A          ROR            ; to low nibble
    0304:6A          ROR
    0305:6A          ROR
    0306:20 0A 03    JSR DrawHexNib ; print high nib in hex
    0309:68          PLA            ; pritn low  nib in hex

    ; FUNC: DrawHexNib() = $030C
    ; PARAM: A = nibble to print as hex char
    030A:         DrawHexNib
    030A:29 0F       AND #F         ; base 16
    030C:AA          TAX            ;
    030D:BD 90 03    LDA NIB2HEX,X  ; nibble to ASCII
                     ; intentional fall into PrintChar

                     ORG $0390
    0390:30 31 32 33 NIB2HEX ASC "0123456789ABCDEF"
    0394:34 35 36 37
    0398:38 39 41 42
    039C:43 44 45 46

Enter in the changes:

1010:20 3C 10
103C:48 20 10 03 68 4C 01 03
0301:   48 6A 6A 6A 6A 20 0A
0308:03 68 29 0F AA BD 90 03
0390:30 31 32 33 34 35 36 37
0398:38 39 41 42 43 44 45 46
1000G

And now we have our own DrawHexByte() function.

**AppleWin** users: Press `F7`, copy & paste the below, press `F7` when done.

sym patch1      = 103C
sym DrawHexByte = 0301
sym DrawHexNib  = 030A
db  Nib2HeX     390:39F
1000L

Character Inspector version 3

Let's add a space after the character but before the hex value to improve readability of the output. The new code is:

                     ORG $1010
    1010:20 37 10    JSR Patch2

                     ORG $1037
    1037:         Patch2
    1037:48          PHA            ; save c
    1038:20 10 03    JSR PrintChar
    103B:A9 20       LDA ' '        ; Draw whitespace

Enter in these changes:

1010:20 37 10
1037:48 20 10 03 A9 20
1000G

Our final version is:

1000:A9 00 85 FE A9 00 85 F5
1008:A9 20 85 F6 A5 FE A0 00
1010:20 37 10 AD 00 C0 10 FB
1018:8D 10 C0 C9 88 D0 0A C6
1020:FE A5 FE 29 7F 85 FE 10
1028:DB C9 95 D0 05 E6 FE 18
1030:90 EF C9 9B D0 DD 60 48

3 1038:20 10 03 A9 20 20 10 03 1040:68 4C 01 03 1000G

(To save this, BSAVE CHAR_INSPECT3.BIN,A$1000,L$68)

**AppleWin** users: Press `F7`, copy & paste the below, press `F7` when done.

sym patch2      = 1037

Y Cursor Position

Right now the line we "print" to is hard-coded since we are using a screen address of $2000 with the pointer at $E5, $E6.

We're going to digress slightly before we fix this.

The secret to getting high speed graphics rendering on the Apple is to use a look-up table. We're going to have a 16-bit address lookup table for Y=0, Y=8, Y=16, .. Y = 184

The HGR screen address is broken up a triad. Every 64 scan lines the offset change by $28.

Table 2: HGR Y Address for every 8 scanlines

Y	Address	Hi	Lo
0	$2000	$20	$00
8	$2080	$20	$80
16	$2100	$21	$00
24	$2180	$21	$80
32	$2200	$22	$00
40	$2280	$22	$80
48	$2300	$23	$00
56	$2380	$23	$80
-	-----	-	-
64	$2028	$20	$28
72	$20A8	$20	$A8
80	$2128	$21	$28
88	$21A8	$21	$A8
96	$2228	$22	$28
104	$22A8	$22	$A8
112	$2328	$23	$28
120	$23A8	$23	$A8
-	-----	-	-
128	$2050	$20	$50
136	$20D0	$20	$D0
144	$2150	$21	$50
152	$21D0	$21	$D0
160	$2250	$22	$50
168	$22D0	$22	$D0
176	$2350	$23	$50
184	$23D0	$23	$D0

We'll split this table of 16-bit addresses into Low and High bytes for easier access. We'll also subtract off the hard-coded graphics page 1 high byte = $20 and instead use relative offsets to make it work with either graphics page 1 or 2.

This is our mini HGR Y Address look-up table. "Funny" that it has 24 entries -- the same height as our text screen. :-)

Enter these bytes (or save hgrtable.bin and bload hgrtable.bin,3A0):

Our HgrLoY table:

03A0:00 80 00 80 00 80 00 80
03A8:28 A8 28 A8 28 A8 28 A8
03B0:50 D0 50 D0 50 D0 50 D0

Our HgrHiY table:

03B8:00 00 01 01 02 02 03 03
03C0:00 00 01 01 02 02 03 03
03C8:00 00 01 01 02 02 03 03

**AppleWin** users: To save this press `F7`, at the debugger console `bsave "hgrtable.bin",3A0:3CF`, press `F7`.

DB HgrLoY 3A0:3B7
DB HgrHiY 3B8:3CF

To select which row to draw at we'll pass that in the X register to our DrawCharColRow() routine:

    ; FUNC: SetCursorRow( row )
    ; PARAM: X = row    to draw at; $0 .. $17 (Rows 0 .. 23) (not modified)
    ; INPUT : $E5,$E6 initial pointer to the destination screen scanline
    ;         Note: Must start at every 8 scanlines.
    ; OUTPUT: $F5,$F5 working pointer to the destination screen scanline
                     ORG $0313
    0313:         SetCursorRow
    0313:BD A0 03    LDA HgrLoY,X   ; HgrLoY[ row ]
    0316:85 F5       STA TmpLo
    0318:BD B8 03    LDA HgrHiY,X   ; HgrHiY[ row ]
    031B:18          CLC
    031C:65 E6       ADC HgrHi
    031E:85 F6       STA TmpHi
    0320:60          RTS

    ; FUNC: DrawCharColRow()
    ; PARAM: A = glyph to draw
    ; PARAM: Y = column to draw at  ; $0 .. $27 (Columns 0 .. 39) (not modified)
    ; PARAM: X = row    to draw at  ; $0 .. $17 (Rows 0 .. 23) (destroyed)
                     ORG $0335
    0335:         DrawCharColRow
    0335:48          PHA
    0336:20 13 03    JSR SetCursorRow
    0339:68          PLA
                                    ; intentional fall into _DrawChar2

Enter in:

313:         BD A0 03 85 F5
318:BD B8 03 18 65 E6 85 F6
320:60
335:48 20 13 03 68

Now we can print a char at any location:

                    ORG $1100
    1100:         PrintAYX
    1100:A9 41       LDA #41        ; A-register = char
    1102:A0 01       LDY #1         ; Y-register = col 1 (2nd column)
    1104:A2 02       LDX #2         ; X-register = row 2 (3rd row)
    1106:4C 35 03    JMP DrawCharColRow

Enter in:

1100:A9 41 A0 01 A2 02 4C 35 03
1100G

**AppleWin** users: You know the drill ...

SYM SetCursorRow   = 313
SYM DrawCharColRow = 335

Natural Params SetCursorColRow()

Unfortunately, our usage of the X and Y registers are not intuitive. This is due to the limited addressing modes of the 6502. :-/ If the 6502 had a symmetrical indirect zero-page X addressing mode:

 LDA ($ZP),X

We could map the X-register to the natural column (x-axis), and the Y-register to the natural row (y-axis). Alas, we're stuck with the X=row and Y=col unless we wanted to add extra code to "swap" the two.

    ; FUNC: SetCursorColRowYX()
    ; PARAM: Y = col
    ; PARAM: X = row
                     ORG $0369
    369:         SetCursorColRowYX
    369:20 13 03     JSR SetCursorRow
    36C:18           CLC
    36D:98           TYA
    36E:65 F5        ADC $F5
    371:85 F5        STA $F5
    373:60           RTS

Or are we stuck? Since we're using a function to calculate the destination address let's fix the order.

We'll need to change the X offset in SetCursorColRowXY() to Y;

    ; FUNC: SetCursorColRow2a( row )
    ; PARAM: Y = row
    ; NOTES: Version 2a !
    0928:            ORG $0928
    0928:         SetCursorColRow2a
    0928:B9 A0 03    LDA HgrLoY,Y   ; changed from: ,X
    092B:18          CLC
    092C:65 E5       ADC HgrLo
    092E:85 F5       STA TmpLo
    0930:B9 B8 03    LDA HgrHiY,Y   ; changed from: ,X
    0933:18          CLC
    0934:65 E6       ADC HgrHi
    0936:85 F6       STA TmpHi
    0938:60          RTS

And change the low byte to add X instead:

    ; FUNC: SetCursorColRow2b( col, row ) = $0379
    ; PARAM: X = col
    ; PARAM: Y = row
    ; NOTES: Version 2b !
                     ORG $0979
    979:         SetCursorColRow2b
    979:20 13 03     JSR SetCursorRow
    97C:18           CLC
    37D:88           TXA            ; changed from: TYA
    97E:65 F5        ADC $F5
    981:85 F5        STA $F5
    983:60

This is a little clunky but it is progress. Let's write the new SetCursorColRow() version with the SetCursorRow() inlined so we don't have to use a JSR.

    ; FUNC: SetCursorColRow( col, row )
    ; PARAM: X = column to draw at; $0 .. $27 (Columns 0 .. 39) (not modified)
    ; PARAM: Y = row    to draw at; $0 .. $17 (Rows 0 .. 23) (not modified)
    ; NOTES: Version 3! X and Y is swapped from earlier version!
    ; [$F5] = HgrLoY[ Y ] + ScreenLo + X
                     ORG $0321
    0321:         SetCursorColRow
    0321:86 F5       STX TmpLo
    0323:B9 A0 03    LDA HgrLoY,Y   ; HgrLoY[ row ]
    0326:18          CLC
    0327:65 F5       ADC TmpLo      ; add column
    0329:85 F5       STA TmpLo
    032B:B9 B8 03    LDA HgrHiY,Y   ; HgrHiY[ row ]
    032E:18          CLC            ; \ could optimize this into
    032F:65 E6       ADC HgrHi      ; / single ORA HgrHi
    0331:85 F6       STA TmpHi
    0333:60          RTS
    0334:EA          NOP            ; pad

Enter in:

321:   86 F5 B9 A0 03 18 65
328:F5 85 F5 B9 B8 03 18 65
330:E6 85 F6 60

DrawString()

Now that we have the basic print char working lets extend it to print a C-style string (one that is zero terminated.)

                  String EQU $F0

    ; FUNC: DrawString( *text )
    ; PARAM: X = High byte of string address
    ; PARAM: Y = Low  byte of string address
    037E:            ORG $037E
    037E:         DrawString
    037E:84 F0       STY String+0
    0380:86 F1       STX String+1
    0382:A0 00       LDY #0
    0384:B1 F0    .1 LDA (String),Y
    0386:F0 07       BEQ .2         ; null byte? Done
    0388:20 10 03    JSR DrawChar   ; or DrawCharCol for speed
    038B:C0 28       CPY #40        ; col < 40?
    038D:90 F5       BCC .1
    038F:60       .2 RTS

**AppleWin**:

SYM SetCursorColRow = 321
SYM DrawString      = 37E
ASC Msg 120E:1219

And our example to verify that it works:

    ; FUNC: DemoDrawString()
                      ORG $1200
    1200:          DemoDrawString
    1200:A2 03        LDX #3        ; col = 3
    1202:A0 02        LDY #2        ; row = 2
    1204:20 21 03     JSR SetCursorColRow
    1207:A2 12        LDX #>Tx      ; High
    1209:A0 0E        LDY #<Tx      ; Low
    120B:4C 7E 03     JMP DrawString
    120E:          Tx ASC "Hello World",0
    120E:48 65 6C 6C 6F 20 57 6F 72 6C 64 00

Enter in:

37E:84 F0 86 F1 A0 00 B1 F0
386:F0 07 20 10 03 C0 28 90 F5 60

1200:A2 03 A0 02 20 21 03
1207:A2 12 A0 0E 4C 7E 03
120E:48 65 6C 6C 6F 20 57 6F 72 6C 64 00
1200G

Note: An easy way to get the hex bytes for a string is to use this tiny JavaScript snippet to convert a text string to hex:

    var txt = "Hello World";
    for( var i=0; i < txt.length; ++i )
        console.log( txt.charCodeAt(i).toString(16) );

Recap

Here are all the routines we've entered in so far:

