Sega CD Game saves (Entire asm library found for BRAM saves)

[lunchbox]

SO i looked around and from my understanding the Internal battery backed memory in the sega cd is known as BRAM? if this is the case, i saw a few places it was said that it would not be implemented due to the fact that it wasn't known how it works. Is this still the case across the homebrew community? Id REALLY love the ability to use sega-cd internal bram to hold a few variables for save games like unlocked items or points and name ect and unlocked levels.

If BRAM is indeed not ever going to be supported is there any other method that could be used? Sega cd Ram Backup cart im guessing would be the next step?

[lunchbox]

Well Well...

Sonic 1 ported to sega CD happened to have bram support coded in

trox.binary-division.com/public/Sonic1_SegaCD_v2.rar

This was found in the /game/ folder in savegame.asm

; =======================================================
; BRAM Manager (Main-CPU handler)
; =======================================================


*****************************************************************************************
; @gf64/Gardeguey:
; this took me some hours to make, not even you
; can do something like this (or maybe yes, I don't know)
;
;
; FORMAT:
; > 4 Save Slots, $20 bytes each one
;   > 0x00: (L) Score
;   > 0x04: (W) Zone & Act
;   > 0x06: (B) Character to use (unused)
;   > 0x07: (B) Lives
;   > 0x08: (b) Continues
;   > 0x09: (b) Emeralds
;   > 0x0A: (l) Emeralds
;   > 0x0E: (l) Emeralds
;   > 0x12: (b) SS number
;   > 0x13 -> 0x1F: blank
;
; - Trox
*****************************************************************************************

; =======================================================================================
; Macros & Equates
; =======================================================================================

BRAM_Enable	macro
		m_SetBank_PRG_RAM	1
		bsr	m_SubHalt
		endm

BRAM_Disable	macro
		bsr	m_SubUnHalt
		bsr	BRAM_Write
		endm
; =======================================================================================
; BRAM's file name
; =======================================================================================

BRAMString:
		;Name (11 bytes), unknown, unknown, size ($20 bytes blocks)
		dc.b	'SONIC1_____',0,0,4
		even
; ---------------------------------------------------------------------------------------

SaveBRAMData:
		BRAM_Enable
		moveq	#0,d0
		move.b	($FFFF8F92).w,d0
		lsl.w	#5,d0
		lea	(m_DLL_align+$3000).l,a6
		lea	(a6,d0.w),a6
		move.l	($FFFF8E26).w,(a6)+
		move.w	($FFFF8E10).w,(a6)+
		move.b	d0,(a6)+
		move.b	($FFFF8E12).w,(a6)+
		move.b	($FFFF8E18).w,(a6)+
		move.b	($FFFF8E57).w,(a6)+
		move.l	($FFFF8E58).w,(a6)+
		move.l	($FFFF8E5C).w,(a6)+
		move.b	($FFFF8E16).w,(a6)
		BRAM_Disable
		bra	BRAM_Write
; ---------------------------------------------------------------------------------------

LoadBRAMData:
		BRAM_Enable
		moveq	#0,d0
		move.b	($FFFF8F92).w,d0
		lsl.w	#5,d0
		lea	(m_DLL_align+$3000).l,a6
		lea	(a6,d0.w),a6
		move.l	(a6)+,($FFFF8E26).w			; load score
		move.w	(a6)+,($FFFF8E10).w			; load zone and act
		move.b	(a6)+,d0				; load character's ID (unused)
		move.b	(a6)+,($FFFF8E12).w			; load lives
		move.b	(a6)+,($FFFF8E18).w			; load continues
		move.b	(a6)+,($FFFF8E57).w			; load emeralds
		move.l	(a6)+,($FFFF8E58).w			; ' '
		move.l	(a6)+,($FFFF8E5C).w			; ' '
		move.b	(a6),($FFFF8E16).w			; load special stage number
		BRAM_Disable
		tst.b	($FFFF8E12).w
		bne	@cont
		move.b	#3,($FFFF8E12).w
@cont:		rts
; ---------------------------------------------------------------------------------------

BRAM_Read:
		moveq	#0,d6
		moveq	#2,d0
		lea	BRAMString-$180000,a1
		bsr	m_CallDLLFunc2_S_RAM
		bsr	CheckError
		rts
; ---------------------------------------------------------------------------------------

BRAM_Write:
		moveq	#0,d6
		moveq	#3,d0
		lea	BRAMString-$180000,a1
		bsr	m_CallDLLFunc2_S_RAM
		bsr	CheckError
		rts
; ---------------------------------------------------------------------------------------

SetupBRAM:
		moveq	#0,d6
		moveq	#0,d0			; Initialize BRAM ('BRMINIT')
		bsr	m_CallDLLFunc_S_RAM	; Process it
		bsr	CheckError		; Check for any errors in the system
		bsr	FindFile		; Check if file already exists
		bra	BRAM_Read		; If everything is ok, we can read the data in our buffer
; ---------------------------------------------------------------------------------------

