Enhancement - AGA Support. by Vairn · Pull Request #268 · AmigaPorts/ACE

Vairn · 2026-04-22T13:13:09Z

Description

This PR merges the AGA branch back into main.

It brings in:

new sprite/advanced sprite manager work
viewport/camera/scrollbuffer updates
palette/extview utility changes
related docs updates
the ability to create code targeting AGA chipsets, including support for screens up to 8bpp

AGA support is enabled by configuring CMake with -DACE_USE_AGA_FEATURES=ON.

Closes #151.

Motivation and Context

The AGA work is now broad enough that keeping it separate is slowing everything down.
Merging now keeps development on one branch and prevents further drift/conflicts.

How Has This Been Tested?

Clean configure + build succeeds
Updated modules compile together (sprite, advancedsprite, viewport, camera, palette, extview)
Manual runtime smoke test completed for sprite rendering and viewport scrolling behavior
Verified AGA-targeted build path using -DACE_USE_AGA_FEATURES=ON, including 8bpp screen support
Docs reviewed against current APIs

Types of changes

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)

Checklist

My code follows the code style of this project.
My change requires a change to the documentation.
I have updated the documentation accordingly.

UNTESTED but I loaded a 256 colour bitmap or 4.

Fixed bitshift for AGA Palette.

finished my comment...

This reverts commit 4468737.

Allocate the ram for the palette depending on the Depth, AGA and ECS use different palette types. Now Germz shows the correct colours again. Free's the palette memory when the view gets destroyed.

removed union, use the two varibles. Added a feof to mini_std. worked for my little test, probs not optimised.

Creating the View, you now need to specify if your using the AGA chipset. Now you can use 24bit colours on all bit plane counts. (in theory)

Added KILLEHB if viewport has AGA enabled and is 6 bpp.

Removed Extra AGA Palette field.

… to setup the display. AGA needs more time so the value needs to be smaller.

This commit refactors the code in palette.cpp to support AGA Palette. Specifically, it adds a new else block that writes alpha and color values for AGA Palette.

The commit fixes a bug where the wrong variable was being passed to logWrite function. The variable name has been corrected to match the intended value.

This commit updates the value of s_pCopperWaitXByBitplanes[8] from 0xC6 to 0xA0. This change is made to optimize fetchmodes for AGA in 8bpp mode.

Enhance README.md to include information about AGA support and its setup. Add AGA support link in docs/README.md and provide build instructions for enabling AGA features in docs/installing/ace.md. Update docs/programming/view.md with details on AGA tags for view/viewport creation.

tehKaiN · 2026-04-22T13:48:42Z

Great! I'll check if my games are breaking on this branch, and if not then we're more or less ready to go.

I also encourage all bystanders to do the same.

kwn · 2026-04-23T23:12:26Z

I want this PR to be merged so badly :D Thanks for adding it!!

Vairn · 2026-04-23T23:19:31Z

I want this PR to be merged so badly :D Thanks for adding it!!

Well you can grab my fork of ACE and do some testing so we know it is stable enough to merge.

More testing faster merging.

https://github.com/Vairn/ACE/tree/AGA

kwn · 2026-04-23T23:33:04Z

The thing is, that I don't have anything to test yet ;) I was wondering if I should rewrite my game using ACE, as I'm struggling to make my scroller working without occasional glitches - however, the game uses AGA features. Up until now I didn't consider ACE, but with that PR I'd need to re-thing it.

Out of curiosity

Does fmode 1 and 3 work with infinite scroller?
Are all 8 hardware sprites available in fmode 1/3? Is it even possible with early DDF?

Vairn · 2026-04-23T23:39:19Z

Well, ACE has a working scroller we have a channel on a discord server if you want any help and advice. https://discord.gg/nuN7wdKEC Edited to remove all the stuff added via replying via email lol.

Vairn · 2026-04-23T23:48:23Z

Out of curiosity

Does fmode 1 and 3 work with infinite scroller? It should, infinite scrollers can have issues with timing and stuff even on Ocs/ECS.

Are all 8 hardware sprites available in fmode 1/3? Is it even possible with early DDF?
Yes and no, yes it is possible, but you may have dma fetch issues with timing.

In FMODE 3:
Bitplane DMA tries to grab multiple adjacent slots
Sprites need specific exact slots
If bitplane DMA “locks” the bus too long: Sprite fetch misses its deadline window.

