FS-UAE for OS X PPC (Not officially supported)

[tlosm]

my suspect was true ...
i had been try to configure the qemu-uae from frode and i had the same errors that come from Qemu sources

wer-mac-g5-di-luigi-burdo:qemu-uae-master luigiburdo$ ./configure
second level name optional in hw.optional.x86_64 is invalid

On qemu the way for start the configure of it is put on configure --cpu=PPC

i try and :

power-mac-g5-di-luigi-burdo:qemu-uae-master luigiburdo$ ./configure --cpu=PPC

ERROR: Unsupported CPU = PPC, try --enable-tcg-interpreter

Probably merge the qemu 2.1 will fix it ... who know ? ... lets try

@snakecoils
how you fixed the autoconf 2.67 ? do you had been installed it thru mac ports or with tiger homebrews ? i had been try to uprgrade it but on my side i have 2.61 and look it is the Xcode last version

[SnakeCoils]

@tlosm
Just checked, the autoconf installed by MacPorts is version 2.69, my qemu binaries were installed through MacPorts too and the package is version 2.1.0 but I can only see x86 guest emulations (i386 and x86_64) no ppc engine.
Do you regularly refresh your MacPorts environment through the usual "selfupdate" and "upgrade outdated" options?

[FrodeSolheim]

Tip: try --cpu=ppc instead of --cpu=PPC (but I don't think it is necessary to specify at all really, host CPU should be autodetected by qemu configure script)

[tlosm]

Originally Posted by SnakeCoils

@tlosm
Just checked, the autoconf installed by MacPorts is version 2.69, my qemu binaries were installed through MacPorts too and the package is version 2.1.0 but I can only see x86 guest emulations (i386 and x86_64) no ppc engine.
Do you regularly refresh your MacPorts environment through the usual "selfupdate" and "upgrade outdated" options?

i Did it every day ..
about Qemu you have to add "sudo port install qemu +target_ppc +target_arm +target_ppc64 and so and so and so :P

... i don't know why i try to start compile ask me autoconf 2.67 i have it already installed too but if i write "autoconf --version" the result is 2.61
... bho?

I forget : Roberto Innocenti (another amigos) made the G4 mini run qemu in KvM under linux ... you can understand what can be if one day Fs-uae PPC will be ported to linux Ppc ?:P:P
@frode
you true
but now it asking for the DTC (libfdt) and i don't know where hell is that
because the submodule command don't work git 2.1.2

[FrodeSolheim]

Originally Posted by tlosm

but now it asking for the DTC (libfdt) and i don't know where hell is that

Yes, but does it not also tell you about a simple command you can type to get it? (something about git submodule...)

[FrodeSolheim]

Also, I'm not sure why you are talking about the QEMU macport, you need this:
https://github.com/FrodeSolheim/qemu-uae
(and run ./configure-qemu-uae, not ./configure)

[tlosm]

Originally Posted by FrodeSolheim

Also, I'm not sure why you are talking about the QEMU macport, you need this:
https://github.com/FrodeSolheim/qemu-uae
(and run ./configure-qemu-uae, not ./configure)

Yes it asking for "git submodule update --init dtc"
but if i do i have :

"not a git repository"


Just for personal colture this is what can do a old G5 quad with Qemu :P

[ Show youtube player ]
Snake i think will be you next machine :P :P :P

[FrodeSolheim]

@SnakeCoils please try the latest git version of FS-UAE now
(for reference, the relevant commit is https://github.com/FrodeSolheim/fs-u...f37783bd78a754)

@tlosm My crystall ball (if it works this time) says that you did not clone the qemu-uae git repo with git (but downloaded a source tarball or something instead). make sure you have properly cloned qemu-uae and run the git submodule command from the qemu-uae root.

[tlosm]

Frode ... this make the difference from a Geek (you) and Me (Lamer) Your crystal ball work good :P

[SnakeCoils]

@Frode
Just downloaded and compiled but the build stopped with an error I never seen before:

	g++ -DHAVE_CONFIG_H -I.  -DMACOSX -DUAE -DFSUAE -I./flac/include -I./gen -I./src -I./src/jit -I./src/include -I./src/od-fs -I./src/od-fs/include -I./src/od-win32 -I./src/od-win32/caps -I./libfsemu/include -I./glee -I./manymouse  -I./src/slirp/include -I./libudis86/include  -I./libmpeg2/include  -I/opt/local/include/freetype2 -I/opt/local/include -I/opt/local/include/libpng16  -D_REENTRANT -I/opt/local/include/glib-2.0 -I/opt/local/lib/glib-2.0/include -I/opt/local/include  -DLUA_USE_POSIX -I/opt/local/Library/Frameworks/OpenAL.framework/Versions/A/Headers  -I/opt/local/include/libpng16  -I/Library/Frameworks/SDL2.framework/Versions/A/Headers -I/opt/local/include    -Wmissing-declarations  -fvisibility=hidden  -g -O2 -MT src/rommgr.o -MD -MP -MF $depbase.Tpo -c -o src/rommgr.o src/rommgr.cpp &&\
	mv -f $depbase.Tpo $depbase.Po
cc1plus: warning: command line option "-Wmissing-declarations" is valid for C/ObjC but not for C++
src/rommgr.cpp:314: error: integer constant is too large for ‘long’ type
src/rommgr.cpp:318: error: integer constant is too large for ‘long’ type
src/rommgr.cpp: In function ‘int decode_cloanto_rom_do(unsigned char*, int, int)’:
src/rommgr.cpp:892: warning: passing negative value ‘-0x00000000000000001’ for argument 3 to ‘romdata* checkromdata(const unsigned char*, int, uae_u32)’
src/rommgr.cpp: In function ‘romdata* getromdatabydata(unsigned char*, int)’:
src/rommgr.cpp:974: warning: passing negative value ‘-0x00000000000000001’ for argument 3 to ‘romdata* checkromdata(const unsigned char*, int, uae_u32)’
src/rommgr.cpp:977: warning: passing negative value ‘-0x00000000000000001’ for argument 3 to ‘romdata* checkromdata(const unsigned char*, int, uae_u32)’
make[3]: *** [src/rommgr.o] Error 1
make[2]: *** [all-recursive] Error 1
make[1]: *** [all] Error 2
make: *** [bundle] Error 2

[FrodeSolheim]

Apply this patch (two lines changed):

diff --git a/fs-uae/src/rommgr.cpp b/fs-uae/src/rommgr.cpp
index 01cdf89..bdbe462 100644
--- a/fs-uae/src/rommgr.cpp
+++ b/fs-uae/src/rommgr.cpp
@@ -311,11 +311,11 @@ static struct romdata roms[] = {
 	ALTROMPN(104, 1, 1, 32768, ROMTYPE_ODD  | ROMTYPE_8BIT, NULL, 0x888da4cf, 0x6ae85f3a, 0x65331ba4, 0xaaba67ae, 0x34763d70, 0x2bde0495)
 	ALTROMPN(104, 1, 2, 32768, ROMTYPE_EVEN | ROMTYPE_8BIT, NULL, 0xaf1f47db, 0x28d5bed0, 0xbc517d46, 0x500e8159, 0x723e0b64, 0x4733c26a)
 	{ _T("A2620/A2630 -07 ROM"), 0, 0, 0, 0, _T("A2620\0A2630\0"), 65536, 105, 0, 0, ROMTYPE_CPUBOARD, 0, 0, _T("390282-07/390283-07"),
-	0x169d80e9, 0x41f518cb,0x41c1dc1f,0xcc6363832,0x0676af5,0x4969010c, NULL, NULL },
+	0x169d80e9, 0x41f518cb,0x41c1dc1f,0xcc636383,0x20676af5,0x4969010c, NULL, NULL },
 	ALTROMPN(105, 1, 1, 32768, ROMTYPE_ODD  | ROMTYPE_8BIT, _T("390282-07"), 0xf2904058, 0x33695119, 0x5fdf5d56, 0x095a696b, 0x0ba2641d, 0x334845df)
 	ALTROMPN(105, 1, 2, 32768, ROMTYPE_EVEN | ROMTYPE_8BIT, _T("390283-07"), 0xf697d458, 0x09fe260b, 0x03784e87, 0x3351dbec, 0x5146a455, 0x814383d1)
 	{ _T("A2620/A2630 -06 ROM"), 0, 0, 0, 0, _T("A2620\0A2630\0"), 65536, 106, 0, 0, ROMTYPE_CPUBOARD, 0, 0, _T("390282-06/390283-06"),
-	0xeb31fd9e, 0x2d6a5c68,0x1040f98d,0x7e63ad089,0x0da9e83,0x2b5c704d, NULL, NULL },
+	0xeb31fd9e, 0x2d6a5c68,0x1040f98d,0x7e63ad08,0x90da9e83,0x2b5c704d, NULL, NULL },
 	ALTROMPN(106, 1, 1, 32768, ROMTYPE_ODD  | ROMTYPE_8BIT, _T("390282-06"), 0xd6ae582c, 0x47b3dea3, 0x31db76e6, 0x1380a3d6, 0x9f191657, 0xdd1cd4b3)
 	ALTROMPN(106, 1, 2, 32768, ROMTYPE_EVEN | ROMTYPE_8BIT, _T("390283-06"), 0xcd379634, 0x65e251e2, 0xf6961c8e, 0x33a86c3d, 0x01248f70, 0xa159823b)

