GCC 6.2 toolchain for AmigaOS 3 - Page 39

Toni Wilen · 10 February 2018, 18:10

Right, only function calls can use PC-relative jumps + extra absolute jump if needed. I would have been quite difficult to do with LEAs

Very tiny observation:

Long word array offset calculation use 2x add.l dn,dn. I think -Os should use shorter lsl.l #2,d0.

bebbo · 10 February 2018, 23:57

Quote:

Originally Posted by Toni Wilen

Right, only function calls can use PC-relative jumps + extra absolute jump if needed. I would have been quite difficult to do with LEAs

you can use -mpcrel on some source files and if ordered correctly it will work.

Quote:

Originally Posted by Toni Wilen

Very tiny observation:

Long word array offset calculation use 2x add.l dn,dn. I think -Os should use shorter lsl.l #2,d0.

I think, I can live with add add.

wawa · 12 February 2018, 21:52

bebbo, dont want to derail your thread any further, just wanted to ask if you could point me to where, if you hat manually enable relative addressing in gcc. like when specifying an offset to address register content here:
bset.b #0,4(a1)
thats what aros gcc seems not to understand.
my thread is here:
http://eab.abime.net/showthread.php?...54#post1219154

bebbo · 02 March 2018, 13:00

Quote:

Originally Posted by Toni Wilen

Long word array offset calculation use 2x add.l dn,dn. I think -Os should use shorter lsl.l #2,d0.

All modes should use lsl.l #2,d0 (12 cycles) instead of 2x add.l dn,dn (2x8cycles).

=> Done

emufan · 07 March 2018, 22:46

i try to build lhasa/lha.exe using -mcrt=clib2 -m68000
same with: -mcrt=clib2 -m68020-40 -msoft-float
and get the linker message:

Code:

lha-list.o(.text+0x63a): undefined reference to `__mulsf3'
lha-list.o(.text+0x654): undefined reference to `__divsf3'

are these missing?

if so, I found something which may help:

Code:

To get gcc to build a floating point library, I think you need to change
the line

softfp_float_modes := tf

in the file gcc/config/i386/t-fprules-softfp to be

softfp_float_modes := sf df tf

I don't know if that will suffice but as far as I can see it is a
necessary step.

#1) it does build using: -mcrt=clib2 -m68020-40 -m68881
so something is missing in the softfloat math lib

bebbo · 08 March 2018, 12:08

Quote:

Originally Posted by emufan

i try to build lhasa/lha.exe using -mcrt=clib2 -m68000
same with: -mcrt=clib2 -m68020-40 -msoft-float
and get the linker message:

Code:

lha-list.o(.text+0x63a): undefined reference to `__mulsf3'
lha-list.o(.text+0x654): undefined reference to `__divsf3'

are these missing?

if so, I found something which may help:

Code:

To get gcc to build a floating point library, I think you need to change
the line

softfp_float_modes := tf

in the file gcc/config/i386/t-fprules-softfp to be

softfp_float_modes := sf df tf

I don't know if that will suffice but as far as I can see it is a
necessary step.

#1) it does build using: -mcrt=clib2 -m68020-40 -m68881
so something is missing in the softfloat math lib

please add -lm

emufan · 08 March 2018, 12:45

Quote:

Originally Posted by bebbo

please add -lm

seriously? I refused to do so, since I could not imagine it's again that simple

#1) of course you are right

Code:

 CFLAGS="-mcrt=clib2 -m68000" LDFLAGS="-lm" ./configure ...

#2) but can you shed some light on the way described in my quote from the gnu website?
is this the way to configure the toolchain build, so it does build softfloat libs?

bebbo · 08 March 2018, 13:24

Quote:

Originally Posted by emufan

seriously? I refused to do so, since I could not imagine it's again that simple

#1) of course you are right

Code:

 CFLAGS="-mcrt=clib2 -m68000" LDFLAGS="-lm" ./configure ...

good to hear!

Quote:

Originally Posted by emufan

#2) but can you shed some light on the way described in my quote from the gnu website?
is this the way to configure the toolchain build, so it does build softfloat libs?

You were referring to the configuration for i386. If you plan to create code for an old cpu like 8086, then you would need code for single and double float.

The configuration for i386 does not affect what's build for the m68k line.

AFAIK the libraries (clib2, libnix) are providing a math lib using the Amiga math libraries. Thus you need -lm for float/double math operations. (And you also need -lm for printing float/double with libnix)

emufan · 08 March 2018, 14:23

Quote:

The configuration for i386 does not affect what's build for the m68k line.

c'mon. :/
of course I saw the i386. it was more a question,
regarding the way those things are build generally.

but thanks for info.

bebbo · 08 March 2018, 15:05

Quote:

Originally Posted by emufan

c'mon. :/
of course I saw the i386. it was more a question,
regarding the way those things are build generally.

but thanks for info.