0301:   48 6A 6A 6A 6A 20 0A
0308:03 68 29 0F AA BD 90 03
0310:4C 3A 03 BD A0 03 85 F5
0318:BD B8 03 18 65 E6 85 F6
0320:60 86 F5 B9 A0 03 18 65
0328:F5 85 F5 B9 B8 03 18 65
0330:E6 85 F6 60 EA 48 20 13
0338:03 68 2A 2A 2A AA 29 F8
0340:8D 53 03 8A 29 03 2A 69
0348:60 8D 54 03 A6 F6 86 FD
0350:A2 00 BD 00 00 91 F5 18
0358:A5 F6 69 04 85 F6 E8 E0
0360:08 D0 EF C8 A6 FD 86 F6
0368:60
0390:30 31 32 33 34 35 36 37
0398:38 39 41 42 43 44 45 46
03A0:00 80 00 80 00 80 00 80
03A8:28 A8 28 A8 28 A8 28 A8
03B0:50 D0 50 D0 50 D0 50 D0
03B8:00 00 01 01 02 02 03 03
03C0:00 00 01 01 02 02 03 03
03C8:00 00 01 01 02 02 03 03

(To save this: BSAVE CODE_0300.BIN,A$300,L$D0)

What's left? Quite a few things actually:

Copy the 40-Column text screen to HGR
Scroll the HGR screen up by 1 pixel
Copy the 80-Column text screen to DHGR (Double High Resolution)
Hook into the COUT so all text appears onto the HGR or DHGR screen

Let's implement those first two.

Copy text screen to HGR

For our final trick we are going to copy the characters off the text screen onto the HGR screen. More magic? Nah, just bit-shuffling.

The text screen, like the HGR screen, is also non-linear, and also broken up into a triad:

Row	Text Address	HGR Address
0	$400	$2000
1	$480	$2080
2	$500	$2100
3	$580	$2180
4	$600	$2200
5	$600	$2280
6	$700	$2300
7	$780	$2380
-	----	-----
8	$428	$2028
9	$4A8	$20A8
10	$528	$2128
11	$5A8	$21A8
12	$628	$2228
13	$6A8	$22A8
14	$728	$2328
15	$7A8	$23A8
-	----	-----
16	$450	$2050
17	$4D0	$20D0
18	$550	$2150
19	$5D0	$21D0
20	$650	$2250
21	$6D0	$22D0
22	$750	$2350
23	$7D0	$23D0

While the Apple's memory layout seems esoteric it has beautiful symmetry. For any given text row notice that:

the low byte of the text address is the same low byte of the HGR address
the high byte of the text address is 0x1C less then the high byte of the HGR address

Technically, to convert the HGR high byte address to a Text high byte address, we only need to map these 4 high bytes:

HGR High Byte	Text High Byte
$20 = %0010_0000	$4 = %0000_0100
$21 = %0010_0001	$5 = %0000_0101
$22 = %0010_0010	$6 = %0000_0110
$23 = %0010_0011	$7 = %0000_0111

Which we could do via:

    LDA HgrHiY, Y     ; Y is row
    AND #7            ; strip off top 6 bits
    OR  #4            ; Set text page 1 = $0400

But we'll save a byte and use the normal subtraction instead:

    LDA HgrHiY, Y     ; Y is row
    CLC               ; Convert HgrHiY to TextHiY byte
    SBC #$1B          ; A -= 0x1C

If we care about absolute speed we could see which one takes the fewer clock cycles.

There is also the reverse conversion -- to convert a Text address to a HGR address which could be done with the same AND #3, OR #20, but since we don't have a Text Y table address and already have a HGR 16-bit address table we will re-use that.

Here's the Pseudo-code to copy the text screen to the HGR Screen:

    for( row = 0; row < 24; row++ )
    {
       SrcTextLo = HgrLoY[ row ];
       SrcTextHi = HgrHiY[ row ] - 0x1C;
    // SetCursorColRow( 0, row ) which does:
       DstHgrLo  = HgrLoY[ row ]
       DstHgrHi  = HgrHiY[ row ]

       for( col = 0; col < 40; col++ )
       {
           c = SrcText[ col ]
           DrawCharCol( c );
       }
    }

And here is the assembly:

    ; FUNC: CopyTextToHGR()
    ; DATA:
    ;    $6000.$63FF  Font 7x8 Data
    ;    $6400.$642F  HgrLoY, HgrHiY table for every 8 scanlines
    1300:         Txt    EQU $F7
                     ORG $1300
    1300:         CopyTextToHgr
    1300:A9 00       LDA #0
    1302:85 F3       STA row
    1304:85 E5       STA $E5
    1306:A9 20       LDA #$20       ; Dest = HGR1 = $2000
    1308:85 E6       STA $E6
    130A:A4 F3    .1 LDY row        ; Y = row
    130C:C0 18       CPY #24        ; 24 rows is #$18
    130E:B0 22       BCS .3         ; Y >= 24
    1310:A2 00       LDX #0
    1312:86 F2       STX col        ; X = col
    1314:20 21 03    JSR SetCursorColRow
    1317:38          SEC            ; A = HgrHiY[ row ]
    1318:E9 1C       SBC #$1C       ; Convert HgrHiY to TextHiY byte
    131A:85 F8       STA Txt+1      ; A -= 0x1C -> TxtHi
    131C:B9 A0 03    LDA HgrLoY, Y  ; A = HgrLoY[ row ]
    131F:85 F7       STA Txt+0      ;           -> TxtLo
    1321:A4 F2       LDY col
    1323:B1 F7    .2 LDA (Txt),Y
    1325:29 7F       AND #$7F
    1327:20 3A 03    JSR DrawCharCol
    132A:C0 28       CPY #$28       ; 40 cols is #$28
    132C:90 F5       BCC .2         ; Y < 40
    132E:E6 F3       INC row
    1330:D0 D8       BNE .1         ; always
    1332:60       .3 RTS

**AppleWin** users:

sym Txt2Hgr.1 = 130A
sym Txt2Hgr.2 = 1323
sym Txt2Hgr.3 = 1332

Enter in:

1300:A9 00 85 F3 85 E5 A9 20
1308:85 E6 A4 F3 C0 18 B0 22
1310:A2 00 86 F2 20 21 03 38
1318:E9 1C 85 F8 B9 A0 03 85
1320:F7 A4 F2 B1 F7 29 7F 20
1328:3A 03 C0 28 90 F5 E6 F3
1330:D0 D8 60

And now for the moment of truth! Don't worry if you can't see what you are typing.

FC58G
1300L
1300G

Voila!

In case you were wondering why I turned 50% scanlines on this is how the HGR screen would normally look like in color:

That's why I turned 50% scanlines on, for better readability:

Using one of the newer emulators with NTSC emulation, unfortunately, doesn't help with readaibility: :-/

NTSC Alpha (with tweaked Palette):

NTSC Sheldon -- which unfortunately has WAY too much ghosting: :-(

And just to prove that it copied the bottom 4 text rows as well:

C052

And to restore the bottom 4 text rows

C053

Exercise 1: ScrollHgrUpPixel()

Hey! Homework? Yes, the only (true) way to demonstrate you understand the theory is with implementation:

Write a function to "scroll" the HGR screen up:

* one "text line" (8 pixels), and

Hint: This is basically a gloried and specialized `memcpy()`.

Exercise 2: ScrollHgrUpLine()

Write a function to "scroll" the HGR screen up:

* one scan line (1 pixel)

Hint: For scrolling up one pixel we can spot the pattern if we inspect
the memory flow of how pixels get shuffled around:

    40 bytes from $2400.$2427 -> $2000.$2027
    40 bytes from $2800.$2827 -> $2400.$2427
    etc

Don't forget that you only need to copy 191 rows, not 192, since the
very bottom scanline should be "blank."

Conclusion

Hope this HGR font tutorial helped you understand the inner workings of a font blitter!

Happy (Apple ][ //e //c) Hacking! Michael "AppleWin Debug Dev"

Solution 1: ScrollHgrUpLine()

Figure it out ! You have all the tools and knowledge.

Solution 2: ScrollHgrUpPixel()

There are many different ways to solve this depending if we want to prioritize space or speed.

We could manually unroll every loop such as this monstrosity (we trade space for speed):

Enter this (or download hgr_scroll_up.bin or BRUN HGR_SCROLL_UP):

1400:A2 27
1402:BD 00 24 9D 00 20
1408:BD 00 28 9D 00 24
140E:BD 00 2C 9D 00 28
1414:BD 00 30 9D 00 2C
141A:BD 00 34 9D 00 30
1420:BD 00 38 9D 00 34
1426:BD 00 3C 9D 00 38
142C:BD 80 20 9D 00 3C
1432:BD 80 24 9D 80 20
1438:BD 80 28 9D 80 24
143E:BD 80 2C 9D 80 28
1444:BD 80 30 9D 80 2C
144A:BD 80 34 9D 80 30
1450:BD 80 38 9D 80 34
1456:BD 80 3C 9D 80 38
145C:BD 00 21 9D 80 3C
1462:BD 00 25 9D 00 21
1468:BD 00 29 9D 00 25
146E:BD 00 2D 9D 00 29
1474:BD 00 31 9D 00 2D
147A:BD 00 35 9D 00 31
1480:BD 00 39 9D 00 35
1486:BD 00 3D 9D 00 39
148C:BD 80 21 9D 00 3D
1492:BD 80 25 9D 80 21
1498:BD 80 29 9D 80 25
149E:BD 80 2D 9D 80 29
14A4:BD 80 31 9D 80 2D
14AA:BD 80 35 9D 80 31
14B0:BD 80 39 9D 80 35
14B6:BD 80 3D 9D 80 39
14BC:BD 00 22 9D 80 3D
14C2:BD 00 26 9D 00 22
14C8:BD 00 2A 9D 00 26
14CE:BD 00 2E 9D 00 2A
14D4:BD 00 32 9D 00 2E
14DA:BD 00 36 9D 00 32
14E0:BD 00 3A 9D 00 36
14E6:BD 00 3E 9D 00 3A
14EC:BD 80 22 9D 00 3E
14F2:BD 80 26 9D 80 22
14F8:BD 80 2A 9D 80 26
14FE:BD 80 2E 9D 80 2A
1504:BD 80 32 9D 80 2E
150A:BD 80 36 9D 80 32
1510:BD 80 3A 9D 80 36
1516:BD 80 3E 9D 80 3A
151C:BD 00 23 9D 80 3E
1522:BD 00 27 9D 00 23
1528:BD 00 2B 9D 00 27
152E:BD 00 2F 9D 00 2B
1534:BD 00 33 9D 00 2F
153A:BD 00 37 9D 00 33
1540:BD 00 3B 9D 00 37
1546:BD 00 3F 9D 00 3B
154C:BD 80 23 9D 00 3F
1552:BD 80 27 9D 80 23
1558:BD 80 2B 9D 80 27
155E:BD 80 2F 9D 80 2B
1564:BD 80 33 9D 80 2F
156A:BD 80 37 9D 80 33
1570:BD 80 3B 9D 80 37
1576:BD 80 3F 9D 80 3B
157C:BD 28 20 9D 80 3F
1582:BD 28 24 9D 28 20
1588:BD 28 28 9D 28 24
158E:BD 28 2C 9D 28 28
1594:BD 28 30 9D 28 2C
159A:BD 28 34 9D 28 30
15A0:BD 28 38 9D 28 34
15A6:BD 28 3C 9D 28 38
15AC:BD A8 20 9D 28 3C
15B2:BD A8 24 9D A8 20
15B8:BD A8 28 9D A8 24
15BE:BD A8 2C 9D A8 28
15C4:BD A8 30 9D A8 2C
15CA:BD A8 34 9D A8 30
15D0:BD A8 38 9D A8 34
15D6:BD A8 3C 9D A8 38
15DC:BD 28 21 9D A8 3C
15E2:BD 28 25 9D 28 21
15E8:BD 28 29 9D 28 25
15EE:BD 28 2D 9D 28 29
15F4:BD 28 31 9D 28 2D
15FA:BD 28 35 9D 28 31
1600:BD 28 39 9D 28 35
1606:BD 28 3D 9D 28 39
160C:BD A8 21 9D 28 3D
1612:BD A8 25 9D A8 21
1618:BD A8 29 9D A8 25
161E:BD A8 2D 9D A8 29
1624:BD A8 31 9D A8 2D
162A:BD A8 35 9D A8 31
1630:BD A8 39 9D A8 35
1636:BD A8 3D 9D A8 39
163C:BD 28 22 9D A8 3D
1642:BD 28 26 9D 28 22
1648:BD 28 2A 9D 28 26
164E:BD 28 2E 9D 28 2A
1654:BD 28 32 9D 28 2E
165A:BD 28 36 9D 28 32
1660:BD 28 3A 9D 28 36
1666:BD 28 3E 9D 28 3A
166C:BD A8 22 9D 28 3E
1672:BD A8 26 9D A8 22
1678:BD A8 2A 9D A8 26
167E:BD A8 2E 9D A8 2A
1684:BD A8 32 9D A8 2E
168A:BD A8 36 9D A8 32
1690:BD A8 3A 9D A8 36
1696:BD A8 3E 9D A8 3A
169C:BD 28 23 9D A8 3E
16A2:BD 28 27 9D 28 23
16A8:BD 28 2B 9D 28 27
16AE:BD 28 2F 9D 28 2B
16B4:BD 28 33 9D 28 2F
16BA:BD 28 37 9D 28 33
16C0:BD 28 3B 9D 28 37
16C6:BD 28 3F 9D 28 3B
16CC:BD A8 23 9D 28 3F
16D2:BD A8 27 9D A8 23
16D8:BD A8 2B 9D A8 27
16DE:BD A8 2F 9D A8 2B
16E4:BD A8 33 9D A8 2F
16EA:BD A8 37 9D A8 33
16F0:BD A8 3B 9D A8 37
16F6:BD A8 3F 9D A8 3B
16FC:BD 50 20 9D A8 3F
1702:BD 50 24 9D 50 20
1708:BD 50 28 9D 50 24
170E:BD 50 2C 9D 50 28
1714:BD 50 30 9D 50 2C
171A:BD 50 34 9D 50 30
1720:BD 50 38 9D 50 34
1726:BD 50 3C 9D 50 38
172C:BD D0 20 9D 50 3C
1732:BD D0 24 9D D0 20
1738:BD D0 28 9D D0 24
173E:BD D0 2C 9D D0 28
1744:BD D0 30 9D D0 2C
174A:BD D0 34 9D D0 30
1750:BD D0 38 9D D0 34
1756:BD D0 3C 9D D0 38
175C:BD 50 21 9D D0 3C
1762:BD 50 25 9D 50 21
1768:BD 50 29 9D 50 25
176E:BD 50 2D 9D 50 29
1774:BD 50 31 9D 50 2D
177A:BD 50 35 9D 50 31
1780:BD 50 39 9D 50 35
1786:BD 50 3D 9D 50 39
178C:BD D0 21 9D 50 3D
1792:BD D0 25 9D D0 21
1798:BD D0 29 9D D0 25
179E:BD D0 2D 9D D0 29
17A4:BD D0 31 9D D0 2D
17AA:BD D0 35 9D D0 31
17B0:BD D0 39 9D D0 35
17B6:BD D0 3D 9D D0 39
17BC:BD 50 22 9D D0 3D
17C2:BD 50 26 9D 50 22
17C8:BD 50 2A 9D 50 26
17CE:BD 50 2E 9D 50 2A
17D4:BD 50 32 9D 50 2E
17DA:BD 50 36 9D 50 32
17E0:BD 50 3A 9D 50 36
17E6:BD 50 3E 9D 50 3A
17EC:BD D0 22 9D 50 3E
17F2:BD D0 26 9D D0 22
17F8:BD D0 2A 9D D0 26
17FE:BD D0 2E 9D D0 2A
1804:BD D0 32 9D D0 2E
180A:BD D0 36 9D D0 32
1810:BD D0 3A 9D D0 36
1816:BD D0 3E 9D D0 3A
181C:BD 50 23 9D D0 3E
1822:BD 50 27 9D 50 23
1828:BD 50 2B 9D 50 27
182E:BD 50 2F 9D 50 2B
1834:BD 50 33 9D 50 2F
183A:BD 50 37 9D 50 33
1840:BD 50 3B 9D 50 37
1846:BD 50 3F 9D 50 3B
184C:BD D0 23 9D 50 3F
1852:BD D0 27 9D D0 23
1858:BD D0 2B 9D D0 27
185E:BD D0 2F 9D D0 2B
1864:BD D0 33 9D D0 2F
186A:BD D0 37 9D D0 33
1870:BD D0 3B 9D D0 37
1876:BD D0 3F 9D D0 3B
187C:A9 00    9D D0 3F
1881:CA 30 03 4C 02 14
1887:60

