ImportedGpuBufferRing.hpp File Reference

Backend-uniform N-slot exportable GPU buffer ring for vendor P2P. More...

Detailed Description

Backend-uniform N-slot exportable GPU buffer ring for vendor P2P.

Captures the QRhi-adopted-buffer ring pattern shared by every GPU-direct video output strategy:

1. Allocate N QRhi StorageBuffers.
2. Extract each underlying native buffer (ID3D11Buffer*, GLuint, ...) via QRhiBuffer::nativeBuffer().
3. Import each into CUDA via the CudaInterop C API.
4. Hand the resulting flat GPU device pointer to the peer device's DMA-pin API (AJA DMABufferLock(inRDMA=true), Magewell MWPinVideoBuffer, Rivermax rmx_register_memory, ...).
5. Per frame: rotate the ring; encoder writes into the current slot's QRhi buffer; peer device DMA-reads from the same slot's GPU device pointer.

Gives capture-card output addons a shared slot ring so each one doesn't re-implement the dance. Vendor-specific concerns (the DMA-lock call, the per-frame submission API, pacing) stay in each vendor's addon.

Backend support matrix:

• D3D11 : real (ID3D11Buffer + cuda_interop_import_d3d11_buffer)
• OpenGL : real (GLuint + cuda_interop_import_gl_buffer)
• Vulkan : stub (QRhi VMA-allocates without VkExportMemoryAllocateInfo; needs either a QRhi private-API patch or a parallel externally-managed buffer ring with beginExternal compute dispatch)
• D3D12 : stub (QRhi D3D12 backend lacks SHARED heap support)

Go to the source code of this file.

Classes
struct	score::gfx::interop::GpuRingBufferSlot
	One slot of the ring. Lifetime owned by ImportedGpuBufferRing. More...
struct	score::gfx::interop::ImportedGpuBufferRingConfig
class	score::gfx::interop::ImportedGpuBufferRing
	Vendor-neutral helper that allocates + manages the slot ring. More...

Namespaces
namespace	score
	Base toolkit upon which the software is built.
namespace	score::gfx
	Graphics rendering pipeline for ossia score.