I think config.host is the right location where soft floating point stuff is enabled:

Code:

m68k-*-amiga*)
    tmake_file="$tmake_file m68k/t-glue m68k/t-floatlib soft-fp"

emufan · 08 March 2018, 15:45

Quote:

Originally Posted by bebbo

I think config.host is the right location where soft floating point stuff is enabled:

Code:

m68k-*-amiga*)
    tmake_file="$tmake_file m68k/t-glue m68k/t-floatlib soft-fp"

ok, thanks

Fook42 · 14 March 2018, 23:40

hello,
i'm on the latest from the git-repo
(commitid: bee61d81df17239be5e158b4264a7470358824cb)
but the build using "toolchain-m68k" fails:

Code:

...
INFO: extract files from ".build-m68k/archives/ira.lha"
DEBUG: makedir ".build-m68k/tmp/tmp2SPBmK/ira"
DEBUG: extract "ira/amiga_hunks.c"
DEBUG: extract "ira/amiga_hunks.h"
DEBUG: extract "ira/atari.c"
DEBUG: extract "ira/atari.h"
DEBUG: extract "ira/binary.c"
DEBUG: extract "ira/binary.h"
DEBUG: extract "ira/config.c"
DEBUG: extract "ira/config.h"
DEBUG: extract "ira/constants.c"
DEBUG: extract "ira/constants.h"
DEBUG: extract "ira/elf.c"
DEBUG: extract "ira/elf.h"
DEBUG: extract "ira/init.c"
DEBUG: extract "ira/init.h"
DEBUG: extract "ira/ira.c"
DEBUG: extract "ira/ira.doc"
DEBUG: extract "ira/ira.exe"
DEBUG: extract "ira/ira.h"
DEBUG: extract "ira/ira.readme"
DEBUG: extract "ira/ira2.doc"
DEBUG: extract "ira/ira_2.c"
DEBUG: extract "ira/ira_2.h"
DEBUG: extract "ira/ira_68k"
DEBUG: extract "ira/ira_config.doc"
DEBUG: extract "ira/ira_mos"
DEBUG: extract "ira/ira_os4"
DEBUG: extract "ira/make.rules"
DEBUG: extract "ira/Makefile"
DEBUG: extract "ira/Makefile.mos"
DEBUG: extract "ira/Makefile.os3"
DEBUG: extract "ira/Makefile.os4"
DEBUG: extract "ira/Makefile.osx"
DEBUG: extract "ira/Makefile.win32"
DEBUG: extract "ira/megadrive.c"
DEBUG: extract "ira/megadrive.h"
DEBUG: makedir ".build-m68k/tmp/tmp2SPBmK/ira/obj"
DEBUG: extract "ira/obj/.dummy"
DEBUG: extract "ira/opcode.c"
DEBUG: extract "ira/opcode.h"
DEBUG: extract "ira/supp.c"
DEBUG: extract "ira/supp.h"
DEBUG: copytree ".build-m68k/tmp/tmp2SPBmK" to ".build-m68k/build/ira"
DEBUG: makedir ".build-m68k/build/ira"
DEBUG: makedir ".build-m68k/build/ira/ira"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/Makefile" to ".build-m68k/build/ira/ira/Makefile"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/Makefile.mos" to ".build-m68k/build/ira/ira/Makefile.mos"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/Makefile.os3" to ".build-m68k/build/ira/ira/Makefile.os3"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/Makefile.os4" to ".build-m68k/build/ira/ira/Makefile.os4"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/Makefile.osx" to ".build-m68k/build/ira/ira/Makefile.osx"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/Makefile.win32" to ".build-m68k/build/ira/ira/Makefile.win32"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/amiga_hunks.c" to ".build-m68k/build/ira/ira/amiga_hunks.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/amiga_hunks.h" to ".build-m68k/build/ira/ira/amiga_hunks.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/atari.c" to ".build-m68k/build/ira/ira/atari.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/atari.h" to ".build-m68k/build/ira/ira/atari.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/binary.c" to ".build-m68k/build/ira/ira/binary.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/binary.h" to ".build-m68k/build/ira/ira/binary.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/config.c" to ".build-m68k/build/ira/ira/config.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/config.h" to ".build-m68k/build/ira/ira/config.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/constants.c" to ".build-m68k/build/ira/ira/constants.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/constants.h" to ".build-m68k/build/ira/ira/constants.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/elf.c" to ".build-m68k/build/ira/ira/elf.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/elf.h" to ".build-m68k/build/ira/ira/elf.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/init.c" to ".build-m68k/build/ira/ira/init.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/init.h" to ".build-m68k/build/ira/ira/init.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira.c" to ".build-m68k/build/ira/ira/ira.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira.doc" to ".build-m68k/build/ira/ira/ira.doc"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira.exe" to ".build-m68k/build/ira/ira/ira.exe"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira.h" to ".build-m68k/build/ira/ira/ira.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira.readme" to ".build-m68k/build/ira/ira/ira.readme"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira2.doc" to ".build-m68k/build/ira/ira/ira2.doc"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira_2.c" to ".build-m68k/build/ira/ira/ira_2.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira_2.h" to ".build-m68k/build/ira/ira/ira_2.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira_68k" to ".build-m68k/build/ira/ira/ira_68k"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira_config.doc" to ".build-m68k/build/ira/ira/ira_config.doc"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira_mos" to ".build-m68k/build/ira/ira/ira_mos"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/ira_os4" to ".build-m68k/build/ira/ira/ira_os4"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/make.rules" to ".build-m68k/build/ira/ira/make.rules"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/megadrive.c" to ".build-m68k/build/ira/ira/megadrive.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/megadrive.h" to ".build-m68k/build/ira/ira/megadrive.h"
DEBUG: makedir ".build-m68k/build/ira/ira/obj"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/obj/.dummy" to ".build-m68k/build/ira/ira/obj/.dummy"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/opcode.c" to ".build-m68k/build/ira/ira/opcode.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/opcode.h" to ".build-m68k/build/ira/ira/opcode.h"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/supp.c" to ".build-m68k/build/ira/ira/supp.c"
DEBUG: copy ".build-m68k/tmp/tmp2SPBmK/ira/supp.h" to ".build-m68k/build/ira/ira/supp.h"
DEBUG: rmtree ".build-m68k/tmp/tmp2SPBmK"
DEBUG: enter directory ".build-m68k/build"
DEBUG: execute "patch -t -p0 -i /home/renef/AMIGA/new/amigaos-cross-toolchain/patches/ira/Makefile.diff"
can't find file to patch at input line 3
Perhaps you used the wrong -p or --strip option?
The text leading up to this was:
--------------------------
|--- ira/Makefile       2015-09-11 09:25:13.000000000 +0200
|+++ ira/Makefile       2015-09-11 09:21:37.000000000 +0200
--------------------------
No file to patch.  Skipping patch.
1 out of 1 hunk ignored
ERROR: command "patch -t -p0 -i /home/renef/AMIGA/new/amigaos-cross-toolchain/patches/ira/Makefile.diff" failed with 1

