New emulator - BuLWiP

[PeteE]

15 hours ago, mr_gw454 said:

I did try to compile this on my Linux Mint 20.1 system but ran into an error:

Looks like it might be related to my version of SDL2? 2.0.10.

Yes, SDL_LockTextureToSurface was added in 2.0.12.  I think I can avoid calling that function.

19 hours ago, TheMole said:

Here's an example cart image and the output from objdump with (some) C code interspersed... For some reason there's large swaths of asm code without C code, but that's probably because I'm including a library that hasn't been compiled/built with the right C flags. Should still work for testing if this type of list format can be made to work with you debugger though...

 

Oh, and the program itself assumes 80 column text mode on an F18A, so won't run correctly in the emulator... I hope that's not an issue for you. I just picked a small-ish project I had lying around.

malloctest8.bin 32 kB · 3 downloads malloctest8.lst 118.18 kB · 1 download

Looks like the listing parser handled it okay.  When the function loads a listing it prints out the segments that it found:

load listing test/malloctest8.lst
list 0-555 pc=6000-6014 bank -1
list 555-55886 pc=2080-2fce bank -1
list 55886-121015 pc=6f6a-7c50 bank 0

The list shows the file offset range and the program counter address range, as well as the assumed bank.  Bank -1 means all banks.  To detect the bank number in GCC listings, it looks for the text "Disassembly of section .bank" followed by the bank number.  I think you have to set this name in your linker script, like this (taken from fcmd)  You probably know about banking in GCC better than I do.

 

  /* overlays */
  OVERLAY : AT (0x10000)
  {
    .bank0 {             /* list the set of modules in bank 0 */
      objects/api.o(.text)
      objects/b0_*.o(.text)
      __LOAD_DATA = .;   /* .data segment is appended here by makefile for final ROM .bin */
    }
    .bank15 { objects/b15_*.o(.text) }  
    .bank14 { objects/b14_*.o(.text) }  
    .bank13 { objects/b13_*.o(.text) }  
    .bank12 { objects/b12_*.o(.text) }  
    .bank11 { objects/b11_*.o(.text) }  
    .bank10 { objects/b10_*.o(.text) }  
    .bank9 { objects/b9_*.o(.text) }  
    .bank8 { objects/b8_*.o(.text) }  
    .bank7 { objects/b7_*.o(.text) }  
    .bank6 { objects/b6_*.o(.text) }  
    .bank5 { objects/b5_*.o(.text) }  
    .bank4 { objects/b4_*.o(.text) }  
    .bank3 { objects/b3_*.o(.text) }  
    .bank2 { objects/b2_*.o(.text) }  
    .bank1 { objects/b1_*.o(.text) }  
  } >bank_rom

 

 

13 hours ago, apersson850 said:

Like this?

Exactly!

Edited March 4, 2023 by PeteE

[+remowilliams]

It seems to be running a bit fast, but is looking very nice!  🙂

 

[+acadiel]

On 3/1/2023 at 12:11 PM, OLD CS1 said:

 

 

Sorry for the OT, but that one never gets old.  

[PeteE]

I pushed some changes to the github repo. The alternate debug key is F12. Use Shift-Insert to paste from the clipboard. Find keys are added. The SDL functions should work on SDL 2.0.10, as well as fixed the CRT filter on Mac, and also added 80-column text mode just for fun.

 

EDIT: Also fixes the bug with XB.  I forgot the post increment on CB instruction.

Edited March 4, 2023 by PeteE

[TheMole]

Nice, I need to figure out how the debugger works, but this all looks extremely promising. I can confirm the new version fixes the crt filter, and the new shortcut works, thanks !

[mr_gw454]

On 3/3/2023 at 11:55 AM, PeteE said:

Yes, SDL_LockTextureToSurface was added in 2.0.12.  I think I can avoid calling that function.

 

I just pulled the latest updates and it now compiles for me (including the Raspberry Pi (ARM).  

 

Thank you very much for making that SDL LockTextureToSurface change.

 

 

[+arcadeshopper]

axel-f music program runs SOOPER FAST 

axelf.BIN

[PeteE]

1 hour ago, arcadeshopper said:

axel-f music program runs SOOPER FAST 

axelf.BIN 32 kB · 0 downloads

Thanks, I'll take a look.  The music player decrements a counter to delay between notes.  Seems like I need to finish checking the instruction cycle timing.

[+arcadeshopper]

4 hours ago, PeteE said:

Thanks, I'll take a look.  The music player decrements a counter to delay between notes.  Seems like I need to finish checking the instruction cycle timing.

Wonder if that's why Donkey Kong is so fast too.. probably 

[PeteE]

30 minutes ago, arcadeshopper said:

Wonder if that's why Donkey Kong is so fast too.. probably 

Found the bug: expansion RAM accesses are not adding their cycle time (2 cycles + 4 for the multiplexer), so they're effectively faster than scratchpad RAM.  Whoops.

[PeteE]

I checked in the instruction cycle count fixes, at least for the instructions I care about.  If there's others, I don't want to know about it.  I made an automated test using XAS99 to generate various combinations of instructions with normal or multiplexed memory addresses, and used the "s+","s-","d+","d-" memory access speed annotations so that the cycle counts in the listing were accurate.  My test harness loads the binary, single steps through it sequentially, and compares the cycles executed to that instruction in the listing.

[+arcadeshopper]

dk runs fine now

[gferluga]

Runs wonderful on MacBook M2 Arm processor ;)-

[gferluga]

How do I map a DSK floppy disk drive image?

Thanks!

 

[PeteE]

2 hours ago, gferluga said:

How do I map a DSK floppy disk drive image?

Thanks!

 

Disk IO hasn't been done yet, sorry 😓

[Willsy]

Compiled fine on my Mint system (once I installed SDL lib!) - it's asking for the rom files - 

Failed to load ROM/GROM files: 994arom.bin 994agrom.bin

Any pointers? Thanks

 

Mark

[PeteE]

4 hours ago, Willsy said:

Compiled fine on my Mint system (once I installed SDL lib!) - it's asking for the rom files - 

Failed to load ROM/GROM files: 994arom.bin 994agrom.bin

Any pointers? Thanks

 

Mark

They can be gotten from here https://ftp.whtech.com/System ROMs/MAME/pre_0.174/ti99_complete.zip nested inside the ti99_4a.zip