Getting back to the Apple II, it would seem that one of the biggest offenders is the handling of interrupts. Documentation of the past shows that the initial deployment didn't go very well at all. improvements were made, but subsequent OS control of interrupts further complicated matters. Ive recently picked up a copy of Assembly lines, a wonderful reprint, thanks Chris, which i've referred to extensively in this project to translate the spoken Commodore into Applese, interesting point though, there is no mention of IRQ's or NMI's in that book. Probably for the aforementioned reasons. Given the book is a reprint, and IRQ kinda rolled out DOA, well..... anyway, to that end, I have slowly worked through old documentation, notes, more PRODOS crap than i'd ever care to know about and lots of debugging to provide a standard setup to build upon. Basically what I have at home, and use this going forward.
The System is an Apple //e enhanced with an 80 column card providing 128K. The 80 column card and the extra 64k is of no consequence to this program. By default, I've included but commented out 80 column mode in the source code. Im using Merlin 8 version 2.58 and added BASIC.SYSTEM to disk 2 so that BRUN could be employed for quicker development. No emulation was used in this endeavor. Final disks were pushed with ADTpro to a linux box, the source extracted for viewing here with Applecommander.
The program is minimalist by design. Made to do as little as possible to meet the requirements set forth. A connection is made to another computer via crossover cable running Minicom, Gtkterm or an equivalent. The program starts with a blank screen. Each key pressed on the Apple appears on both the apple and the distant end computer. With the Apple, starting at the top left and advancing with each character. When the distant end presses a key in will appear on the bottom row of the Apple starting at the bottom left and advancing to the right until the 15th character at which point it will start updating the bottom left again. A quasi visible ring buffer. If you don't see a feature in this program its because it was not necessary. SSC location detection, enhanced apple detection, split screen vector graphics support.... beyond the scope of this exercise. WORKING SEND RECEIVE SSC ASSEMBLY LANGUAGE PROGRAM WITH PRODOS IRQ SUPPORT. THATS IT.
The actual buffer that is offloaded from the SSC ACIA is in the zero page at $09. The offloading portion is done during the IRQ. That IRQ signal gets kicked off because of the SSC card receiving a byte, it doesn't mean the computer knows this program owns it. Because the status register indicates such activity on the SSC itself, it can be tested in the IRQ code for ownership, acted upon or dismissed. Nothing special about that, run of the mill ACIA stuff. The fact that the IRQ code drops it in the BUFFER allows for the READBUFFER routine in the main program to test that zero page byte for new data and act upon it, keeping the heavy lifting out of the IRQ.
So START clears the screen and initialized the SSC. It also conducts some PRODOS voodoo to onboard the IRQ with its own identifier into one of 4 IRQ placeholders into PRODOS. Ill admit i'm not a fan, this is not for the purist at heart. But if you want mouse and disk drive support with your IRQ, well then....
MAIN JSR'S to READBUFFER to see if its not #$00 to act upon it. If its #$00, wait for a keyboard event by jumping back to MAIN and getting stuck in this loop forever.
There are only 2 ways out of this loop. Press a key on the keyboard is the first. If you press it it will appear on the screen, have the 7th bit removed from the byte and then a JSR to the SEND routine where it will check the STATUSREGISTER until it is ready to accept the byte, once ready will send it off to the ACIA touch STROBE to clear that KEYSTROKE and loop back to MAIN
The other way is for the IRQ to have left something in the BUFFER, that would make it no longer #$00. If that is the case READBUFFER will pull that value out of the BUFFER and print it to $0750, AKA BOTTOM LEFT, offset by the COUNTER. Once that value is printer the BUFFER is set back to #$00 and the COUNTER is incremented. Should the counter reach #$15, the COUNTER will be set back to 0, completing the loop. In either case , reset counter or not, the program will loop back to main once the receive byte is printed.
I could further reduce the size of this program at the expense of readability and decide that it violated the goal of this exercise so I chose not to. The parameters of which have already been stated. So if anyone wants to provide a means for a new feature or make this smaller I have no interest whatsoever. Should you find an outright bug of course, thats a different matter. Furthermore, If anyone would like to provide some context with regard to the PRODOS IRQ onboarding voodoo as well as the bottom area labeled IRQLIST, that would be most welcomed.
So, did I need the IRQ. Probably not. I probably could have polled the STATUSREG if the timing was acceptable and program flow allowed up to a given speed. Mileage varies. At some point in a far extreme a case is to be made for NMI's which is a whole other business. Another case is looking at the benefit of constant polling for data versus the SSC informing of an event and acting upon it external from the program.
DSK file is included below in google drive link as well as highlighted source.
Thank you and enjoy,
Jordan
Click here for PRODOS DSK file including Merlin source and executable
CONTENTS of SSCTEST.S
******************************
* **** COMMODORE 64 BASIC V2 **** *
* 64K RAM SYSTEM 38911 BASIC BYTES FREE *
* READY. *
* LOAD"SSCTEST",8,1 *
******************************
* Super Duper Simple SSC IRQ TEST *
* Written and compiled natively on Merlin 8 *
* No Emulators used...... TASM style formatting *
* Requires Apple IIe, most likely enhanced with *
* working interrupts and this code must also be *
* run from PRODOS to work... *
* execute with ---> BRUN SSCTEST *
* Do not execute from the monitor *
* Change SLOT value as necessary , set for 2 *
* Default values are for N-8-1 9600 baud conn. *
* Transmit appears on top, RX buffer is a 15 byte *
* rotating buffer within text memory on bottom of *
* the screen. *
* Free for all to use - Jordan Rubin 2017 *
* More at www.youtube.com/c/
******************************
ORG $2000 ; HIRES area unused and safe
COUNTER = $08 ; Rotating counter for RXbyte
BUFFER = $09 ; Where IRQ will store RXbyte
SLOT = $20 ; This is slot 2 where SSC is
DATAREG = $C088+SLOT
STATUSREG = $C089+SLOT
COMMANDREG = $C08A+SLOT
CONTROLREG = $C08B+SLOT
HOME = $FC58 ; CLS & Cursor to top POS
COUT = $FDED ; Like C64 $FFD2
KYBD = $C000 ; Keyboard map point
STROBE = $C010 ; Keyboard strobe map point
MLI = $BF00 ; PRODOS MLI entry point
************** Start of program
START
; JSR $C300 ; Uncomment for 80 COL
JSR HOME ; Clear screen
JSR INIT ; Run init for SSC Card
************** Main loop of program
MAIN
JSR READBUFFER ; Read from where IRQ stores incoming
LDA KYBD ; Poll keyboard pointer
CMP #$80 ; Check keyboard for activity
BCC MAIN ; Keep reading buffer till TXkey event
JSR COUT ; print TX locally on top of screen
AND #$7F ; shut off 7th bit for ESC issue
JSR SEND ; Send the byte to the ACIA outbound
STA STROBE ; Remove event by accessing
JMP MAIN
************** CHECKS FOR CONTENT OF BUFFER AND PRINTS IT
READBUFFER
LDA BUFFER ; Pull Buffer content
CMP #$00 ; Change since last initialization?
BNE :PRINT ; If so, go to PRINT routine
RTS
:PRINT
LDY COUNTER ; Load COUNTER into Y
INY ; Increment it
STA $0750,Y ; Store BUFFER content to our screen bottom
LDA #$00 ; .......and offset by counter
STA BUFFER ; Reset Buffer to Null
TYA ; copy y to a
CMP #$15 ; Are we on the Hex15 rotation
BEQ RESETCOUNTER ; if so Reset the counter to 0
STY COUNTER ; if not store the counter
RTS
RESETCOUNTER
LDA #$00
STA COUNTER ; COUNTER is 0 again.....
RTS
************** INIT Sets up the SSC and IRQ stuff
INIT
LDA #$00
STA STATUSREG
STA COUNTER ; set counter to 0
JSR MLI ; PRODOS ENTRY
DB $40 ; MORE PRODOS
DW IRQLIST ; EVEN MORE PRODOS
********* LOAD VALUES INTO CONTROLREG
* BIT [7] 0 - ONE STOP BIT / 1 TWO STOP BITS
* BIT [6-5] LENGTH - 8 (00) / 7 (01) / 6 (10) / 5 (11)
* BIT [4] CLK - EXTERNAL (0) / INTERNAL (1)
* BIT [3-0] BAUD - 9600 (1110) / 19.2K (1111)
* 4800 (1100) / 2400 (1010)
******* USING N81 INTERNAL AT 9600 BAUD
LDA #011110
STA CONTROLREG
********* LOAD VALUES INTO COMMANDREG
* BIT [7-5] PARITY- (000)DIS (001)O (011)E (101)M (111)S
* BIT [4] ECHO - (0) NORM / (1) ON
* BIT [3-2] TXINT- (00)DIS:RTSH (01)EN:RTSL
* (10)DIS:RTSL (11)DIS:SBRK
* BIT [1] RXINT - 0 ENABLED / 1 DISBLED
* BIT [0] DTR - 0 DISABLE RXTX,DTR HIGH / 1 ENABLE RXTX DTR LOW
********* USING RX INTERRUPTS ENABLED DTR LOW ENABLE RXTX
LDA #001001
STA COMMANDREG
LDA DATAREG ; Pull from DATAREG on init just because
RTS
************** IRQ This is the IRQ code
IRQ
CLD ; Required by PRODOS
LDA STATUSREG ; Was it us who caused this interrupt
BMI :ACTION ; If so, Action!!!
SEC ; If not inform, not us....
RTS
:ACTION
LDA DATAREG ; Load byte directly from dataregister
STA BUFFER ; and drop it in the ZP BUFFER
CLC ; clear interrupt
RTS
************** SEND Check status and send a byte to ACIA
SEND
TAY
:CHECK
LDA STATUSREG ; Load status register
AND #$10 ; mask for ready bit
BEQ :CHECK ; not ready? keep checking
STY DATAREG ; ready?? store byte
RTS
************** IRQLIST For PRODOS, overhead requirements
IRQLIST
DB 2,0
DW IRQ
No comments:
Post a Comment