Snes ASM cpu core

[yoyofr]

Please, report here if the "ASM cpu" option for SNES module works on your device or not.
I'll need to have the following informations :

• your device
  
• hack used (udmh, ...)
  
• special launcher used or not (ZLauncher, ....)
  
• free memory available
  
• stability : does it always crash at startup or not ?
  
• if you get a "Fatal Error", then please give me all the following information (number, ...)
  
• any other information you think is useful
  

thx

[Vitreal]

Palm Tungsten C.

ASM core does not work.

Reports Fatal Error. (No other information. Just fatal error then reset button)

Hacks/Progs Used

Launcher: Ispin launcher

Powernet (to promote the always on enviro)

12mb heap prior
10mb after the fact.
-----------------------------------------

Extra Info:

Unrelated: keys when pressed on C core caused dramatic slowdown.
Audio to to quite. Suggest increasing the band for audio output (something like how most media players extend the volume)
non-standard keys cannot be mapped. EX: All keyboard keys on the Palm T|C

[Tinnus]

# Works
# Tungsten|E
# UDMH
# ZLauncher
# 12+MB
# stability: crashes once in a while when starting and/or when trying to view the in-game menu... but I think this happens with C too (I just use ASM because it's faster ;D)

[snempaa]

Vitreal said:

non-standard keys cannot be mapped. EX: All keyboard keys on the Palm T|C

What you can do is map all the keys to one button, ex: when you want to map the y button, just press a non standard key and all the non standard keys will be used for the y button... so you have 1 extra button;)

Here on T|C asm core doesn't work, no error msg, just a black screen.

Heap before loading the snes.lem 12m. after 10m.

Hacks used:

Lightspeed to speed T|C up to 600mhz

Also after a couple of mins of playing the screen still goes black (and i cant get it on again, part from a softreset).

[yoyofr]

Heap before loading the snes.lem 12m. after 10m.

if you've got 2 messages before the emu is launched with "Load PNO Free Ram xxx" & co, then you don't have the latest beta. Please report here only with latest beta results.

[scott0968]

Dosen't Work

Device: T3

No Hacks (Although I did try it with UDMH and it still failed)

Launcher: Launcher X

Memory: 10958K

Always crashes after I see the Loading: (Insert Rom name here) Screen then the palm resets.

I used to get a fatal error but with the latest version of LJP I don't get an error anymore, it just resets the palm.

I'm not sure if this is useful or not but I have Palm OS 5.2.1

[NumbSkullMD]

UX50
ASM Works
No hack used
Launched with ZLauncher
I need about 9Mb free RAM in order to not get "Out of Memory" error at startup

[JeremySmith]

Zire72, 4998k Ram free without UDMH: Everytime I try to use the Asm core I get a black screen and then my palm locks up. Sorry this post isn't more helpful.

Zire72, 22843k Ram free with UDMH: Same problem as without UDMH.

[Reaper]

Hi

Asm works

- Using a Palm tungsten E
- UDMH hack used
- 20 mb free memory availble
- Stabilty: does not chrash on startup
- so far works great

[Genesis]

# Zire 72
# hack used UDMH
# Zlauncher v5.10.3
# Free Memory 12.2M
# I've got a black screen on all roms, have to do a soft reset.

Great work Yoyofr, i like a lot this new full screen option. Thanks!

[StuBee]

Palm: TC
hack used : PowerNet
special launcher: Facer (no memory features)
stability : SNES ASM: Black screen on startup

Thanks

[yoyofr]

I've a possible clue : after some research, it seems the devices with motorola or sony engine arm cpu are working.
devices with other arm cpu (TI OMAP & Intel XScale) do not work. I'll try to have the assembly sources compiled with prc-tools and armv5te architecture on next release.
does anyone here have knowledge in this area ?

right now everything is compiled with arm7(or arm9, don't remember now and cannot check before tonight) architecture.

[scott0968]

Actually it appears that the OMAP works. The T|E has an OMAP processor.

However it does appear that the Intel does not. The T5, T3, T|C, Treo 650, and the Zire 72 all use Intel processors.

[hackit98]

Even though it looks like the proc's might be to blame, here is my info anyway.

For TJ37:

• Device -> Has work on TJ37 since the very beginning, just couldn't see it at first
  
• UDMH -> Using UDMH does make any impact on the emulation. It runs great with or without it.
  
• Launcher -> I'm using ZLauncher.
  
• Free mem -> Approx 7.9MB before emulation. If I remember correctly, the PNO box that came up with the other version only dropped a meg or slightly more, I can load an old version to get the exact amount if you need it.
  
• Stability -> Never any problems with the emulation itself.
  

[Vitreal]

Procs and if that doesnt fix it. i have been looking intot he source code to see how input is handled. drawing a blank.

where have all the // gone.

hehehe. its very ahrd to read code

[Vitreal]

i dunno if your gonna do this or not or aleady have it planned. but ont he test deployments of the snes numbering 1,2,3 would help so we could know when tests are out and which test we are using

[yoyofr]

sure, it's quite a mess currently....
anyway, I think I've found the issue this morning.
Angel helped me to locate the issue : it's the asm version of the snes io functions.
I think the Intel proc does not have the same value in PC than other one.
here is an extract of the code :

.macro S9xGetWord
// in : rscratch (0x00hhmmll)
// out : rscratch (0xhhll0000)
#ifdef _C_GW_
STR regCycles,[regCPUvar,#Cycles_ofs]
PREPARE_C_CALL
BL asm_S9xGetWord
RESTORE_C_CALL
MOV R0, R0, LSL #16
LDR regCycles,[regCPUvar,#Cycles_ofs]
#else
STMFD R13!,{PC} //Push return address
B asmS9xGetWord
MOV R0,R0
MOV R0, R0, LSL #16
#endif
.endm

I think the STMFD R13!,{PC] causes the issue, I'll try
to delete it and replace the B asmS9xGetWord by a BL
it should fix the thing.
A testing build will probably be available when I'll be back home tonight or tomorrow.

after thinking a bit about it, it should not cause any issue. I'll build a more detailed debug version asap and do more testing...

[daigorobr]

Device: Tungsten E
Memory: 21,220,660 bytes free with 16,777,204 max chunk using UDMH
Launcher: ZLauncher

ROM/result:
Final Fanatsy 2 (U) - works great, no flaw at all
Mario RPG (U) - doesn't show anything but emulator keeps running

[yoyofr]

super mario rpg works using C cpu core and sound emulation on "SIM" or "ON". But it will be slow....
in fact it needs to emulate the standard cpu (65c816) of the snes, as other games, and a special chip (SA1) which us a second 65c816 cpu, clocked at doubled speed!