Even though priority says:
Sprites > Bitplanes
In reality:
FMODE 3 packs bitplane DMA tightly
Arbitration granularity isn't infinite
If timing margins are exceeded:
Later sprites lose fetch slots
Typical failure pattern:
Sprite
Stability
SPR0–2
Solid
SPR3–5
Risky
SPR6–7
First to break.

With everything Amiga there is always a compromise.

I normally think if you want fmode 3, then you might as well think you've lost the last two sprites.

But as with anything mileage may vary.

Btw, I haven't really tested fmode 1, just mode 3.

kwn · 2026-04-24T06:34:02Z

Thanks for the response. Based on my experience, even with fmode = 1 I'm only able to use up to 5 slots on both FS-UAE and real Amiga. The remaining ones don't get the DMA cycles due to the early DDF and pre-fetch.

Thanks for sharing the link to the Discord. I'm a real boomer and never managed to use it. I'll take a look.

…e UWORD Change the parameter types from UBYTE to UWORD for maximum color length in paletteLoadFromPath, paletteLoadFromFd, paletteDim, and paletteDump functions. This improves consistency and allows for a broader range of color counts. Update related logic to accommodate the new types.

kwn · 2026-04-26T20:47:04Z

Hey @Vairn . I tried using the AGA branch in my project and I think I found a bug. I opened a PR with a fix. Would you mind taking a look and merge if it makes sense to you, please?

Vairn · 2026-04-26T21:19:23Z

Hey @Vairn . I tried using the AGA branch in my project and I think I found a bug. I opened a PR with a fix. Would you mind taking a look and merge if it makes sense to you, please?

Awesome, I'll have a look later once I get to Work.

Fix fmode 1 and 2

Fix aga bplcon0 bpu3

Added a new function to calculate DDF step based on the viewport's bitplane mode. Updated scroll buffer processing and reset logic to accommodate AGA features, ensuring correct handling for FMODE 1 and 2. Cleaned up whitespace in simple buffer initialization.

Updated the viewLoad function to streamline AGA feature support. Simplified the logic for setting bplcon0 and bplcon2 based on the viewport's bit depth, ensuring correct handling for 64-color modes. Removed redundant conditions and improved code clarity.

Refined the logic for setting bplcon0 and bplcon2 in the viewLoad function to improve AGA feature support. Added checks for VP_FLAG_HIRES and VP_FLAG_AGA to ensure correct configuration based on viewport flags and bit depth. This update enhances clarity and correctness in handling 64-color modes and AGA-specific settings.

…unctions Updated the calculation for colour banks in viewUpdateGlobalPalette to ensure correct handling of bit depth. Enhanced comments in viewLoad to clarify the configuration of bplcon0 and bplcon2, specifically regarding VP_FLAG_HIRES and VP_FLAG_AGA. This improves the clarity and correctness of AGA feature support.

I was testing with AGA enabled, which also enabled ECS.

tehKaiN

first minor and non-exhaustive review - I'd get rid of palette v1 support completely and at most leave detection of v1 and print the error on debug targets. User is expected to use latest convert tools at all times.

Not sure if we even need to document the old format. Having docs for new one is a nice to have.

…ures - Removed support for legacy (v1) `.plt` files; users must reconvert assets to v2 format. - Updated documentation to reflect changes in palette loading and saving functions. - Enhanced AGA palette handling in `palette_conv` and related functions to ensure compatibility with v2 formats only. - Improved error handling for unsupported legacy formats in the palette loading process.

- Updated documentation for spriteSetOddColourPaletteBank and spriteSetEvenColourPaletteBank to improve clarity. - Refined the logic for setting the palette banks in sprite functions to ensure correct handling of AGA features. - Improved logging in viewCreate to include AGA-related flags for better debugging and visibility of feature usage. - Fixed the deletion of the palette to check the views AGA flag, not the first VPORT.

tehKaiN

Code looks kinda okay, now needs testing in my games.

The diff would be way way smaller if you haven't wrecked the formatting. Please fix it - extra spaces near parentheses are okay if you feel attached to them, other stuff must go.

tehKaiN · 2026-05-04T06:44:10Z

+	print("\tplt\tACE palette (default)\n");
 	print("\tpng\tPalette preview\n");
-	print("extraOpts:\n");
-	print("\t-cc\tConvert colors. Truncate non-OCS colors to OCS precision if necessary\n");


That's required to stay. The intended behavior is that palette_conv for ocs catches bad colors in input palette and only if you don't care about them you add a switch that will auto-downgrade colors to 12-bit.

