Development build. This page describes main, not a released package. The latest published Lodestar.Gpu is 0.2.0 — read its documentation.

Home › Gpu › GPU kernels

GpuContext.Create

Opens the best available device, or ILGPU's CPU accelerator.

public static GpuContext Create(bool preferCpu = false)

Parameters — preferCpu forces the CPU accelerator. That is what a machine with no GPU runs, and what proves a kernel correct where the 5–10× gate cannot be evaluated at all.

Returns — GpuContext, owning an accelerator the caller disposes.

Exceptions — none of its own. A device that fails to open surfaces ILGPU's own exception, after the context opened for it is disposed.

Example — the shape a caller writes.

using Lodestar.Gpu.Compute;

using var forced = GpuContext.Create(preferCpu: true);

bool graphics = forced.IsHardwareGpu;  // => False

Remarks — Context.GetPreferredDevice is not trusted to pick a GPU. Measured: asked for a non-CPU device it returned an OpenCL runtime executing on the processor, ahead of a CUDA card, so every figure a benchmark produced would have been the CPU's. This orders the devices itself — CUDA first, then OpenCL on hardware whose device type is a GPU — and asks GetPreferredDevice only for the CPU accelerator, with preferCPU: true.

Where no graphics device exists the CPU accelerator is opened rather than throwing, because a correctness suite has to run on a machine without one. Check IsHardwareGpu before reading a timing; a functional result is the same either way.

Applies to — net10.0, netstandard2.1.

See also — GpuContext.