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

[SnakeCoils]

Originally Posted by eastone

Welcome back SnakeCoils and thank you for the new version
This 2.9.8 is a bit faster than 2.9.7 and of course very stable on my G5 2.3

Glad to see it worked well also on G5 machines, on my G4 I din't noticed a big speed up but all the previous programs works as usual, it also a good thing it can still be compiled for PPC hosts because some recent changes in source indicates the minimum OSX should be 10.9 so I think at a certain point the changes will became so deep that a build under 10.5 would not be more possible. Let's see what is going to happen with incoming updates :-)

[FrodeSolheim]

Configure with

--disable-x86

as well (disables bridgeboard emulation)

[SnakeCoils]

Originally Posted by FrodeSolheim

Configure with

--disable-x86

as well (disables bridgeboard emulation)

Thank you very much Frode, however declaring

./configure --disable-jit --disable-x86

was not sufficient to avoid a break in the compile, this is what happened:

g++-mp-7 -DHAVE_CONFIG_H -I.  -DMACOSX -DMACOS -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./manymouse -I./glad/include   -I./libudis86/include  -I./prowizard/include   -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/include/libpng16 -I/opt/local/include -D_THREAD_SAFE -I/opt/local/include/SDL2 -I/opt/local/include  -fno-strict-overflow -fno-strict-aliasing -Wmissing-declarations -Waddress -fvisibility=hidden -I/opt/local/include/mpeg2dec -fpermissive -O3 -maltivec -mabi=altivec -MT src/libuae_a-x86.o -MD -MP -MF src/.deps/libuae_a-x86.Tpo -c -o src/libuae_a-x86.o `test -f 'src/x86.cpp' || echo './'`src/x86.cpp
In file included from src/x86.cpp:56:0:
src/pcem/codegen.h:12:2: error: #error Dynamic recompiler not implemented on your platform
 #error Dynamic recompiler not implemented on your platform
  ^~~~~
src/pcem/codegen.h: In function 'void addbyte(uint8_t)':
src/pcem/codegen.h:317:26: error: 'BLOCK_MAX' was not declared in this scope
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
src/pcem/codegen.h:317:26: note: suggested alternative: 'LONG_MAX'
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
                          LONG_MAX
src/pcem/codegen.h: In function 'void addword(uint16_t)':
src/pcem/codegen.h:327:26: error: 'BLOCK_MAX' was not declared in this scope
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
src/pcem/codegen.h:327:26: note: suggested alternative: 'LONG_MAX'
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
                          LONG_MAX
src/pcem/codegen.h: In function 'void addlong(uint32_t)':
src/pcem/codegen.h:337:26: error: 'BLOCK_MAX' was not declared in this scope
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
src/pcem/codegen.h:337:26: note: suggested alternative: 'LONG_MAX'
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
                          LONG_MAX
src/pcem/codegen.h: In function 'void addquad(uint64_t)':
src/pcem/codegen.h:347:26: error: 'BLOCK_MAX' was not declared in this scope
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
src/pcem/codegen.h:347:26: note: suggested alternative: 'LONG_MAX'
         if (block_pos >= BLOCK_MAX)
                          ^~~~~~~~~
                          LONG_MAX

With a bit of luck, since I am not familiar with C programming at all, I was able to overcome this error editing the src/pcem/codegen.h part replacing this line:

#error Dynamic recompiler not implemented on your platform

with this other:

#define BLOCK_MAX 0

additionally the Tar command on OSX ppc seems to dislike the "J" option so in the /dist/macos/Makefile.in I have simply modified the argument in "Jcfv" in "zcfv" in this way:

    tar zcfv ../../$(dname).tar.xz FS-UAE.app

With these modification I was able to compile the actual 4210 branch and now I am testing it extensively. From a first impression it is somewhat slower than latest 2.9.8 branch and Picasso96 does not work (but this was expected). I am going to publish in the Zone now but I stronlgy suggest the other OSX PPC users to not throw away the previous binaries because this is only a WIP release.

[FrodeSolheim]

A slighly better fix is to move the WITH_X86 ifdef in src/x86.cpp further up so that src/pcem/codegen.h is not included at all; it can be moved up to below #include "sysdeps.h" for example:

#include "sysconfig.h"
#include "sysdeps.h"

#ifdef WITH_X86
...

(Will move that up there in the git repo soon, as well).

As for the tar command, using z instead of J, which is fine, produces a tar.gz file instead of tar.xz, so you want to change the archive extension as well ;-) :

tar zcfv ../../$(dname).tar.gz FS-UAE.app

(Yes, I expect FS-UAE to run a bit slower. At least the display rendering code is running a bit slower, I had to remove some optimizations I did in FS-UAE earlier, so for now, UAE re-renders the screen every frame even when nothing changes, but at least it displays correctly ).

