Experimenting with Voxel Filters — Handmade Hero — Episode Guide

0:01Recap and set the stage for the day making the ray caster produce a stable set of inputs to pass to the light field interpolater, with thoughts on multiple light fields

🗩

0:01Recap and set the stage for the day making the ray caster produce a stable set of inputs to pass to the light field interpolater, with thoughts on multiple light fields

🗩

0:01Recap and set the stage for the day making the ray caster produce a stable set of inputs to pass to the light field interpolater, with thoughts on multiple light fields

🗩

5:25Show the current state of the lighting

🏃

5:25Show the current state of the lighting

🏃

5:25Show the current state of the lighting

🏃

7:11Show the current state of the lighting, without filling in negative light values

🏃

7:11Show the current state of the lighting, without filling in negative light values

🏃

7:11Show the current state of the lighting, without filling in negative light values

🏃

7:35Consult our lighting TODOs

📖

7:35Consult our lighting TODOs

📖

7:35Consult our lighting TODOs

📖

8:03Show the absence of light flickering, thanks to our switch to floats

🏃

8:03Show the absence of light flickering, thanks to our switch to floats

🏃

8:03Show the absence of light flickering, thanks to our switch to floats

🏃

9:02Reorganise and consult our lighting TODOs

9:28Demo our need to flood fill the voxel so that unsampled portions get filled in from nearby samples

🏃

9:28Demo our need to flood fill the voxel so that unsampled portions get filled in from nearby samples

🏃

9:28Demo our need to flood fill the voxel so that unsampled portions get filled in from nearby samples

🏃

12:03Consider where is most appropriate to perform the flood filled interpolation

🗩

12:03Consider where is most appropriate to perform the flood filled interpolation

🗩

12:03Consider where is most appropriate to perform the flood filled interpolation

🗩

13:04Consider whether or not the reconstructed edge lighting values should persist across frames

🏃

13:04Consider whether or not the reconstructed edge lighting values should persist across frames

🏃

13:04Consider whether or not the reconstructed edge lighting values should persist across frames

🏃

15:22Consider our options for flood filling the lighting voxel: 1) Persistently in lighting_solution; 2) In EndLightingComputation() when: a) pushing the cells to the graphics card each frame; b) accumulating light contribution over time

📖

15:22Consider our options for flood filling the lighting voxel: 1) Persistently in lighting_solution; 2) In EndLightingComputation() when: a) pushing the cells to the graphics card each frame; b) accumulating light contribution over time

📖

15:22Consider our options for flood filling the lighting voxel: 1) Persistently in lighting_solution; 2) In EndLightingComputation() when: a) pushing the cells to the graphics card each frame; b) accumulating light contribution over time

📖

19:13Determine to make EndLightingComputation() flood fill the lighting voxel when pushing the cells to the graphics card

📖

19:13Determine to make EndLightingComputation() flood fill the lighting voxel when pushing the cells to the graphics card

📖

19:13Determine to make EndLightingComputation() flood fill the lighting voxel when pushing the cells to the graphics card

📖

20:36Seperable Filters

🖌

20:36Seperable Filters

🖌

20:36Seperable Filters

🖌

25:03Cache implications of our light voxel¹

📖

25:03Cache implications of our light voxel¹

📖

25:03Cache implications of our light voxel¹

📖

30:44Propose constraining our flood fill filter to cells predetermined to require filtering

📖

30:44Propose constraining our flood fill filter to cells predetermined to require filtering

📖

30:44Propose constraining our flood fill filter to cells predetermined to require filtering

📖

32:35Abortively augment game_render_commands with a FloodIndices array

35:54Begin to make EndLightingComputation() flood fill the lighting voxel with purple when pushing the cells to the graphics card, asserting that we're well inside the hot region

38:17Try unsuccessfully to trigger our hot region assertions

🏃

38:17Try unsuccessfully to trigger our hot region assertions

🏃

38:17Try unsuccessfully to trigger our hot region assertions

🏃

39:02Make EndLightingComputation() flood fill out to neighbour cells, introducing HasLight() and FloodLight()

