Small IRA Tutorial - Page 5

phx · 17 December 2023, 13:45

Quote:

Originally Posted by hop

Does this sound like a worthwhile addition?

Quite easy to add, by testing the new option in

itoasc()

. Not sure if it will help much, as there will always be displacements which are better viewed in decimal or in hex, in similar amounts. But I can add it for a test.
Myself, I used the

-a

option in the past, and looked at the hexdump in the comment.

hop · 17 December 2023, 16:13

Quote:

Originally Posted by phx

Quite easy to add, by testing the new option in

itoasc()

. Not sure if it will help much, as there will always be displacements which are better viewed in decimal or in hex, in similar amounts. But I can add it for a test.

If you mean that you can implement this yourself, then thanks very much. If I have misunderstood you then let me know!

Quote:

Originally Posted by phx

Myself, I used the

-a

option in the past, and looked at the hexdump in the comment.

I do this too. Hopefully the new option will remove this layer of indirection

Regarding CMakeLists.txt. I have tested the attached CMakeLists.txt with ira_211beta.lha on Windows (Visual Studio 2022), Linux (Ubuntu 18.04, 20.04, 22.04 and Windows WSL Ubuntu 22.04), and Mac OSX Montgomery. It works just fine. Unfortunately it looks like Amiga support for CMake is lacking, but I still think this would make a worthwhile addition to the package. Usage:

Code:

Windows: Open "x64 Native Tools Command Prompt for VS 20xx" from the Start menu. Run gensln.bat. Open generated solution with Visual Studio and build.
Linux/OSX: mkdir build && cd build && cmake .. && make

hop · 18 December 2023, 12:26

Are the TRAP equates in ira off by one and missing an entry?

From constants.c

Code:

const x_adr_t x_adrs[] = {
{"ABSEXECBASE", 0x0004},
{"BUS_ERROR",   0x0008},
{"ADR_ERROR",   0x000C},
{"ILLEG_OPC",   0x0010},
{"DIVISION0",   0x0014},
{"CHK",         0x0018},
{"TRAPV",       0x001C},
{"PRIVILEG",    0x0020},
{"TRACE",       0x0024},
{"LINEA_EMU",   0x0028},
{"LINEF_EMU",   0x002C},
{"INT_NOINI",   0x003C},
{"INT_WRONG",   0x0060},
{"AUTO_INT1",   0x0064},
{"AUTO_INT2",   0x0068},
{"AUTO_INT3",   0x006C},
{"AUTO_INT4",   0x0070},
{"AUTO_INT5",   0x0074},
{"AUTO_INT6",   0x0078},
{"NMI",         0x007C},
{"TRAP_01",     0x0080},
{"TRAP_02",     0x0084},
{"TRAP_03",     0x0088},
{"TRAP_04",     0x008C},
{"TRAP_05",     0x0090},
{"TRAP_06",     0x0094},
{"TRAP_07",     0x0098},
{"TRAP_08",     0x009C},
{"TRAP_09",     0x00A0},
{"TRAP_10",     0x00A4},
{"TRAP_11",     0x00A8},
{"TRAP_12",     0x00AC},
{"TRAP_13",     0x00B0},
{"TRAP_14",     0x00B4},
{"TRAP_15",     0x00B8},
{"CIAB_PRA",    0xBFD000},
...

Shouldn't that be:

Code:

const x_adr_t x_adrs[] = {
{"ABSEXECBASE", 0x0004},
{"BUS_ERROR",   0x0008},
{"ADR_ERROR",   0x000C},
{"ILLEG_OPC",   0x0010},
{"DIVISION0",   0x0014},
{"CHK",         0x0018},
{"TRAPV",       0x001C},
{"PRIVILEG",    0x0020},
{"TRACE",       0x0024},
{"LINEA_EMU",   0x0028},
{"LINEF_EMU",   0x002C},
{"INT_NOINI",   0x003C},
{"INT_WRONG",   0x0060},
{"AUTO_INT1",   0x0064},
{"AUTO_INT2",   0x0068},
{"AUTO_INT3",   0x006C},
{"AUTO_INT4",   0x0070},
{"AUTO_INT5",   0x0074},
{"AUTO_INT6",   0x0078},
{"NMI",         0x007C},
{"TRAP_00",     0x0080},
{"TRAP_01",     0x0084},
{"TRAP_02",     0x0088},
{"TRAP_03",     0x008C},
{"TRAP_04",     0x0090},
{"TRAP_05",     0x0094},
{"TRAP_06",     0x0098},
{"TRAP_07",     0x009C},
{"TRAP_08",     0x00A0},
{"TRAP_09",     0x00A4},
{"TRAP_10",     0x00A8},
{"TRAP_11",     0x00AC},
{"TRAP_12",     0x00B0},
{"TRAP_13",     0x00B4},
{"TRAP_14",     0x00B8},
{"TRAP_15",     0x00BC},
{"CIAB_PRA",    0xBFD000},
...

phx · 19 December 2023, 18:20

Quote:

Originally Posted by hop

If you mean that you can implement this yourself, then thanks very much

Tried that today. New option is called

-radix=n

, with n = 10 (default, decimal), 16 for hex and 0 for hex-positive, decimal-negative.
Unfortunately I realised that

itoasc()

is not called for

move

immediate addressing modes, which always call

itohex()

with a fixed number of digits (2, 4, 8). Maybe this makes sense, as, unlike displacements, you don't know whether you are dealing with a signed value.

Quote:

Regarding CMakeLists.txt. I have tested the attached CMakeLists.txt with ira_211beta.lha on Windows (Visual Studio 2022), Linux (Ubuntu 18.04, 20.04, 22.04 and Windows WSL Ubuntu 22.04), and Mac OSX Montgomery. It works just fine.

How does it work with Linux and OSX, when I only see a section for MSVC in it? (Being a Cmake noob.)

Quote:

Unfortunately it looks like Amiga support for CMake is lacking

That's already one of the major problems I have here. Why should I use something which is less portable and adds more dependencies (cmake is not standard in a development environment)?

I really like the minimalistic approach in all my portable projects: ANSI-C90 or C99 (maximum), together with a standard Makefile. Nothing is more portable!

Quote:

Originally Posted by hop

Are the TRAP equates in ira off by one and missing an entry?

Indeed. Thanks. Fixed!

Current snapshot is here (verified): http://sun.hasenbraten.de/~frank/TEST/ira_211beta.lha

zenox98 · 19 December 2023, 18:43

@phx
Any reason for not including a ready-built exe in your releases?

hop · 19 December 2023, 18:51

Quote:

Originally Posted by phx

Tried that today. New option is called

-radix=n

, with n = 10 (default, decimal), 16 for hex and 0 for hex-positive, decimal-negative.
Unfortunately I realised that

itoasc()

is not called for

move

immediate addressing modes, which always call

itohex()

with a fixed number of digits (2, 4, 8). Maybe this makes sense, as, unlike displacements, you don't know whether you are dealing with a signed value.

Great. Are these changes supposed to be in the linked package? The contents look no different to the last one you shared.

EDIT: Sorry - my fault. The package has the changes.

Quote:

Originally Posted by phx

How does it work with Linux and OSX, when I only see a section for MSVC in it? (Being a Cmake noob.)

That block you are referring to is only required to disable a specific MSVC warning. Linux and Mac build fine with the default compiler options. If we wanted to add other platform/compiler specific options, this is where they would go. (btw Professional CMake is a very good introductory read if you are ever inclined).

Quote:

Originally Posted by phx

That's already one of the major problems I have here. Why should I use something which is less portable and adds more dependencies (cmake is not standard in a development environment)?

I really like the minimalistic approach in all my portable projects: ANSI-C90 or C99 (maximum), together with a standard Makefile. Nothing is more portable!

