Contributor bounties / feature requests

[skmp]

Issues for bounty hunters

Issues that have asked for bounties

BountySource Issues view

BountySource Team View

---

Want to back this issue? Post a bounty on it! We accept bounties via Bountysource.

[skmp]

@twinaphex do you have any links handy?

[inactive123]

Sure thing.

https://www.bountysource.com/issues/43656021-make-libretro-reicast-work-on-arm-ios-android-and-32bit-pc

I'm completely willing to throw more bounty funds behind specific Reicast milestones. Some bounties I can think of that I would be completely OK with throwing some money behind -

• Fog table emulation in the GL renderer. This can be an optional feature, it can even require certain GLES 3 and/or GL3 extensions, as long as it's something that can be toggled on/off.

• An optional order independent transparency codepath for the most accurate transparency handling (can be DX11/GL 4.5-specific, it can be slower, as long as it is optional it doesn't matter to me personally at what performance cost it comes. It would be nice to have an accurate but slow codepath for this)

• Depth buffer solutions - this can still be a pretty big problem right now. Some game still need near/far Z tweaking in order for the depth rendering to appear correctly, and worse, this seems to be GPU-specific. I hear that 'depth peeling' could be a potential solution there?

• Bounties for fixing some of the self modifying code bugs such as the crashes in Shenmue.

• Improved Naomi / Windows CE emulation.

Let me know if any of this is stuff that would be interesting to you and other Reicast members, I can start allocating some funds behind it then. I am sure a lot of users would be willing to throw their weight behind such things too if we created enough coverage for it.

[JoeMatt]

Why is the bounty for 3 different platforms though? I can definitely make it run on iOS (does it need to be jailbreak free?) but wouldn't know to start with the others.

[AbandonedCart]

@JoeMatt That is exactly why it is a bounty, though. The idea is to provide enough motivation to have it fixed for every platform, not just one.

[JoeMatt]

Yeah but who has enough experience in all those disciplines in some of the most niche and difficult aspects of each. CoreAudio alone is a major pain, I've written about 5 commercial core audio based apps and it still gives me nightmares.

[AbandonedCart]

I don't know what to tell you, honestly. It (being a pain) is the same reason I don't bother with it lol

[JoeMatt]

I agree, hate to get the people’s hope up that such an enigma exists though. Maybe set some realistic expectations would be more engaging for the community.

Who knows, maybe there is such a person lurking in the shadows

[inactive123]

You dont have to touch Coreaudio at all in a libretro core, which is what the bounty is for. Honestly, a lot of the codebase doesnt cater to platform specific APIs when making a libretro core.

[blackman91]

I would definitely contribute to a bounty to make cannon spike playable on reicast:
#898

It works on nulldc.

[skmp]

@twinaphex can people submit bounties somehow?

On Thu, 17 May 2018 at 02:46, blackman91 ***@***.***> wrote: I would definitely contribute to a bounty to make cannon spike playable on reicast: #898 <#898> It works on nulldc. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1144 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAYAMukh1EdjXBW3rAklJySKE3H_t3jnks5tzLpugaJpZM4TzhXN> .

-- ~skmp

[inactive123]

Sure, I just made one for the Cannon Spike compatibility. I will be asking @hunterk to drop some money into it just to get the ball rolling, hoping for others to chip in too.

Users can create a Github issue here -

https://github.com/libretro/reicast-emulator/issues

I can later attach the bounty label to it then. Users can then go to Bountysource, input the URL of the Github issue, and then attach money to it.

[inactive123]

A question before I start creating bounties and adding some initial seed money -

Since a lot of the visual problems with these games come down to unemulated GPU features, what should be the nature of these bounties?

Enduser-oriented where the main subject is game-specific and trying to basically get it to render correctly which could be part of a much larger problem? This could bring unrealistic expectations with each and every bounty though and a moving target.

Or technically oriented ones, more vague stuff that an enduser would not necessarily understand that are nevertheless approachable and could be managed in bite sized chunks by contributors.

For instance, individual bounties for -

• implement volume shadows
• implement clipping
• implement fog tables
• etc.

Etc. You guys let me know which one you prefer. I could start creating separate bounties then for these individual elements where the GL renderer is currently lacking.