This is due to bitmap_conv not auto-fitting the images to palette, and you want to catch bad colors where they really are - with auto-fixed palette bitmap_conv will complain about palette-image color mismatch even though the image might be correctly 12-bit rgb.

tehKaiN · 2026-05-04T06:47:00Z

-	print("\t-cc\tConvert colors. Truncate non-OCS colors to OCS precision if necessary\n");
 }

-int main(int lArgCount, const char *pArgs[])


I really need to create a clangformat thingy. Plz minimize the diff. Extra whitespace is okay but "stickiness" of ptr/ref should be the same.

tehKaiN · 2026-05-04T06:47:34Z

 tools/*/*.png
 tools/*/*.bm
 tools/*/*.plt
-


what's this for?

tehKaiN · 2026-05-04T06:50:10Z

+/** First byte of .plt v2: ECS/OCS packed entries (2 bytes per colour). */
+#define PLT_NEW_ECS 0
+/** First byte of .plt v2: AGA entries (4 bytes per colour, alpha+R+G+B). */
+#define PLT_NEW_AGA 1


does it need to be in .h file and visible in other files that include them? or can we just have it in palette.c?

tehKaiN · 2026-05-04T06:51:03Z

+ * @brief Saves ECS/OCS palette into .plt v2 file (PLT_NEW_ECS + BE count + packed colours).
 */
-void paletteSave(UWORD *pPalette, UBYTE ubColorCnt, char *szPath);
+void paletteSave(const UWORD *pPalette, UWORD uwColorCnt, char *szPath);


Suggested change

void paletteSave(const UWORD *pPalette, UWORD uwColorCnt, char *szPath);

void paletteSaveOcs(const UWORD *pPalette, UWORD uwColorCnt, char *szPath);

tehKaiN · 2026-05-04T07:13:18Z

@@ -92,7 +92,17 @@ long ftell(FILE *pStream) {

 int feof(UNUSED_ARG FILE *pStream) {
 	// FIXME: implement


it's fixme'd, you can now remove this comment.

Have you confirmed that it returns true only after the last byte was read?

tehKaiN · 2026-05-04T07:15:53Z

+	LONG lPos = Seek((BPTR)pStream, 0L, OFFSET_CURRENT);
+	Seek((BPTR)pStream, 0L, OFFSET_END);
+
+	LONG lEnd = Seek((BPTR)pStream, 0L, OFFSET_CURRENT);


Ain't Seek() return the current position at all times?

Suggested change

LONG lPos = Seek((BPTR)pStream, 0L, OFFSET_CURRENT);

Seek((BPTR)pStream, 0L, OFFSET_END);

LONG lEnd = Seek((BPTR)pStream, 0L, OFFSET_CURRENT);

LONG lPos = Seek((BPTR)pStream, 0L, OFFSET_END);

LONG lEnd = Seek((BPTR)pStream, 0L, OFFSET_CURRENT);

tehKaiN · 2026-05-04T07:17:35Z

+	if (lPos == lEnd)
+		return 1;
+
+	return 0;


Suggested change

if (lPos == lEnd)

return 1;

return 0;

return lPos == lEnd;

tehKaiN · 2026-05-04T07:19:53Z

+constexpr std::uint8_t PLT_NEW_ECS = 0;
+constexpr std::uint8_t PLT_NEW_AGA = 1;
+
+static void writeUwordBE(std::ofstream &Dest, std::uint16_t uwValue) {


same as in main ace code - how about reading/writing a 2-byte thingy and using endian.h fns to swap them?

American spelling of Color where I use the Colour version. Fixes for badly formatted code.

Vairn and others added 30 commits June 3, 2021 10:24

WIP Testing AGA

1ee276f

UNTESTED but I loaded a 256 colour bitmap or 4.

Update extview.c

4468737

Fixed bitshift for AGA Palette.

Update extview.h

80cd224

finished my comment...

Revert "Update extview.c"

b03e8c1

This reverts commit 4468737.

Dynamically alloc mem for palette

780dd53

Allocate the ram for the palette depending on the Depth, AGA and ECS use different palette types. Now Germz shows the correct colours again. Free's the palette memory when the view gets destroyed.

Fix for Union

4f8fbe2

removed union, use the two varibles. Added a feof to mini_std. worked for my little test, probs not optimised.

Added AGA Tag, and Viewflag.

9bf9215

Creating the View, you now need to specify if your using the AGA chipset. Now you can use 24bit colours on all bit plane counts. (in theory)

Merge branch 'ptplayer_c' into AGA

4a32578

Merge branch 'master' into AGA

20a5980

Update extview.c

4c843e0

Added KILLEHB if viewport has AGA enabled and is 6 bpp.

