|
Post by yoyofr on Jan 29, 2005 21:10:59 GMT 1
Here is the email I sent to arab, since I believe it can help others too, in order to send useful file :
|
|
|
Post by yoyofr on Jan 29, 2005 21:13:13 GMT 1
btw, angel, did you got a crash or did you reset by yourself ?
|
|
|
Post by angel on Jan 29, 2005 21:37:15 GMT 1
I reseted it myself.
actually my palm run since 20 minutes and not "crashed" with the "fatal exception" thing, but i cant make any key working or light it off.
It doesn't "light off" by itself also...
Edit : about 50 minutes with still freeze and no "official" crash.
|
|
|
Post by angel on Jan 29, 2005 22:44:04 GMT 1
|
|
|
Post by StuBee on Jan 30, 2005 1:58:31 GMT 1
I don't believe my TC crashes with any SNES ASM version. It just hangs at the black screen. I tried using the test lem and I just get a hang forever (black screen) as well.
It never gets a chance to display any image after the rom has loaded (with or without the test ASM lem)
|
|
arab
New Member
Posts: 10
|
Post by arab on Jan 30, 2005 9:19:59 GMT 1
tried both roms for one hour each, nothing happened. only difference is that DOOM actually loaded into first screen after about 20 mins.
|
|
|
Post by yoyofr on Jan 30, 2005 12:30:37 GMT 1
sorry for the trouble building a debug version, it's always hard when you dont have the real device and when the bug is so "device dependant". I've build a debug4 version, which will show stuff on screen, instead of writing to card. Moreover it should be a little bit faster to process debug data. please try it and report what you see if you can get a crash (fatal error is fine) to occur. yoyofr.fr.st/ljp/files/ljp-snes-debug4.zip
|
|
arab
New Member
Posts: 10
|
Post by arab on Jan 30, 2005 13:54:45 GMT 1
I think it's not working. Shows nothing on the screen (black screen) after "loading rom". I've tried it for more than half an hour.
|
|
|
Post by yoyofr on Jan 30, 2005 14:08:24 GMT 1
it's really strange... and did you tried on more than one rom ? does the 'C cpu core' mode work ? btw, what's your device already ?
|
|
arab
New Member
Posts: 10
|
Post by arab on Jan 30, 2005 14:14:30 GMT 1
tested 3 roms device is zire 72, mentioned it in the 1st post works fine in C-mode
|
|
|
Post by yoyofr on Jan 30, 2005 16:16:33 GMT 1
ok ,thx for the infos there's a new test version, with a new way of handling the problematic code section : yoyofr.fr.st/ljp/files/ljp-snes-debug5.zipNothing special is done (no output on screen neither in a file), so basically there's only 2 possibilites : it works/doesn't. BTW, if someone can do some testing now and have msn, you can join me at ymagnien@hotmail.com
|
|
j5350
New Member
Posts: 13
|
Post by j5350 on Jan 30, 2005 16:40:31 GMT 1
I've got a Treo 650. When I try it, it just shows a image on screen with Out of memory ...
|
|
Vanya
New Member
Posts: 3
|
Post by Vanya on Jan 30, 2005 16:54:46 GMT 1
Hi. I've tested the latest beta5 on my Palm T3, and i could say - it works fine. At last you made it ! The ASM version works faster (55 fps VS 48 fps). So, could you say - is there any way to achieve more speed ?
|
|
|
Post by phoinix on Jan 30, 2005 17:37:52 GMT 1
Work on my Zire72, but only with UDMH and without any speed increase (E.V.O.).
|
|
|
Post by yoyofr on Jan 30, 2005 18:05:07 GMT 1
|
|
Vanya
New Member
Posts: 3
|
Post by Vanya on Jan 30, 2005 19:33:13 GMT 1
Hello. Here is my result for beta6. As beta5 this one works well. About speed: Game: Aladdin Palm T3 (overclocking - none). Level2 - sound OFF ASM - 58 fps C - 49 fps Level1 - sound ON (44khz) ASM - 38 fps C - 35 fps. The results of Beta5 and Beta6 are equal ! So i haven't found any speed improvement. Any way - good work. Thanks. (I am ready to test for new version :-)
|
|
|
Post by angel on Jan 30, 2005 19:35:38 GMT 1
Test on T|T5 Debug 4 : Loading secret of mana => Fatal exception Debug 5 : Loading secret of mana => the game launch and play normally, a bit faster than usual i would say. System crash at the second attempt to pop up the gui LJP menu - 50 fps (no resize, no smooth...) Debug 6 : Loading secret of mana => the game start, seem a bit little slower than on debug 5 but nothing sure... System crash at the second attempt to pop up the gui LJP menu - 50 fps (no resize, no smooth, nothing ) Note : i have the exact same results (fps) with C core using the Januray 27, 2005 beta 6b version
|
|
|
Post by yoyofr on Jan 30, 2005 20:02:16 GMT 1
Last build, last test : all asm functions activated (but speed's differences between debug6 & this one should be low). yoyofr.fr.st/ljp/files/ljp-snes-debug7.zip On fast devices, C & ASM cpu cores give mostly the same result, since the cpu core execution time is a LOT lower then gfx core & sound core. So it's normal!
|
|
Vanya
New Member
Posts: 3
|
Post by Vanya on Jan 30, 2005 20:22:50 GMT 1
I have tested beta7. I am not sure, but speed inceased for 1 fps (in my case). Any way, by using 1-2 frameskip, almost every game is playable on Palm T3!
