More Tube ULA fun

[samskivington-poq]

Hello.

I know you've said you're not completely familiar with the Tube in the previous issue I raised, but Is there a reason why jsBeeb's implementation forces me to write code that introduces a delay between repeatedly polling the ULA to know when the R1 FIFO has space and then writing to it, and doing the same again, whereas on a real machine that is not required? It's slowing my parasite code down unnecessarily in jsBeeb, for something that needs to run as fast as possible.

The result of running the code in the .else block is that spurious values are received on the host side of R1 that are nothing to do with the actual values written on the parasite side, at a rate of roughly 500ms. Hoping the 500ms points you to the cause.

In this instance, the parasite OS and Acorn host code have both been completely eliminated and this is raw, tight loop communication between both sides. This code snippet is run as part of a NMI handler on the parasite side.

	;	write the command queue
_loopCommandQueue
	lda	ZP_COMMAND_QUEUE_INPUT_PTR
	cmp	ZP_COMMAND_QUEUE_OUTPUT_PTR
	bne	_doCommandQueueByte
	lda	ZP_COMMAND_QUEUE_INPUT_PTR + 1
	cmp	ZP_COMMAND_QUEUE_OUTPUT_PTR + 1
	beq	_done

_doCommandQueueByte
	lda	(ZP_COMMAND_QUEUE_OUTPUT_PTR)

.ifdef JSBEEB_BROKEN_TUBE
	;	RAISE-ISSUE: Tube ULA timing
	;	the code in the .else block waits until there is space in the FIFO
	;	and is the exact same code as in the function TubeParasiteWriteR1
	;	I shouldn't need to introduce a delay like this
	;	I don't on a real machine or in BeebEm
	;	but I do in b-Em and jsBeeb
	;	this is parasite side and interrupts are enabled
	; 	but nothing is even attempting to trigger them at this point
	beq	_send
	cmp	#0
	bcs	_send
_send
	jsr	TubeParasiteWriteR1
.else

_loopWaitUntilR1HasSpace
	bit	TUBE_STATUS_REGISTER_1_PARASITE
	bvc	_loopWaitUntilR1HasSpace
	sta	TUBE_DATA_REGISTER_1_PARASITE

.endif

	inc	ZP_COMMAND_QUEUE_OUTPUT_PTR
	bne	_loopCommandQueue
	inc	ZP_COMMAND_QUEUE_OUTPUT_PTR + 1
	bne	_loopCommandQueue

_done

Thanks
Sam

[Matt Godbolt]

Hi Sam! Honeslty, nothing in the Tube stuff is "on purpose" :) I copy pasted from various emulators until Tube Elite "sort of" worked. I can't imagine what's causing those 500ms pulses BUT I'll try and take a look.

Thanks for taking the time to file this!

[samskivington-poq]

If it helps (could be random) the spurious values in my case were either 3 or 4, whereas I was only sending 1 or 0.