cinera_handmade.network/cmuratori/hero/code/code350.hmml

74 lines
7.4 KiB
Plaintext
Raw Permalink Normal View History

[video output=day350 member=cmuratori stream_platform=twitch stream_username=handmade_hero project=code title="Multithreaded World Simulation" vod_platform=youtube id=tWzslFE9Qvg annotator=Miblo]
[0:35][On two-hour blocks enabling a more continuous workflow]
[1:43][Run the game and set the stage for the day]
[2:48][handmade_world_mode.cpp: Consider reworking UpdateAndRenderSimRegion() for state visibility minimisation]
[7:54][Run the game and consider how we want UpdateAndRenderSimRegion() to operate differently in the debug and the regular view]
[11:32][On when and why to pass in a superset of arguments to a function]
[14:56][handmade_world_mode.cpp: Introduce CheckForJoiningPlayers() to contain the player joining code from UpdateAndRenderSimRegion()]
[18:06][handmade_world_mode.cpp: Consider how to split up UpdateAndRenderSimRegion()]
[21:45][handmade_world_mode.cpp: Break out UpdateAndRenderSimRegion() functionality into UpdateAndRenderWorld()]
[23:31][handmade_sim_region.cpp: Consider how the simulation and the packing sides are two different functions]
[24:47][handmade_sim_region.cpp: Rename BeginSim() to BeginWorldChange() and EndSim() to EndWorldChange()]
[25:50][handmade_world_mode.cpp: Introduce new, higher level BeginSim() and EndSim() functions, and Simulate()]
[32:57][handmade_world_mode.cpp: Make UpdateAndRenderWorld() call our new functions]
[37:52][Run the game to see that we're still getting the same results]
[38:26][Consider reworking the passing of input to ExecuteBrain() and Simulate()]
[41:48][handmade_entity.cpp: Introduce DEBUGPickEntity()]
[50:04][handmade_debug_interface.h: Add MouseP to the debug_table struct and introduce SET_DEBUG_MOUSE_P() and GET_DEBUG_MOUSE_P()]
[53:44][Run the game to see that's all good now too]
[54:13][Q&A][:speech]
[55:03][@asafgartner][Can you try compiling with debug off?]
[55:41][handmade_world_mode.cpp: Enable DEBUGPickEntity() and DEBUG_PLATFORM_GET_MEMORY_STATS() to be compiled out]
[57:39][@badflydog][CTime stats. I know I know. It's irrelevant for the code]
[58:31][Run CTime]
[59:48][@chaoscom][What does the "..." do in the define?]
[1:00:06][handmade.cpp: Demo varargs]
[1:09:26][@flirtychair][How about documentation? I see you don't have too much comments in your code]
[1:12:45][@tomisqi][I sometimes receive comments from colleagues during code reviews that I should put long pieces of code into functions for "cleanliness". What parameters go into your thought process when deciding if a piece of code should be placed under a function? How do you make the decision?]
[1:18:16][@lvias][How much code is debug code vs real code?]
[1:20:15][@uplinkcoder][Expand the files and take the diffs?]