FindFile:
		moveq	#6,d0			; Search file ('BRMSERCH')
		lea	BRAMString-$180000,a1	; Set filename
		bra	m_CallDLLFunc2_S_RAM	; Process it; if not found, will be created (checks the size first)
; ---------------------------------------------------------------------------------------

CheckError:
		move.l	m_stat8,d6		; Get DLL's returned data
		swap	d6
		cmpi.w	#'ER',d6		; Did any error happen?
		beq	ProcessBRAMError	; If yes, branch
		rts

ProcessBRAMError:
		addq.w	#8,sp			; Keep stack correctly...
		moveq	#0,d0
		bra	m_CallDLLFunc
; ---------------------------------------------------------------------------------------

[lunchbox]

Does anyone have something like this implemented for bex? I only ask because id rather not go down the path of trying to make the above code work if someone already has something working that they might share.

[nathan999]

I've made experiments and it's relatively easy to write to the BBRAM, just check this simple code:

[

[nathan999]

I've made experiments and it's relatively easy to write to the BBRAM, just check out this simple code:

'' SRAM test!

option SEGACD

Const #BASEBBRAM = &H600000

' Read 3 values from BBRAM...

PRINT "READING 3 VALUES FROM BBRAM"
Print BBRamRead(0)
Print BBRamRead(1)
Print BBRamRead(2)

Print "WRITING 3 VALUES TO BBRAM"
BBRAMWrite 0, 22
BBRAMWrite 1, 33
BBRAMWrite 2, 44

Declare Function BBRamRead(offs&)
Return Peek(#BASEBBRAM + 1 + offs& + offs&)
End Function

Declare Sub BBRamWrite(offs&, val%)
Poke &H7FFFFF, 1 ' Write enable
Poke #BASEBBRAM + 1 + offs& + offs&, val%
Poke &H7FFFFF, 0 ' Write protect
End Sub

This works. The first time it will read/print 0, 0, 0; then write 22, 33, 44. Then you reset the emulator (I use Kega Fusion), run it again, and it reads/prints 22, 33 44.

Problem is, there has to be any kind of already established format for binary blocks writen to the bram, so you can search for free space and not corrupt already existing saves. I can't find anything about this subject anywhere. I'll study the code provided above and do some tests in case I can figure out what's going on.

[nathan999]

I've examined the code (Sonic 1 for Sega CD) further and it seems that everything is achieved calling the Sega CD BIOS. My time is scarce, but maybe I can make a simple savegame handler. But don't expect anything anytime soon.

[nathan999]

This is what I've come up with, but it doesn't work.

I have to investigate this further. I know that the main CPU and the SUB CPU have different addressing maps, maybe that's the problem. I'm using main CPU addresses and this may need adjustment.

'' BRAM test!
'' Will attempt to use SEGACD BIOS (somehow)

' Code has been generated using "Sonic 1 for Sega CD" as a basis & reference.

option SEGACD

Const #BASEHEADER = &HFFFFA0
Const #BASEDATA = &HFFFF00

' BIOS entry point:
' _BURAM            EQU   $00005F16

' BIOS functions:
' BRMINIT           EQU   $0000
' BRMSTAT           EQU   $0001
' BRMSERCH          EQU   $0002
' BRMREAD           EQU   $0003
' BRMWRITE          EQU   $0004
' BRMDEL            EQU   $0005
' BRMFORMAT         EQU   $0006
' BRMDIR            EQU   $0007
' BRMVERIFY         EQU   $0008

' For VANE engine, we need to save 128 flags
' plus some other data.

' It seems that SEGA CD BURAM blocks are 32 bytes.
' So we need 1 block 'header' and 4 blocks 'flags'

' For the tests, let's imagine we have everything
' set up in a buffer in main RAM in &HFFFF00

' We also need the 'header' somewhere, that's
' 11 chars - game name
' 1 byte 0
' 1 byte 0 - block size is $20 bytes
' 1 byte 5 - block count.

' Header will be at $HFFFFA0

' To call a BURAM BIOS function you move
' the function code to d0, then jsr _BURAM

' write some data
Print "WRITE HEADER"
writeHeader

Print "WRITE DATA"
writeData

'Print "CLEAR DATA"
'clearData

Print "SHOW DATA"
showData

Print "INIT"
buramInit

Print "SAVE"
buramSave

Print "LOAD"
buramLoad

Print "SHOW DATA"
showData

Print "OK"

Declare Sub showData
For i&=0 to 31
Print Peek(#BASEDATA + i&);" ";
Next i&
Print
End Sub

Declare Sub writeHeader
Reload headerData
For i& = 0 To 13
Read d%
Poke #BASEHEADER + i&, d%
Next i&
End Sub

Declare Sub writeData
For i& = 0 To 159
Poke #BASEDATA + i&, i&
Next i&
End Sub

Declare Sub clearData
For i& = 0 To 159
Poke #BASEDATA + i&, 0
Next i&
End Sub

' Init. Supposedly you have to call this first.
' For some reason it needs 640 bytes of scratch data.
' Okay.

