1. A system for rollback support in an exposed-pipeline processing element, the system comprising:
the exposed-pipeline processing element comprising rollback support logic; and
the rollback support logic configured to perform a method comprising:
detecting an error associated with execution of an instruction in the exposed-pipeline processing element;
determining whether the exposed-pipeline processing element supports replay of the instruction for a predetermined number of cycles; and
based on determining that the exposed-pipeline processing element supports replay of the instruction, performing a rollback action in the exposed-pipeline processing element to attempt recovery from the error.
2. The system of claim 1, wherein the rollback support logic is further configured to perform:
based on determining that the exposed-pipeline processing element does not support replay of the instruction, triggering an exception to restore the exposed-pipeline processing element to a previously stored checkpoint.
3. The system of claim 1, wherein determining whether the exposed-pipeline processing element supports replay of the instruction further comprises checking a state of an instruction bit of the instruction by decode logic, the instruction bit configured to indicate whether the instruction supports rollback and replay.
4. The system of claim 1, further comprising: a result buffer, an effective address buffer, and an operand buffer, wherein the rollback support logic is further configured to perform:
storing results of an arithmetic logic unit pipeline of the exposed-pipeline processing element to the result buffer;
storing results of a load-store unit pipeline of the exposed-pipeline processing element to the effective address buffer; and
storing operands to the operand buffer of the exposed-pipeline processing element, wherein performing the rollback action comprises selecting one or more of: the result buffer, the effective address buffer, and the operand buffer to replay the instruction for the predetermined number of cycles.
5. The system of claim 4, wherein storing to the result buffer is performed selectively based on determining that a value of the arithmetic logic unit pipeline is used as an operand; storing to the effective address buffer is performed based on determining that a value of the load-store unit pipeline is used as an operand; and storing operands to the operand buffer is based on a value of an instruction bit set to activate the operand buffer.
6. The system of claim 1, wherein the rollback support logic is further configured to perform:
storing current instruction addresses to a current instruction address buffer of the exposed-pipeline processing element; and
storing vector length counter values to a vector length counter buffer of the exposed-pipeline processing element, wherein performing the rollback action comprises restoring a current instruction address with a value stored in the current instruction address buffer, and restoring a vector length counter with a value stored in the vector length counter buffer.
7. The system of claim 1, wherein the rollback support logic is further configured to perform:
storing a previous version of index counters associated with register files in the exposed-pipeline processing element to shadow counters, wherein performing the rollback action comprises restoring the index counters with values from the shadow counters.
8. The system of claim 7, wherein the rollback support logic is further configured to perform:
passing counter control signals used to control the index counters through a multi-cycle delay to control the shadow counters.
9. The system of claim 8, wherein the rollback support logic is further configured to perform:
storing the counter control signals in delay registers, the delay registers updated based on detecting a change in the counter control signals.
10. The system of claim 9, further comprising:
a circular buffer configured to advance based on detecting the change in the counter control signals, the circular buffer advancing a one-hot ring counter head and a one-hot ring counter tail, the one-hot ring counter head combined with a past value of the circular buffer to select one of the delay registers to store the counter control signals, and the one-hot ring counter tail configured to select one of the delay registers to control the shadow counters.
11. A system for rollback support in an exposed-pipeline processing element in an active memory device, the system comprising:
memory in the active memory device; and
the exposed-pipeline processing element configured to communicate with the memory, the exposed-pipeline processing element comprising rollback support logic and configured to perform a method comprising:
saving a checkpoint of the exposed-pipeline processing element to memory of the active memory device;
detecting an error associated with execution of an instruction in the exposed-pipeline processing element;
determining whether the exposed-pipeline processing element supports replay of the instruction for a predetermined number of cycles;
based on determining that the exposed-pipeline processing element supports replay of the instruction, performing a rollback action in the exposed-pipeline processing element to attempt recovery from the error; and
based on determining that the exposed-pipeline processing element does not support replay of the instruction, triggering an exception to restore the exposed-pipeline processing element to the checkpoint from the memory of the active memory device.
12. The system of claim 11, wherein the active memory device is a three-dimensional memory cube with the memory divided into three-dimensional blocked regions as memory vaults.
13. The system of claim 11, wherein determining whether the exposed-pipeline processing element supports replay of the instruction further comprises checking a state of an instruction bit of the instruction by decode logic, the instruction bit configured to indicate whether the instruction supports rollback and replay.
14. The system of claim 11, further comprising: a result buffer, an effective address buffer, and an operand buffer, wherein the rollback support logic is further configured to perform:
storing results of an arithmetic logic unit pipeline of the exposed-pipeline processing element to the result buffer;
storing results of a load-store unit pipeline of the exposed-pipeline processing element to the effective address buffer; and
storing operands to the operand buffer of the exposed-pipeline processing element, wherein performing the rollback action comprises selecting one or more of: the result buffer, the effective address buffer, and the operand buffer to replay the instruction for the predetermined number of cycles.
15. The system of claim 14, wherein storing to the result buffer is performed selectively based on determining that a value of the arithmetic logic unit pipeline is used as an operand; storing to the effective address buffer is performed based on determining that a value of the load-store unit pipeline is used as an operand; and storing operands to the operand buffer is based on a value of an instruction bit set to activate the operand buffer.
