cinera_handmade.network/cmuratori/hero/code/code586.hmml

129 lines
12 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

[video output=day586 member=cmuratori stream_platform=twitch stream_username=handmade_hero project=code title="Finishing Indirect Diffuse Sampling" vod_platform=youtube id=BoWkXfSgJdE annotator=Miblo]
[0:02][Mention the 'Handmade Collectives Worldwide" thread in the handmade.network forums[ref
author="Casey Muratori"
publisher=Twitter
title="It looks like @handmade_net folks are starting a list of low-level programmers by geography…"
url=https://twitter.com/cmuratori/status/1238588103280832512][ref
site="handmade.network Forums"
page="Handmade Collectives around the world"
url=https://handmade.network/forums/t/7212-handmade_collectives_around_the_world][ref
site="Google Sheets"
page="Handmade Collectives Worldwide"
url=https://docs.google.com/spreadsheets/d/1FzUm4A3ZB5XDtegEy4PocR7_nU1XLMOVfGWYZvhocnw/edit#gid=0]][:research]
[3:44][Fix typo in CompileZBiasProgram()][:hardware :lighting]
[5:17][Show our current specular indirect-bounce :lighting][:run]
[6:14][Embark on the switch to looking up into the diffuse light atlas, guided by the removal of light_voxel_cell from lighting_solution][:"data structure" :lighting]
[6:56][Consider relieving the :lighting system of keeping stale data around][:lighting :research]
[8:36][Respecify light_voxel_cell as light_atlas_tile][:"data structure" :lighting]
[9:57][Update LookUpVoxelClamped() to use the light_atlas_tile, rename it to GetTileClamped(), and introduce GetTileRowStride()][:lighting]
[14:01][Augment light_atlas with a TileRowStride for MakeLightAtlas() to set and LIGHT_ATLAS_OFFSET() to use][:"data structure" :lighting]
[17:44][Remove Stride from light_atlas_tile, rename its Base to Texel and update ComputeVoxelIrradianceAt() to sample from the DiffuseAtlas as a light_atlas_tile][:"data structure" :lighting]
[24:13][Set up ComputeVoxelIrradianceAt() to sample based on the surface reflection direction][:lighting]
[26:04][Update UpdateVoxel() to use our light_atlas_tile, respecifying the latter as light_atlas_texel containing a Value, introducing OffsetFromTexel()][:"data structure" :lighting]
[30:28][Update TestCast(), TestSphere(), FullCast() and ComputeLightPropagationWork() to use our light_atlas_texel][:lighting]
[34:08][Update ComputeLightPropagationWork() to blur from the SpecAtlas to the DiffuseAtlas, introducing a version of OffsetFromTexel() that takes a Tx and Ty][:lighting]
[39:23][Respecify BlockCopyVoxel() as BlockCopyAtlas()][:lighting]
[44:41][Introduce CopyTile(), a light_atlas_texel version of AdvanceRow(), and ZeroTile()][:lighting]
[49:56][Update BeginLightingComputation() to use BlockCopyAtlas() on both atlases][:lighting]
[50:40][Consider adding assertions in our new code to aid testing][:lighting :research]
[51:23][Hit a read access violation in FillLightAtlasBorder()][:lighting :run]
[52:05][:Run in debug mode without crashing][:lighting]
[52:49][Hit a read access violation in FillLightAtlasBorder() in release mode][:lighting :run]
[53:27][Introduce GetTileUnclamped() to CopyTile() and ZeroTile() to use, asserting that we are looking up within the bounds of the atlas][:lighting]
[56:04][@somebody_took_my_name][It's all just moving :memory around]
[56:26][Fix compile error][:lighting]
[56:31][Hit a read access violation in ComputeLightPropagationWork(), and determine to hunt for classes of bugs][:lighting :run]
[57:32][Assert in GetTileUnclamped() that Tx and Ty are in bounds, and there and also in OffsetFromTexel() and AdvanceRow() that the resulting pointer is in bounds, introducing PointerIsInBounds()][:lighting]
[1:01:42][Hit our PointerIsInBounds() assertion in GetTileUnclamped()][:lighting :run]
[1:02:24][Remove the LIGHT_LARGE_VOXEL_DIM definitions from lighting_solution, and propagate this change][:"data structure" :lighting]
[1:05:05][Embark on updating BeginLightingComputation() to place our hot voxel, removing LIGHT_CHUNK_COUNT, augmenting lighting_solution with LightingEnabled and turning off the :lighting for now]
[1:08:28][See no :lighting][:run]
[1:08:58][Disable the :lighting in CompileZBiasProgram()][:hardware]
[1:09:08][See the orphanage with no :lighting, hop to the neighbouring room and hit our PointerIsInBounds() assertion in GetTileUnclamped()][:run]
[1:10:17][:Run in debug mode to investigate our PointerIsInBounds() assertion hit in GetTileUnclamped()][:lighting]
