As far as I understand, Casey originally had his memory allocation setup so that he could just separate the allocated memory block into separate 'memory arenas'. So you could have a memory arena allocated for all world objects you'll create, an arena for all assets, an arena for a render buffer, etc. But later on he added the possibility of a 'sub-arena' to sub allocate an arena even further. I was wondering what the purpose of the sub-arena's were and if they are really necessary? I ask this because I was thinking of creating a memory scheme based off his but I figured it would look something like:
.......
Basically I would be partitioning off memory arena's and specifying what type of allocation scheme I want to run on each of them when allocating memory. I thought this would handle most situations but I'm wondering what I might be missing and if I need to add in the possibility to sub-allocate an arena even further?
1 2 3 4 5 6 | i32 generalHeap_id = CreateParitionFromMemoryBlock(gameMemoryBlock, Megabytes(100), DYNAMIC); i32 assets_id = CreateParitionFromMemoryBlock(gameMemoryBlock, Megabytes(200), FRAME); i32 renderBuffer_id = CreateParitionFromMemoryBlock(gameMemoryBlock, Megabytes(10), LINEAR); i32* thing = Alloc(gernalHeap_id, i32, 1); Render_Command* command = Alloc(renderBuffer_id, Render_Command_Type, 1); |
.......
Basically I would be partitioning off memory arena's and specifying what type of allocation scheme I want to run on each of them when allocating memory. I thought this would handle most situations but I'm wondering what I might be missing and if I need to add in the possibility to sub-allocate an arena even further?
Edited by Jason
on
Reason: Initial post
SubArena is just a simple way to reserve a block of memory. It's equivalent to your CreatePartitionFromMemoryBlock function (you wrote parition instead of partition, unless parition means something).