And let's write a little demo ...

    13F7:A0 C0        LDY #C0
    13F9:20 00 14  .1 JSR ScrollHgrUpPixel
    13FC:88           DEY
    13FD:D0 FA        BNE .1
    13FF:60           RTS

Enter in:

13F7:A0 C0 20 00 14 88 D0 FA 60

(To save to disk type BSAVE HGR_SCROLL_UP.BIN,A$13F7,L$490)

And let's try it out:

And for the finale:

13F7G

Sweet !

Here's the (non-standard) assembly to scroll the HGR screen up one pixel: (I'm using the non-conventional : as an assembler end-of-statement sepearator to logically group the byte copies together.)

    ; FUNC: ScrollHgrUpPixel()
              ORG $1400
    1400:     LDX #27 ; 39 columns       ; Src Y    Dst Y
    1402:  .1 LDA $2400,X : STA $2000,X  ; [  1] -> [  0]
    1408:     LDA $2800,X : STA $2400,X  ; [  2] -> [  1]
    140E:     LDA $2C00,X : STA $2800,X  ; [  3] -> [  2]
    1414:     LDA $3000,X : STA $2C00,X  ; [  4] -> [  3]
    141A:     LDA $3400,X : STA $3000,X  ; [  5] -> [  4]
    1420:     LDA $3800,X : STA $3400,X  ; [  6] -> [  5]
    1426:     LDA $3C00,X : STA $3800,X  ; [  7] -> [  6]
    142C:     LDA $2080,X : STA $3C00,X  ; [  8] -> [  7]
    1432:     LDA $2480,X : STA $2080,X  ; [  9] -> [  8]
    1438:     LDA $2880,X : STA $2480,X  ; [ 10] -> [  9]
    143E:     LDA $2C80,X : STA $2880,X  ; [ 11] -> [ 10]
    1444:     LDA $3080,X : STA $2C80,X  ; [ 12] -> [ 11]
    144A:     LDA $3480,X : STA $3080,X  ; [ 13] -> [ 12]
    1450:     LDA $3880,X : STA $3480,X  ; [ 14] -> [ 13]
    1456:     LDA $3C80,X : STA $3880,X  ; [ 15] -> [ 14]
    145C:     LDA $2100,X : STA $3C80,X  ; [ 16] -> [ 15]
    1462:     LDA $2500,X : STA $2100,X  ; [ 17] -> [ 16]
    1468:     LDA $2900,X : STA $2500,X  ; [ 18] -> [ 17]
    146E:     LDA $2D00,X : STA $2900,X  ; [ 19] -> [ 18]
    1474:     LDA $3100,X : STA $2D00,X  ; [ 20] -> [ 19]
    147A:     LDA $3500,X : STA $3100,X  ; [ 21] -> [ 20]
    1480:     LDA $3900,X : STA $3500,X  ; [ 22] -> [ 21]
    1486:     LDA $3D00,X : STA $3900,X  ; [ 23] -> [ 22]
    148C:     LDA $2180,X : STA $3D00,X  ; [ 24] -> [ 23]
    1492:     LDA $2580,X : STA $2180,X  ; [ 25] -> [ 24]
    1498:     LDA $2980,X : STA $2580,X  ; [ 26] -> [ 25]
    149E:     LDA $2D80,X : STA $2980,X  ; [ 27] -> [ 26]
    14A4:     LDA $3180,X : STA $2D80,X  ; [ 28] -> [ 27]
    14AA:     LDA $3580,X : STA $3180,X  ; [ 29] -> [ 28]
    14B0:     LDA $3980,X : STA $3580,X  ; [ 30] -> [ 29]
    14B6:     LDA $3D80,X : STA $3980,X  ; [ 31] -> [ 30]
    14BC:     LDA $2200,X : STA $3D80,X  ; [ 32] -> [ 31]
    14C2:     LDA $2600,X : STA $2200,X  ; [ 33] -> [ 32]
    14C8:     LDA $2A00,X : STA $2600,X  ; [ 34] -> [ 33]
    14CE:     LDA $2E00,X : STA $2A00,X  ; [ 35] -> [ 34]
    14D4:     LDA $3200,X : STA $2E00,X  ; [ 36] -> [ 35]
    14DA:     LDA $3600,X : STA $3200,X  ; [ 37] -> [ 36]
    14E0:     LDA $3A00,X : STA $3600,X  ; [ 38] -> [ 37]
    14E6:     LDA $3E00,X : STA $3A00,X  ; [ 39] -> [ 38]
    14EC:     LDA $2280,X : STA $3E00,X  ; [ 40] -> [ 39]
    14F2:     LDA $2680,X : STA $2280,X  ; [ 41] -> [ 40]
    14F8:     LDA $2A80,X : STA $2680,X  ; [ 42] -> [ 41]
    14FE:     LDA $2E80,X : STA $2A80,X  ; [ 43] -> [ 42]
    1504:     LDA $3280,X : STA $2E80,X  ; [ 44] -> [ 43]
    150A:     LDA $3680,X : STA $3280,X  ; [ 45] -> [ 44]
    1510:     LDA $3A80,X : STA $3680,X  ; [ 46] -> [ 45]
    1516:     LDA $3E80,X : STA $3A80,X  ; [ 47] -> [ 46]
    151C:     LDA $2300,X : STA $3E80,X  ; [ 48] -> [ 47]
    1522:     LDA $2700,X : STA $2300,X  ; [ 49] -> [ 48]
    1528:     LDA $2B00,X : STA $2700,X  ; [ 50] -> [ 49]
    152E:     LDA $2F00,X : STA $2B00,X  ; [ 51] -> [ 50]
    1534:     LDA $3300,X : STA $2F00,X  ; [ 52] -> [ 51]
    153A:     LDA $3700,X : STA $3300,X  ; [ 53] -> [ 52]
    1540:     LDA $3B00,X : STA $3700,X  ; [ 54] -> [ 53]
    1546:     LDA $3F00,X : STA $3B00,X  ; [ 55] -> [ 54]
    154C:     LDA $2380,X : STA $3F00,X  ; [ 56] -> [ 55]
    1552:     LDA $2780,X : STA $2380,X  ; [ 57] -> [ 56]
    1558:     LDA $2B80,X : STA $2780,X  ; [ 58] -> [ 57]
    155E:     LDA $2F80,X : STA $2B80,X  ; [ 59] -> [ 58]
    1564:     LDA $3380,X : STA $2F80,X  ; [ 60] -> [ 59]
    156A:     LDA $3780,X : STA $3380,X  ; [ 61] -> [ 60]
    1570:     LDA $3B80,X : STA $3780,X  ; [ 62] -> [ 61]
    1576:     LDA $3F80,X : STA $3B80,X  ; [ 63] -> [ 62]
    157C:     LDA $2028,X : STA $3F80,X  ; [ 64] -> [ 63]
    1582:     LDA $2428,X : STA $2028,X  ; [ 65] -> [ 64]
    1588:     LDA $2828,X : STA $2428,X  ; [ 66] -> [ 65]
    158E:     LDA $2C28,X : STA $2828,X  ; [ 67] -> [ 66]
    1594:     LDA $3028,X : STA $2C28,X  ; [ 68] -> [ 67]
    159A:     LDA $3428,X : STA $3028,X  ; [ 69] -> [ 68]
    15A0:     LDA $3828,X : STA $3428,X  ; [ 70] -> [ 69]
    15A6:     LDA $3C28,X : STA $3828,X  ; [ 71] -> [ 70]
    15AC:     LDA $20A8,X : STA $3C28,X  ; [ 72] -> [ 71]
    15B2:     LDA $24A8,X : STA $20A8,X  ; [ 73] -> [ 72]
    15B8:     LDA $28A8,X : STA $24A8,X  ; [ 74] -> [ 73]
    15BE:     LDA $2CA8,X : STA $28A8,X  ; [ 75] -> [ 74]
    15C4:     LDA $30A8,X : STA $2CA8,X  ; [ 76] -> [ 75]
    15CA:     LDA $34A8,X : STA $30A8,X  ; [ 77] -> [ 76]
    15D0:     LDA $38A8,X : STA $34A8,X  ; [ 78] -> [ 77]
    15D6:     LDA $3CA8,X : STA $38A8,X  ; [ 79] -> [ 78]
    15DC:     LDA $2128,X : STA $3CA8,X  ; [ 80] -> [ 79]
    15E2:     LDA $2528,X : STA $2128,X  ; [ 81] -> [ 80]
    15E8:     LDA $2928,X : STA $2528,X  ; [ 82] -> [ 81]
    15EE:     LDA $2D28,X : STA $2928,X  ; [ 83] -> [ 82]
    15F4:     LDA $3128,X : STA $2D28,X  ; [ 84] -> [ 83]
    15FA:     LDA $3528,X : STA $3128,X  ; [ 85] -> [ 84]
    1600:     LDA $3928,X : STA $3528,X  ; [ 86] -> [ 85]
    1606:     LDA $3D28,X : STA $3928,X  ; [ 87] -> [ 86]
    160C:     LDA $21A8,X : STA $3D28,X  ; [ 88] -> [ 87]
    1612:     LDA $25A8,X : STA $21A8,X  ; [ 89] -> [ 88]
    1618:     LDA $29A8,X : STA $25A8,X  ; [ 90] -> [ 89]
    161E:     LDA $2DA8,X : STA $29A8,X  ; [ 91] -> [ 90]
    1624:     LDA $31A8,X : STA $2DA8,X  ; [ 92] -> [ 91]
    162A:     LDA $35A8,X : STA $31A8,X  ; [ 93] -> [ 92]
    1630:     LDA $39A8,X : STA $35A8,X  ; [ 94] -> [ 93]
    1636:     LDA $3DA8,X : STA $39A8,X  ; [ 95] -> [ 94]
    163C:     LDA $2228,X : STA $3DA8,X  ; [ 96] -> [ 95]
    1642:     LDA $2628,X : STA $2228,X  ; [ 97] -> [ 96]
    1648:     LDA $2A28,X : STA $2628,X  ; [ 98] -> [ 97]
    164E:     LDA $2E28,X : STA $2A28,X  ; [ 99] -> [ 98]
    1654:     LDA $3228,X : STA $2E28,X  ; [100] -> [ 99]
    165A:     LDA $3628,X : STA $3228,X  ; [101] -> [100]
    1660:     LDA $3A28,X : STA $3628,X  ; [102] -> [101]
    1666:     LDA $3E28,X : STA $3A28,X  ; [103] -> [102]
    166C:     LDA $22A8,X : STA $3E28,X  ; [104] -> [103]
    1672:     LDA $26A8,X : STA $22A8,X  ; [105] -> [104]
    1678:     LDA $2AA8,X : STA $26A8,X  ; [106] -> [105]
    167E:     LDA $2EA8,X : STA $2AA8,X  ; [107] -> [106]
    1684:     LDA $32A8,X : STA $2EA8,X  ; [108] -> [107]
    168A:     LDA $36A8,X : STA $32A8,X  ; [109] -> [108]
    1690:     LDA $3AA8,X : STA $36A8,X  ; [110] -> [109]
    1696:     LDA $3EA8,X : STA $3AA8,X  ; [111] -> [110]
    169C:     LDA $2328,X : STA $3EA8,X  ; [112] -> [111]
    16A2:     LDA $2728,X : STA $2328,X  ; [113] -> [112]
    16A8:     LDA $2B28,X : STA $2728,X  ; [114] -> [113]
    16AE:     LDA $2F28,X : STA $2B28,X  ; [115] -> [114]
    16B4:     LDA $3328,X : STA $2F28,X  ; [116] -> [115]
    16BA:     LDA $3728,X : STA $3328,X  ; [117] -> [116]
    16C0:     LDA $3B28,X : STA $3728,X  ; [118] -> [117]
    16C6:     LDA $3F28,X : STA $3B28,X  ; [119] -> [118]
    16CC:     LDA $23A8,X : STA $3F28,X  ; [120] -> [119]
    16D2:     LDA $27A8,X : STA $23A8,X  ; [121] -> [120]
    16D8:     LDA $2BA8,X : STA $27A8,X  ; [122] -> [121]
    16DE:     LDA $2FA8,X : STA $2BA8,X  ; [123] -> [122]
    16E4:     LDA $33A8,X : STA $2FA8,X  ; [124] -> [123]
    16EA:     LDA $37A8,X : STA $33A8,X  ; [125] -> [124]