Requirements for these advanced features could be - optional features that require a GL featureset beyond GLES 2.0. Fog table could be implemented through an UBO for instance.

Once it's decided which direction we are going to go in, I'm going to start allocating $50 to a $100 worths on initial seed money on some of these bounties.

[skmp]

I think we can start with game-specific ones, and we can move the money to issue-specific ones once the initial research is done. We can use a bug thread per bounty once work on them begins, and we can track them via tickets, merge them to other tickets etc. How does that sound?

[inactive123]

Sounds good. Tell me some issues that I should either create or that already exist which you would want the $100 to go towards. You can also tell me if I should give it all to one bounty or spread it across several ones.

[AbandonedCart]

I agree with @twinaphex's initial point in that game-specific bounties could spiral out of control. Someone fixes one game and then the person solving it unaware it solved another, or someone working on another game is out the bounty because it is no longer available. Unless there is some huge rush to immediately start spamming bounties, it would probably be best to figure out at least enough of the issue to categorize it instead of just throwing cash at titles.

[skmp]

@LoungeKatt I think people should track the approximate hours worked, so that we can fairly split the bounties. How does that sound?

Eg, If you solve multiple issues you get extra cash, if you want to, but you can leave it to the pool for other issues if you think it was "easy" or didn't take long enough

[skmp]

@twinaphex I created the "bounty requested" and "bountry granted" labels so we can experiment with the tracking. To users: Please mark bounty requests with the "bounty requested" label

[AbandonedCart]

It's going to be fun to watch this unfold. All I can picture is that scene in Office Space where the two guys interview everyone to determine their role in the company.

[skmp]

Lol. Well, ideally the game specific stuff will be moved to the gamedb sub project. But that takes time.

[AbandonedCart]

"Yeahhhhh... I'm gonna need you to come in on Saturday....."

[skmp]

/me passes @loungekatt some pop corn

[blackman91]

Idea: Can we create a bounty to fix reicast closing the game if you press the recent apps or home buttons on Android? That seems more doable.

[inactive123]

Alright, here is how I currently spent the first $66 -

Spread it across two issues -

Shenmue 1/2 crashes (dynarec-related) ($50)

libretro#28

Cannon Spike graphical glitches ($10)

libretro#27

I will soon be spending another $50 entirely dedicated to Reicast issues. Let me know if it should be added to new/existing issues, or added to these existing bounties in order to make the amount worth a bit more. Entirely up to you guys and community feedback.

[inactive123]

Added two more bounties -

Implement volume shadows ($40)

https://www.bountysource.com/issues/58559302-implement-volume-shadows

Add order independent transparency codepath ($15)

https://www.bountysource.com/issues/58561426-order-independent-transparency-codepath

I can add some more funds at the start of June.

[AbandonedCart]

@blackman91 I think the idea is to motivate fixing more long-time, universal, deep-seeded issues.

[AbandonedCart]

@flyinghead Well whatever motivated you, just don't forget to circle back and sign the CLA.

[AbandonedCart]

Speaking of which, there should probably be some plan for how to handle the reicast CLA when one of these is submitted directly to libretro. @skmp @twinaphex

[flyinghead]

CLA signed

[rcaridade145]

Concerning "Depth buffer solutions" i remember redream did something on that department.

Depth Buffer Fixes
On the Dreamcast, vertices are submitted for rendering in screen space, with the z component being equal to 1/w. These values are not normalized to any particular range, which is fine for the PowerVR's 32-bit floating-point depth buffer.

When rendering in OpenGL, these vertices must be converted back to normalized device coordinates. While unprojecting the x and y components is trivial, getting a z value that maintains the same depth ordering is not.

Originally, redream linearly scaled the z-component to the range of [0.0, 1.0] with the equation z = (z-zmin)/(zmax-min) and passed it off to the depth buffer. This worked ok for many games, but some games had an extremely large z range (e.g. zmin of 0.000001 and zmax of 100000.0) which caused a serious precision loss when normalizing, especially after the value was quantized to a 24-bit integer to be written to OpenGL's depth buffer.