So, what are you going to do next with your emulator ?
|
|
|
Post by angel on Jan 30, 2005 21:05:59 GMT 1
T|T5
7th test :
Run well, but when i try to do things in LJP gui (open the gui 2 times makes the palm crash.
49 fps stable (the olders versions was running from 46 to 50, mostly 50)
The Gui is very unstable in asm mode since the test 5 i should say.
|
|
arab
New Member
Posts: 10
|
Post by arab on Jan 30, 2005 21:40:36 GMT 1
Just to confirm that debug7 works fine on Zire72 I have same problems angel has + some issues with save option.
But I guess the main aim was achived - ASM works. Well Done. Congrats.
|
|
|
Post by StuBee on Jan 30, 2005 21:45:32 GMT 1
Test Lem Debug 7
On the TC any use of the D-pad still results in a drop of 50% in speed/frame, which makes the games kinda unplayable (14 frames/sec)
This is true both ASM and C core.
Is anyone else experiencing this?
|
|
|
Post by yoyofr on Jan 30, 2005 22:06:41 GMT 1
I'm aware of this, but it seems to be related to the os! I didn't find a way to block the key event to be generated, so every button pressed send an event to the os. If this event isn't handled then the application will crash after a short time. On zodiac the keymask function is properly handled, but it's not the case on most devices. I have to find a way to solve the issue, but right now I've no solution.
|
|
|
Post by JeremySmith on Jan 31, 2005 1:52:52 GMT 1
Beta 7, Zire72, without UDMH: Asm core works great. There is no slowdown when a dialog box comes up on FFII. ;D The only problem I have is if a tap the screen (to hit a virtual button or the menu) It gives me a fatal exception. Sometimes with smaller roms I can hit the virtual buttons and sometimes I get the error MemoryManager.c, line 3632 Null Handle
|
|
|
Post by Vitreal on Jan 31, 2005 7:43:56 GMT 1
Okay here are my thoughts on the Device.
The issue is not how the CPU core handles I/O functions but its related to LJP itself.
Whenever a key is clicked. there is a sound. and an action takes place. Seeing how your keys are managed. They take precedence over the OS way the key is handled. SO there is going to be a conflict. The function isnt activated yet the OS asks the question bla bla bla confusion.
Now i have a theory. Just a theory. BUT if you let the OS handle the keys and have the program process the input such as if the navigator is pressed up then you use this event generated by the OS to produce the results instead of overwriting the results produced by the OS.
im saying try to take advantage of the natural Keymap instead of overriding the keymap. it might just work.
|
|
|
Post by yoyofr on Jan 31, 2005 11:44:23 GMT 1
no, it's definitely of problem with palmos. you cannot send the keydown event to the os, or you'll have a big impact on speed (more than what you can experience now). here is my event handler :
I already bypass the 68k land by using yahm & armlib, so the speed is faster. The SysGetEvent & SysEventAvail are the same that the usual EvtGetEvent & EvtEventAvail. The issue is some devices do not implement the standard palmos keysetmask function, which is used to tell the os not to generate event for the buttons corresponding to the mask sent as parameter. So perhaps I need to handle more event in my handler in order to prevent them from being processed by the system handler, but I have to find more docs then...
|
|
|
Post by Vitreal on Jan 31, 2005 13:14:09 GMT 1
well d**n.
such a mess then.
|
|
|
Post by Vitreal on Jan 31, 2005 13:17:23 GMT 1
|
|
|
Post by yoyofr on Jan 31, 2005 13:45:34 GMT 1
I already searched a lot of palm-dev forum (and pno too), but there's no answer to the problem. Moreover, sources of frodo & phoenix show the same way of handling events as I already do.... I'll try to find another way to speed up things, but I'm afraid it's system limitation...
|
|
|
Post by haoyong on Jan 31, 2005 19:30:26 GMT 1
Running Debug 7 with Beta 6b on my T|E. 11.1 mb of memory free, using UDMH. Tested with Chrono Trigger.
Emulation speed seems up, and emulation seems more smooth. Only drawback is fatal errors that keep popping up when I try and access the GUI. This is the error,
"Application 'Little John for Palm OS' has requested a 409624-byte non-movable chunk from the dynamic heap, and the system is unable to satisfy that request." And the reset button pops up.
Hope this helps.
|
|