    16F0:     LDA $3BA8,X : STA $37A8,X  ; [126] -> [125]
    16F6:     LDA $3FA8,X : STA $3BA8,X  ; [127] -> [126]
    16FC:     LDA $2050,X : STA $3FA8,X  ; [128] -> [127]
    1702:     LDA $2450,X : STA $2050,X  ; [129] -> [128]
    1708:     LDA $2850,X : STA $2450,X  ; [130] -> [129]
    170E:     LDA $2C50,X : STA $2850,X  ; [131] -> [130]
    1714:     LDA $3050,X : STA $2C50,X  ; [132] -> [131]
    171A:     LDA $3450,X : STA $3050,X  ; [133] -> [132]
    1720:     LDA $3850,X : STA $3450,X  ; [134] -> [133]
    1726:     LDA $3C50,X : STA $3850,X  ; [135] -> [134]
    172C:     LDA $20D0,X : STA $3C50,X  ; [136] -> [135]
    1732:     LDA $24D0,X : STA $20D0,X  ; [137] -> [136]
    1738:     LDA $28D0,X : STA $24D0,X  ; [138] -> [137]
    173E:     LDA $2CD0,X : STA $28D0,X  ; [139] -> [138]
    1744:     LDA $30D0,X : STA $2CD0,X  ; [140] -> [139]
    174A:     LDA $34D0,X : STA $30D0,X  ; [141] -> [140]
    1750:     LDA $38D0,X : STA $34D0,X  ; [142] -> [141]
    1756:     LDA $3CD0,X : STA $38D0,X  ; [143] -> [142]
    175C:     LDA $2150,X : STA $3CD0,X  ; [144] -> [143]
    1762:     LDA $2550,X : STA $2150,X  ; [145] -> [144]
    1768:     LDA $2950,X : STA $2550,X  ; [146] -> [145]
    176E:     LDA $2D50,X : STA $2950,X  ; [147] -> [146]
    1774:     LDA $3150,X : STA $2D50,X  ; [148] -> [147]
    177A:     LDA $3550,X : STA $3150,X  ; [149] -> [148]
    1780:     LDA $3950,X : STA $3550,X  ; [150] -> [149]
    1786:     LDA $3D50,X : STA $3950,X  ; [151] -> [150]
    178C:     LDA $21D0,X : STA $3D50,X  ; [152] -> [151]
    1792:     LDA $25D0,X : STA $21D0,X  ; [153] -> [152]
    1798:     LDA $29D0,X : STA $25D0,X  ; [154] -> [153]
    179E:     LDA $2DD0,X : STA $29D0,X  ; [155] -> [154]
    17A4:     LDA $31D0,X : STA $2DD0,X  ; [156] -> [155]
    17AA:     LDA $35D0,X : STA $31D0,X  ; [157] -> [156]
    17B0:     LDA $39D0,X : STA $35D0,X  ; [158] -> [157]
    17B6:     LDA $3DD0,X : STA $39D0,X  ; [159] -> [158]
    17BC:     LDA $2250,X : STA $3DD0,X  ; [160] -> [159]
    17C2:     LDA $2650,X : STA $2250,X  ; [161] -> [160]
    17C8:     LDA $2A50,X : STA $2650,X  ; [162] -> [161]
    17CE:     LDA $2E50,X : STA $2A50,X  ; [163] -> [162]
    17D4:     LDA $3250,X : STA $2E50,X  ; [164] -> [163]
    17DA:     LDA $3650,X : STA $3250,X  ; [165] -> [164]
    17E0:     LDA $3A50,X : STA $3650,X  ; [166] -> [165]
    17E6:     LDA $3E50,X : STA $3A50,X  ; [167] -> [166]
    17EC:     LDA $22D0,X : STA $3E50,X  ; [168] -> [167]
    17F2:     LDA $26D0,X : STA $22D0,X  ; [169] -> [168]
    17F8:     LDA $2AD0,X : STA $26D0,X  ; [170] -> [169]
    17FE:     LDA $2ED0,X : STA $2AD0,X  ; [171] -> [170]
    1804:     LDA $32D0,X : STA $2ED0,X  ; [172] -> [171]
    180A:     LDA $36D0,X : STA $32D0,X  ; [173] -> [172]
    1810:     LDA $3AD0,X : STA $36D0,X  ; [174] -> [173]
    1816:     LDA $3ED0,X : STA $3AD0,X  ; [175] -> [174]
    181C:     LDA $2350,X : STA $3ED0,X  ; [176] -> [175]
    1822:     LDA $2750,X : STA $2350,X  ; [177] -> [176]
    1828:     LDA $2B50,X : STA $2750,X  ; [178] -> [177]
    182E:     LDA $2F50,X : STA $2B50,X  ; [179] -> [178]
    1834:     LDA $3350,X : STA $2F50,X  ; [180] -> [179]
    183A:     LDA $3750,X : STA $3350,X  ; [181] -> [180]
    1840:     LDA $3B50,X : STA $3750,X  ; [182] -> [181]
    1846:     LDA $3F50,X : STA $3B50,X  ; [183] -> [182]
    184C:     LDA $23D0,X : STA $3F50,X  ; [184] -> [183]
    1852:     LDA $27D0,X : STA $23D0,X  ; [185] -> [184]
    1858:     LDA $2BD0,X : STA $27D0,X  ; [186] -> [185]
    185E:     LDA $2FD0,X : STA $2BD0,X  ; [187] -> [186]
    1864:     LDA $33D0,X : STA $2FD0,X  ; [188] -> [187]
    186A:     LDA $37D0,X : STA $33D0,X  ; [189] -> [188]
    1870:     LDA $3BD0,X : STA $37D0,X  ; [190] -> [189]
    1876:     LDA $3FD0,X : STA $3BD0,X  ; [191] -> [190]
    187C:     LDA #00     : STA $3FD0,X  ; zero  -> [191]
    1881:     DEX
    1882:     BMI .2   ; x < 0 ?
    1884:     JMP .1
    1887:  .2 RTS

The bulk of the ScrollHgrUpPixel() was generated with this JavaScript program scroll_hgr_up_pixel.html:

    var hgr = [];
    for( var y = 0; y < 193; ++y ) // Intentional 1 scanline too many!
        hgr[ y ] = 0x2000 + ((y/64)|0)*0x28 + ((y%8)|0)*0x400 + ((y/8)&7)*0x80;

    for( var y = 0; y < 192; ++y )
        console.log( "[" + y + "]: " + hgr[y].toString(16).toUpperCase() );

    function byte2hex$( byte )
    {
        return ("0" + byte.toString(16)).toUpperCase().substr(-2)
    }

    var address = 0x1402, out = "";
    for( var y = 0; y < 192/8; ++y )
        for( var x = 0; x < 8; ++x )
        {
            var row = y*8 + x; // Assumes hgr[] has a dummy 193rd scanline!
            var src = hgr[ row + 1 ];
            var dst = hgr[ row + 0 ];
            var mem = "BD "
                    + byte2hex$( (src >> 0) & 0xFF ) + " "
                    + byte2hex$( (src >> 8) & 0xFF ) + " "
                    + "9D "
                    + byte2hex$( (dst >> 0) & 0xFF ) + " "
                    + byte2hex$( (dst >> 8) & 0xFF ) + " ";
            var txt = "  "
                 + "  LDA $" + src.toString(16).toUpperCase() + ",Y "
                 + ": STA $" + dst.toString(16).toUpperCase() + ",Y "
                 + " ; ["   + ("  " + (row+1)).substr(-3) 
                 + "] -> [" + ("  " + (row  )).substr(-3) + "]"  + "\n";
            if (row != 191)
                out += address.toString(16).toUpperCase() + ":" + mem + txt;
            address += 6; // 6 bytes per line
        }
    console.log( out );

And who said JavaScript was a useless language? :-)

That's all folks! Now go write some cool font blitter code.

References

Misc. Utilities and Files

Convert font image to C array
Convert C array to binary font
Raw Binary Font Within AppleWin's debugger: bload font.bin,6000

TODO:

Screenshots!
Cleanup all assembly for consistent indentation and alignment
Binary code for 300.bin and 1000.bin so you can load it directly into the emulator
Disk image: HGR_FONT.DSK (In progress)
Count cycles of old and new DrawCharCol Address Calculation
Alternative fonts
Re-engineer Codepage 437 Font to 7x8 cells:
Double Hi-Res
PDF of this document (As a work-around use Chrome and Print to PDF)

README.md

Table of Contents

Introduction

The Problem

The Solution:

Hard-Coded: A

Functions we want & will write

Quirks of the Apple HGR screen

Non-Linear Memory

No FONT data in ROM

HGR bytes are reversed

Half-pixel shift

Font Data

Raw Font Data

Image to Font Data (JavaScript)

DrawChar()

Font -> Screen Memory Trace

DrawChar() version 1

X Cursor Position

CursorCol( col )

Introduction to Optimization

DrawChar() version 2

DrawChar() version 3

Character Inspector

Character Inspector version 2

Character Inspector version 3

Y Cursor Position

Natural Params SetCursorColRow()

DrawString()

Recap

Copy text screen to HGR

Exercise 1: ScrollHgrUpPixel()

Exercise 2: ScrollHgrUpLine()

Conclusion

Solution 1: ScrollHgrUpLine()

Solution 2: ScrollHgrUpPixel()

References

Misc. Utilities and Files

TODO: