Friday, February 02, 2024
Making it to the Orange Site
A previous post made it to Lobster only to later show up at the Orange Site. How about that?
Making it to the Orange Site
A previous post made it to Lobster only to later show up at the Orange Site. How about that?
Making it to the Orange Site
A previous post made it to Lobster only to later show up at the Orange Site. How about that?
It should be obvious by now what day it is, but just in case
I want to ensure everybody that yes, I intentionally made the previous three posts. And while they are similar, they are not in fact, identical. You know, much like Ground Hog Day.
Monday, February 05, 2024
The difficulties in supporting “write-only memory” in assembly
When I last wrote about this, I had one outstanding problem with static analysis of read-only/write-only memory, and that was with hardware that could be input or output only. It was only after I wrote that that I realized the solution—it's the same as a hardware register having different semantics on read vs. write—just define two labels with the semantics I want. So for the MC6821, I could have:
org $FF00 PIA0.A rmb/r 1 ; read only org $FF00 PIA0.Adir rmb/w 1 ; write only, to set the direction of each IO pin PIA0.Acontrol rmb 1 ; control for port A
So that was a non-issue. It was then I started looking over some existing code I had to see how it might look. I didn't want to just jump into an implementation without some forethought, and I quickly found some issues with the idea by looking at my maze generation program. The code in question initializes the required video mode (in this case 64×64 with four colors). Step one involves writing a particular value to the MC6821:
lda #G1C.PIA ; 64x64x4 sta PIA1.B
So far so good. I can mark PIA1.B as write-only
(technically, it also has some input pins so I really can't,
but in theory I could).
Now, the next bit requires some explaining. There's another 3-bit value that needs to be configured on the MC6883, but it's not as simple as writing the 3-bit value to a hardware register—each bit requires writing to a different address, and worse—it's a different address if the bit is 0 or 1. So that's six different addresses required. It's not horrible though—the addresses are sequential:
| bit | 0/1 | address |
|---|---|---|
V0 | 0 | $FFC0 |
V0 | 1 | $FFC1 |
V1 | 0 | $FFC2 |
V1 | 1 | $FFC3 |
V2 | 0 | $FFC4 |
V2 | 1 | $FFC5 |
Yeah,
to a software programmer,
hardware can be weird.
To set bit 0 to 0,
you do a write
(and it does not matter what the value is)
to address $FFC0.
If bit 0 is 1,
then it's a write to $FFC1.
So with that in mind,
I have:
sta SAM.V0 + (G1C.V & 1<<0 <> 0) sta SAM.V1 + (G1C.V & 1<<1 <> 0) sta SAM.V2 + (G1C.V & 1<<2 <> 0)
OOh. Yeah.
I wrote it this way so I wouldn't have to look up the appropriate value and write the more opaque (to me):
sta $FFC1 sta SFFC2 sta $FFC4
The expression (G1C.V & 1<<n <> 0) checks bit n to see if it's set or not,
and returns 0 (for not set) or 1 (for set).
This is then added to the base address for bit n,
and it all works out fine. I can change the code for,
say, the 128×192 four color mode by using a different constant:
lda #G6C.PIA sta PIA1.B sta SAM.V0 + (G6C.V & 1<<0 <> 0) sta SAM.V1 + (G6C.V & 1<<1 <> 0) sta SAM.V2 + (G6C.V & 1<<2 <> 0)
But I digress.
This is a bit harder to support. The address being written is part of an expression, and only the label (defining the address) would have the read/write attribute associated with it. At least, that was my intent. I suppose I could track the read/write attribute by address, which would solve this particular segment of code.
And the final bit of code to set the address of the video screen (or frame buffer):
ldx #SAM.F6 ; point to frame buffer address bits lda ECB.grpram ; get MSB of frame buffer mapframebuf clrb lsla rolb sta b,x ; next bit of address leax -2,x cmpx #SAM.F0 bhs mapframebuf
Like the VDG Address Mode bits,
the bits for the VDG Address Offset have unique addresses,
and because the VDG Address Offset has seven bits,
the address is aligned to a 512 byte boundary.
Here,
the code loads the X register with the address of the upper end of the VDG Address Offset,
and the seven top most bits of the video address is sent,
one at a time,
to the B register,
which is used as an offset to the X register to set the appropriate address for the appropriate bit.
So now I would have to track the read/write attributes via the index registers as well.
That is not so easy.
I mean, here, it could work, as the code is all in one place, but what if instead it was:
ldx #SAM.F6 lda ECB.grpram jsr mapframebuf
Or an even worse example:
costmessage fcc/r "A constant message" ; read only text buffer rmb 18 ldx #constmessage ldy #buffer lda #18 jsr memcpy
The subroutine memcpy might not even be in the same source unit,
so how would the read/write attribute even be checked?
