[CSDb] - User Forums

You are not logged in - nap

CSDb User Forums

Forums > C64 Productions > SIDBlaster

2025-05-04 19:33

Raistlin

Registered: Mar 2007
Posts: 722

SIDBlaster

A little early perhaps to put here on CSDb as a full release .. but for anyone interested, here's a new tool from me:-

https://github.com/RobertTroughton/SIDBlaster

I have big plans with this.. main features right now (may be buggy!) are:-

- ability to add simple players to SIDs <-- later on of course this will allow full visualisers, SID information and more

- relocation of SIDs. The process used for this is very different to other relocators that I've seen. I essentially do a 10mins emulation of the CPU, running through the SID, checking what's data, what's code, what's read from, what's written to, etc... detecting indirect reads, and where the vectors for those come from ... dealing with "backups" of those vectors and all that nastiness. It's far from perfect .. but it works already with many SIDs and I'll look at other SIDs to see how easy those will be to fixup.

Next tasks:-
- exposing more information to players/visualisers
- allowing import of information/defines from txt files to pass to the visualisers
- adding an "insanity check" to the relocator - so I'll do 10mins' emulation of the SID before and after relocation, recording all the $d4xx writes and making sure they tally up.

Much more to be done ... feedback is of course welcome.

Windows/Ubuntu/Mac releases are there .. though I've only tested Windows - hopefully the others "just work"??

... 7 posts hidden. Click here to view all posts....

2025-05-05 16:52

iAN CooG

Registered: May 2002
Posts: 3229

RSID are not supported, it's not told anywhere, add in the readme a warning at least. Trying to relocate \MUSICIANS\T\Turbocracker_87\Crockets_Theme.sid the message is a bit vague

SIDBlaster.exe -relocate=$1000 Crockets_Theme.sid Crockets_1000.sid

Not a valid SID file!
[2025-05-05 16:50:52] [ERROR] Failed to load file: Crockets_Theme.sid

other than that, if the exe needs java to run kickass, I'm outta here. I won't install java to test if further

2025-05-05 17:13

Raistlin

Registered: Mar 2007
Posts: 722

Quoting acrouzet

the name SIDBlaster is already used

God damn, I knew the name reminded me of something! I'll change it "eventually"... I'm not sure yet how easy it'll be with GitHub etc etc .. or maybe I just start a new repo with a clean copy of the codebase :-)

Quoting Burglar

I'm wondering about the relocating bit, I'm a huge fan of lft's Sidreloc V1.0, which hasn't failed me yet. So, why a new tool?

So, long-term, if I can get more SIDs to be -fully- relocatable, to the point where the generated ASM can be hand-edited without problems.. then that opens up a truckload of options. Optimisation. Direct hooks for visualisers. Etc etc.

Right now, I can blank out unused parts of SID files .. I've seen some SIDs with 8-10% of memory that is never touched .. including branches that are never taken.

By comparison, LFT's relocator is bloody cool. It works very differently though.

Worth noting as well .. what needs to be done to get byte-wise relocation working, full ASM editing, is HARD on many SIDs. Some have, for example, self-modifying code where there might be such as a jump - but where only the low-byte of the jump is changed. Eg:-

ldy $fe
lda $1323, y
sta $1041

$1040: jsr $1100
...
$1323: .byte $00, $03, $09, $10, $18, $20

So basically it will change that $1100 to $1100, $1103, $1109, ..., $1120.

Because the high-byte is never changed, it becomes difficult. Not -impossible- but definitely <i>difficult</i>. It might need the code to be expanded, to do some math to actually calculate the low and high byte.

It's for these sort of reasons, I assume, that LFT chose to only allow whole-page relocation in SIDreloc.

Quoting Mixer

I wonder if this could be taken a step further to map out the execution path from the routine call to each sid write and perhaps figure out a way to squeeze more cycles or bytes out of the routines.

Yes! Absolutely. I need the relocation to be working such that the ASM can be changed, as mentioned above, and then everything like this becomes possible.

Quoting iAN CooG

RSID are not supported, it's not told anywhere, add in the readme a warning at least.

Thanks, Ian! I'll grab that and give it a go later. I'm not quite sure where the code's at for the various formats .. I think I saw somewhere that PSID had v2, v3 and v4.. if so, I'll need to check each if they're still in circulation. That's why I PM'ed you on FB, I thought you'd be a good person to speak to about getting the files and header specs etc :-)

I'm adding stuff to this all the time.. I'm adding the post-relocation SID-write verifying code now.. and will test on some new SIDs ... and then probably get the various other SID formats added in before doing a pass over large chunks of HVSC.

I'll only move this tool to v1.0 once I know that it can cope with 90%+ of HVSC (and emitting errors/warnings for the other <10%).

2025-05-05 17:25

Stone

Registered: Oct 2006
Posts: 176

Quoting hedning

My thought exactly. I actually own one. :) https://cbmretro.fi/product/sidblaster-usb-nano/

I even made a driver for it back in the day, available here on csdb SIDBlaster USB driver BETA

I think this version is obsolete though.

Quoting Raistlin

relocation of SIDs. The process used for this is very different to other relocators that I've seen. I essentially do a 10mins emulation of the CPU, running through the SID, checking what's data, what's code, what's read from, what's written to, etc... detecting indirect reads, and where the vectors for those come from

Not sure how LFT does it in Sidreloc V1.0 but this sounds similar to what I did in SIDdecompiler V0.8 I basically create (through emulation) a graph for each value that is written to the SID chip and by traversing that and applying some heuristics, I try to determine which are vectors that should be relocated. I found that some drivers need to be page-aligned to some extent, e.g because of self-modifying code where only the low-byte is changed in the vector. It's a very interesting project, for sure!