1:02:58Directional Flood Filling

🖌

1:02:58Directional Flood Filling

🖌

1:02:58Directional Flood Filling

🖌

1:04:28Fix FloodLight() to correctly fill in positive and negative directions, before welding its code in to EndLightingComputation()

1:08:22Check out our partially working flood filling

🏃

1:08:22Check out our partially working flood filling

🏃

1:08:22Check out our partially working flood filling

🏃

1:08:52Make EndLightingComputation() flood fill with purple

1:09:23Check out our purple flood filling, and spot some erroneously unlit / interpolated areas

🏃

1:09:23Check out our purple flood filling, and spot some erroneously unlit / interpolated areas

🏃

1:09:23Check out our purple flood filling, and spot some erroneously unlit / interpolated areas

🏃

1:10:28Make CompileZBiasProgram() clamp the light to 0, introducing Clamp0Inf()²

1:14:37Still see dead zones

🏃

1:14:37Still see dead zones

🏃

1:14:37Still see dead zones

🏃

1:15:49Make CompileZBiasProgram() normalise the LightN

1:16:12Still see dead zones

🏃

1:16:12Still see dead zones

🏃

1:16:12Still see dead zones

🏃

1:17:21Prevent EndLightingComputation() from flood filling with purple

1:17:27Still see dead zones

🏃

1:17:27Still see dead zones

🏃

1:17:27Still see dead zones

🏃

1:17:46Toggle on the purple in EndLightingComputation()

1:17:48Change HasLight() to base its decision on LightC, rather than LightN

1:18:22Still see dead zones

🏃

1:18:22Still see dead zones

🏃

1:18:22Still see dead zones

🏃

1:19:13Make EndLightingComputation() flood fill with yellow the cells that have light

1:19:49Find that all cells are lit either purple or yellow

🏃

1:19:49Find that all cells are lit either purple or yellow

🏃

1:19:49Find that all cells are lit either purple or yellow

🏃

1:20:11Comment out the yellow in EndLightingComputation()

1:20:18See our dead zones return

🏃

1:20:18See our dead zones return

🏃

1:20:18See our dead zones return

🏃

1:20:45Make HasLight() less permissive with a threshold

1:21:50See our dead zones flood to purple (i.e. unlit)

🏃

1:21:50See our dead zones flood to purple (i.e. unlit)

🏃

1:21:50See our dead zones flood to purple (i.e. unlit)

🏃

1:22:36Toggle off the purple in EndLightingComputation()

1:23:00Check out the interpolation across the light boundaries

🏃

1:23:00Check out the interpolation across the light boundaries

🏃

1:23:00Check out the interpolation across the light boundaries

🏃

1:23:40Scrutinise the directional flood fill interpolation code in EndLightingComputation()

📖

1:23:40Scrutinise the directional flood fill interpolation code in EndLightingComputation()

📖

1:23:40Scrutinise the directional flood fill interpolation code in EndLightingComputation()

📖

1:25:12Directional Flood Fill Interpolation

🖌

1:25:12Directional Flood Fill Interpolation

🖌

1:25:12Directional Flood Fill Interpolation

🖌

1:26:33Scrutinise the light value lookup and clamping in CompileZBiasProgram()

📖

1:26:33Scrutinise the light value lookup and clamping in CompileZBiasProgram()

📖

1:26:33Scrutinise the light value lookup and clamping in CompileZBiasProgram()

📖

1:28:41Closely look at the apparent interpolation bug

🏃

1:28:41Closely look at the apparent interpolation bug

🏃

1:28:41Closely look at the apparent interpolation bug

🏃

1:30:19Make EndLightingComputation() flood fill purple if the LightWeight is 0

1:30:42See some erroneously purple areas

🏃

1:30:42See some erroneously purple areas

🏃

1:30:42See some erroneously purple areas

🏃

1:31:31Scrutinise EndLightingComputation() for gather bugs

📖

1:31:31Scrutinise EndLightingComputation() for gather bugs

📖

1:31:31Scrutinise EndLightingComputation() for gather bugs

📖

1:32:12Closely look at the erroneously purple areas

🏃

1:32:12Closely look at the erroneously purple areas

🏃

1:32:12Closely look at the erroneously purple areas

🏃

1:32:41Remove the threshold from HasLight() and ensure that EndLightingComputation() can blend down to 0

1:34:35Still see our interpolation artifacts

🏃

1:34:35Still see our interpolation artifacts

🏃

1:34:35Still see our interpolation artifacts

🏃

1:36:10Toggle off the final flood filling in EndLightingComputation()

1:36:23Still see our interpolation artifacts

🏃

1:36:23Still see our interpolation artifacts

🏃

1:36:23Still see our interpolation artifacts

🏃

1:36:58Try making EndLightingComputation() regularise our light normals

1:37:20See no interpolation bugs

🏃

1:37:20See no interpolation bugs

🏃

1:37:20See no interpolation bugs

🏃

1:37:50Toggle on the final flood filling in EndLightingComputation()

1:38:06See some very bizarre half-lit artifacts

🏃

1:38:06See some very bizarre half-lit artifacts

🏃

1:38:06See some very bizarre half-lit artifacts

🏃

1:39:19Investigate our half-lit artifacts

📖

1:39:19Investigate our half-lit artifacts

📖

1:39:19Investigate our half-lit artifacts

📖

1:40:26Toggle off the whole flood filling code in EndLightingComputation()

1:40:40Check out our lighting without flood filling, and spot some discontinuity across a wall

🏃

1:40:40Check out our lighting without flood filling, and spot some discontinuity across a wall

🏃

1:40:40Check out our lighting without flood filling, and spot some discontinuity across a wall

🏃

1:41:29Try making EndLightingComputation() regularise our light normals as 0, 0, 1

1:42:04Find that this doesn't help us

🏃

1:42:04Find that this doesn't help us

🏃

1:42:04Find that this doesn't help us

🏃

1:42:11Try making EndLightingComputation() regularise our light normals as 0, -1, 0

1:42:23Find that our discontinuity has gone

🏃

1:42:23Find that our discontinuity has gone

🏃

1:42:23Find that our discontinuity has gone

🏃

1:42:33Let EndLightingComputation() compute each voxel's normal

1:42:44Wonder if the discontinuity is due to the normal interpolation being unable to properly swing

🏃

1:42:44Wonder if the discontinuity is due to the normal interpolation being unable to properly swing

🏃

1:42:44Wonder if the discontinuity is due to the normal interpolation being unable to properly swing

🏃

1:42:57Make CompileZBiasProgram() visualise the normal

1:44:11Determine that we are getting a 0, 0, 0 normal

🏃

1:44:11Determine that we are getting a 0, 0, 0 normal

🏃

1:44:11Determine that we are getting a 0, 0, 0 normal

🏃

1:44:51Prevent CompileZBiasProgram() from normalising the normal

1:45:13Find that the hard-edged 0, 0, 0 normal remains

🏃

1:45:13Find that the hard-edged 0, 0, 0 normal remains

🏃

1:45:13Find that the hard-edged 0, 0, 0 normal remains

🏃

1:45:21Make GetIrradiance() in CompileZBiasProgram() always visualise the normals

1:46:34Find that the hard-edged 0, 0, 0 normal is gone

🏃

1:46:34Find that the hard-edged 0, 0, 0 normal is gone

🏃

1:46:34Find that the hard-edged 0, 0, 0 normal is gone

🏃

1:46:51Explain the problem when the light and world normals are orthogonal

📖

1:46:51Explain the problem when the light and world normals are orthogonal

📖

1:46:51Explain the problem when the light and world normals are orthogonal

📖

1:47:43Make GetIrradiance() in CompileZBiasProgram() modify its result by the LightDC, and not visualise the normal

1:48:16Find that our discontinuity remains

🏃

1:48:16Find that our discontinuity remains

🏃

1:48:16Find that our discontinuity remains

🏃

1:48:44Make GetIrradiance() in CompileZBiasProgram() visualise the normal, modified by the LightDC

1:49:29Find that our discontinuity remains

🏃

1:49:29Find that our discontinuity remains

🏃

1:49:29Find that our discontinuity remains

🏃

1:49:58Make GetIrradiance() in CompileZBiasProgram() visualise the LightDC

1:50:36See our discontinuity

🏃

1:50:36See our discontinuity

🏃

1:50:36See our discontinuity

🏃

1:50:49Ponder our discontinuity

📖

1:50:49Ponder our discontinuity

📖

1:50:49Ponder our discontinuity

📖

1:51:50Disable trilinear interpolation in OpenGLInit()

1:52:20Check out our non-interpolated LightDC

🏃

1:52:20Check out our non-interpolated LightDC

🏃

1:52:20Check out our non-interpolated LightDC

🏃

1:53:20Scrutinise GetIrradiance()

📖

1:53:20Scrutinise GetIrradiance()

📖

1:53:20Scrutinise GetIrradiance()

📖

1:54:27Make GetIrradiance() in CompileZBiasProgram() visualise the normal

1:54:55See smooth interpolation

🏃

1:54:55See smooth interpolation

🏃

1:54:55See smooth interpolation

🏃

1:55:01Make CompileZBiasProgram() visualise just the normal, without the textures

1:55:25Check out our normals, with no – or at least less – discontinuity

🏃

1:55:25Check out our normals, with no – or at least less – discontinuity

🏃

1:55:25Check out our normals, with no – or at least less – discontinuity

🏃

1:57:30Let CompileZBiasProgram() use the textures and not visualise the normals

1:57:39Check out our discontinuity

🏃

1:57:39Check out our discontinuity

🏃

1:57:39Check out our discontinuity

🏃

1:58:03Temporarily make CompileZBiasProgram() hard set the LightDC to 0.5

1:58:22Find that our discontinuity has gone

🏃

1:58:22Find that our discontinuity has gone

🏃

1:58:22Find that our discontinuity has gone

🏃

1:59:25Let CompileZBiasProgram() gather in some light from perpendicular sources

1:59:52Find that our discontinuity has really gone

🏃

1:59:52Find that our discontinuity has really gone

🏃

1:59:52Find that our discontinuity has really gone

🏃

2:01:10Re-enable the flood filling code in EndLightingComputation()

2:01:29See that our flood filling brightens the light

🏃

2:01:29See that our flood filling brightens the light

🏃

2:01:29See that our flood filling brightens the light

🏃

2:01:48Prevent EndLightingComputation() from hard setting the normal to 0, 0, 1

2:02:30See that our flood filling fills in some areas

🏃

2:02:30See that our flood filling fills in some areas

🏃

2:02:30See that our flood filling fills in some areas

🏃

2:03:23Take a close look at our flood filling

🏃

2:03:23Take a close look at our flood filling

🏃

2:03:23Take a close look at our flood filling

🏃

2:04:29Respecify HasLight() as a smoothly varying GetLightAmount() for EndLightingComputation() to use as a multiplier

2:07:58Find that our flood fill's popping artifacts have been fixed

🏃

2:07:58Find that our flood fill's popping artifacts have been fixed

🏃

2:07:58Find that our flood fill's popping artifacts have been fixed

🏃

2:08:47Make SpamVoxel() spam out to all voxel slices

2:09:27Check out the spamming (with flood filling)

🏃

2:09:27Check out the spamming (with flood filling)

🏃

2:09:27Check out the spamming (with flood filling)

🏃

2:09:58Toggle off the flood filling in EndLightingComputation()

2:10:08Check out the spamming (without flood filling)

🏃

2:10:08Check out the spamming (without flood filling)

🏃

2:10:08Check out the spamming (without flood filling)

🏃

2:11:17Find that the spamming does not capture negative light properly in the outdoor area

🏃

2:11:17Find that the spamming does not capture negative light properly in the outdoor area

🏃

2:11:17Find that the spamming does not capture negative light properly in the outdoor area

🏃

2:11:38Prevent SpamVoxelSlice() from clamping the light's distance falloff to 0 to 1