[BastyCDGS]

Originally Posted by FrodeSolheim

(Yes, I expect FS-UAE to run a bit slower. At least the display rendering code is running a bit slower, I had to remove some optimizations I did in FS-UAE earlier, so for now, UAE re-renders the screen every frame even when nothing changes, but at least it displays correctly ).

Regardings display optimizations, you (and perhaps Toni, too) could consider my FFmpeg IFF-ILBM decoding routines, which can be found here:
https://github.com/FFmpeg/FFmpeg/blo...bavcodec/iff.c

They use quite large lookup tables, but are blazing fast.
Even HAM decoding doesn't need conditional branches in the inner main decoding loop.
However, it hasn't the HAM6-AGA fix (mul color 0x11 instead). Maybe worth a try!

[SnakeCoils]

Originally Posted by FrodeSolheim

A slighly better fix is to move the WITH_X86 ifdef in src/x86.cpp further up so that src/pcem/codegen.h is not included at all; it can be moved up to below #include "sysdeps.h"

Much more elegant than my quick and dirty fix :-)

Originally Posted by FrodeSolheim

As for the tar command, using z instead of J, which is fine, produces a tar.gz file instead of tar.xz, so you want to change the archive extension as well ;-)

Yes, philologically is correct, I admit I didn't mind about the extension since it was working as it was. Before made further PPC builds I will wait for the next commits to 4210 branch.

[SnakeCoils]

Originally Posted by BastyCDGS

They use quite large lookup tables, but are blazing fast.
Even HAM decoding doesn't need conditional branches in the inner main decoding loop.
However, it hasn't the HAM6-AGA fix (mul color 0x11 instead). Maybe worth a try!

Any speedup in graphics subsytem would be a godsend for PowerMacs since it is where FS-UAE show its weak point against E-UAE-JIT.
A good example of this can be seen in the Nexus 7 AGA demo in the part of the morphing face (pure HAM8 rendering) where everything chokes also on faster G5 machines.

[SnakeCoils]

The very latest OSX PPC binary of 4210 branch is online and ready to download in The Zone. Picasso 96 works again and overall seems a bit faster than previous build.
Compile went smooth and without errors, a truly good job Frode, thank you VERY VERY much! :-)

[eastone]

Originally Posted by SnakeCoils

The very latest OSX PPC binary of 4210 branch is online and ready to download in The Zone.

Unfortunately this version is ~20% slower than previous one on G5.

[SnakeCoils]

Originally Posted by eastone

Unfortunately this version is ~20% slower than previous one on G5.

This is the reason because I warned to not trash the previous FS-UAE binaries, at present the code (according to what Frode wrote some post above) still lack optimizations, maybe in the final version the emulator will run much smoother but do not expect a dramatic boost in performance: any advance in accuracy means more needed computation power.

[eastone]

@SnakeCoils
No problem
It was just info

[Methanoid]

With a 1.25 Ghz G4 Mini which version is gonna give me best features with 100% Amiga speed??

Sent from my Redmi Note 7 using Tapatalk

[SnakeCoils]

Originally Posted by Methanoid

With a 1.25 Ghz G4 Mini which version is gonna give me best features with 100% Amiga speed??

FS-UAE has the accuracy and versatility as main goals, therefore the requirements for a nice experience are a bit high even for the old versions.
The fastest and reasonabily complete version of FS-UAE for PowerMacs you can execute is the 2.4.1 and on a MacMini you can run the most part of games and demos without major slowdowns (declaring the parameter Accuracy=-1), what really could make the difference would be a Level 3 cache on the CPU but it was not present in the processors of MacMini product line.
If you are going to accept some (many) compromises in accuracy but you absolutely want the fastest emulated Amiga possible on PowerMacs you have one only choice: E-UAE-JIT PPC from the developer Amos Rajnai.
The development is almost stopped now but the binary is available both for G4 and G5 machines: of course forget the CD32 and Picasso96 emulation but for the rest it is quite usable.
The hard work of Amos on the JIT recompiler for PPC is invaluable and so the efforts of Frode in mantain a project so "clean" and "open" to can be still compiled on PPC machines now, in 2019!
From what I have seen using both emulators I have understood that the major slowdowns of FS-UAE (on PPC hosts) are not related to CPU tasks but chipset emulation, any improvement in this direction (also from other developers with their own FS-UAE-PPC branch) would really make a world of difference.
In the meantime, if you are interested, I can re-upload the 2.4.1 OSX PPC binary in The Zone.