[1:11:13][Fix PointerIsInBounds() to find the atlas extents using GetLightAtlasSize()][:lighting]
[1:11:36][Hit our PointerIsInBounds() assertion in AdvanceRow(), called by ZeroTile()][:lighting :run]
[1:12:59][Remove the PointerIsInBounds() assertion from AdvanceRow(), instead asserting the same within CopyTile() and ZeroTile()][:lighting]
[1:14:04][Hit no assertions in the light_atlas_tile code][:lighting :run]
[1:14:25][Make BeginLightingComputation() draw the voxels][:"debug visualisation" :lighting]
[1:15:13][Hit an assertion in GetTileUnclamped(), called by BlockCopyAtlas() within BeginLightingComputation()][:lighting :run]
[1:22:05][Fix BlockCopyAtlas() to correctly operate within bounds][:lighting]
[1:25:12][:Run without hitting any assertions in GetTileUnclamped(), with a few words on paranoid bounds checking][:lighting]
[1:26:42][Check out the unexpectedly large and correctly centred :lighting voxel][:"debug visualisation" :run]
[1:27:30][Re-enable the :lighting in BeginLightingComputation()]
[1:27:38][Double-check the :lighting distribution code in EndLightingComputation()][:research]
[1:27:56][:Run in debug mode without hitting any assertions][:lighting]
[1:28:23][:Run in release mode and hit a read access violation in ComputeLightPropagationWork()][:lighting]
[1:29:58][Toggle off the speculardiffuse atlas blurring code in ComputeLightPropagationWork()][:lighting]
[1:30:14][:Run successfully in release mode][:lighting]
[1:30:54][Toggle on the speculardiffuse atlas blurring code in ComputeLightPropagationWork()][:lighting]
[1:31:08][Consider removing LIGHT_ATLAS_OFFSET()][:lighting :research]
[1:32:03][Scour for bugs the speculardiffuse atlas blurring code in ComputeLightPropagationWork()][:lighting :research]
[1:34:51][Toggle off the store calls in ComputeLightPropagationWork()][:lighting]
[1:35:11][Break on the final Transpose() call in ComputeLightPropagationWork() and check the :asm to ensure the blurring loop has not been optimised out][:lighting :run]
[1:35:50][Continue to :run without storing, but also without crashing][:lighting]
[1:36:05][Fix ComputeLightPropagationWork() to use the DiffuseTexel.Value pointer itself, rather than its address][:language :lighting]
[1:36:38][:Run successfully with the speculardiffuse blurring and bouncing][:lighting]
[1:37:01][Re-enable the :lighting in CompileZBiasProgram()][:hardware]
[1:37:20][Admire our diffuse :lighting][:run]
[1:38:33][Try hard-setting the SampleRefColor to 70% in FullCast()][:lighting]
[1:39:41][See the light brighten and brighten][:lighting :run]
[1:40:04][Try decreasing the SampleRefColor from 70% to 25% in FullCast()][:lighting]
[1:40:15][See the light's brightness remain stable][:lighting :run]
[1:40:30][Try increasing the SampleRefColor from 25% to 50% in FullCast()][:lighting]
[1:40:35][Hit a write access violation in the debug arena][:"debug system" :lighting :run]
[1:40:51][Admire our diffuse :lighting][:run]
[1:41:06][Remove the commented out reflectance code from ComputeVoxelIrradianceAt(), and the SampleRefColor hard-setting from FullCast()][:lighting]
[1:42:19][Try modifying the W by ×0.25 in BuildDiffuseLightMaps()][:lighting]
[1:43:09][See our :lighting bounce less][:run]
[1:43:17][Try increasing the W modification from ×0.25 to ×0.5 in BuildDiffuseLightMaps()][:lighting]
[1:43:19][Hit a write access violation in DEBUGGetArena()][:"debug system" :run]
[1:43:42][Check out our ×0.5 bounced :lighting][:run]
[1:44:01][Try increasing the W modification from ×0.5 to ×0.75 in BuildDiffuseLightMaps()][:lighting]
[1:44:03][Check out our more correct ×0.75 bounced :lighting][:run]
[1:45:29][Try increasing the W modification from ×0.75 to ×0.85 in BuildDiffuseLightMaps()][:lighting]
[1:45:38][Check out our ×0.85 bounced :lighting, with the determination to tighten our lighting equation][:run]
[1:46:54][Try increasing the W modification from ×0.85 to ×0.9 in BuildDiffuseLightMaps()][:lighting]
[1:46:56][Check out our ×0.9 bounced :lighting][:run]
[1:47:39][Try increasing the W modification from ×0.9 to ×0.95 in BuildDiffuseLightMaps()][:lighting]
[1:47:43][Check out our still stable ×0.95 bounced :lighting][:run]
[1:48:58][Remove the W modification from BuldDiffuseLightMaps()][:lighting]
[1:49:07][Check out our (slowly destabilising) bounced :lighting][:run]
[1:49:54][Q&A][:speech]
[1:50:30][@picklebobdogflog][Q: What is the cause of the flickering in the :lighting? Will you do some de-noising or cast more rays in the future to try to fix it?]
