Custom ROM creation log.

[sickle]

Edit:
This sort of turned into a log, so instead of wasting a fresh forum post...
Scroll down?


Original Text:

Hey guys, I've been having a little bit of bother getting any useful code running on my XplorerFX cart as a custom .ROM.

I've studied the Multirom source pretty thoroughly and quite successfully patched various ROMs and EXEs together with bits of multirom, myAR and the romrunner ROM by hand in the hex editor - so I've got a fairly good idea of the layout at least, and that the code is at least attempting to run.

To some extent the code is working fine - for example Silpheed's InitGPU routines from the GreenTro get at least as far as visibly altering the screen size, and I've gotten a simple test program to navigate some jumps and gracefully return control to Caetla after preserving the ra register.

However, a few things just won't play ball:

The WaitGPU function from GreenTro (called from InitGPU, while it's setting things up) hangs.
-Compiled with Spasm as .EXE - Works perfectl.y
-Compiled with ASMPSX as .EXE - Hangs at the first WaitGPU call (visible change)
-Compiled with Spasm as .BIN (spasm -b whatever.s,whatever.rom) hangs (visible change)
-Compiled with ASMPSX as .BIN (asmpsx /p whatever.s,whatever.rom) hangs (visible change)

Or using a simple program running HitMOD (asm), with a 0x20000 cycle For-Loop simulating the v sync (so it's the only other code running):

-Spasm .EXE works perfectly
-ASMPSX .EXE works perfectly
-Spasm .BIN hangs immediately with a blank screen
-ASMPSX .BIN hangs immediately with a blank screen

Is there a specific reason certain calls just won't work when ran from the cart?

Here's my code if it helps any:

Code: Select all

;Attempt at getting Hitmod (with Greentro BGM) running on Xplorer FX cart.
;Hitmod init, poll and includes borrowed from Greentro.

	opt	c+


	org	$1F000000
	dw	entryPoint2
	db	"Licensed by Sony Computer Entertainment Inc."


	org	$1F000080
	dw	entryPoint1
	db	"Licensed by Sony Computer Entertainment Inc."

	org	$1F000100

;Compiles correctly to 0x80
entryPoint1:
	
	j initStuff
	nop;

;Compiles correctly to 0x00 with header
entryPoint2:


	j initStuff
	nop;
	

initStuff:

		;Store RA on the stack. 
                addiu     sp, sp, $FFF4
                sw        ra, $0(sp)
                sw        t1, $4(sp)
                sw        t2, $8(sp)

            
            	la a0, module
            	jal HM_Init                 ; init the mod player
           	 nop
		

mainLoop:

            	jal HM_Poll                 ; call mod player
            	nop	


		li t1,$00020000
innerLoop:
		bltz t1,innerLoopEnd;
		nop
		subiu t1,t1,$1
		j innerLoop
		nop		

innerLoopEnd:
	

		j mainLoop
		nop



exitCode:

		;Return RA to its former glory, and allow to bask once again in the sweet sunlight.
                lw        ra, $0(sp)
                lw        t1, $4(sp)
                lw        t2, $8(sp)
		addiu     sp, sp, $C
	
		nop
		jr ra
		nop


;;---------------- Includes n Stuff.



	nop
	
module incbin thesong.hit
	
	nop

	include hitmod.inc	

	cnop	0,1024


	end

WaitGPU / InitGPU snippet:

Code: Select all

; WaitGPU - waits until GPU ready to receive commands
            
WaitGPU                     
            lw t1, GP1(k1)             
            li t0, $10000000
            and t1, t1, t0
            beqz t1, WaitGPU
            nop
            jr ra
            nop  


;;And a bit further down:


; InitGPU - basic GPU init routine
; in:  a0 - display mode
            
InitGPU:     



	move t3,ra	

             sw zero, GP1(k1)            ; reset

            li t2, $03000001
            sw t2, GP1(k1)              ; disable display
  
            li t2, $06c40240            ; horizontal start/end
            sw t2, GP1(k1)
            nop
  
            li t2, $07049025            ; vertical start/end 
            sw t2, GP1(k1)              
            nop              
                        
            li t2, $e1000685            ; draw mode, texture page = (8bit,320,0)
            jal WaitGPU                  ;Causes the Hang
            nop

            sw t2, GP0(k1)
            nop                                    
	    jal WaitGPU
    	    nop

As you can see it's pretty basic, but the ORG statements are set up.
(Padded them out manually with nops in spasm 'cause it only supports a single org)
I was initially pretty convinced that it was something to do with T3 failing to store RA properly, but even pushing it onto the stack, or other random registers just in case didn't make much difference. The final BIN is quite small too as you'd imagine so it's not spanning a huge amount of memory.

I've only really been learning this over about 2 days, so I'm probably missing something very obvious here - so any help would be pretty awesome - even if it's just a link to an old obscure document.

Cheers

[inc^lightforce]

I hope this will help you:
http://rapidshare.com/files/2701055793/GREEN.rar

read *.inc Files

[sickle]

"; NOTE: register k1 is expected to contain $1f800000 (hardware base) " - Not sure how I missed that one...

Thanks for hunting that out man

Also, I hate to be an arse, but I just can't source out the Macros.inc and Equates.inc files. I'm pretty sure I remember reading through them for interesting memory locations and such a few years ago I just can't find them. Any ideas?

[Type 79]

"... I just can't source out the Macros.inc and Equates.inc files. I'm pretty sure I remember reading through them for interesting memory locations and such a few years ago I just can't find them. Any ideas?

Found those from harddrive.

EquatesMacros.zip

[sickle]

Thankyou muchly =) I was elbow deep in a pile of ancient CDRs!