the ira-files (including the missing ira/Makefile) are here .build-m68k/build/ira/ira/ .. instead of .build-m68k/build/ira/ ...

how can we fix this?
just moving the files from the ira/ira/-directory to ira/ is not working completely (several other patches fail)

Marlon_ · 15 March 2018, 00:28

Use this repo instead: https://github.com/bebbo/amiga-gcc

alkis · 27 March 2018, 17:56

So, I have a small project (like the command "ls" in linux) and uses nostartup and thus it has a custom startup in assembly.

The exe size is at ~19k, and if I add -fbaserel -msmall-code goes at ~15k. Impressive and all but....what is the inner workings for a4?

I probably have to provide a "geta4()" in my startup, right? (and also call it at start, to set it once)
But to what value should I set it to? Does linker expose a symbol for that purpose when -fbaserel is used?

Amiga1992 · 27 March 2018, 18:24

OK so let's say (because it's true :P) that I am a n00b, but I know C, and want to try this out.

What is exactly all I need to be able to code on my modern (PC/Win) computer and make binaries for Amiga? Is it just this? > https://github.com/bebbo/amigaos-cross-toolchain
Do I need an Amiga, emulated or not, at all for this? (ideally no).

Samurai_Crow · 27 March 2018, 18:26

Quote:

Originally Posted by Akira

OK so let's say (because it's true :P) that I am a n00b, but I know C, and want to try this out.

What is exactly all I need to be able to code on my modern (PC/Win) computer and make binaries for Amiga? Is it just this? > https://github.com/bebbo/amigaos-cross-toolchain
Do I need an Amiga, emulated or not, at all for this? (ideally no).

That repo is frozen in favor of the link in Marlon_'s post.

Amiga1992 · 27 March 2018, 18:29

Quote:

Originally Posted by Samurai_Crow

That repo is frozen in favor of the link in Marlon_'s post.

OK I see, and these are tools that run in a PC right?
But not in Windows?
Also: Is there any tutorial to get started?

Samurai_Crow · 27 March 2018, 18:30

Quote:

Originally Posted by Akira

OK I see, and these are tools that run in a PC right?
But not in Windows?
Also: Is there any tutorial to get started?

