Scorpion Engine

[Lemming880]

@earok: Congrats on releasing 2022.8 Final! That took a while.

[earok]

Cheers @Lemming880. It really did - I should try and stick to a monthly release cycle.

---

Checked in a quick fix to experimental for projectile->actor collisions (reported by Lemming880). There are a few more issues I need to work through.

[skyzoo73]

I noticed a possible bug in the 128-color aga palette:
if one range of colors is selected for sprites (spr) and
then another range, perhaps the next one in the palette,
all colors are marked as SPR even if the range is later,
changing between ranges leads to a situation where almost all
can be marked as SPR.

For example I am now with the range 17-31 but only the colors before 17
are normal, the later ones up to 128 are marked SPR.

[earok]

Oh whoops, that's an embarrassing error.

Latest experimental fixes the bug reported by Skyzoo above where colors at the end of a long palette would be incorrectly flagged as sprite.

More than that, the colors that are used by each sprite channel are labelled as such (and the colors that are only available to 16 color sprites are also labelled that way - note that 16 color sprites can use ANY of the colors labelled as SPR).


There's still some serious bugs in experimental reported by Lemming880 that I haven't yet investigated.

---

I probably should note the reason why I never noticed the bug before - I always put the sprite color range at the end of the palette. You can't move it on OCS of course, but on AGA I'd typically put the sprite colors beyond the bitplane colors. So on a 7 bitplane (128 color) game, I'd put it beyond the 128th color, essentially upgrading the palette to 144 color without any additional overhead.


---

Edit: Another update. Latest experimental is a bit smarter about memory usage on panels.

Still major outstanding issues from Lemming880 (and now BasementApe) to investigate.

[earok]

Up way too late but I've done a check in to experimental: Fix for issue reported by Lemming880 where certain actors would float through walls in certain positions.

Still more issues from Lemming880 and BasementApe to investigate.

[earok]

Latest check in fixes issue reported by Lemming880 and BasementApe where sometimes there can be graphic corruption if an actor is blitted on both sides of the graphics buffer

Also, fix for issue reported by Lemming880 - block and animated tiles would often freeze in the middle of their animation

I think this solves all of the major reported bugs with the new experimental but I'm not totally certain of that

---

I have to admit a major error with the previous test scores - the actor vs projectile collision was initially broken, and that's a major part of the performance test. As such, the real scores are:

LATEST SCORES
A500 = 471 (55% improvement on base, 9% improvement on last)
A1200 = 684 (40% improvement on base, 4% improvement on last)
A1200+Fast = 1176 (20% improvement on base, 2% decrease on last)

For the most part, there's still some considerable performance gains, especially on lower end devices - the test is definitely more fluid on A500 than it's been in the past.

I can't quite explain why it seems to be marginally slower on A1200+Fast (I guess maybe some of the optimisations I made were only good for when there's no FastRAM available). It's possible that the previous test scores were also bad for some reason, or there could have been something that slipped in unrelated to the recent optimisations that caused that slowdown. (The performance test is a bit odd in how it stress tests collisions, so it may not be a good indicator of actual game performance on A1200+Fast anyway)

Having said that - I'm still experimenting, and I'm hoping to boost the test scores at every level (as long as there's a single line that could be converted to ASM, there's room for improvement)

[domkid]

Originally Posted by earok

- Issue reported above by SkyZoo and others where editing the palette inside Scorpion Engine would break the PNG in some image editors (specifically, it would result in an invalid CRC). Simply editing a corrupted palette by changing a color within Scorpion should fix it for external editors.

- Fix for long time issue where 256 pixel wide sprite AGA backgrounds (see: Gothicvania, Vertical Parallax AGA) would occasionally dip one pixel vertically.

thank you for the 2 fixes, I didn't want to request it

[earok]

No worries!

---

Latest experimental:

* Fix for issue reported by DomKid - Actors can leave the left side of the map
* Added missing command to actors menu for changing the controller (eg, if you want to set an actor to use a different joystick to normal in codeblocks)
* Another round of optimizations and overhauls (in particular, A500 speed is again now more than it used to be under A1200). Though again - the performance upgrades are likely to come with some breaking bugs.

LATEST SCORES
A500 = 509 (67% improvement on base, 8% improvement on last)
A1200 = 761 (56% improvement on base, 11% improvement on last)
A1200+Fast = 1348 (39% improvement on base, 15% improvement on last)

[earok]

Latest experimental:

- Fix for Basement Ape: Unable to use "player_animation" etc in IF statements.
- Friendlier error message if attempting to compile when WinUAE has the cartridge file locked (I think AcidBottle reported this issue originally).
- Various actor fields don't show in the "set variable" drop down anymore (these ones can all be set via the actor menus).
- Removed redundant "data" folder from floppy drive export

[earok]

Latest experimental:

- Fix for Basement Ape (I hope!). Was an issue where, after changing an actor's type, it didn't immediately set the new animation. Hopefully that's okay now.
- Some improvements for codeblocks 68K compilation - your mileage may vary, but on my test case the virtual machine compiled to be 5% smaller than it was.

[earok]

Latest experimental brings a fix for a bug Basement Ape reported (maybe related to some other issues reported by BitBeamCannon). Commands for manipulating the "new" actor were broken, they should be fine now.

[ImmortalA1000]

I hope to have time soon to be able to sit down and give this a go, looking for games to port as a simple part level demo to get used to this awesome tool already.