Haha. This is an interesting conundrum. CMake is quite ubiquitous in modern build environments, but unfortunately Amiga is not as well supported as Windows, Linux and Mac. I would perhaps argue that many users will be cross-compiling for Amiga and developing on a PC/Mac and will benefit from CMake support. I certainly benefitted from it when I wanted to prototype a new feature. The CMakeLists.txt will should not have to be modified unless a file is added or removed.

Quote:

Originally Posted by phx

Current snapshot is here (verified): http://sun.hasenbraten.de/~frank/TEST/ira_211beta.lha

Can you please confirm that this does actually contain some new changes? It still seems to have the var = var warnings present and I can't see any additional changes since the last package.

EDIT: Sorry - my fault. The package has the changes.

Thanks very much for your support.

hop · 19 December 2023, 19:45

Quote:

Originally Posted by phx

Tried that today. New option is called

-radix=n

, with n = 10 (default, decimal), 16 for hex and 0 for hex-positive, decimal-negative.
Unfortunately I realised that

itoasc()

is not called for

move

immediate addressing modes, which always call

itohex()

with a fixed number of digits (2, 4, 8). Maybe this makes sense, as, unlike displacements, you don't know whether you are dealing with a signed value.

Many thanks for this.

I've tested it and it seems to be almost working. I think there may be a problem with null-terminated strings in TEXT regions - the trailing zero ends up becoming a single $ with radix=16. Attached a repro with output in included txt file.

-radix=10

Code:

LAB_1105:
	;01105
	;DC.B	$20,$20,$20,$20,$20,$20,$20,$20,$20,$20,$20,$45,$4e,$54,$45,$52
	;DC.B	$20,$57,$4f,$52,$4c,$44,$20,$43,$4f,$44,$45,$00
	DC.B	"           ENTER WORLD CODE",0
	DC.B	$ff			;01121

-radix=16

Code:

LAB_1105:
	;01105
	;DC.B	$20,$20,$20,$20,$20,$20,$20,$20,$20,$20,$20,$45,$4e,$54,$45,$52
	;DC.B	$20,$57,$4f,$52,$4c,$44,$20,$43,$4f,$44,$45,$00
	DC.B	"           ENTER WORLD CODE",$
	DC.B	$ff			;01121

phx · 20 December 2023, 12:13

Quote:

Originally Posted by zenox98

Any reason for not including a ready-built exe in your releases?

I cannot compile for any possible target platform in the world. So I only include the portable source and let the user compile!

(Especially, I don't have a development environment for Windows, if you are refering to that.)

Quote:

Originally Posted by hop

I think there may be a problem with null-terminated strings in TEXT regions - the trailing zero ends up becoming a single $ with radix=16.

Indeed. There is a stupid assumption for the length of decimal numbers in the text output. Fixed.

I updated yesterday's download link with it.

zenox98 · 20 December 2023, 12:29

Quote:

Originally Posted by phx

I cannot compile for any possible target platform in the world. So I only include the portable source and let the user compile!

(Especially, I don't have a development environment for Windows, if you are refering to that.)

Thanks for reply.

I will investigate how to compile it for windows.

hop · 20 December 2023, 12:39

Quote:

Originally Posted by zenox98

I will investigate how to compile it for windows.

It's really easy. Open the x64 Native Tools Command Prompt for Visual Studio 2022 from the start menu. Change directory to the ira folder and type:

Code:

nmake -f Makefile.win32

Alternatively, use the CMakeLists.txt and gensln.bat file in the archive I shared earlier up this thread to generate native VS solution and project files and allow debug or release build config to be built easily.

Visual Studio Community edition is free and works fine.

hop · 20 December 2023, 12:41

Quote:

Originally Posted by phx

Indeed. There is a stupid assumption for the length of decimal numbers in the text output. Fixed.

I updated yesterday's download link with it.

Many thanks. This will be very convenient. I'll test ASAP.

hop · 22 December 2023, 11:15

Quote:

Originally Posted by phx

There is a stupid assumption for the length of decimal numbers in the text output. Fixed.

I updated yesterday's download link with it.

Thank you very much for the changes. I see what you've done - I thought that the hex/decimal numbering would just apply to offsets (in addressing modes), but it applies more globally, which makes sense.

