Systems for two-dimensional main memory including memory modules with read-writeable non-volatile memory devices
View Patent ↗In one embodiment of the invention, a system is disclosed including a master memory controller and a plurality of memory modules coupled to the master memory controller. Each memory module includes a plurality of read-writeable non-volatile memory devices in a plurality of memory slices to form a two-dimensional array of memory. Each memory slice in each memory module includes a slave memory controller coupled to the master memory controller. When the master memory controller issues a memory module request, it is partitioned into a slice request for each memory slice.
1. A system including
a master memory controller,
a plurality of memory modules arranged into N rows by M columns of memory modules coupled to the master memory controller to form a two-dimensional memory array of Z memory slices over the M columns of memory modules,
wherein each memory module includes
a plurality of read-writeable non-volatile memory devices in a plurality of memory slices to form a portion of the two-dimensional memory array of Z memory slices,
wherein each of the plurality of memory slices in each memory module includes
a slave memory controller coupled to and between one or more of the plurality of read-writeable non-volatile memory devices and the master memory controller;
and
wherein the master memory controller forms a compound memory request comprising a plurality of slice requests that is partitioned into a slice request for each respective memory slice of the Z memory slices over the M columns of memory modules.
2. The system of claim 1 wherein
the compound memory request is a read request that expects a response of data thereto, and
the plurality of memory slices to generate a plurality of memory slice responses that are packed together to form a compound memory response to the compound memory request.
3. The system of claim 1 , wherein
each memory slice can access the plurality of read-writeable non-volatile memory devices autonomously to selectively access one or more read-writeable non-volatile memory devices and selectively access different memory locations in each.
4. The system of claim 1 wherein
a memory slice responds with known null data if its memory access operation is incomplete or if the memory slice does not participate in the compound memory request.
5. The system of claim 1 , wherein
the memory controller to pack a plurality of memory requests for the read-writeable non-volatile memory devices to form the compound memory request and
each slave memory controller to remap the addresses for the two-dimensional array into linear memory accesses into the read-writeable non-volatile memory devices in its respective memory slice.
6. The system of claim 3 , wherein
the address to each read-writeable non-volatile memory device differs in the memory slice.
7. The system of claim 3 , wherein
the master memory controller synchronously forms the compound memory request to randomly access blocks, one block per slice, across memory slices in the two-dimensional array over a memory channel;
the master memory controller synchronously receives a compound memory response including packed partial block responses from each memory slice, and
the master memory controller includes a two-dimensional transpositional buffer to store the partial block responses from each memory slice, and
the master memory controller to read data from the two dimensional transpositional buffer to concatenate the partial block responses into a full response from the respective memory slices for a system processor to increase data bandwidth over the memory channel.
8. The system of claim 7 , wherein
the partial block responses from each memory slice are stored row-wise into the two-dimensional transpositional buffer; and
the full response is read out from the two dimensional transpositional buffer column-wise to a system processor.
9. The system of claim 3 , wherein
the master memory controller includes a two-dimensional transpositional buffer to store data;
the master memory controller writes a plurality of blocks of data from a system processor associated with one or more memory requests into the two-dimensional transpositional buffer;
the two dimensional transpositional buffer to read out a plurality of blocks one block per slice across memory slices into the two-dimensional memory array over the memory channel; and
the master memory controller schedules a compound memory request to begin writing blocks of data over the memory channel one block per slice into random memory block locations within the respective memory slices of the two-dimensional memory array.
10. The system of claim 9 , wherein
the plurality of blocks of data from the system processor are stored column-wise into the two-dimensional transpositional buffer; and
blocks of data are read out from the two dimensional transpositional buffer row-wise into the two-dimensional memory array.
11. The system of claim 3 , wherein
the master memory controller includes command queues associated each memory slice to concatenate a plurality of memory requests from software applications into one compound memory request for one or more memory slices into the two-dimensional memory array.
12. The system of claim 11 , wherein
at least one memory slice of the two-dimensional memory array is not accessed by the compound memory request.
13. The system of claim 12 , wherein
the at least one memory slice not accessed responds with known null data.
14. The system of claim 1 , wherein
the compound memory request activates a plurality of memory modules in the two dimensional memory array.
15. The system of claim 14 , wherein
the compound memory request activates differing memory modules in the two dimensional memory array to perform the plurality of slice requests in the compound memory request.
16. A system comprising:
a processor,
a master memory controller coupled to the processor, the master memory controller including a two-dimensional transpositional buffer, the two-dimensional transpositional buffer to access partial block data;
a plurality of memory modules coupled to the master memory controller, each memory module including a plurality of read-writeable non-volatile memory devices in a plurality of memory slices to form a two-dimensional memory array of a plurality of memory slices over two or more columns of memory modules, wherein each memory slice in each memory module includes
a slave memory controller coupled to and between the master memory controller and one or more of the plurality of read-writeable non-volatile memory devices in each memory module; and
wherein the master memory controller forms a compound memory request comprising a plurality of slice requests that is partitioned into a slice request for each respective memory slice of the plurality of memory slices, and
the two-dimensional transpositional buffer transposes data between a column-wise access with the processor and a row-wise access with the plurality of memory slices in the two-dimensional memory array.