2025-05-05 18:47

Raistlin

Registered: Mar 2007
Posts: 722

Quoting Stone

e.g because of self-modifying code where only the low-byte is changed in the vector

Yep, exactly this. I mentioned this oddity just above. I do plan to tackle this one eventually as well though so that relocation still works without restricting to whole pages .. but I'm not there yet.

2025-05-05 18:52

Raistlin

Registered: Mar 2007
Posts: 722

Quoting Stone

this sounds similar to what I did in SIDdecompiler V0.8

Interesting! That does sound very similar indeed, actually..! We look to have very similar commandline options too.

It's a fun project to work on so I'm definitely going to push ahead and see if I can get the relocation working solidly.

2025-05-05 21:43

Bansai

Registered: Feb 2023
Posts: 51

Quoting Raistlin

Quoting Burglar
I'm wondering about the relocating bit, I'm a huge fan of lft's Sidreloc V1.0, which hasn't failed me yet. So, why a new tool?

So, long-term, if I can get more SIDs to be -fully- relocatable, to the point where the generated ASM can be hand-edited without problems.. then that opens up a truckload of options. Optimisation. Direct hooks for visualisers. Etc etc.

SIDdecompiler has worked great for me. Relocation cases that I've come across somewhat often that need human intervention are optimized pointer math where the upper byte of a 16-bit pointer is skipped. Addresses used as data are another fun one. Fortunately, it's not typical. :-) This is from Beyond.sid (Michael Hendriks) using the FAME player:

 957                   L_D_0EFD_TEST
 958                   .ifne L_D_0EFD_TEST-$0EFD
 959                   L_D_0EFD_RELO_PATCH ; L_ARPTAB_LO data were abused to provide filtering enables, and we need these constants, otherwise the code isn't relocatable
 960                                       ; $11 gives low pass filtering, bad values will write 1 to bit 7 of $d418 turning off voice 3
 961                    .byte $11,$14,$17,$1A,$1D
 962                   .endif
 963                   .ifeq L_D_0EFD_TEST-$0EFD
 964                    L_D_0EFD_RELO_PATCH
 965                   .endif
 966
 967                   L_ARPTAB_LO ; from 60-7D block command, writes into EFFTAB[0] table!
 968 0efd 11 14 17 1a   .byte <L_ARP_037,<L_ARP_03A,<L_ARP_035,<L_ARP_03C,<L_ARP_0180
     0f01 1d
 969                   L_ARPTAB_HI
 970 0f02 0f 0f 0f 0f   .byte >L_ARP_037,>L_ARP_03A,>L_ARP_035,>L_ARP_03C,>L_ARP_0180
     0f06 0f
 971                   L_ARPTAB_LEN
 972 0f07 03 03 03 03   .byte $03,$03,$03,$03,$03
     0f0b 03
 973                   L_ARPTAB_RPT
 974 0f0c 00 00 00 00   .byte $00,$00,$00,$00,$00
     0f10 00
 975
 976                   L_ARP_037
 977 0f11 00 03 07      .byte $00,$03,$07
 978                   L_ARP_03A
 979 0f14 00 03 0a      .byte $00,$03,$0A
 980                   L_ARP_035
 981 0f17 00 03 05      .byte $00,$03,$05
 982                   L_ARP_03C
 983 0f1a 00 03 0c      .byte $00,$03,$0C
 984                   L_ARP_0180
 985 0f1d 00 18 00      .byte $00,$18,$00
 986

2025-05-05 21:58

Burglar

Registered: Dec 2004
Posts: 1129

@raistlin, cheers for the explanation, will follow this project :)

2025-05-05 22:06

Stone

Registered: Oct 2006
Posts: 176

Quoting Raistlin

It's a fun project to work on so I'm definitely going to push ahead and see if I can get the relocation working solidly.

I Hope you do one of your write-ups on this one too. It's always interesting.

2025-05-06 19:33

Raistlin

Registered: Mar 2007
Posts: 722

Quoting Stone

I Hope you do one of your write-ups on this one too. It's always interesting.

Well, now that you mention it..!

https://c64demo.com/welcome-to-sidblaster/

And, also, while I'm here .. I added some new features today - WIP and subject to massive change - to make it easier to verify relocation work etc. It's not 100% yet but it works well enough for now.

https://github.com/RobertTroughton/SIDBlaster

PS. I will still change the name of the tool .. maybe .. but probably as I get to v1.0.

2025-05-09 17:20

Raistlin

Registered: Mar 2007
Posts: 722

Quote: RSID are not supported, it's not told anywhere, add in the readme a warning at least. Trying to relocate \MUSICIANS\T\Turbocracker_87\Crockets_Theme.sid the message is a bit vague

SIDBlaster.exe -relocate=$1000 Crockets_Theme.sid Crockets_1000.sid

Not a valid SID file!
[2025-05-05 16:50:52] [ERROR] Failed to load file: Crockets_Theme.sid

other than that, if the exe needs java to run kickass, I'm outta here. I won't install java to test if further

Hey Ian,

I've added detection for RSID on the head version now and will put in a later release .. but .. I don't suppose you can give example SIDs for the various other versions, could you..? If they exist in HVSC - maybe everything's already updated to the latest PSID version?

I might come back to RSID if there's any way to emulate it - but it sounds like it would be tricky?

Previous - 1 | 2 | 3 - Next

Refresh

Subscribe to this thread: