Apparatus and method to improve memory access performance between shared local memory and system global memory

1. A non-transitory machine readable storage medium having instructions stored thereon, that when executed, cause a processor to perform a method comprising:

grouping two or more work groups to form a super-workgroup, wherein the number of work groups forming the super-workgroup is a dynamic number, which is not a fixed number of workgroups during operation; and

partitioning a portion of a memory space into one or more super-shared local memories (Super-SLMs), wherein memory space, which is shared within the super-workgroup forms at least one Super-SLM of the one or more Super-SLMs.

2. The non-transitory machine readable storage medium of claim 1 having further instructions stored thereon that when executed cause the processor to perform the method comprising: writing data to a global memory by the super-workgroup.

3. The non-transitory machine readable storage medium of claim 1 , wherein at least one work group includes a plurality of work items.

4. The non-transitory machine readable storage medium of claim 3 , wherein memory space within a work group from among the two or more work groups comprises a shared local memory (SLM).

5. The non-transitory machine readable storage medium of claim 4 , wherein the SLM is separate from the super-SLM.

6. The non-transitory machine readable storage medium of claim 4 , wherein the SLM is a subset of the super-SLM.

7. The non-transitory machine readable storage medium of claim 1 having further instructions that when executed cause the processor to perform a further method which comprises: determining, for the super-workgroup, a last work group in a sub-slice.

8. The non-transitory machine readable storage medium of claim 7 having further instructions that when executed cause the processor to further perform a method which comprises: generating the last work group when all other work groups in the super-workgroup complete their operations.

9. The non-transitory machine readable storage medium of claim 7 having further instructions that when executed cause the processor to further perform a method which comprises: writing, by the last work group, data back to a global memory from the Super-SLM.

10. The non-transitory machine readable storage medium of claim 1 , wherein the portion of the memory space is at least one of: a cache; a First-in-First-Out (FIFO); or a logical partitioned memory.

11. The non-transitory machine readable storage medium of claim 1 , wherein the work groups that form the super-workgroup is generated by a scheduler.

12. The non-transitory machine readable storage medium of claim 1 , wherein the dynamic number of work groups depends on a number of workgroups in a sub-slice.

13. The non-transitory machine readable storage medium of claim 1 , wherein the dynamic number of work groups depends on when a last workgroup is determined or generated.

14. An apparatus comprising:

a plurality of execution units;

a cache memory having a portion which operates as a shared local memory (SLM), which is shared with the plurality of execution units, at least one of the execution units being configured to operate on a work group of a sub-slice, wherein the SLM is shared within the work group; and

at least one Super-SLM to provide shared memory accessible by a number of different work groups in the sub-slice, wherein the at least one of the execution units is configured to operate on the different work groups, wherein the different work groups are grouped to form a super-workgroup, and wherein the number of the different work groups that forms a super-work group is a dynamic number of work groups, which is not a fixed number of workgroups during operation.

15. The apparatus of claim 14 , wherein the super-workgroup is configured to write data to a global memory of the apparatus.

16. The apparatus of claim 14 further comprising a scheduler to determine, for the super-workgroup, a last work group in a sub-slice.

17. The apparatus of claim 16 , wherein the scheduler is operable to service the last work group when all other work groups in the super-workgroup complete their operations.

18. The apparatus of claim 16 wherein the last work group is configured to write data back to a global memory from the Super-SLM.

19. The apparatus of claim 14 , wherein the Super-SLM comprises a super set of the SLM.

20. A system comprising:

an input/output (I/O) interface;

a memory;

a graphics processor coupled to the memory via the I/O interface, the graphics processor including:

a plurality of execution units;

a cache memory having a portion which operates as a shared local memory (SLM), which is shared with the plurality of execution units, at least one of the execution units being configured to operate on a work group of a sub-slice, wherein the SLM is shared within the work group; and

at least one Super-SLM to provide shared memory accessible by different work groups in the sub-slice, wherein the at least one of the execution units is configured to operate on the different work groups, wherein the number of the different work groups that forms a super workgroup is a dynamic number of work groups, which is not a fixed number of workgroups during operation; and

a wireless interface to allow the graphics processor to communicate with another device.

21. The system of claim 20 , wherein the I/O interface comprises a Double Data Rate (DDR) compliant interface.