Switched to GCC6 with LibNix for Amiga, disabled the timidity, mp3 and vorbis codecs by BSzili · Pull Request #27 · sezero/uhexen2

BSzili · 2022-03-16T08:31:20Z

I started working on the AmigaOS3.x port of uhexen2 again to make it playable on 68060 accelerators. As a first step I fixed the game to properly build with Bebbo's GCC6 toolchain which produces faster code:
https://github.com/bebbo/amiga-gcc
I also switched back to LibNix as the one comes with that toolchain is actively maintained and fixes many of the original bugs. The Vorbis, MP3 and Timidity codecs are disabled for now, as these are too slow on the real hardware. I plan to restore MP3 support later using MHI which uses hardware decoders on sounds cards and external decoders like the MAS Player.
Let me know if I missed something or there's something else I should change in the PR.

…d vorbis codecs

sezero · 2022-03-16T11:00:54Z

 #ifndef PLATFORM_AMIGAOS3
 #include <sys/param.h>
+#else
+#define __NO_NET_API


This is universal I guess?

EDIT: Or is it related to commenting out of -I$(OSLIBS)/amigaos/netinclude down below?

I added it to prevent some function prototypes from declared in arpa/inet.h that would clash with the bsdsocket.library GCC inlines. I just noticed that __NO_NET_API is already in q_endian.h but for CLIB2 only. Maybe I should add libnix to q_endian.h instead and remove this one?

Can't really remember why I did that in q_endian.h (was for RoadShow SDK support??), so I don't know what to do with it for now :(

I can just stick to using q_endian.h for this purpose, the important thing is that __NO_NET_API is defined.

That didn't work out, as amigaos is not defined there. I think I'll leave it in net_sys.h as it's a network-related macro.

sezero · 2022-03-16T11:02:27Z

 INCLUDES += -I$(OSLIBS)/amigaos/include
 # AmiTCP SDK
-NET_INC   = -I$(OSLIBS)/amigaos/netinclude
+#NET_INC   = -I$(OSLIBS)/amigaos/netinclude


Why are we not using the RoadShow SDK?

The original SDK had some compilation issues with some errno codes, so I opted to use the modified Roadshow SDK that comes with the GCC6 toolchain. I could probably work around this in a different way to keep the original SDK.

I modified the Makefiles to use the BEBBO_TOOLCHAIN switch for this as well.

sezero · 2022-03-16T11:04:20Z

@@ -1,4 +1,4 @@
-#ifdef __CLIB2__
+#if defined(__CLIB2__) || defined(__libnix__)


Old libnix doesn't have stdint.h - how to handle that, I don't know.

I first thought about using __has_include, but that is introduced in gcc5 - sigh.

The new toolchain's libnix v4 has stdint.h, so I had to do this to prevent redefining the existing types. Maybe it could be done the other way? E.g. #if (defined(__has_include) && __has_include(<stdint.h>))

__has_include is supported as of gcc5, so not an option for gcc4.6

I see, then don't have a good idea for a workaround for now.

stdlib.h has __MY_INLINE__, but that's obviously not reliable.

ixemul seems to have added stdint.h as of v50.0. https://github.com/bebbo/ixemul
seems to have started with v60 something, but then has gone back to v48.2 while
keeping stdint.h, sigh...

On the other hand, we are using the same header guard as the stdint.h from the
bebbo ixemul repo, i.e. __STDINT_H, so there should be no conflicts.

Ixemul.library is a bit of a mess, v50 and v60 were closed source forks, so many people stayed with v4x. Anyway, I'll try to figure out the root cause of the header clash.

If it helps, this is the diff between stdint.h from bebbo's ixemul and ours:

--- ixemul-git/include/stdint.h +++ uhexen2-git/oslibs/amigaos/include/stdint.h @@ -1,12 +1,14 @@ +#ifdef __CLIB2__ +#include_next <stdint.h> /* clib2 provides a stdint.h already */ + +#else + #ifndef __STDINT_H #define __STDINT_H 1 +#include <machine/types.h> /* defines signed variants of int8_t..int64_t */ #include <limits.h> -typedef signed char int8_t; -typedef signed short int16_t; -typedef signed int int32_t; -typedef signed long long int64_t; #define INT8_MIN SCHAR_MIN #define INT8_MAX SCHAR_MAX #define INT16_MIN SHRT_MIN @@ -16,19 +18,19 @@ #define INT64_MIN LLONG_MIN #define INT64_MAX LLONG_MAX -typedef unsigned char uint8_t; -typedef unsigned short uint16_t; -typedef unsigned int uint32_t; -typedef unsigned long long uint64_t; +typedef u_int8_t uint8_t; +typedef u_int16_t uint16_t; +typedef u_int32_t uint32_t; +typedef u_int64_t uint64_t; #define UINT8_MAX UCHAR_MAX #define UINT16_MAX USHRT_MAX #define UINT32_MAX UINT_MAX #define UINT64_MAX ULLONG_MAX -typedef signed char int_least8_t; -typedef signed short int_least16_t; -typedef signed int int_least32_t; -typedef signed long long int_least64_t; +typedef int8_t int_least8_t; +typedef int16_t int_least16_t; +typedef int32_t int_least32_t; +typedef int64_t int_least64_t; #define INT_LEAST8_MIN SCHAR_MIN #define INT_LEAST8_MAX SCHAR_MAX #define INT_LEAST16_MIN SHRT_MIN @@ -38,19 +40,19 @@ #define INT_LEAST64_MIN LLONG_MIN #define INT_LEAST64_MAX LLONG_MAX -typedef unsigned char uint_least8_t; -typedef unsigned short uint_least16_t; -typedef unsigned int uint_least32_t; -typedef unsigned long long uint_least64_t; +typedef u_int8_t uint_least8_t; +typedef u_int16_t uint_least16_t; +typedef u_int32_t uint_least32_t; +typedef u_int64_t uint_least64_t; #define UINT_LEAST8_MAX UCHAR_MAX #define UINTLEAST16_MAX USHRT_MAX #define UINTLEAST32_MAX UINT_MAX #define UINTLEAST64_MAX ULLONG_MAX -typedef int8_t int_fast8_t; -typedef int int_fast16_t; -typedef int int_fast32_t; -typedef long long int_fast64_t; +typedef int32_t int_fast8_t; +typedef int32_t int_fast16_t; +typedef int32_t int_fast32_t; +typedef int64_t int_fast64_t; #define INT_FAST8_MIN INT_MIN #define INT_FAST8_MAX INT_MAX #define INT_FAST16_MIN INT_MIN @@ -60,22 +62,22 @@ #define INT_FAST64_MIN LLONG_MIN #define INT_FAST64_MAX LLONG_MAX -typedef uint8_t uint_fast8_t; -typedef unsigned int uint_fast16_t; -typedef unsigned int uint_fast32_t; -typedef unsigned long long uint_fast64_t; +typedef u_int32_t uint_fast8_t; +typedef u_int32_t uint_fast16_t; +typedef u_int32_t uint_fast32_t; +typedef u_int64_t uint_fast64_t; #define UINT_FAST8_MAX UINT_MAX #define UINT_FAST16_MAX UINT_MAX #define UINT_FAST32_MAX UINT_MAX #define UINT_FAST64_MAX ULLONG_MAX -typedef int intptr_t; -#define INTPTR_MIN INT_MIN -#define INTPTR_MAX INT_MAX +typedef long intptr_t; +#define INTPTR_MIN LONG_MIN +#define INTPTR_MAX LONG_MAX -typedef unsigned int uintptr_t; -#define UINTPTR_MAX UINT_MAX +typedef unsigned long uintptr_t; +#define UINTPTR_MAX ULONG_MAX typedef long long intmax_t; #define INTMAX_MIN LLONG_MIN @@ -108,3 +110,5 @@ #define UINTMAX_C(x) x##ULL #endif /* __STDINT_H */ + +#endif /* __CLIB2__ */

The problem appears to be sys/types.h which includes sys/_stdint.h. Fortunately that one has _..._T_DECLARED macro guards, so I can work this around properly in the oslibs stdint.h

sezero · 2022-03-16T11:04:50Z

 # dynamic loading of ogl functions doesn't work
 LINK_GL_LIBS=yes

+USE_CODEC_TIMIDITY=no


These need mirroring in hexenworld side too

Oops, I'll update the hexenworld makefiles too.

sezero · 2022-03-16T11:08:19Z


 ifndef DEBUG
-CFLAGS  += -O3
+#CFLAGS  += -O3


-O3 is bad with gcc6?

Not gcc6 specifically, but I found that O3 bloats up the code in places, so it's probably better to stick with O2 which fits in the cache better.

We do -O3 only for gcc2.95 / aos3, I wonder why we did that back at the time..

It was probably an attempt to make it to run faster, but with O3 you start to hit diminishing returns, and in some cases it can do more harm than good.

sezero · 2022-03-16T11:08:22Z

-CFLAGS  += -fno-omit-frame-pointer
+#CFLAGS  += -fno-omit-frame-pointer
+# these break the game with GCC6
+CFLAGS  += -fbbb=- -fno-tree-forwprop


These aren't supported by old gcc2.95: We need some flag or some makefile variable to support it?

I can add an variable e.g. GCC = 6, but I found that GCC6 usually produces better code than older versions, so I thought it's enough to keep the old flags commented out in case someone really wants to use GCC2.

Some AMIGA_GCC6, or BEBBO_TOOLCHAIN, or something may be good. (I don't have bebbo's toolchain available on my setup yet, that why I am whining :))

I'll probably go with BEBBO_TOOLCHAIN then, since he improved a lot of things besides GCC (e.g. libnix).

…HexenWorld

sezero · 2022-03-18T12:46:20Z

This is fairly good now. Attaching a tidy-up patch (on top of yours): 27a_patch.txt

(I wonder whether we need those optimizer flags out of the engine directory, but that's ultra low priority.)

BSzili · 2022-03-18T18:33:44Z

Sure thing, I applied the patch.

Tied to the new Makefile variable BEBBO_TOOLCHAIN,which defaults to yes. Also disabled the timidity, mp3 and vorbis codecs as those are too slow on the real hardware. Original discussion is at #27

sezero · 2022-03-18T19:13:47Z

Patchset squashed, commit message edited, and applied as 531d625
Thanks!

switched to GCC6 with LibNix for Amiga, disabled the timidity, mp3 an…

4399df1

…d vorbis codecs

sezero reviewed Mar 16, 2022

View reviewed changes

BSzili added 2 commits March 17, 2022 06:46

added the BEBBO_TOOLCHAIN to the Makefile, propagated the changes to …

ea120b0

…HexenWorld

better stdint.h workaround for the new libnix SDK

0bff52e

Makefile tidy-up

63e2411

sezero closed this Mar 18, 2022

		@@ -1,4 +1,4 @@
		#ifdef __CLIB2__
		#if defined(__CLIB2__) \|\| defined(__libnix__)

Conversation

BSzili commented Mar 16, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sezero commented Mar 18, 2022

Uh oh!

BSzili commented Mar 18, 2022

Uh oh!

sezero commented Mar 18, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants