kraken_engine/kraken
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

KRDevice now creates transfer queues for each GPU. Implemented algorithm to keep graphics, compute, and transfer queue families independent when possible.

Browse Source
This commit is contained in:
Kearwood Gilbert
2022-07-11 23:45:08 -07:00
parent 7f3ceebd5f
commit 738cfd29ef
2 changed files with 119 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
kraken/KRDevice.h
View File

@@ -65,10 +65,14 @@ public:
VkQueue m_graphicsQueue;
uint32_t m_computeFamilyQueueIndex;
VkQueue m_computeQueue;
uint32_t m_transferFamilyQueueIndex;
VkQueue m_transferQueue;
VkCommandPool m_graphicsCommandPool;
VkCommandPool m_computeCommandPool;
VkCommandPool m_transferCommandPool;
std::vector<VkCommandBuffer> m_graphicsCommandBuffers;
std::vector<VkCommandBuffer> m_computeCommandBuffers;
std::vector<VkCommandBuffer> m_transferCommandBuffers;
VmaAllocator m_allocator;
// Staging buffer for uploading with the transfer queue