[1:50:54][Try modifying the W by ×0.975 in BuildDiffuseLightMaps()][:lighting]
[1:52:17][@somebody_took_my_name][Q: I think you might have flipped the tx / /ty in TestSphere() while rewriting the code][:lighting]
[1:52:33][@xxthebigfoxx][Q: Could it be useful to have it print a value like the average texel color of the diffuse and / or specular maps on the screen?][:"debug visualisation" :lighting]
[1:53:05][@vaualbus][Q: Are you using different level on assert also for release build? Or you only use asserts on debug builds?]
[1:53:10][Disable HANDMADE_INTERNAL and HANDMADE_SLOW in build.bat][:language]
[1:53:24][Hit a read access violation in OpenGLEndFrame(), and identifier not found compile errors][:"debug system" :hardware :rendering]
[1:53:46][Make HANDMADE_INTERNAL define the macros that HANDMADE_SLOW within HANDMADE_INTERNAL would have been able to define][:"debug system" :language]
[1:55:24][:Run with both HANDMADE_INTERNAL and HANDMADE_SLOW disabled, but asserts enabled][:lighting]
[1:56:53][@yesyesyourmother][Can't find anything with DuckDuckGo[ref
site=DuckDuckGo
url=https://duckduckgo.com/]]
[1:57:27][@sahfortv][Q: How slow are we? Is there any significant change in speed from the changes today?][:lighting :performance]
[1:57:30][Check out the :performance of our :lighting][:run]
[2:01:06][@sahfortv][Q: And was the goal is to get to 30FPS in debug mode?][:performance]
[2:01:16][@vaualbus][Q: So we are faster than the previous solution? I don't remember the :performance of the old system][:lighting]
[2:01:48][Consider removing all mentions of LIGHT_LOOKUP_VOXEL_DIM][:lighting :research]
[2:02:49][Temporarily reduce LIGHT_LOOKUP_VOXEL_DIM_POW_2 from 5 to 4][:lighting]
[2:02:59][Admire our much faster :lighting][:performance]
[2:03:42][Revert our LIGHT_LOOKUP_VOXEL_DIM_POW_2 to 5][:lighting]
[2:04:09][@mattiamanzati][Q: Can you still parameterise the voxel size? I'm curious if decreasing voxel size will remove some cubeness][:lighting]
[2:04:42][@pragmascrypt][Q: For what are you still using voxels?][:lighting]
[2:05:40][@data_pi][Q: Off-topic: Do you see Rust replacing C / C++ in the game industry in the next 510 years?][:language]
[2:11:05][@ravarix][Interesting that you think Rust doesn't help with swizzling. There are many libs for it just like C. It's not really a :language problem]
[2:12:03][@s0imn][Q: Are there any :language features in C++ (that is not borrowed from C) that you like, and are there other language features than :metaprogramming, that you want in a new language?]
[2:13:39][@ravarix][Have you looked at the macro power of Rust? You can manipulate the AST as desired][:language]
[2:14:02][@etheraxx][Functional languages like Haskell allow developing embedded DSLs and you can write compiler extensions to do symbol replacement][:language]
[2:16:22][@fierydrake][Procedural Macros?[ref
site="The Rust Reference"
page="Procedural Macros"
url=https://doc.rust-lang.org/reference/procedural-macros.html]][:language]
[2:17:46][@ravarix][@handmade_hero Oh sorry, I wasn't asking for specific detail. I was just curious if you had looked at it at all][:language]
[2:18:49][@maliusarth][Q: Where can I find Moustache?]
[2:19:01][@maliusarth][Q: What do you think of clang tooling to create such AST manipulation tools as part of the build pipeline?][:language]
[2:19:29][Wrap this up][:speech]
[/video]