The results are good, but I think one more tiny tweak could improve the readability of the output:

Code:

if (val > -10 && val < 10)
    always format as decimal

This would eliminate all the awkward looking $ on small numbers (including zeros) that coders would almost never write. For example

From:

Code:

MOVE.L	A3,$4(A4)
MOVEQ	#$0,d0
BSET	#$6,$E00(A0)
MOVEQ	#$FFFFFFFF,D0
DC.B	"graphics.library",$0
DS.L	$1

To:

Code:

MOVE.L	A3,4(A4)
MOVEQ	#0,d0
BSET	#6,$E00(A0)
MOVEQ	#-1,D0
DC.B	"graphics.library",0
DS.L	1

phx · 23 December 2023, 00:28

Quote:

Originally Posted by hop

Code:

if (val > -10 && val < 10)
    always format as decimal

No problem. Please try the update: http://sun.hasenbraten.de/~frank/TEST/ira_211beta.lha

hop · 23 December 2023, 15:47

Quote:

Originally Posted by phx

No problem. Please try the update: http://sun.hasenbraten.de/~frank/TEST/ira_211beta.lha

That is great thanks. Should EQU values be covered by the RADIX setting too?

phx · 25 December 2023, 00:06

Quote:

Originally Posted by hop

Should EQU values be covered by the RADIX setting too?

I don't understand. Can you give an example?

hop · 25 December 2023, 10:58

Quote:

Originally Posted by phx

I don't understand. Can you give an example?

No problem. Sorry for not being clearer. Please find a minimal repro attached.

test_equ_radix: 80fc1800

Code:

ira -a -compat=bi -binary -preproc test_equ_radix

Generates

Code:

; IRA V2.11 (Dec 23 2023) (c)1993-1995 Tim Ruehsen
; (c)2009-2023 Frank Wille, (c)2014-2019 Nicolas Bastien




	ORG	$0

SECSTRT_0:
	DIVU	#$1800,D0		;0: 80fc1800
	END

Add equate to config file:

Code:

MACHINE 68000
ENTRY $00000000
OFFSET $00000000
CODE $00000000 - $00000004

; The following line results in decimal value in equate:
; TRACK_SIZE_BYTES	EQU	6144
; even if -RADIX=16 is used
EQU TRACK_SIZE_BYTES $2.W

END

Results in this output, even if -RADIX=16 option is used:

Code:

; IRA V2.11 (Dec 23 2023) (c)1993-1995 Tim Ruehsen
; (c)2009-2023 Frank Wille, (c)2014-2019 Nicolas Bastien

; Custom equates (from config file)
TRACK_SIZE_BYTES	EQU	6144




	ORG	$0

SECSTRT_0:
	DIVU	#TRACK_SIZE_BYTES,D0	;0: 80fc1800
	END

Note that the EQU value is always displayed in decimal. Perhaps the -RADIX=16 should format this value as decimal:

Code:

ira -a -compat=bi -radix=16 -config test_equ_radix

Manually modified sample output:

Code:

; IRA V2.11 (Dec 23 2023) (c)1993-1995 Tim Ruehsen
; (c)2009-2023 Frank Wille, (c)2014-2019 Nicolas Bastien

; Custom equates (from config file)
TRACK_SIZE_BYTES	EQU	$1800




	ORG	$0

SECSTRT_0:
	DIVU	#TRACK_SIZE_BYTES,D0	;0: 80fc1800
	END

phx · 26 December 2023, 01:27

Quote:

Originally Posted by hop

Code:

; Custom equates (from config file)
TRACK_SIZE_BYTES    EQU    6144

I see. Can be done (try the update at the usual URL).

But there are two types of EQUs. The config-file definitions and the internal ones (interrupt vectors, custom chips, etc.). The latter will always be printed in hexadecimal, which is still unchanged. Should they follow the selected radix as well?

hop · 26 December 2023, 12:33

Quote:

Originally Posted by phx

I see. Can be done (try the update at the usual URL).

Thanks very much - that's looking much better. I was reluctant to use EQU before, because it often obfuscated the code with decimal!

Code:

diff --git a/Game.asm b/Game.asm
index ba9f8dc..b891fbb 100644
--- a/Game.asm
+++ b/Game.asm
@@ -1,4 +1,4 @@
-; IRA V2.11 (Dec 23 2023) (c)1993-1995 Tim Ruehsen
+; IRA V2.11 (Dec 26 2023) (c)1993-1995 Tim Ruehsen
 ; (c)2009-2023 Frank Wille, (c)2014-2019 Nicolas Bastien

        INCLUDE "hardware/custom.i"
@@ -60,8 +60,8 @@ SPR7PTH               EQU     $DFF13C
 COLOR00                EQU     $DFF180

 ; Custom equates (from config file)
-ciaicr EQU     3328
-TRACK_SIZE_BYTES       EQU     6144
+ciaicr EQU     $D00
+TRACK_SIZE_BYTES       EQU     $1800

Quote:

Originally Posted by phx

But there are two types of EQUs. The config-file definitions and the internal ones (interrupt vectors, custom chips, etc.). The latter will always be printed in hexadecimal, which is still unchanged. Should they follow the selected radix as well?

This is tricky - it's a global option, but it doesn't make any sense for hardware addresses to be printed in decimal, so I think we should avoid the temptation to convert these to decimal.

phx · 26 December 2023, 23:56

Quote:

Originally Posted by hop

This is tricky - it's a global option, but it doesn't make any sense for hardware addresses to be printed in decimal, so I think we should avoid the temptation to convert these to decimal.

I think you are right. Already changed it, but reverted that change now. There is an updated source at http://sun.hasenbraten.de/~frank/TEST/ira_211beta.lha

Now I also finished my work on PC-relative references, which point outside of the originating section. Instructions are generated with constant displacements instead of

DC.W

.

hop · 27 December 2023, 18:23

Quote:

Originally Posted by phx

I think you are right. Already changed it, but reverted that change now. There is an updated source at http://sun.hasenbraten.de/~frank/TEST/ira_211beta.lha

Now I also finished my work on PC-relative references, which point outside of the originating section. Instructions are generated with constant displacements instead of

DC.W

.

There seems to be a problem with the latest changes.

Code:

diff --git a/Game.asm b/Game.asm
index b891fbb..ac62fea 100644
--- a/Game.asm
+++ b/Game.asm
@@ -1,4 +1,4 @@
-; IRA V2.11 (Dec 26 2023) (c)1993-1995 Tim Ruehsen
+; IRA V2.11 (Dec 27 2023) (c)1993-1995 Tim Ruehsen
 ; (c)2009-2023 Frank Wille, (c)2014-2019 Nicolas Bastien

        INCLUDE "hardware/custom.i"
@@ -60,8 +60,8 @@ SPR7PTH               EQU     $DFF13C
 COLOR00                EQU     $DFF180

 ; Custom equates (from config file)
-ciaicr EQU     $D00
-TRACK_SIZE_BYTES       EQU     $1800
+ciaicr EQU     $89698DC8
+TRACK_SIZE_BYTES       EQU     $89698DC8

I think I have already shared these files, but let me know if you need a repro.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
ira for Windows	jotd	Coders. General	63	12 December 2023 10:53
New tutorial on sprites	Yragael	Coders. Tutorials	8	04 September 2023 21:00
Debugging messages in serial (small tutorial, mainly for cross-dev)	alkis	Coders. Asm / Hardware	7	22 February 2016 14:16
68020 to 68000 code convertion using Ira and PhxAss	gulliver	Coders. Asm / Hardware	18	12 April 2014 01:09
Looking for IRA v1.07 or newer :-)	voxel	request.Apps	7	30 July 2008 01:39

19 December 2023, 18:43	#85
zenox98 Joy Division Join Date: Nov 2006 Location: East Yorkshire Age: 60 Posts: 241	@phx Any reason for not including a ready-built exe in your releases?

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)