[1:21:14][Restart, recap the first block of work and set the stage for the second block]
[1:23:14][handmade_world_mode.cpp: Delineate the primary and non-primary region simulation routines in UpdateAndRenderWorld() and consider how to thread them]
[1:25:15][Run the game and consult the profiler to see the expense of GetClosestTraversable()]
[1:27:34][handmade_render.cpp: Recap how RenderCommandsToBitmap() creates work orders to distribute to our threads]
[1:30:13][handmade_world_mode.cpp: Introduce DoWorldSim() in order to thread the simulation]
[1:32:32][Note the fact that the arenas are not multithreaded constructs]
[1:33:53][handmade_world_mode.cpp: Make BeginSim() take a TempArena that we pass in, to enable it to be multithreaded]
[1:36:23][Consider reworking memory allocaton in terms of multithreading]
[1:40:47][Blackboard: Memory and Multiple Threads]
[1:46:33][handmade_world_mode.cpp: Make DoWorldSim() allocate memory that exists only for the lifetime of the work]
[1:50:31][handmade_world_mode.cpp: Enable UpdateAndRenderWorld() to create world_sim_work orders]
[1:56:06][Run the game and consult the profiler to see a lot of debug events being generated]
[1:57:32][handmade_world_mode.cpp: Toggle UpdateAndRenderWorld() to take the multithreaded path, :run the game and consider why it doesn't work]
[2:00:21][handmade_world_mode.cpp: Consider why Simulate(), ExecuteBrain() and UpdateAndRenderEntities() take the WorldMode]
[2:02:46][handmade_world_mode.cpp: Remove AddCollisionRule() and ClearCollisionRulesFor(), and prevent functions from unnecessarily taking the WorldMode]
[2:07:13][handmade_world_mode.cpp: Make Simulate() take a particle_cache]
[2:08:44][handmade_brain.cpp: Make ExecuteBrain() take a random_series Entropy]
[2:09:53][handmade_entity.cpp: Make UpdateAndRenderEntities() take the TypicalFloorHeight directly for now]
[2:11:46][handmade_sim_region.cpp: Prevent BeginWorldChange() and EndWorldChange() from taking WorldMode, and consider how better to deal with the camera]
[2:15:03][handmade_world_mode.cpp: Make DoWorldSim() lock a mutex around BeginSim() and EndSim()]
[2:18:21][Run the game on the multithreaded path, and investigate why the World is null]
[2:18:38]["The world is null... the world is null!"][quote 534]
[2:19:52][handmade_world_mode.cpp: Make DoWorldSim() call Platform.CompleteAllWork(), run the game and consult the profiler]
[2:21:56][Consider how to improve this multithreading of the simulation, with a few words on functional programming]
[2:23:42][todo.txt: Update the TODO list]
[2:24:42][Q&A][:speech]
[2:25:15][@Kknewkles][(vicariously asked via @Miblo) o7 Casey! Two questions: 1) Have you finished Sausage Roll? I have! >:D 2) I have been Shilling Knight for you to play Shovel Knight, and now I have to add Undertale to the list. That game a gift from gods on high]
[2:26:17][@a_pulsing_mage][Isn't simulating the world in several threads, can't that make the game non deterministic? Isn't the point of the RNG having a starting kernel for deterministic play for competition and stuff?]
[2:29:14][Blackboard: RNG in Handmade Hero]
[2:32:32][@thegeekpirate][When do you go back and complete "TODO" tasks you've given yourself in the code? Is each one going to be completed before you're done, or are they simply wishlists?]
[2:33:14][@sumsar10171][Is it important for a game programmer to have a portfolio in the same way artists do and, if so, what would be important to include?[ref
site="Molly Rocket"
page="HandmadeCon 2015 - Mike Acton"
url="https://mollyrocket.com/news_0035.html"]]
[2:34:53][@garryjohanson][In a tight inner loop do you think these memory copies for the temporary scratch memory could become an overhead concern? If I understood the stream correctly, that is]
[2:35:31][@longboolean][Currently what is the behavior when walking several rooms over, into another chunk?]
[2:36:13][@williambundy][Save scumming?]
[2:36:21][@jim0_o][Will there be streams from the job fair people?]
[2:36:48][@marcramhim][What's the relation between IQ and success at programming?]
[2:37:20][@ginger_bill][Off-topic question which can be answered later if necessary. I've been developing a new programming language to replace my needs for C/C++ and pretty much every other language. The main areas of research for me at the moment are: Metaprogramming, SSA optimizations, Concurrency. You being a metaprogramming "expert", other than Compile Time Execution (CTE) that is Jon's language, what other forms of metaprogramming would you like?]
[2:37:50][@williambundy][Sorry, it was in context: if I open a chest, and don't get what I want, reloading a save game and playing slightly differently might get me a different RNG for the chest]
[2:40:36][@ginger_bill][Will the game have different "levels" like the Binding of Isaac, or will the entire game be a single "level"?]
[2:42:13][Wind it down][:speech]
[/video]