e-uaes from os4depot, fresh recompile (post #10)

@Samo
Hm! But works on me, and i am on beta.. What newlib.library version on you setup ?

@kas1e

The one included in Final Edition Update 1 for Sam440 -->

newlib.library 53.30 (12.02.2014)

@kas1e

tested on my SAM460 and works fine.

BTW, can't remember, but if I enable JIT, shouldn't it show on bottom-leds 'JIT'?


-UAE 1.0.0
Build date: Nov 22 2018 23:26:52
SDLGFX: Initialized.
SDLGFX: Using SDL version 1.2.15.
SDLGFX: Display is 32 bits deep.
SDLGFX: Found screenmode: 1280x1024.
SDLGFX: Found screenmode: 1024x768.
SDLGFX: Found screenmode: 800x600.
SDLGFX: Found screenmode: 640x480.
Found 0 joystick(s)
Found amiga raw keyboard mapping
Using raw keymap
Opening cfgfile '.uaerc'...okay.
unknown config entry: 'amiga.screen_type=public'
unknown config entry: 'amiga.publicscreen='
unknown config entry: 'amiga.use_dither=false'
EClock frequency:1155.000010 MHz
Found amiga raw keyboard mapping
Using raw keymap
Rexx port "UAE" installed.
Found amiga raw keyboard mapping
Using raw keymap
SDLGFX: Added P96 mode: 640x480x32
SDLGFX: Added P96 mode: 640x480x8
SDLGFX: Added P96 mode: 800x600x32
SDLGFX: Added P96 mode: 800x600x8
SDLGFX: Added P96 mode: 1024x768x32
SDLGFX: Added P96 mode: 1024x768x8
SDLGFX: Added P96 mode: 1280x1024x32
SDLGFX: Added P96 mode: 1280x1024x8
Enabling argb32 byte-swapping for P96.
Building CPU table for configuration: 68000 24-bit addressing
1866 CPU functions
Building CPU function table, 44736 opcodes (0 0 1).
JIT: Building compiler function table.
JIT: Allocation of translation cache...
JIT: Selected processor type has no cache, leaving
reset at 0
JIT: Compiling reset
PAL mode, 50Hz (h=227 v=312)
chipmem cleared
Illegal instruction: 4e7b at 00fc0564 -> 00fc0582
SERIAL: period=372, baud=9600, hsyncs=16 PC=fc2234
PAL mode, 56Hz (h=227 v=275)
PAL mode, 50Hz (h=227 v=312)

@jabirulo
You probably on beta ? What your newlib.library ? As for leds: there special option "show_leds=true" , if you set it and didnt have button "jit" at bottom of the right part of screen then imho jit didnt enabled for you , it enables by that block:

@kas1e

Yes, you're right I'm on beta.
I tought I had already such options enabled, will check it. THX

Retested with public AOS4.1FE upd1 and got a GR too:
Crash log for task "uae"
Generated by GrimReaper 53.19
Crash occured in module uae at address 0x7F8AC1BC
Type of crash: DSI (Data Storage Interrupt) exception
Alert number: 0x80000003

Register dump:
GPR (General Purpose Registers):
0: 7F8AC194 5CEB8B20 00000000 5D810950 00000000 00000001 00000000 1CFE7380
8: 00000001 00000000 5D81096F 0181A0A4 26842422 5F0AF780 618B4AC0 5D58A340
16: 7F888500 00000000 5D5D19D0 5CECB1B8 02340000 02340000 00000000 0000003F
24: 5D887DF0 5CECB4F8 5D810950 00000000 5BB6C840 5D020000 5CECB1D0 00000000


FPR (Floating Point Registers, NaN = Not a Number):
0: nan 2.2 0.45455 1.80454e+267
4: -1.54858e+297 517 840 0
8: 0 74 4.5036e+15 4.5036e+15
12: 66 65536 -1.00582e+16 -5.18821e+34
16: -8.44502e+286 -6.84646e+303 1.15632e+140 -3.12957e-82
20: 1.16567e+304 3.07933e+227 -5.65167e+227 -4.61495e+227
24: -2.8935e-24 -6.01285e-51 -9.87375e+298 -2.07549e+294
28: 6.22079e+219 4.35442e+305 -8.14839e-10 -7.01169e+144