[inc^lightforce]

now you have all files and information. let's see some creative WORK made by you
but why ASM? huaaaaaaaaaaaaa. hardcore. i only use a few lines of ASM when coding Cracks and Patches. But not for a whole demo thats too crazy ha ha

[sickle]

Heh, I've always had a fascination with ASM being so close to the machine. 10 years ago or so I would have been all over PsyQ if I could find it anywhere, but suddenly it seems like it's not what I was craving. I've found some pretty cool things already.

For example, did you know that a jumping to 0x1f000280 (well 0x1f009f88 if you adjust the gp) in caetla0.34 brings up a testcard style comm screen? I've never seen that, but it's handy even if just for visual confirmation.
(I compiled a quick .exe with org at 0x1f000000 and pasted some .ROMs into the .exe at 0x800 - r3kDis can disassemble it quite well after that. Few issues, but it's an insight into the ROMs' inner workings).

On the ROM side of things, I got sick of trashing my XP cart and figured I'd try and find a way to break to the system bios when the switch is off, so I don't need to plug/unplug it each time it needs an X-Flash. (Which is quite often right now).
That or I'll have to figure out which pin disables the cart completely and add a switch...
Some of the routines are working though - the GPUInit, WaitGPU and WaitIdle functions for example. Seems I have some figuring out to do for the WaitVSync function and the Hitmod Init but it's all good fun


Thanks again - I'll make a wee post if anything fruitful comes from this.

[Administrator]

In theory, you should be able to disable the XP cart by adding a switch on the /EN line of the EEPROM

[sickle]

Hehe, yeah I found some old pinout diagrams, but uhmm... when they started producing the FX carts they cut a few corners.
The new EEPROM for example isn't a regular chip with .1 inch pin spacing, it's one of those little fellers, sort of smushed in the middle of where the old chip was, with extra lines etched to the old pin spaces. Quite clever really, but no part of me really wants to be desoldering that! Thanks though man

Anyway, project update:
-Figured out the return to bios from EP1 and EP2.
-Graphical mode change is working pretty spiffingly. (not doing anything yet though)
-Got most of RunRom disassembled and rewritten - focusing on other areas first.
-Intergrated the Comms routines by Foo (rudimentary parallel port support).
-Switch now lets you boot regular, embedded .EXE, or graphical portion of the rom.
(toggle mid-boot to trigger the embedded .EXE)

Todo:
Transfer a ton of variables to user RAM - VSync Routines, HitMod, etc.
Like, everything else.

Basically, getting the off switch to allow XFlash to boot normally saved so much time and development is speeding ahead
This is so much fun... I'll post the source tomorrow.

[sickle]

Excuse the double post, but this is feeling like a work log.

Just a little release - like RunRom, but you don't (always) need a disc to boot.
Some .EXE will work just fine off the bat, some need a little wiggle of the switch and some just take a bloody long time, but hey, it's coming together.

http://www.psxdev.net/forum/viewtopic.php?f=60&t=393

@Type97 - Already checked - it sorta works with the last preview of your CD player
(Start psx with cart off, toggle cart on and hit reset, then wait 10 seconds... yeah I'm working on it!)

[Type 79]

@Type97 - Already checked - it sorta works with the last preview of your CD player
(Start psx with cart off, toggle cart on and hit reset, then wait 10 seconds... yeah I'm working on it!)

have to test this later

[Type 79]

Works nicely with Xplorer using reset/switch/reset.
With GameShark-cartridge you have to first start PSX without cartridge, plug cartridge in and then reset PSX.

Something in my cdplayer causes about 30s delay on start, as there is no delay with program that just clears screen.

Great work!

[Administrator]

http://www.psxdev.net/forum/viewtopic.p ... 2811#p2811

[sickle]

Hoot cheers again. At the risk of cross posting - I figured I'd update here.
(It feels like the right thing to do).

Fixed Booting:
http://www.psxdev.net/forum/viewtopic.php?f=60&t=393

Not sure what to fiddle with next.... converting hitmod, dual booting custom exe + XFlash (timesaver!), some basic PC comms... etc. Would be nice to write something fresh though, most of this is bodged together from other sources >.<

[Type 79]

Now, with 0.2, cd-player starts from GameShark without any tricks, just power on, wait about 15s and music starts playing.

[Type 79]

Combine romloader + serial-loader, flash it to old AR/GS-cart = instaboot serial-loader, poorman's Xplorer
Unfortunately Shadow's serial-loader is too fat to fit in GameShark, but old Hitmen's loader is slim enough.

[sickle]

Lol, might be a good way to get some use out of my pile of gameshark clones - I don't have a serial cable or anything though Guess I could just use an Arduino? I'll think of something amusing to do

[sickle]

Just a little project update:

-Got pad input working pretty nicely
-Menu options! Circle to return to BIOS for example, Triangle to boot xFlash or *.exe
-Got a bit of a gui up and running, just some lines and gouraud shaded quads
-Got some sprites working (as above, needs to be copied to RAM)
-Sprites' pixel and CLUT data are read directly from a TIM (no need for separate files)
-Added a sin table for some spinning sprites, helps track down freezes, looks cute.
-Switch off currently boots XFlash (basically to make coding the rom easier)

Todo:
-Do something with all those comms routines.
-Work out why linked ordering tables aren't working when copied to RAM.
-Clean up the graphics a bit
-Pack the spritesheet a bit better, could save 20kb or so

Here's a pic:



And I've included the ROM with xFlash if someone wants to play - source needs a good cleanup before I release it, but having the bailout to XFlash has made it so much faster to work with - switch off, hit reset and reflash