This is for static analysis,
not runtime.
I have one variation on the maze generation program that generates multiple mazes at the same time, on the same screen (it's fun to watch) and as such, I have the data required for each “maze generator” stored in a structure:
explorec equ 0 ; read-only backtrackc equ 1 ; read-only xmin equ 2 ; read-only ymin equ 3 ; read-only xstart equ 4 ; read-only ystart equ 5 ; read-only xmax equ 6 ; read-only ymax equ 7 ; read-only xpos equ 8 ; read-write ypos equ 9 ; read-write color equ 10 ; read-write func equ 11 ; read-write
This is from the source code, but I've commented each “field” as being “read-only” or “read-write.” That's another aspect of this that I didn't consider:
lda explorec,x ; this is okay sta explorec,x ; this is NOT okay
Not only would I have to track read/write attributes for addresses, but for field accesses to a structure as well. I'm not saying this is impossible, it's just going to take way more thought than I thought. I don't think I'll have this feature done any time soon …
Tuesday, February 06, 2024
So you want to amplify my SEO
- From
- Krystal XXXXXXX <XXXXXXXXXXXXXXXXXXXX@gmail.com>
- To
- sean@conman.org
- Subject
- Amplify Your SEO with Strategic Link Inserts
- Date
- Wed, 7 Feb 2024 01:16:35 +0300
Hi Content Team,
It’s Krystal XXXXXXX here from Next Publisher, your next potential partner in digital storytelling. We're thrilled about the idea of featuring both guest posts and link insertions on your dynamic website.
We would like to know your fee structure for hosting guest posts and link insertions. Our team aims to create compelling content that is tailored to your site’s audience and enhances your overall content strategy.
A quick note: this initial email is only for starting our dialogue. All more detailed communications, including agreements and transactions, will be carried out through our official Next Publisher email.
We admire the quality of your platform and are excited to explore how we can work together for mutual benefit.
Looking forward to your prompt reply.
Warm wishes,
Krystal XXXXXXX
Hello Krystal.
Since you neglected to include a link
(in an email sent as HTML no less!)
it's hard for me judge the value Next Publisher will provide for my site,
so I'm going to have to adjust my prices accordingly.
My fee for both guest posts and link insertions is $10,000 (US) per.
So if a guest post also includes a link insertion,
that would be a total of $20,000 (US).
If I'm going to be whoring selling out what Google Page Rank I have,
it's going to cost.
I look forward to hearing back from you.
Sean.
Okay! I'll answer your question, LinkedIn. Also, Orange Site! Orange Site! Orange Site!
Today's “you're one of the few experts invited to add this collaborative article” from LinkedIn is “How do you become a senior application developer?” My answer? Stay in a programming job for three years. Boom! You're a senior application developer. I know, I know, that's a big ask these days when everybody is jumping ship every two years. But hey, if you want to be a “senior application developer,” you got to make sacrifices.
Oh, and the title to today's post? I found out that LinkedIn really liked when I mentioned the Orange Site in the title of my post. Almost two orders of magnitude more. So I'm doing a test to see if I can game the system there.
Wednesday, February 07, 2024
Instead of “write-only memory” assembly support, how about floating point support?
You might think it odd to add support for floating point constants for an 8-bit CPU, but Motorola did development on the MC6839 floating point firmware for the MC6809, an 8K ROM of thread-safe, position-independent 6809 code that implements the IEEE Standard for Floating-Point Arithmetic. It was never formally released by Motorola as a product, but from what I understand, it was released later under a public domain license. At the very least, it's quite easy to MC6839 find both the ROM image and the source code on the Intarwebs. So that's one reason.
Another reason is that the Color Computer BASIC supports floating point operations, and while not IEEE-754, as it was written before the IEEE-754 standard become a standard, it still floating point, and there are only minor differences between it and the current standard, namely the exponent bias, number of fractional bits supported, and where the sign bit is stored. It really comes down to some bit manipulations to massage a standard float into the Color Computer BASIC float format. There are some differences, but the differences are small (literally, on the scale of 0.0000003) probably due to parsing differences, and small enough that it should be “good enough.” Especially since the Color Computer BASIC float format doesn't support infinity or NaN.
So if you specify a backend other than the rsdos backend,
you get IEEE-754,
and if you do specify rsdos as a backend,
you get the Color Computer BASIC float format.
And yes,
I added support for floating point expressions
(but not for the test backend—I'm still thinking on how to support it),
and one interesting feature I added is the factorial operator “!”.
Factorials are used in Talor series,
which the Color Computer BASIC uses for the sin() function,
so I can literally write:
; Oh! '**' is exponentiation by the way! taylor_series .float -((2 * 3.14159265358979323846) ** 11) / 11! .float ((2 * 3.14159265358979323846) ** 9) / 9! .float -((2 * 3.14159265358979323846) ** 7) / 7! .float ((2 * 3.14159265358979323846) ** 5) / 5! .float -((2 * 3.14159265358979323846) ** 3) / 3! .float 2 * 3.14159265358979323846
and have it generate the correct values. I personally don't know of any language that has a factorial operator (maybe APL? I don't know).
I think I'm having more fun writing the assembler than I am writing assembly code.
Sunday, February 11, 2024
An extensible programming language
A few days ago I wrote about adding a factorial operator to my assembler, and I noted that I knew not of any other languages that had such a feature. So imagine my surprise as I'm reading about XL (via Lobsters) and the second example is factorial! Not only that, but that was an example of extending the language itself! The last time I was this excited about software was reading about Synthesis OS, a JIT-based operating system where you could create your own system calls.
How it handles precedence is interesting. In my assembler, I have left or right associativity as an explicit field, whereas in XL, it's encoded in the precedence level itself—even if its left, odd if its right. I'm not sure how I feel about that. On the one hand it feels nice and it's one less field to carry around; on the other, being explicit as I did makes it clear if something is left or right. But on the gripping hand, it sounds like matching precedence on a left and right operator could lead to problems, so I still may have an explicitness problem.
But I digress.
It's a very simple language with only one keyword “is” and a user-definable precedence table. The parser generates a parse tree of only eight types, four leaf nodes (integer, real, text, name (or symbol)) and four non-leaf nodes (prefix, infix, postfix and block). And from there, you get XL.
This is something I definitely want to look into.
Wednesday, February 14, 2024
Notes from an overheard conversation from a car attempting a right turn
“Oh! Now what?”
“Sir, see the lit sign up there? You cannot turn right.”
“But did you not see the car right in front of me turning right?”
“Sir, if a person jumped off a bridge, would you follow?”
“Yes.”
“You must be very smart then.”
“And selective enforcement of the laws leads to distrust of the police.”
“Sir—”
“Oh look! The ‘No Right Turn’ sign is off now! Gotta go! Bye!”
“I don't think it's wise to taunt the poice like that.”
“Down with the Man! Power to the people! Yo!”
Wednesday, February 28, 2024
Converting IEEE-754 floating point to Color BASIC floating point
I'm still playing around with floating point on the 6809—specifically, support for floating point for the Color Computer. The format for floating point for Color BASIC (written by Microsoft) predates the IEEE-754 Floating Point Standard by a few years and thus, isn't quite compatible. It's close, though. It's defined as an 8-bit exponent, biased by 129, a single sign bit (after the exponent) and 31 bits for the mantissa (the leading one assumed). It also does not support ±∞ nor NaN. This differs from the IEEE-754 single precision that uses a single sign bit, an 8-bit exponent biased by 127 and 23 bits for the mantissa (which also assumes a leafing one) and support for infinities and NaN. The IEEE-754 double precision uses a single sign bit, an 11-bit exponent biased by 1023 and 52 bit for the mantissa (leading one assumed) plus support for infinities and NaN.
So the Color BASIC is about halfway between single precision and double precision. This lead me to use IEEE-754 double precision for the Color Computer backend (generating an error for inifinities and NaN) then massaging the resulting double into the proper format. I double checked this by finding some floating point constants in the Color BASIC ROM as shown in the book Color BASIC Unravelled II, (available on the Computer Computer Archives), like this table:
4634 * MODIFIED TAYLOR SERIES SIN COEFFICIENTS 4635 BFC7 05 LBFC7 FCB 6-1 SIX COEFFICIENTS 4636 BFC8 84 E6 1A 2D 1B LBFC8 FCB $84,$E6,$1A,$2D,$1B * -((2*PI)**11)/11! 4637 BFCD 85 28 07 FB F8 LBFCD FCB $86,$28,$07,$FB,$F8 * ((2*PI)**9)/9! 4638 BFD2 87 99 68 89 01 LBFD2 FCB $87,$99,$68,$89,$01 * -((2*PI)**7)/7! 4639 BFD7 87 23 35 DF E1 LBFD7 FCB $87,$23,$35,$DF,$E1 * ((2*PI)**5)/5! 4640 BFDC 86 A5 5D E7 28 LBFDC FCB $86,$A5,$5D,$E7,$28 * -((2*PI)**3)/3! 4641 BFE1 83 49 0F DA A2 LBFE1 FCB $83,$49,$0F,$DA,$A2 * 2*PI
Then using the byte values to populate a variable and printing it inside BASIC (this is the expression -2π3/3!):
X=0 ' CREATE A VARIABLE Y=VARPTR(X) ' GET ITS ADDRESS POKE Y,&H86 ' AND SET ITS VALUE POKE Y+1,&HA5 ' THE HARD WAY POKE Y+2,&H5D POKE Y+3,&HE7 POKE Y+4,&H28 PRINT X ' LET'S SEE WHAT IT IS -41.3417023
Then using that to create a floating point value:
org $1000 .float -41.3417023 end
Checking the resulting bytes that were generated:
| FILE ff.a
1 | org $1000
1000: 86A55DE735 2 | .float -41.3417023
3 | end
And adjusting the floating point constant until I got bytes that matched:
| FILE ff.a
1 | org $1000
1000: 86A55DE728 2 | .float -41.341702110
3 | end
I figure it's “close enough.” The parsing code in the Color BASIC ROM is old and predates the IEEE-754 floating point standard, so a few different digits at the end I think is okay.
As a final check, I wrote the following bit of code to calculate and display -2π3/3!, display the pre-calculated result, as well as display the pre-calculated value of 2π:
include "Coco/basic.i" include "Coco/dp.i" CB.FSUBx equ $B9B9 ; FP0 = X - FP0 ; addresses for CB.FSUB equ $B9BC ; FP0 = FP1 - FP0 ; these routines CB.FADDx equ $B9C2 ; FP0 = X + FP0 ; from CB.FADD equ $B9C5 ; FP0 = FP1 + FP1 ; Color BASIC Unravelled II CB.FMULx equ $BACA ; FP0 = X * FP0 CB.FMUL equ $BAD0 ; FP0 = FP0 * FP1 CB.FDIVx equ $BB8F ; FP0 = X / FP0 CB.FDIV equ $BB91 ; FP0 = FP1 / FP0 CB.FP0fx equ $BC14 ; FP0 = X CB.xfFP0 equ $BC35 ; X = FP0 CB.FP1f0 equ $BC5F ; FP1 = FP0 CB.FP0txt equ $BDD9 ; result in X, NUL terminated org $4000 start ldx #tau ; point to 2*pi jsr CB.FP0fx ; copy to FP0 ldx #tau ; 2PI * 2PI jsr CB.FMULx ldx #tau ; 2PI * 2PI * 2PI jsr CB.FMULx jsr CB.FP1f0 ; copy fp acc to FP1 ldx #fact3 ; point to 3! jsr CB.FP0fx ; copy to FP0 jsr CB.FDIV ; FP0 = FP1 / FP0 neg CB.fp0sgn ; negate result by flippping FP0 sign jsr CB.FP0txt ; generate string bsr display ; display on screen ldx #answer ; point to precalculated result jsr CB.FP0fx ; copy to FP0 jsr CB.FP0txt ; generate string bsr display ; display ldx #tau ; now display 2*pi jsr CB.FP0fx ; just to see how close jsr CB.FP0txt ; it is. bsr display rts display.char jsr [CHROUT] ; display character display lda ,x+ ; get character bne .char ; if not NUL byte, display lda #13 ; go to next line jsr [CHROUT] rts tau .float 6.283185307 fact3 .float 3! answer .float -(6.283185307 ** 3 / 3!) end start
The results were:
-41.3417023 -41.3417023 6.23418531
The calculation results in -41.3417023 and the direct result stored in answer also prints out -41.3417023,
so that matches and it reinforces my approach to this nominally right.
But I think Microsoft had issues with either generating some of the floating point constants for the larger terms, or transcribing the byte values of the larger terms. Take for instance -2π11/11!. The correct answer is -15.0946426, but the bytes in the ROM define the constant -14.3813907, a difference of .7. And it's not like Color BASIC can't calculate that correctly—when I typed in the expression by hand, it was able to come up with -15.0946426.
Or it could be that Walter K. Zydhek, the
author of Color BASIC Unravelled II,
is wrong in his interpretation of the expressions used to generate the values,
or his interpretation of what the values are used for.
I'm not sure who is at fault here.
Update on Friday, March 1st, 2024
I was wrong about the authorship of Color BASIC Unravelled II. It was not Walter K. Zydhek, but some unknown author of Spectral Associates, a company that is no longer in business. All Zydhek did was to transcribe a physical copy of the book (which is no longer available for purchase anywhere) into a PDF and make it available.
