Interactive symbolic bitop calculator

[paraj]

I've converted a small tool I've been using (originally written in macro-ridden C++) to run on this newfangled interwebs-gizmo all the cool kids are talking about, and figured some of you might find it useful even in its current work-in-progress state.

It's a small javascript testbed with some m68k-like functions built-in.
The intended purpose is to test / debug small snippets of assembly code. Working with scrambled C2P's is what prompted me to develop the C++ tool - the pen and paper method got too unwieldy.

Click on the "interactive calculator" link at https://github.com/mras0/symbitops (it uses githubs HTML preview feature). It requires a modern browser (I have no idea how modern. I test with Chrome 58, Firfox 53 and Microsoft Edge 40).

The code window contains some hints to get started.

As an example, consider a routine to convert 32-bit RGB data in little endian format to good old 12-bit RGB. Clear the input area and copy the following:

// Set D0 = symbolic version of our input data
MOVE.L('B7B6B5B4B3B2B1B0 G7G6G5G4G3G2G1G0 R7R6R5R4R3R2R1R0 ????????????????', D0)
// Now run our routine
LSR.W(4, D0)
MOVE.L(D0, D1)
SWAP(D0)
AND.W(0xF0F0, D0)
MOVE.W(D0, D2)
ROL.W(4, D2)
OR.B(D2, D0)
MOVE.B(D0, D1)
// Check that the lower word of D1 is as expected
assert.equal(state[D1].get(16)+'', '........R7R6R5R4 G7G6G5G4B7B6B5B4')

And then press "Run!" (or ctrl+enter), the output should appear below shortly thereafter.

Note that only a subset of 68K instructions are supported.

Oh yeah, and just because I could, there is also an automatic 4-bit C2P code generator. You supply symbolic input and output and it figures out the steps necessary It's a bit slow though (especially in firefox) so give it a minute (literally).

PS: This is my first real web "app", so the code is probably poison to knowledgeable developers.

[ross]

Great!

Link saved

[paraj]

Continuing my consolidation of various small tools into one, I've added a very (very) basic assembler and cycle counter (probably lacking and buggy). I've also tried making the editor slightly less user-hostile.

The following is now possible (larger examples also work):

MOVE.B('a3a2a1a0',D0)

lines = parse_lines(`
	lsl.w	#6, d0
	lsr.b	#6, d0
`);

to_code(lines)()
print_cost(lines)

Which gives the following output:

	MOVE.B	#magic, D0      	; D0 = ???????????????? ???????????????? ???????????????? ........a3a2a1a0
	LSL.W	#$6, D0         	; D0 = ???????????????? ???????????????? ????........a3a2 a1a0............
	LSR.B	#$6, D0         	; D0 = ???????????????? ???????????????? ????........a3a2 ............a1a0
	LSL.W	#6, d0            	18
	LSR.B	#6, d0            	18
	Total (estimated) cost: 36

The "assembler" supports way more instructions than the symbolic evaluater (in particular the evaluater doesn't handle anything involving condition codes), but there still might be missing useful instructions.

As always everything is very much work in progress, so don't expect miracles. If y'all have any reasonable suggestions/requests I'll be happy to look at them.

To be clear, say we want to count (68000) cycles in the 5 instruction loop from https://amycoders.org/opt/innerloops.html. Enter the following in the code text area and press ctrl+enter:

print_cost(parse_lines(`
.pixel
	move.w	d0,d4		; d4 = VVvv
	move.b	d2,d4		; d4 = VVUU
	move.b	(a0,d4.w),(a5)+
	add.l	d1,d0
	addx.b	d3,d2
	dbf	d7,.pixel
`));

Expected output:

	MOVE.W	d0, d4           	4
	MOVE.B	d2, d4           	4
	MOVE.B	(a0,d4.w), (a5)+ 	18
	ADD.L	d1, d0            	8
	ADDX.B	d3, d2           	4
	DBF	d7, .pixel          	10
	Total (estimated) cost: 48