[SnakeCoils]

The (unofficial) OSX PPC binary of 2.9.10dev FS-UAE source is online and ready to download in The Zone.

Noticeably faster, on par with previous dev releases before the recent WinUAE merge. Very nice! :-)

[SnakeCoils]

The (unofficial) OSX PPC binary of 2.9.12dev FS-UAE source is online and ready to download in The Zone. Enjoy!

[SnakeCoils]

The (unofficial) OSX PPC binary of 3.0.0 stable FS-UAE source is online and ready to download in The Zone. Enjoy!

[vulture]

Awesome, thx!

[SnakeCoils]

Just tried to compile the 3.0.2 release but I have hit this error:

depbase=`echo src/fsemu/fsemu-sdlwindow.o | sed 's|[^/]*$|.deps/&|;s|\.o$||'`;\
	gcc-mp-7 -DHAVE_CONFIG_H -I.  -DMACOSX -DMACOS -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./manymouse -I./glad/include   -I./libudis86/include  -I./prowizard/include    -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/include/libpng16 -I/opt/local/include  -D_THREAD_SAFE -I/opt/local/include/SDL2 -D_THREAD_SAFE -I/opt/local/include/SDL2 -I/opt/local/include  -fno-strict-overflow -fno-strict-aliasing -Wno-format-truncation -Waddress -Werror=missing-declarations -fvisibility=hidden -I/opt/local/include/mpeg2dec -std=gnu99 -O3 -maltivec -mabi=altivec -MT src/fsemu/fsemu-sdlwindow.o -MD -MP -MF $depbase.Tpo -c -o src/fsemu/fsemu-sdlwindow.o src/fsemu/fsemu-sdlwindow.c &&\
	mv -f $depbase.Tpo $depbase.Po
src/fsemu/fsemu-sdlwindow.c:208:8: error: unknown type name 'SDL_HitTestResult'
 static SDL_HitTestResult fsemu_sdlwindow_hit_test_2(SDL_Window *window,
        ^~~~~~~~~~~~~~~~~
src/fsemu/fsemu-sdlwindow.c: In function 'fsemu_sdlwindow_hit_test_2':
src/fsemu/fsemu-sdlwindow.c:218:16: error: 'SDL_HITTEST_NORMAL' undeclared (first use in this function); did you mean 'SDL_HINT_NORMAL'?
         return SDL_HITTEST_NORMAL;
                ^~~~~~~~~~~~~~~~~~
                SDL_HINT_NORMAL
src/fsemu/fsemu-sdlwindow.c:218:16: note: each undeclared identifier is reported only once for each function it appears in
src/fsemu/fsemu-sdlwindow.c:229:20: error: 'SDL_HITTEST_DRAGGABLE' undeclared (first use in this function); did you mean 'SDL_HITTEST_NORMAL'?
             return SDL_HITTEST_DRAGGABLE;
                    ^~~~~~~~~~~~~~~~~~~~~
                    SDL_HITTEST_NORMAL
src/fsemu/fsemu-sdlwindow.c:233:20: error: 'SDL_HITTEST_RESIZE_BOTTOMLEFT' undeclared (first use in this function); did you mean 'SDL_HITTEST_DRAGGABLE'?
             return SDL_HITTEST_RESIZE_BOTTOMLEFT;
                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                    SDL_HITTEST_DRAGGABLE
src/fsemu/fsemu-sdlwindow.c:235:16: error: 'SDL_HITTEST_RESIZE_LEFT' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_BOTTOMLEFT'?
         return SDL_HITTEST_RESIZE_LEFT;
                ^~~~~~~~~~~~~~~~~~~~~~~
                SDL_HITTEST_RESIZE_BOTTOMLEFT
src/fsemu/fsemu-sdlwindow.c:238:20: error: 'SDL_HITTEST_RESIZE_BOTTOMRIGHT' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_BOTTOMLEFT'?
             return SDL_HITTEST_RESIZE_BOTTOMRIGHT;
                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                    SDL_HITTEST_RESIZE_BOTTOMLEFT
src/fsemu/fsemu-sdlwindow.c:240:16: error: 'SDL_HITTEST_RESIZE_RIGHT' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_LEFT'?
         return SDL_HITTEST_RESIZE_RIGHT;
                ^~~~~~~~~~~~~~~~~~~~~~~~
                SDL_HITTEST_RESIZE_LEFT
src/fsemu/fsemu-sdlwindow.c:247:16: error: 'SDL_HITTEST_RESIZE_BOTTOM' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_BOTTOMLEFT'?
         return SDL_HITTEST_RESIZE_BOTTOM;
                ^~~~~~~~~~~~~~~~~~~~~~~~~