2:11:57See that we're still not getting negative light interpolation

🏃

2:11:57See that we're still not getting negative light interpolation

🏃

2:11:57See that we're still not getting negative light interpolation

🏃

2:12:50Prevent EndLightingComputation() from clamping the light values

2:13:53See that we're still not getting negative light interpolation

🏃

2:13:53See that we're still not getting negative light interpolation

🏃

2:13:53See that we're still not getting negative light interpolation

🏃

2:14:24Scan the code for any more 0-clamping

📖

2:14:24Scan the code for any more 0-clamping

📖

2:14:24Scan the code for any more 0-clamping

📖

2:15:40Make CompileZBiasProgram() draw negative values in red

2:16:22See no negative values

🏃

2:16:22See no negative values

🏃

2:16:22See no negative values

🏃

2:16:44Scan the code for any more 0-clamping

📖

2:16:44Scan the code for any more 0-clamping

📖

2:16:44Scan the code for any more 0-clamping

📖

2:18:06Add an assertion in SpamVoxelSlice that LightDF >= 0.0f

2:18:37Fail to hit that assertion

🏃

2:18:37Fail to hit that assertion

🏃

2:18:37Fail to hit that assertion

🏃

2:18:43Scrutinise the LightDF calculation in SpamVoxelSlice()

📖

2:18:43Scrutinise the LightDF calculation in SpamVoxelSlice()

📖

2:18:43Scrutinise the LightDF calculation in SpamVoxelSlice()

📖

2:20:50Run the game afresh and successfully hit our assertion in SpamVoxelSlice()

🏃

2:20:50Run the game afresh and successfully hit our assertion in SpamVoxelSlice()

🏃

2:20:50Run the game afresh and successfully hit our assertion in SpamVoxelSlice()

🏃

2:20:53Comment out the assertion in SpamVoxelSlice()

2:21:07Take another look at our non-interpolation of negative light

🏃

2:21:07Take another look at our non-interpolation of negative light

🏃

2:21:07Take another look at our non-interpolation of negative light

🏃

2:21:44Make TestCastFromProbes() fall-off the light more steeply

2:21:57Admire our steeper light falloff, and wonder where our stair-step pattern is coming from

🏃

2:21:57Admire our steeper light falloff, and wonder where our stair-step pattern is coming from

🏃

2:21:57Admire our steeper light falloff, and wonder where our stair-step pattern is coming from

🏃

2:23:03Prevent CompileZBiasProgram() from drawing negative values in red

2:23:12Consider our entire falloff to not be that smooth

🏃

2:23:12Consider our entire falloff to not be that smooth

🏃

2:23:12Consider our entire falloff to not be that smooth

🏃

2:23:35Try making SpamVoxelSlice() set the light's alpha to the LightDF squared

2:24:33Admire our ring of light falloff

🏃

2:24:33Admire our ring of light falloff

🏃

2:24:33Admire our ring of light falloff

🏃

2:24:51Make SpamVoxelSlice() preserve the sign of the LightDF

2:25:03Check out our hard light falloff

🏃

2:25:03Check out our hard light falloff

🏃

2:25:03Check out our hard light falloff

🏃

2:25:18Try making SpamVoxelSlice() set the light's alpha to the LightDF square-root

2:25:32Check out our hard light falloff

🏃

2:25:32Check out our hard light falloff

🏃

2:25:32Check out our hard light falloff

🏃

2:25:38Revert SpamVoxelSlice() to setting the light's alpha to the LightDF

2:25:51Check out our hard light falloff

🏃

2:25:51Check out our hard light falloff

🏃

2:25:51Check out our hard light falloff

🏃

2:26:04Determine that we are doing ShaderSimTexWriteSRGB

📖

2:26:04Determine that we are doing ShaderSimTexWriteSRGB

📖

2:26:04Determine that we are doing ShaderSimTexWriteSRGB

📖

2:27:45Temporarily prevent CompileZBiasProgram() from square-rooting the SurfaceReflect

2:28:12See that our light falloff is much smoother

🏃

2:28:12See that our light falloff is much smoother

🏃

2:28:12See that our light falloff is much smoother

🏃

2:28:29Revert CompileZBiasProgram() to square-rooting the SurfaceReflect, and try making SpamVoxelSlice() set the LightDist to LightN squared

