biospud wrote:The closest thing to a HID device in the gzdoom code might be the joystick detection, but I'm unsure exactly how the joystick stuff works, nor whether it would be a good example of what you need.
Existing APIs might be enough and I played around with it a bit but couldn't figure out how to read a page from the device.
The procedure itself is really simple. The C SDK was just released yesterday so I'll need to review it.
biospud wrote:Do I understand correctly that all 45 views are offset only horizontally, and there is no relative vertical offset? If so, the use of the term "light field" for this display might be a bit of an exaggeration. When there is only one dimension of variation, this would be equivalent to putting a calibrated lenticular screen over a monitor, yes?
Yes, horizontal only. It's still technically a lightfield with all the properties that come with it - just a very thin one.
Practically it is just a tilted lenticular lens, but also has a huge glass brick in front of it that also works to augment the optics.
Autostereoscopic is possibly more descriptive of the horizontal-only nature, but sells short that you can still actually move forward/back, left/right within that volume and peek around corners etc.
If there was a cracked door, you could lean forward and peek through the crack.
But if there was a keyhole, you could lean forward and peek through the keyhole, but it would be vertically squished, unlike a full lightfield.
biospud wrote:I understand that you did not exactly follow the flow of the two-eye stereo modes, because one shader cannot access 45 different textures at once. It looks like you are still using 45 little eye framebuffers plus textures for the eye buffers. That seems inefficient.
Hmm, looks like i need to rereview the eye render code path.
biospud wrote:I don't understand why items at distance would be blurry. Is it because you place your face very close to the display? In that case, it seems reading glasses would help, but I guess that sorta defeats the purpose.
Blurry is maybe the wrong word.
The pixel offset at distance becomes larger than a single lenticular cell visible angle and it jumps onto the next pixel over, which is a visual artifact and effectively looks blurry.
Doom depth is also centered around the gun/menu and most of the scene is in negative depth only, wasting half of the screen's lenticular depth range.
biospud wrote:I still think it would be useful to have this mode in the menus. Most folks cannot use most of the other modes as they are; yet the presence of the different modes teach something about what is possible. And in my opinion this one looks rather interesting on an ordinary display. More importantly, I like to use the menu to quickly scan through and sanity check the various 3D modes for testing and maintenance, even when I'm not using the exact eyewear or display.
Yea fair enough.
[quote="biospud"]The closest thing to a HID device in the gzdoom code might be the joystick detection, but I'm unsure exactly how the joystick stuff works, nor whether it would be a good example of what you need.[/quote]
Existing APIs might be enough and I played around with it a bit but couldn't figure out how to read a page from the device.
The procedure itself is really simple. The C SDK was just released yesterday so I'll need to review it.
[quote="biospud"]Do I understand correctly that all 45 views are offset only horizontally, and there is no relative vertical offset? If so, the use of the term "light field" for this display might be a bit of an exaggeration. When there is only one dimension of variation, this would be equivalent to putting a calibrated lenticular screen over a monitor, yes?[/quote]
Yes, horizontal only. It's still technically a lightfield with all the properties that come with it - just a very thin one. :)
Practically it is just a tilted lenticular lens, but also has a huge glass brick in front of it that also works to augment the optics.
Autostereoscopic is possibly more descriptive of the horizontal-only nature, but sells short that you can still actually move forward/back, left/right within that volume and peek around corners etc.
If there was a cracked door, you could lean forward and peek through the crack.
But if there was a keyhole, you could lean forward and peek through the keyhole, but it would be vertically squished, unlike a full lightfield.
[quote="biospud"]I understand that you did not exactly follow the flow of the two-eye stereo modes, because one shader cannot access 45 different textures at once. It looks like you are still using 45 little eye framebuffers plus textures for the eye buffers. That seems inefficient.[/quote]
Hmm, looks like i need to rereview the eye render code path.
[quote="biospud"]I don't understand why items at distance would be blurry. Is it because you place your face very close to the display? In that case, it seems reading glasses would help, but I guess that sorta defeats the purpose.[/quote]
Blurry is maybe the wrong word.
The pixel offset at distance becomes larger than a single lenticular cell visible angle and it jumps onto the next pixel over, which is a visual artifact and effectively looks blurry.
Doom depth is also centered around the gun/menu and most of the scene is in negative depth only, wasting half of the screen's lenticular depth range.
[quote="biospud"]I still think it would be useful to have this mode in the menus. Most folks cannot use most of the other modes as they are; yet the presence of the different modes teach something about what is possible. And in my opinion this one looks rather interesting on an ordinary display. More importantly, I like to use the menu to quickly scan through and sanity check the various 3D modes for testing and maintenance, even when I'm not using the exact eyewear or display.[/quote]
Yea fair enough.