Merge branch 'NTSC' into AGA

beaf4a4

Removed Extra AGA Palette field.

a7b8330

Removed Extra AGA Palette field.

Merge branch 'AmigaPorts:master' into AGA

8e73348

Merge branch 'master' into AGA

715e93d

Fix for AGA depth modes and how long the copper waits before it stars…

07780bd

… to setup the display. AGA needs more time so the value needs to be smaller.

Merge remote-tracking branch 'upstream/master' into AGA

dcfacad

Merge branch 'COPPER_X_WAIT_FIX' into AGA

c9c89bc

Colour Dimming for AGA

033db93

Merge remote-tracking branch 'upstream/master' into AGA

1e2578d

Merge branch 'master' of https://github.com/AmigaPorts/ACE into AGA

c1e6e57

Update Palette.c - added volatile Keyword

ddd5ded

Merge branch 'AmigaPorts:master' into AGA

e9c9bea

WIP Updating Tools

dcb4f2d

AGA Palette Loading

cbba258

Merge branch 'AGA' of https://github.com/Vairn/ACE into AGA

7e1e879

Refactor palette.cpp to support AGA Palette

9703496

This commit refactors the code in palette.cpp to support AGA Palette. Specifically, it adds a new else block that writes alpha and color values for AGA Palette.

Merge branch 'master' into AGA

b0873e2

Fix logWrite argument in paletteLoad function

3303a24

The commit fixes a bug where the wrong variable was being passed to logWrite function. The variable name has been corrected to match the intended value.

Merge remote-tracking branch 'upstream/master' into AGA

b6d2349

Update s_pCopperWaitXByBitplanes for 8bpp mode

ed424f0

This commit updates the value of s_pCopperWaitXByBitplanes[8] from 0xC6 to 0xA0. This change is made to optimize fetchmodes for AGA in 8bpp mode.

fix(extview): set safe default for fmode when no view is loaded

be4de5b

Vairn and others added 2 commits April 25, 2026 15:10

Fix AGA bplcon0-bpu3

7e4504f

Fix fmode 1 and 2

4e114cc

Vairn and others added 9 commits April 27, 2026 08:57

Merge pull request #8 from kwn/fix-fmode-1-and-2

6ac2a41

Fix fmode 1 and 2

Merge pull request #7 from kwn/fix-aga-bplcon0-bpu3

d05622f

Fix aga bplcon0 bpu3

Merge branch 'AmigaPorts:main' into AGA

b3a7b8c

Merge branch 'AGA' of https://github.com/Vairn/ACE into AGA

96bffc8

Revert those changes, was a user error :/

591ab6c

I was testing with AGA enabled, which also enabled ECS.

tehKaiN requested changes Apr 27, 2026

View reviewed changes

Comment thread docs/programming/aga.md Outdated

Comment thread docs/programming/aga.md Outdated

Comment thread docs/programming/view.md Outdated

Comment thread src/ace/utils/palette.c Outdated

Vairn added 4 commits April 28, 2026 11:17

Fixed the double application of the Modulo modifier.

ff2118e

Merge branch 'AGA' of https://github.com/Vairn/ACE into AGA

079b192

tehKaiN requested changes May 4, 2026

View reviewed changes

Peer Review Pass

a871883

American spelling of Color where I use the Colour version. Fixes for badly formatted code.

	void paletteSave(const UWORD pPalette, UWORD uwColorCnt, char szPath);
	void paletteSaveOcs(const UWORD pPalette, UWORD uwColorCnt, char szPath);

		@@ -92,7 +92,17 @@ long ftell(FILE *pStream) {

		int feof(UNUSED_ARG FILE *pStream) {
		// FIXME: implement

Conversation

Vairn commented Apr 22, 2026

Description

Motivation and Context

How Has This Been Tested?

Types of changes

Checklist

Uh oh!

tehKaiN commented Apr 22, 2026

Uh oh!

kwn commented Apr 23, 2026

Uh oh!

Vairn commented Apr 23, 2026

Uh oh!

kwn commented Apr 23, 2026

Uh oh!

Vairn commented Apr 23, 2026 via email • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Vairn commented Apr 23, 2026

Uh oh!

kwn commented Apr 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kwn commented Apr 26, 2026

Uh oh!

Vairn commented Apr 26, 2026

Uh oh!

tehKaiN left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tehKaiN left a comment

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!

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!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Vairn commented Apr 23, 2026 via email •

edited

Loading

kwn commented Apr 24, 2026 •

edited

Loading