2:29:32See a bizarrely octagonal light falloff

🏃

2:29:32See a bizarrely octagonal light falloff

🏃

2:29:32See a bizarrely octagonal light falloff

🏃

2:29:41Our problem: The falloff interpolation is being interpreted linearly

📖

2:29:41Our problem: The falloff interpolation is being interpreted linearly

📖

2:29:41Our problem: The falloff interpolation is being interpreted linearly

📖

2:31:13Revert SpamVoxelSlice() to not square the LightN, and instead make GetIrradiance() in CompileZBiasProgram() square the LightV when setting the LightColor, introducing Square()

2:32:53Make SpamVoxelSlice() square the LightN when setting the LightDist, and GetIrradiance() in CompileZBiasProgram() preserve the sign³

2:35:12See our brighter light

🏃

2:35:12See our brighter light

🏃

2:35:12See our brighter light

🏃

2:36:30Fix the sign-preservation in GetIrradiance() in CompileZBiasProgram()

2:36:44See our falloff

🏃

2:36:44See our falloff

🏃

2:36:44See our falloff

🏃

2:37:13Make SpamVoxelSlice() square-root the LightC0, introducing SignedSquareRoot()

2:38:23Admire our softer falloff

🏃

2:38:23Admire our softer falloff

🏃

2:38:23Admire our softer falloff

🏃

2:38:38Make TestCastFromProbes() fall-off the light more gradually

2:39:03Admire our softer, yet wobbly falloff

🏃

2:39:03Admire our softer, yet wobbly falloff

🏃

2:39:03Admire our softer, yet wobbly falloff

🏃

2:39:36Make TestCastFromProbes() darken the LightC

2:39:45Admire our darker light, but doubt that the falloff is softer

🏃

2:39:45Admire our darker light, but doubt that the falloff is softer

🏃

2:39:45Admire our darker light, but doubt that the falloff is softer

🏃

2:41:00Make TestCastFromProbes() further darken the LightC

2:41:03Admire our darker light

🏃

2:41:03Admire our darker light

🏃

2:41:03Admire our darker light

🏃

2:41:26Toggle off the square-rooting / squaring

2:42:03Check out our less gradual light falloff

🏃

2:42:03Check out our less gradual light falloff

🏃

2:42:03Check out our less gradual light falloff

🏃

2:42:52Make TestCastFromProbes() brighten the LightC

2:43:04Check out our brighter and less gradual light falloff

🏃

2:43:04Check out our brighter and less gradual light falloff

🏃

2:43:04Check out our brighter and less gradual light falloff

🏃

2:44:16Q&A

🗩

2:44:16Q&A

🗩

2:44:16Q&A

🗩

2:44:49@cemuzunlar Q: Solution->HasCornerI.x + X + 2*dZ

🗪

2:44:49@cemuzunlar Q: Solution->HasCornerI.x + X + 2*dZ

🗪

2:44:49@cemuzunlar Q: Solution->HasCornerI.x + X + 2*dZ

🗪

2:45:10Fix typo in EndLightingComputation() and toggle on the flood filling

2:45:25Check out the blotchy flood filling

🏃

2:45:25Check out the blotchy flood filling

🏃

2:45:25Check out the blotchy flood filling

🏃

2:46:15Toggle off flood filling

2:46:23Consider our flood filling not to be close to working

🏃

2:46:23Consider our flood filling not to be close to working

🏃

2:46:23Consider our flood filling not to be close to working

🏃

2:47:04@0lpbm Damn, it wasn't that

🗪

2:47:04@0lpbm Damn, it wasn't that

🗪

2:47:04@0lpbm Damn, it wasn't that

🗪

2:48:29Wrap it up

🗩

2:48:29Wrap it up

🗩

2:48:29Wrap it up

🗩

Keyboard Navigation

Global Keys

Menu toggling

In-Menu and Index Controls

Quotes and References Menus and Index

Quotes, References and Credits Menus

Filter Menu

Filter and Link Menus

Credits Menu