xlog_cuda/
lib.rs

1//! GPU kernel provider for XLOG
2
3pub mod arrow_device;
4pub mod cuda_compat;
5pub mod cuda_graph;
6pub mod device;
7pub mod device_pool;
8pub mod device_runtime;
9pub mod dlpack;
10pub mod kernel_manifest_data;
11pub mod launch;
12pub mod memory;
13pub mod multi_gpu_memory;
14pub mod provider;
15pub mod type_seam;
16pub mod wcoj_metadata;
17#[cfg(feature = "wcoj-phase-timing")]
18pub mod wcoj_phase_timing;
19
20pub(crate) mod embedded_kernel_data {
21    include!(concat!(env!("OUT_DIR"), "/embedded_kernel_data.rs"));
22}
23
24pub use arrow_device::{ArrowDeviceArray, ArrowDeviceArrayOwned, ARROW_DEVICE_CUDA};
25pub use cuda_compat::{
26    sys, AsKernelParam, CudaFunction, CudaSlice, CudaStream, CudaView, CudaViewMut, DevicePtr,
27    DevicePtrMut, DeviceRepr, DeviceSlice, DriverError, IntoKernelParamStorage, KernelParamStorage,
28    KernelScalar, LaunchAsync, LaunchConfig, ValidAsZeroBits,
29};
30pub use device::CudaDevice;
31pub use device_pool::GpuDevicePool;
32pub use dlpack::{DLManagedTensor, DlpackManagedTensor, DlpackTable};
33pub use memory::{CudaBuffer, CudaColumn, GpuMemoryManager, RuntimeAllocBlock};
34pub use multi_gpu_memory::MultiGpuMemoryManager;
35pub use provider::{
36    circuit_kernels, dedup_kernels, filter_kernels, groupby_kernels, ilp_kernels, join_kernels,
37    pack_kernels, pir_kernels, scan_kernels, set_ops_kernels, sort_kernels, CompareOp,
38    CudaKernelProvider, FjDeltaCols, FjNode, FjPlan, FjSubAtom, JoinIndexV2, JoinType,
39    CIRCUIT_MODULE, DEDUP_MODULE, FILTER_MODULE, GROUPBY_MODULE, ILP_MODULE, JOIN_MODULE,
40    PACK_MODULE, PIR_MODULE, SCAN_MODULE, SET_OPS_MODULE, SORT_MODULE,
41};
42pub use wcoj_metadata::{
43    HeatDist, LayoutSignature, RootMetadata, VertexId, WcojCycle4HgWorkPlanU32,
44    WcojCycle4HgWorkPlanU64, WcojRelationMetadata, WcojTriangleHgWorkPlanU32,
45    WcojTriangleHgWorkPlanU64,
46};
xlog_cuda/lib.rs

xlog_cuda/
lib.rs
