pystencils.codegen.config.GpuIndexingScheme

pystencils.codegen.config.GpuIndexingScheme#

class pystencils.codegen.config.GpuIndexingScheme(value)#

Available index translation schemes for GPU kernels.

Attributes

Linear3D

Map coordinates to global thread indices.

GridstridedLinear3D

Extension to GpuIndexingScheme.Linear3D scheme which introduces additional loops with strides over the execution configuration's grid size.

Blockwise4D

On a 3D grid of 1D blocks, map the fastest coordinate onto the intra-block thread index, and slower coordinates onto the block index.