Multi-tile memory management mechanism
AssigneeIntel Corporation
InventorsZACK S. WATERS, TRAVIS SCHLUESSLER, MICHAEL APODACA, ANKUR SHAH
Life of this patent
Estimating the term…
Abstract
Graphics processors for implementing multi-tile memory management are disclosed. In one embodiment, a graphics processor includes a first graphics device having a local memory, a second graphics device having a local memory, and a graphics driver to provide a single virtual allocation with a common virtual address range to mirror a resource to each local memory of the first and second graphics devices.
Claims25
A graphics processor for a multi-tile architecture, comprising:
- a first graphics device having a local memory;
- a second graphics device having a local memory;
- and
- a graphics driver to provide a single virtual allocation with a common virtual address range to mirror a resource to each local memory of the first and second graphics devices.
A graphics processor for a multi-tile architecture, comprising:
- a first graphics device having a local memory;
- a second graphics device having a local memory;
- and
- a graphics driver to provide a single virtual allocation with a common virtual address range to interleave physical pages of a shared resource to local memory of the first and second graphics devices.
A graphics processor for a multi-tile architecture, comprising:
- a first graphics device having a local memory;
- a second graphics device having a local memory;
- and
- a graphics driver to provide a heap with a common virtual address range for mapping virtual addresses of at least one resource into the heap and to interleave the at least one resource into local memory of the first and second graphics devices.
A graphics processor, comprising:
- a first graphics device having a local memory;
- a second graphics device having a local memory;
- a communication link to couple the first and second graphics devices;
- and
- a memory management unit is configured to utilize a single virtual allocation with a common virtual address range to mirror a resource or to interleave physical pages of the resource to the local memory of the first and second graphics devices.
Dependencies are inferred from the claim text.