PC, Mac or Linux. Preferably Mac or Linux.

alkis · 27 March 2018, 18:32

Quote:

Originally Posted by Akira

OK so let's say (because it's true :P) that I am a n00b, but I know C, and want to try this out.

What is exactly all I need to be able to code on my modern (PC/Win) computer and make binaries for Amiga? Is it just this? > https://github.com/bebbo/amigaos-cross-toolchain
Do I need an Amiga, emulated or not, at all for this? (ideally no).

that repository is frozen.

Now, you should use https://github.com/bebbo/amiga-gcc

Someone with windows experience should tackle this, but if you were in linux you would do:
git pull https://github.com/bebbo/amiga-gcc
cd amiga-gcc
export PREFIX=/home/username/gcc6
make update
make all -j8

Amiga1992 · 27 March 2018, 18:59

It's just that my only Linux machines are Raspberry Pis. I don't want to run Linux on my Windows PC, although I usually have a virtual machine ready for special cases.

07 March 2018, 22:46	#765
emufan Registered User Join Date: Feb 2012 Location: #DrainTheSwamp Posts: 4,545	i try to build lhasa/lha.exe using -mcrt=clib2 -m68000 same with: -mcrt=clib2 -m68020-40 -msoft-float and get the linker message: Code: lha-list.o(.text+0x63a): undefined reference to `__mulsf3' lha-list.o(.text+0x654): undefined reference to `__divsf3' are these missing? if so, I found something which may help: Code: To get gcc to build a floating point library, I think you need to change the line softfp_float_modes := tf in the file gcc/config/i386/t-fprules-softfp to be softfp_float_modes := sf df tf I don't know if that will suffice but as far as I can see it is a necessary step. #1) it does build using: -mcrt=clib2 -m68020-40 -m68881 so something is missing in the softfloat math lib Last edited by emufan; 07 March 2018 at 23:03.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
New GCC based dev toolchain for AmigaOS 3.x	cla	Coders. Releases	8	24 December 2017 10:18
Issue with photon/xxxx WinUAE Toolchain	arpz	Coders. Asm / Hardware	2	26 September 2015 22:33
New 68k gcc toolchain	arti	Coders. C/C++	17	31 July 2015 03:59
Hannibal's WinUAE Demo Toolchain 5	Bobic	Amiga scene	1	23 July 2015 21:04
From gcc to vbcc.	Cowcat	Coders. General	9	06 June 2014 14:45

10 February 2018, 18:10	#761
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,502	Right, only function calls can use PC-relative jumps + extra absolute jump if needed. I would have been quite difficult to do with LEAs Very tiny observation: Long word array offset calculation use 2x add.l dn,dn. I think -Os should use shorter lsl.l #2,d0.

12 February 2018, 21:52	#763
wawa Registered User Join Date: Aug 2007 Location: berlin/germany Posts: 1,054	bebbo, dont want to derail your thread any further, just wanted to ask if you could point me to where, if you hat manually enable relative addressing in gcc. like when specifying an offset to address register content here: bset.b #0,4(a1) thats what aros gcc seems not to understand. my thread is here: http://eab.abime.net/showthread.php?...54#post1219154

15 March 2018, 00:28	#773
Marlon_ AmigaDev.com Join Date: Mar 2016 Location: Stockholm, Sweden Age: 35 Posts: 625	Use this repo instead: https://github.com/bebbo/amiga-gcc

27 March 2018, 17:56	#774
alkis Registered User Join Date: Dec 2010 Location: Athens/Greece Age: 53 Posts: 719	So, I have a small project (like the command "ls" in linux) and uses nostartup and thus it has a custom startup in assembly. The exe size is at ~19k, and if I add -fbaserel -msmall-code goes at ~15k. Impressive and all but....what is the inner workings for a4? I probably have to provide a "geta4()" in my startup, right? (and also call it at start, to set it once) But to what value should I set it to? Does linker expose a symbol for that purpose when -fbaserel is used?

27 March 2018, 18:24	#775
Amiga1992 Registered User Join Date: May 2001 Location: ? Posts: 19,645	OK so let's say (because it's true :P) that I am a n00b, but I know C, and want to try this out. What is exactly all I need to be able to code on my modern (PC/Win) computer and make binaries for Amiga? Is it just this? > https://github.com/bebbo/amigaos-cross-toolchain Do I need an Amiga, emulated or not, at all for this? (ideally no).

27 March 2018, 18:59	#780
Amiga1992 Registered User Join Date: May 2001 Location: ? Posts: 19,645	It's just that my only Linux machines are Raspberry Pis. I don't want to run Linux on my Windows PC, although I usually have a virtual machine ready for special cases.

Currently Active Users Viewing This Thread: 3 (0 members and 3 guests)