FPSCR (Floating Point Status and Control Register): 0x82004000


SPRs (Special Purpose Registers):
Machine State (msr) : 0x0002F030
Condition (cr) : 0x5D5DC37C
Instruction Pointer (ip) : 0x7F8AC1BC
Xtended Exception (xer) : 0x5CE4C018
Count (ctr) : 0x6FF714D0
Link (lr) : 0x7FC16900
DSI Status (dsisr) : 0x5CE4BDD0
Data Address (dar) : 0x01A7DD08



680x0 emulated registers:
DATA: 6182F91E 00000003 00000000 00000000 00000000 00000000 00000000 00000000
ADDR: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5CEB8A90
FPU0: 0 0 0 0
FPU4: 0 0 0 0



Symbol info:
Instruction pointer 0x7F8AC1BC belongs to module "uae" (HUNK/Kickstart)

Stack trace:
module uae at 0x7F8AC1BC (section 0 @ 0x346198)
module uae at 0x7F8AC194 (section 0 @ 0x346170)
module uae at 0x7F8AD06C (section 0 @ 0x347048)
module uae at 0x7F8ABAA0 (section 0 @ 0x345A7C)
module uae at 0x7F896150 (section 0 @ 0x33012C)
module uae at 0x7F896274 (section 0 @ 0x330250)
module uae at 0x7F5675FC (section 0 @ 0x15D8)
module uae at 0x7F88862C (section 0 @ 0x322608)
native kernel module newlib.library.kmod+0x000020ac
native kernel module newlib.library.kmod+0x00002d14
native kernel module newlib.library.kmod+0x00002ef0
uae:_start()+0x170 (section 1 @ 0x16C)
native kernel module dos.library.kmod53175+0x00027914
native kernel module kernel+0x0003bc38
native kernel module kernel+0x0003bc80

PPC disassembly:
7f8ac1b4: 913a001c stw r9,28(r26)
7f8ac1b8: 83fdc80c lwz r31,-14324(r29)
*7f8ac1bc: 813f005c lwz r9,92(r31)
7f8ac1c0: 7fe3fb78 mr r3,r31
7f8ac1c4: 83df040c lwz r30,1036(r31)

System information:

CPU
Model: AMCC PPC460EX V1.2
CPU speed: 1155 MHz
FSB speed: 231 MHz
Extensions:
...

@Jabirulo
Right, it crashes because on beta we have newer newlib.library which had statvfs() implemented. So to not wait when another OS update will come up, i can just build euae without statvfs() usage (as probably mickjt do before when build versions which on os4depot).

@all
Is it make any sense for anyone ? Imho having latest SDL be build with , and have alt+enter for switch between fullscreen/window worth it. At least for me is :) i can also add Salass00 fix for >2gb support, if anyone will test it before, of course :) maybe in meantime Capehill will add (?:) ) iconify support to sdl1 :)

It can be also good idea to switch from sdl1 to sdl2 in euae.

@kas1e

Without _ARCH_PWR4 defined (which also gets defined by -mcpu=970), then the JIT is very slow on X1000.

See http://euaejit.blogspot.com/2012/10/jinxed-it-more-apples.html and http://euaejit.blogspot.com/2014/07/p ... dreds-and-xthousands.html

Note that -mno-powerpc64 will be needed if you're going to use -mcpu=970, otherwise you'll get weird stuff like the mouse movements being inverted (see this link).

As for the stuff I said in the e-mail about -maltivec, it just causes strange behaviour if linked with the current binutils. Not necessarily crashes, but just incorrect code execution. Using -N when linking fixes it, or manually using amigaos.x from an older binutils (i.e the one included in SDK 53.30) with -Wl,-T /path/to/file.x

> And you probably build and generic version and x1000 one before with that flag ?

That flag/definition is the difference between an X1000 and non-X1000 version, so there's no such thing as building a "generic" version with that flag. I probably confused you by talking about altivec flags in the email.

There's no altivec code in e-uae, but I think the -maltivec flag was possibly used in Rachy's X1000 builds regardless, and I think it does have some minor effect even on code that has no specific altivec instructions, so I thought I'd let you know about any problems that might arise from using that.

@samo
Can you run plz that binary on your machine (that one with disabled statvfs() stuff): http://kas1e.mikendezign.com/aos4/uae ... _ppcjit_2018_11_24_v3.zip

Just run it the same from ram:, without any config, to see if it stop crashing

@Mick
About altivec problems: did you compare those 2 linker scripts, to see what the differences is, and so to report bug?

@kas1e

No crash

And this is the output:

Quote:

@samo79
You may try to check it with configs now , and alt+enter should works too

@kas1e

https://github.com/AmigaPorts/SDL/tree/1.2-iconification

Not much tested, just testwm run some times.

@kas1e

Yeah i put this exe instead of the one of RunInUAE and no problem for now, alt+enter seems working fine too

However amigfx exe runs faster on my low end machine so for now i will stick with it

@Capehill
Yeah !
(press to open image in new tab for fullscreen)



It also works ! I can iconify, and it puts where should, and uniconify back. Thorugh there is little issue as it was with sdl2 at begining: icon didn't takes. I.e. it take some default os4 icon, but not the app's name icon.

But in others all works ! Probably adding the code as in sdl2 to take the app's icon, and then we can pack it all, wrote readme, and release it ? Also can be released as normal and as debug versions as well.

@samo79
Yeah, on slow machines overhead of SDL on top probabaly can be noticed.

@all
What is intersting, is that probabaly i find some bug in SDK (or around). I.e. in newlib.library support of statvfs() function was added in 53.23 version of newlib.library. But it definately not works how it should even on 53.30. But works on beta, so something was fixed in that regards since newlib.library v53.30.

But, when one compile code with SDK 53.30 which use statvfs(), then he have warning, that newlib.library 53.23 is must and minimum (i.e. when statvfs() was added), but then, euae's code based on statvfs() du not works even on 53.30 , but works on later ones. Probabaly that need to be changed in the SDK, to say not "53.23" minimum, but whatever version are where it all start to works as expected.

@kas1e

Cool for iconify :)
As you here can you complete the full version name in titlebar, aka something like "E-UAE 1.0.0 (xxx.xx.2018)"
Just "E-UAE" text is not enough having 3 different versions in test so have a more visible version number there will help when in window mode

@kas1e

I've compared them using diff, but I don't really know what I'm looking at to make a proper bug report. I don't know which change it is. All I can say is it's one of them, and my best guess is the change from rev 335 of adtools on sourceforge.

I'm not the best person to make a bug report about this. I have no altivec enabled machine, and therefore no way of testing binaries myself to create an easily reproducible test case, nor do I know whether it's only specific instructions that cause a problem. I've tracked it down to the linker scripts (internal & external) in binutils 2.23.2, but to find the exact instruction(s), and a proper test case, is too tedious for me and annoying for someone else if I have to send them countless binaries to test for me. Not to mention I don't think I'd even know how to create a proper test case. I don't know assembly, or even C for that matter. Best someone else debugs it :)

This is a bit off-topic now and doesn't relate to e-uae. As mentioned earlier, e-uae has no altivec specific code and so -maltivec is not required, but I only wanted to mention any problem that might arise if it was used. I think there is an effect when using that flag even when compiling code that has no specific instructions for it.

@samo
Quote:

Something like that ? (open in new tab for full screen):

@kas1e

Tested your v2 versión in my X5000, works fine. And also faster, but that can be a psychological issue.

Virginio

@RIBDEVIL
In theory it can be faster, as new SDL, new 8.2.0 compiler, -O3 optimisation, in theory. But even if, it will be faster not on emulation level of course, just general stuff around emulation (windowses, keys, inputs, threads, etc).

@samo79
Or maybe like that is better (as usual RMB in new tab for full size):

@kas1e

Branch updated.

By the way, is scaling in those images done by CPU or OpenGL?