                SDL_HITTEST_RESIZE_BOTTOMLEFT
src/fsemu/fsemu-sdlwindow.c: At top level:
src/fsemu/fsemu-sdlwindow.c:252:8: error: unknown type name 'SDL_HitTestResult'
 static SDL_HitTestResult fsemu_sdlwindow_hit_test(SDL_Window *window,
        ^~~~~~~~~~~~~~~~~
src/fsemu/fsemu-sdlwindow.c: In function 'fsemu_sdlwindow_create':
src/fsemu/fsemu-sdlwindow.c:334:9: warning: implicit declaration of function 'SDL_SetWindowHitTest'; did you mean 'SDL_SetWindowTitle'? [-Wimplicit-function-declaration]
     if (SDL_SetWindowHitTest(window, fsemu_sdlwindow_hit_test, NULL) == -1) {
         ^~~~~~~~~~~~~~~~~~~~
         SDL_SetWindowTitle
src/fsemu/fsemu-sdlwindow.c: In function 'fsemu_sdlwindow_handle_cursor':
src/fsemu/fsemu-sdlwindow.c:545:5: error: unknown type name 'SDL_HitTestResult'; did you mean 'GTestResult'?
     SDL_HitTestResult hit_test =
     ^~~~~~~~~~~~~~~~~
     GTestResult
src/fsemu/fsemu-sdlwindow.c:547:21: error: 'SDL_HITTEST_RESIZE_LEFT' undeclared (first use in this function); did you mean 'SDL_WINDOW_RESIZABLE'?
     if (hit_test == SDL_HITTEST_RESIZE_LEFT ||
                     ^~~~~~~~~~~~~~~~~~~~~~~
                     SDL_WINDOW_RESIZABLE
src/fsemu/fsemu-sdlwindow.c:548:21: error: 'SDL_HITTEST_RESIZE_RIGHT' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_LEFT'?
         hit_test == SDL_HITTEST_RESIZE_RIGHT) {
                     ^~~~~~~~~~~~~~~~~~~~~~~~
                     SDL_HITTEST_RESIZE_LEFT
src/fsemu/fsemu-sdlwindow.c:550:28: error: 'SDL_HITTEST_RESIZE_BOTTOM' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_LEFT'?
     } else if (hit_test == SDL_HITTEST_RESIZE_BOTTOM) {
                            ^~~~~~~~~~~~~~~~~~~~~~~~~
                            SDL_HITTEST_RESIZE_LEFT
src/fsemu/fsemu-sdlwindow.c:552:28: error: 'SDL_HITTEST_RESIZE_BOTTOMLEFT' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_BOTTOM'?
     } else if (hit_test == SDL_HITTEST_RESIZE_BOTTOMLEFT) {
                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                            SDL_HITTEST_RESIZE_BOTTOM
src/fsemu/fsemu-sdlwindow.c:554:28: error: 'SDL_HITTEST_RESIZE_BOTTOMRIGHT' undeclared (first use in this function); did you mean 'SDL_HITTEST_RESIZE_BOTTOMLEFT'?
     } else if (hit_test == SDL_HITTEST_RESIZE_BOTTOMRIGHT) {
                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                            SDL_HITTEST_RESIZE_BOTTOMLEFT
make[3]: *** [src/fsemu/fsemu-sdlwindow.o] Error 1
make[2]: *** [all-recursive] Error 1
make[1]: *** [all] Error 2
make: *** [bundle] Error 2

Is there something I can do for going ahead? The rest of source compiled flawlessy so far. The maximum version of SDL2 I can use on PPC OSX is 2.0.3

[FrodeSolheim]

This looks like code from the git master branch, not 3.0.02 ;-)

(For reference, the errors are probably due to too old SDL2)

[SnakeCoils]

Yes, I'm gone back to the 3.0.2 source code archive and the compile went good: the new release is now available in The Zone for download.

If there will be no more updates of SDL2 libraries for OSX PPC from TigerPorts this likely will be the final FS-UAE release for PowerMacs because the new modification on the source code rely on new instructions preset from 2.0.4
I've seen in his GitHub repository a preliminary study for a 2.0.5 port but the project is on hold from some months, we can only hope...
Anyway thank you very much for your support so far Frode, having this emulator running on our beloved G4/G5 machines wouldn't be possible without the attention you deserved to ALL the FS-UAE users including this little niche :-)