17. The system of claim 16 , wherein
each memory slice can access the plurality of read-writeable non-volatile memory devices autonomously to selectively access one or more read-writeable non-volatile memory devices and selectively access different memory locations in each.
18. The system of claim 17 , wherein
a read request is issued by the processor to the master memory controller,
the master memory controller synchronously makes a compound memory request to randomly access blocks, one block per slice, across memory slices in the two-dimensional array over the memory channel;
the master memory controller synchronously receives a compound memory response including packed partial block data responses from each memory slice and stores them in the two dimensional transpositional buffer, and
the master memory controller to read data from the two dimensional transpositional buffer to concatenate the partial block data responses into a full response from the respective memory slices for the processor.
19. The system of claim 18 , wherein
the partial block responses from each memory slice are stored row-wise into the two-dimensional transpositional buffer; and
the full response is read out from the two dimensional transpositional buffer column-wise to the processor.
20. The system of claim 17 , wherein
a write request with full write data is issued by the processor to the master memory controller,
the master memory controller synchronously makes a compound memory request to randomly access blocks one block per slice across memory slices in the two-dimensional array over the memory channel; and
the master memory controller synchronously issues packed partial block data to each memory slice.
21. The system of claim 20 , wherein
the full write data from the processor is written column-wise into the two dimensional transpositional buffer; and
the packed partial block data for each memory slice is read out row-wise from the two-dimensional transpositional buffer.
22. The system of claim 16 , wherein
the compound memory request activates a plurality of memory modules in the two dimensional memory array.
23. The system of claim 22 , wherein
the compound memory request activates differing memory modules in the two dimensional memory array to perform the plurality of slice requests in the compound memory request.
24. A system comprising:
a plurality of memory modules arranged into a plurality of columns and a plurality of rows to form a two-dimensional memory array of a plurality of memory slices over the plurality of columns of memory modules, each memory module including a plurality of read-writeable non-volatile memory devices to form a portion of the plurality of memory slices, wherein each memory slice in each memory module further includes
a slave memory controller coupled to the plurality of read-writeable non-volatile memory devices of the memory module;
a master memory controller coupled to each slave memory controller of the plurality of memory modules, the master memory controller forms a compound memory request comprising a plurality of slice requests for each respective memory slice of the plurality of memory slices;
a processor coupled to the master memory controller, the processor to generate the plurality of slice requests; and
wherein the master memory controller includes a two-dimensional transpositional buffer coupled between the processor and the plurality of memory modules, the two-dimensional transpositional buffer to provide row by row data access to the two-dimensional memory array of the plurality of memory slices and to provide column by column data access to the processor.
25. The system of claim 24 , wherein the two-dimensional transpositional buffer provides data access to partial blocks of data.
26. The system of claim 24 , wherein
the master memory controller further includes
a memory request buffer coupled to the processor, the memory request buffer to buffer memory requests from the processor, and
a request aggregator coupled to the memory request buffer, the request aggregator to aggregate a plurality of memory requests in the memory request buffer to form the compound memory request of the plurality of slice requests for each respective memory slice of the plurality of memory slices.
27. The system of claim 24 , wherein
the master memory controller further includes
a plurality of command queues respectively associated with each memory slice, the plurality of command queues to concatenate a plurality of memory requests from one or more software applications into one compound memory request for one or more memory slices in the two-dimensional memory array.
28. The system of claim 24 , wherein
the two-dimensional transpositional buffer stores data;
the master memory controller writes a plurality of blocks of data from the processor associated with one or more memory requests into the two-dimensional transpositional buffer;
the two dimensional transpositional buffer reads out a plurality of blocks one block per slice across the plurality of memory slices into the two-dimensional memory array over a memory channel; and
the master memory controller schedules a compound memory request to begin writing blocks of data over the memory channel, one block per slice, into random memory block locations within the respective memory slices of the two-dimensional memory array.
29. The system of claim 28 , wherein
the plurality of blocks of data from the processor are stored column-wise into the two-dimensional transpositional buffer; and
the blocks of data are read out from the two dimensional transpositional buffer row-wise into the two-dimensional memory array.
30. The system of claim 24 , wherein
the compound memory request activates a plurality of memory modules in the two dimensional memory array.
31. The system of claim 30 , wherein
the compound memory request activates differing memory modules in the two dimensional memory array to perform the plurality of slice requests in the compound memory request.