Would appreciate any tutorials (sites or videos) for people who have never used anything like this to read up about.

[malko]

My only regret is that it's not working on x86 W7.

[Mixel]

Originally Posted by ImmortalA1000

I hope to have time soon to be able to sit down and give this a go, looking for games to port as a simple part level demo to get used to this awesome tool already.

Would appreciate any tutorials (sites or videos) for people who have never used anything like this to read up about.

So far I think the only thing even remotely like tutorials are Wei-Ju Wu's video..
[ Show youtube player ]
Actually that's the only real tutorial probably?

Mike from Bitbeamcannon and Erik's series where they talk about scorpionboy dev
[ Show youtube player ]

And Earoks various streams and feature videos..
[ Show youtube player ] - its worth watching all of them.. Though the further back you go the more different to the current version things get.

You want to install Tiled to edit maps, something to do .csv spreadsheets (libreoffice?) .. Something to do pixel art ( https://libresprite.github.io/#!/ is a free fork of Aseprite, but there are soo many other options) - And probably Audacity for audio.. OpenMPT for making music? Some of these are totally to your preference but that's a good starter pack.

[earok]

Welcome ImmortalA1000! Unfortunately yes, tutorials/documentation is a bit sparse right now, but Mixel's suggestions are all good ones. There's also a lot of demos to toy around with, also happy to help out however we can.

---

Latest experimental:

- Fix for bug reported by Mixel wherein blocks above 128 would be broken
- Fix for bug reported by BitBeamCannon wherein player Melee->Block collision was't being detected
- Fix for bug reported by BitBeamCannon wherein setting the Actor_Lookdir (etc) variable directly wasn't working
- Some minor optimizations relating to blocks

[ImmortalA1000]

Thank you earok and Mixel. And of course thank you for making this awesome tool earok!

I was a little worried about minimum PC specs when I found out it was written with Unity, I have an AMD A10-9700 APU 8gb PC so hopefully that will be OK.

I have been using Paintshop Pro 8 for well over a decade so I got that covered, also HAMconvert lets me do copperlist tests, which I usually split to 16+16 colours with 'sprites' in 16 colours and backgrounds in 16+copperlist for side scrollers like Robocop 2 arcade. Just want to take things a stage further than proving the Amiga 1000 could technically display nice renditions of certain arcade games.

I have spent about a month looking for a source project to convert, my first idea was a single level silent demo of Castle of Illusion but not all the graphic assets have been ripped. There are some great website with entire arcade, and other system, game background ripped (but not always as separate foreground/background parallax planes). What I do know is it has to be a game I love playing, I think that will help with motivation for me.

My hard drive is full of all sorts of graphic assets from arcade or 16 bit console games thanks to VGmaps and spriters-resource websites.

I am also not sure in general about certain technical considerations, like the performance impact of having a repeating background of 4x 15 colour sprites (saves me wasting palette colours when using 8x 3 colour sprites etc) so there is 100% of learning to do on my part.

I did watch some impressive demos for things like Amiga the Fox and also the Bitbeamcannon video which has some interesting aspects pointed out but I am at ground zero, never even used SEUCK etc in my life.

[acidbottle]

Take some of the Scorpion demos, pull them apart, change things, discover how the UI works then evolve from there.

Then come on here and ask questions when you get stuck. Will all eventually click, though of course we all are learning the system as we go along

Have said it a hundred times to, its an awesome community and very very helpful. Good luck, ports are a nice way to get to grips with this.

[earok]

Agreed with @AcidBottle, I'm very fortunate that there's such a mutually supportive community around Scorpion. Good luck with your Scorpion endeavors @ImmortalA1000, and we'll help how we can.

---

Latest experimental fixes an issue discovered by Basement Ape:

If a load level is triggered multiple times in a single frame (for example, if the player overlaps multiple "load next level" blocks), it may cause the level load to be triggered multiple times.

This didn't used to be a problem because Scorpion would check every time custom code was run if the level needed to be reloaded, though this added substantial overhead to projects that extensively use codeblocks so I've removed it.

The fix is actually slightly more optimal than the last build was, delaying all level loads to the next frame (and if multiple level loads are triggered, one will overwrite the other). It should be rock solid now, but do let me know if there's any issues loading levels (particularly if there's something like a level loaded after a yield or a gosub)

[domkid]

Originally Posted by ImmortalA1000

I have spent about a month looking for a source project to convert, my first idea was a single level silent demo of Castle of Illusion ...

I had made a small demo, some time ago...

https://discord.com/channels/8793381...07799672098917

[earok]

Latest experimental

* Panel redraw is split across frames, rather than all at once (which is to say, the panel is wiped on every second frame and drawn on every other frame). This is something mandatory that I may make optional in future (incase bars emptying or filling is noticeably less smooth, for example), but it should give at least a small boost on games where variables on panel elements are changed frequently.

* Fixed bug wherein at least some of the time, loading a level without a level already being loaded would crash (not fully clear on the circumstances behind this).

* Fixed bug wherein Quit command would exit without fading out level

* Redid a previously cancelled optimization to do with collision detection (I removed it originally because it broke collisions with the left side of the map, but I worked out the cause and fixed that).

* New feature - Ability to set which channels are reserved only for music (and thus, SFX will never use those channels). I'm not totally sure it's working as I expected, but I did confirm that if I reserve all channels for music, no SFX will play.