Declare Asm Sub buramInit
; #BASEHEADER == &H00FFFFA0
; a0 pointer to $640 bytes scratch data.
move.l #$00FFFFA0 - 1600, a0
; a1 pointer to 12 bytes string buffer
move.l #$00FFFFA0 - 1600 - 12, a1
; d0 API function #0, INIT
move.w #0, d0
; Call BIOS
jsr $00005F16

; cs means not formatted...
bcs.w BRAM_Format
bra BRAM_Init_End

BRAM_Format:
; d0 API function #6, FORMAT
move.w #$6, d0
; Call BIOS
jsr $00005F16

BRAM_Init_end:
End Sub

' Save needs a pointer to the header, and a pointer
' to the data itself.

Declare Asm Sub buramSave
; #BASEHEADER == &H00FFFFA0
; #BASEDATA = &H00FFFF00
; a0 pointer to header (file name, header data)
move.l #$00FFFFA0, a0
; a1 pointer to save data (enough data, 5 blocks as described)
move.l #$00FFFF00, a1
; d0 API function #4, WRITE
move.w #$4, d0
; Call BIOS
jsr $00005F16
End Sub

' Load needs a pointer to the header, and a pointer
' to the data itself
Declare Asm Sub buramLoad
; #BASEHEADER == &H00FFFFA0
; #BASEDATA = &H00FFFF00
; a0 pointer to header (file name, header data)
move.l #$00FFFFA0, a0
; a1 pointer to save data (enough data, 5 blocks as described)
move.l #$00FFFF00, a1
; d0 API function #3, READ
move.w #$4, d0
; Call BIOS
jsr $00005F16
End Sub

headerData:
Data $45,$53,$50,$5F,$54,$49,$45,$5F,$52,$5F,$5F
Data 0, 0, 5


I'm only posting this in case somebode seems something.

[nathan999]

Despite the lack of interest, this is interesting for me, so I've been researching futher. I'm calling the API using the right entry points, but it seems that I'm not passing the parameters right. I'm starting to believe that the registers involved are the second 68000 registers, not the main one's. There seems to be a gate array to communicate both processors.

With this knowledge in mind, I'll try to re-read the sources and try to understand how this all works.

[lunchbox]

Jul 23, 2015 2:34:45 GMT -5 nathan999 said:

Despite the lack of interest, this is interesting for me, so I've been researching futher. I'm calling the API using the right entry points, but it seems that I'm not passing the parameters right. I'm starting to believe that the registers involved are the second 68000 registers, not the main one's. There seems to be a gate array to communicate both processors.

With this knowledge in mind, I'll try to re-read the sources and try to understand how this all works.

Nice!!!

I'm certainly interested in it. i havnt had a bunch of time to follow lately but with some progress being made, im following now!!

Ive been super busy myself writing code for another console but when i have some time i might be able to lend a hand in this project.

[nathan999]

Cool

I am still clueless. The lack of proper documentation is overwhelming :-(

[Deleted]

Jul 28, 2015 7:15:30 GMT -5 nathan999 said:

The lack of proper documentation is overwhelming :-(

I assume you already have ( read ) the following pages from the official manual?

[lunchbox]

Bump:

Ive seen the docs myself but havnt played with anything in relation to them. Would be nice to enable the sub cpu and play around with that via bex too. Hell ultimately scaling via segacd would be awesome!!

Also, cant mention scaling and segacd and not mention batman returns. Examples of how its done here in the SegaCD Batman returns source code....

www.shrigley.com/codearc/Batman_Returns_SegaCD_Project_Package.zip

[landeel]

I'm very interested in this. Any news?

[lunchbox]

Nathan, you make any headway on this? if not where did you leave off?

Maybe post what you were calling and see if we can debug this stuff here. Might need a snippet of asm to call or or something. I would assume if the sub cpu registers are being referenced that something has to be toggled prior to being able to read them.

[nathan999]

Sorry, I'm currently busy with completely different affairs. This is the code I was testing:

'' BRAM test!
'' Will attempt to use SEGACD BIOS (somehow)
 
' Code has been generated using "Sonic 1 for Sega CD" as a basis & reference.

  option SEGACD

  Const #BASEHEADER = &HFFFFA0
  Const #BASEDATA = &HFFFF00

  ' BIOS entry point:
  ' _BURAM            EQU   $00005F16
 
  ' BIOS functions:
  ' BRMINIT           EQU   $0000
  ' BRMSTAT           EQU   $0001
  ' BRMSERCH          EQU   $0002
  ' BRMREAD           EQU   $0003
  ' BRMWRITE          EQU   $0004
  ' BRMDEL            EQU   $0005
  ' BRMFORMAT         EQU   $0006
  ' BRMDIR            EQU   $0007
  ' BRMVERIFY         EQU   $0008

  ' For VANE engine, we need to save 128 flags
  ' plus some other data.
 
