Glossary
Key terminology and definitions for DotCompute development.
🚧 Documentation In Progress - Glossary is being developed with comprehensive terminology.
A
Accelerator
TODO: Define accelerator term
AOT (Ahead-of-Time)
TODO: Define Native AOT compilation
B
Backend
TODO: Define compute backend
CUDA/Compute Capability
TODO: Define CUDA and compute capability levels
C
Cooperative Groups
TODO: Define cooperative groups concept
D
Device
TODO: Define compute device
G
GPU
TODO: Define GPU terminology
Kernel
TODO: Define compute kernel
H
Host
TODO: Define host system
L
LINQ
TODO: Define LINQ integration
M
Memory Pooling
TODO: Define memory pooling
Metal
TODO: Define Apple Metal API
N
Native AOT
TODO: Define Native AOT (Native Ahead-of-Time)
NCCL
TODO: Define NCCL collective communications
O
Orchestrator
TODO: Define compute orchestrator
P
P2P (Peer-to-Peer)
TODO: Define GPU P2P transfers
PTX
TODO: Define PTX parallel thread execution
R
Ring Kernel
TODO: Define Ring Kernel programming model
ROCm
TODO: Define AMD ROCm platform
S
SIMD
TODO: Define SIMD vectorization
T
Thread Block
TODO: Define CUDA thread blocks
U
Unified Memory
TODO: Define unified memory concept
Unified Buffer
TODO: Define DotCompute unified buffers
W
Warp
TODO: Define GPU warp concept