(An error with the checksums of two ROMs. Will have to check with Toni if the fixed checksums are correct, but it will nevertheless compile for you)

[tlosm]

@frode in compiling Qemu-uae

-c -o qemu-log.o qemu-log.c
qemu-log.c:24:21: error: uae/log.h: No such file or directory
qemu-log.c: In function ‘qemu_log’:
qemu-log.c:38: warning: implicit declaration of function ‘UAE_LOG_VA_ARGS’
qemu-log.c:38: warning: nested extern declaration of ‘UAE_LOG_VA_ARGS’
make: *** [qemu-log.o] Error 1

[FrodeSolheim]

@tlosm you need to have fs-uae *alongside* qemu-uae (i.e. projects/fs-uae, project/qemu-uae). See here: http://eab.abime.net/showpost.php?p=976709&postcount=26 and here: http://fs-uae.net/plugins if you manage to get it compiled.

(If you now use the latest git version of qemu-uae, you also need the latest git version of fs-uae to go with it...)

[tlosm]

Ok Frode ... Lets hope
but for sure SnakeCoils will do better than me :P

[SnakeCoils]

It looks like things with AIBB are MUCH better that before (see attached screenshot), the system seems now a little more slower, I have to check this with other programs.

Thank you again Frode! Great work :-)

The patched FPU version is uploaded in The Zone! and ready for testing for everyone. Enjoy!

[SnakeCoils]

Originally Posted by tlosm

Snake i think will be you next machine :P :P :P

OT Mode: ON

I had two different G5 machines, one dual 2 GHz CPU "early 2005" and later one 2.3 GHz "late 2005" that died some years ago.
They were nice number-cruncher and great looking computers but when I decided to reduce my collection I preferred the versatilty over other aspect so I choosed the MDD (first version, not 2003, because it is less picky about RAM modules) and the MacPro 2009 (not the previous versions because of the slow and obsolete architecture). The purpose was to cover all main MacOSes from 9.2.2 to 10.5.8 for PPC and everything from 10.6.8 to actual 10.9.5 for Intel.
They are side-by-side on my desktop, sharing the same keyboard, CRT display and mouse (I was lucky enough to found the right Aluminium keybord that is recognized by both machines including the special keys) and they help each other in many tasks that alone could not complete, a real dream team :-)
I am not nostalgic about the G5, the MacPro has almost the same look and it is much less energy hungry but the PowerPC is (and will be forever) in my hearth, the only serious weak point about G5 is that it does not natively support OS9 (Classic mode is not low-level enough to handle all OS9 applications), for this reason I have choosen the MDD tuning it almost to the max to reduce the performance gap with the G5 machines.
However your benchmarks on the Quad are impressive, on that machine FS-UAE can really fly! :-)

OT Mode: OFF

[tlosm]

ot mode on

Mate i dont know im using classic on Tiger 10.4.11 and is working good ... some incompatiblity with virtualpc or x86 emulator but for the others software i had been tested are working good. i had been installed the 3dfx wrapper and is really fast there.

ot mode off
now is time to check the fpu results with aibb :P

[tlosm]

Confirm Snake the performances are about 75% more slower.
Right now i have the same performance on G5 like euae not jit on my pegasos II G4 1 ghz.

[FrodeSolheim]

Originally Posted by SnakeCoils

the system seems now a little more slower, I have to check this with other programs.

Do you mean that this version of FS-UAE is slightly slower than the previous dev version? This is possible. Another change included in the git version is using waiting blits instead of immediate blits (and some other changes to how the "fastest possible" Amiga models are configured). It is possible these changes has affected the performance to some degree.

You can test if this matters by reverting these changes via config:

uae_immediate_blits = true
uae_waiting_blits = false

[SnakeCoils]

You can test if this matters by reverting these changes via config

It seems to not make any effect, the speed hit is still here. The big difference, for what I can see on my machine, is not much related to pure computational power, the most demanding demos are running more or less the same as before: where the speed loss is noticeable is in the Workbench environment while browsing the virtual HDD drawers, the icons appears slowly if compared to previous release. My virtual HD is a dedicated drawer in the /Documents/FS-UAE/Hard Drives/ where all files can be seen and manipulated from Mac side too. Just for completeness this is my A4000 / Workbench 3.1 configuration:

[config]
amiga_model = A4000/040
fast_memory = 8192
accuracy = -1
uae_immediate_blits = true
uae_waiting_blits = false
floppy_drive_volume = 0
video_format = bgra
video_sync = off
fullscreen = 1
viewport = * * * * => 0 0 752 574
hard_drive_0 = A4000_HD/

No hacks are loaded on the Amiga side, it is just a flat, fresh install from Workbench 3.1 disk set.