  ' It seems that SEGA CD BURAM blocks are 32 bytes.
  ' So we need 1 block 'header' and 4 blocks 'flags'
 
  ' For the tests, let's imagine we have everything
  ' set up in a buffer in main RAM in &HFFFF00
 
  ' We also need the 'header' somewhere, that's
  ' 11 chars - game name
  ' 1 byte 0
  ' 1 byte 0 - block size is $20 bytes
  ' 1 byte 5 - block count.
 
  ' Header will be at $HFFFFA0
 
  ' To call a BURAM BIOS function you move
  ' the function code to d0, then jsr _BURAM
 
  ' write some data
  Print "WRITE HEADER"
  writeHeader
 
  Print "WRITE DATA"
  writeData
 
  'Print "CLEAR DATA"
  'clearData
 
  Print "SHOW DATA"
  showData
 
  Print "INIT"
  buramInit
 
  Print "SHOW DATA"
  showData
 
  Print "SAVE"
  buramSave
 
  Print "CLEAR DATA"
  clearData
 
  Print "SHOW DATA"
  showData
 
  Print "LOAD"
  buramLoad
 
  Print "SHOW DATA"
  showData
 
  Print "OK"
 
Declare Sub showData
  For i&=0 to 31
     Print Peek(#BASEDATA + i&);" ";
  Next i&
  Print
End Sub
 
Declare Sub writeHeader
  Reload headerData
  For i& = 0 To 13
     Read d%
     Poke #BASEHEADER + i&, d%
  Next i&
End Sub
 
Declare Sub writeData
  For i& = 0 To 159
     Poke #BASEDATA + i&, i&
  Next i&
End Sub

Declare Sub clearData
  For i& = 0 To 159
     Poke #BASEDATA + i&, 0
  Next i&
End Sub

  ' Init. Supposedly you have to call this first.
  ' For some reason it needs 640 bytes of scratch data.
  ' Okay.
 
Declare Asm Sub buramInit
  ; #BASEDATA = &H00FFFF00
  ; a0 pointer to $640 bytes scratch data.
  move.l #$00FFFF00 - 1600, a0
  ; a1 pointer to 12 bytes string buffer
  move.l #$00FFFF00 - 1600 - 12, a1
  ; d0 API function #0, INIT
  move.w #0, d0
  ; Call BIOS
  jsr $00005F16
 
  ; cs means not formatted...
  bcs.w BRAM_Format
  bra BRAM_Init_End
 
BRAM_Format:
  ; d0 API function #6, FORMAT
  move.w #$6, d0
  ; Call BIOS
  ;jsr $00005F16  
 
BRAM_Init_end:
End Sub
 
  ' Save needs a pointer to the header, and a pointer
  ' to the data itself.
 
Declare Asm Sub buramSave
  ; BRAM_Enable
  bsr   m_SubUnHalt  
  ; #BASEHEADER == &H00FFFFA0
  ; #BASEDATA = &H00FFFF00
  ; a0 pointer to header (file name, header data)
  move.l #$00FFFFA0-$180000, a0
  ; a1 pointer to save data (enough data, 5 blocks as described)
  move.l #$00FFFF00, a1
  ; d0 API function #4, WRITE
  move.w #$4, d0
  ; Call BIOS
  jsr $00005F16  
End Sub  

  ' Load needs a pointer to the header, and a pointer
  ' to the data itself
Declare Asm Sub buramLoad
  ; BRAM_Enable
  bsr   m_SubUnHalt
  ; #BASEHEADER == &H00FFFFA0
  ; #BASEDATA = &H00FFFF00
  ; a0 pointer to header (file name, header data)
  move.l #$00FFFFA0-$180000, a0
  ; a1 pointer to save data (enough data, 5 blocks as described)
  move.l #$00FFFF00, a1
  ; d0 API function #3, READ
  move.w #$3, d0
  ; Call BIOS
  jsr $00005F16  
End Sub

Declare Asm Sub haltStuffContainer
m_SubHalt:
     move   #$2700,sr
     bset   #1,($A12001).l
     beq   m_SubHalt
     rts

m_SubUnHalt:
     bclr   #1,($A12001).l
     bne   m_SubUnHalt
     rts
End Sub

headerData:
  Data $45,$53,$50,$5F,$54,$49,$45,$5F,$52,$5F,$5F
  Data 0, 0, 5
 

Maybe you can build upon this simple foundations?