After writing a small tool to measure the accuracy of the results of different normalization methods, the previous linear scaling was replaced with the logarithmic equation z = log2(1.0 + w) / 17.0. Using this method, the accuracy of the depth ordering went from as low as 30% to 99% in every problematic scene I could get my hands on.

Wonder if this would fix Cannon Spyke.

[AbandonedCart]

Keep in mind, redream also converted to closed source. From the atmosphere in their Discord, using the older, public versions of their code for other projects is not something they endorse. Out of respect for their project, it is best to contact them and request permission to use that code with a clear explanation of how it will be used and proper credit for using it. As for their current source, they may not want that to be shared with public projects and are fully qualified to decide that.

[rcaridade145]

True and i totally agree. I only remembered reading about that on one of their progress reports.

[flyinghead]

It does.
Before:

After:

As well as many many other games.

[rcaridade145]

Nice :) @flyinghead

[AbandonedCart]

https://github.com/reicast/reicast-emulator/projects/2

In case the idea of tracking status via drag and drop sounds nice...

[anothername99]

What do you all think about software rendering?

ParaLLEl (N64 emulator for libretro) recently received a multi-threaded software renderer. It's the most accurate renderer that I know of and it's pretty fast on Coffee Lake CPUs. A software renderer also eliminates any portability problems related to GPUs/graphics APIs on the host.

If anyone expresses interest in writing one for Reicast, I'll be very willing to pay.

[AbandonedCart]

I think the idea was to post a new issue explaining what you wanted and tag it with "Bounty Requested" (and possibly point to it here).

Also, there was some work put into adding software rendering. It just needs a lot of attention.
https://github.com/reicast/reicast-emulator/blob/master/core/rend/soft/softrend.cpp

[rcaridade145]

I for one would replace the X code with SDL. Is that desirable?
At least it would reduce the flickering (double buffering) and be more portable imho.

[baka0815]

+1 for changing X specific code to SDL.

[AbandonedCart]

I vaguely remember it being used by (at least) Dolphin. If that is any example, they had to keep raising the hardware requirements for mobile until it almost seemed like a carrot on a stick.
I'm not sure even they still use it, but proprietary code is always more effective than using a generic wrapper. It may solve a few issues, but at the cost of creating exponentially more limitations.

[skmp]

I prefer to keep the minimal X backend, but I'm not opposed towards also having an SDL oslib implementation. @flyinghead the bounties are yours, if you don't want them, please donate them back to the project :p. @twinaphex donating back the bounties for other use or further bounties is possible, no?

@anothername99 there is already a software renderer, though written in sse2+ intrinsics and not fully portable or accurate. Would be great to aim for a pixel perfect one, and then work from that on. Can you check if we have a ticket for software rendering?

[skmp]

@LoungeKatt lol, love the bounty tracking view :D

[anothername99]

@skmp I didn't see any open issues about improved software rendering, so I opened one here: #1166

If you think this is ok, then I'll put in some money right away.

[skmp]

@rcaridade145 want to create an issue for an update of the SDL frontend with a few specific tasks? I vaguely remember having an SDL frontend, but I don't think it was anywhere near complete. Add the "Bounty Requested" label so that we know it is on the bounty list

[skmp]

Alright, tried installing the gh integration and it's still not working :/ Will poke the bountysource people later in the week to see what's up with that

[baka0815]

@Holzhaus tried to create some kind of a frontend if I remember correctly (https://github.com/reicast/reicast-emulator/tree/holzhaus/gui).

[skmp]

(Updated first post with relevant links)

[skmp]

A simple guide on how to take on bounties (Proposal)
——

Find an issue that already has a bounty, or an issue that looks interesting to you.
Start posting on that ticket that you’re willing to look at it
Colaborate with the bounty posters and the rest of the team to get the issue resolved
If the issue is proving too difficult, it is up to you to advertise and look for further backers. The reicast team is always willing to help spread the word about work on tickets, but you should not depend only on this. The reicast forums, facebook page, ngemu, dcemulation, assembla, etc are places to interact with people that might be willing to sponsor you are working on.

For the starting months, @skmp will provide support to help locate the actual issues, as the codebase is extremely convoluted and hard to work with. The nature of the project doesn’t make matters easy, either. But it shouldn’t be too hard.

—

How do you like it?