16. The system of claim 11, wherein the rollback support logic is further configured to perform:
storing current instruction addresses to a current instruction address buffer of the exposed-pipeline processing element; and
storing vector length counter values to a vector length counter buffer of the exposed-pipeline processing element, wherein performing the rollback action comprises restoring a current instruction address with a value stored in the current instruction address buffer, and restoring a vector length counter with a value stored in the vector length counter buffer.
17. The system of claim 11, wherein the rollback support logic is further configured to perform:
storing a previous version of index counters associated with register files in the exposed-pipeline processing element to shadow counters, wherein performing the rollback action comprises restoring the index counters with values from the shadow counters.
18. The system of claim 17, wherein the rollback support logic is further configured to perform:
passing counter control signals used to control the index counters through a multi-cycle delay to control the shadow counters.
19. The system of claim 18, wherein the rollback support logic is further configured to perform:
storing the counter control signals in delay registers, the delay registers updated based on detecting a change in the counter control signals.
20. The system of claim 19, further comprising:
a circular buffer configured to advance based on detecting the change in the counter control signals, the circular buffer advancing a one-hot ring counter head and a one-hot ring counter tail, the one-hot ring counter head combined with a past value of the circular buffer to select one of the delay registers to store the counter control signals, and the one-hot ring counter tail configured to select one of the delay registers to control the shadow counters.
The claims below are in addition to those above.
All refrences to claim(s) which appear below refer to the numbering after this setence.
1. A method for controlling storage of data blocks in a plurality of storage devices, the method comprising:
buffering a plurality of write requests associated with the data blocks for a single write transaction;
defining a group of storage blocks, the group comprising a plurality of storage blocks in each of the plurality of storage devices; and
associating each data block with a respective one of the storage blocks across the plurality of stripes to generate an association; and
transmitting the association to a storage device manager for processing of the single write transaction.
2. The method of claim 1, further comprising receiving by a file system the plurality of write requests, and transmitting from the file system to the storage device manager the association for processing of the single write transaction, wherein the storage device manager comprises a RAID layer.
3. The method of claim 1, wherein associating comprises associating each data block of at least one of the write requests with storage blocks of only one of the plurality of storage devices.
4. The method of claim 1, wherein defining the group of storage blocks comprises receiving by the file system from the storage device manager configuration information of the plurality of storage devices.
5. The method of claim 4, wherein the configuration information comprises disk topology information.
6. The method of claim 1, wherein the group of storage blocks comprises more than one stripe in a RAID group.
7. The method of claim 6, wherein the group of storage blocks comprises more than one contiguous stripe.
8. The method of claim 7, further comprising transmitting the association and the data to the plurality of storage devices to store each data block at the respective one of the storage blocks of the more than one contiguous stripe.
9. The method of claim 6, wherein the plurality of storage devices comprises more than one RAID group, and defining the group of storage blocks comprises defining the group of storage blocks from storage blocks in a first RAID group.
10. The method of claim 9, further comprising defining a second group of storage blocks in a second RAID group for association with data associated with a second plurality of write requests for processing of a second write transaction.
11. The method of claim 1, further comprising queuing a plurality of associations by the storage device manager for processing of a plurality of write transactions.
12. The method of claim 11, further comprising processing the plurality of write transactions by the storage manager in an order derived from a plurality of priorities associated with the plurality of write transactions.
13. The method of claim 1, wherein the group of storage blocks comprises a same quantity of storage blocks in each one of the plurality of storage devices.
14. The method of claim 13, wherein each of the plurality of storage blocks comprises contiguous storage blocks.
15. The method of claim 1, wherein the group of storage blocks comprises allocated and unallocated storage blocks.
16. The method of claim 15, wherein defining the group further comprises selecting an unallocated storage block having a lowest block identifier of any unallocated storage block as a storage block having a lowest storage block identifier of any storage block in the group.
17. The method of claim 1, wherein the write requests comprise data to be written.
18. A device for controlling storage of data in a plurality of storage devices each having storage blocks configured in a plurality of parallel stripes, the device comprising:
a buffer collecting write requests, each request associated with data blocks;
a processor defining a group of storage blocks, the group having a plurality of storage blocks in each of the plurality of storage devices, the processor associating each data block with a respective one of the storage blocks across the plurality of stripes for a single write transaction; and
a transmitter transmitting an association to a storage device manager for processing of the single write transaction.
19. A device for controlling storage of data in a plurality of storage devices each having storage blocks configured in a plurality of parallel stripes, the device comprising:
means for buffering a plurality of write requests associated with data blocks for a single write transaction;
means for defining a group of storage blocks, the group having a plurality of storage blocks in each of the plurality of storage devices; and
means for associating each data block with a respective one of the storage blocks across the plurality of stripes, for transmitting an association to a storage device manager for processing of the single write transaction.
20. A computer readable media, comprising: the computer readable media containing instructions for execution in a processor for the practice of the method of,
buffering a plurality of write requests associated with data blocks for a single write transaction;
defining a group of storage blocks, the group comprising a plurality of storage blocks configured in a plurality of parallel stripes in each of a plurality of storage devices; and
associating each data block with a respective one of the storage blocks across the plurality of stripes, for transmitting an association to a storage device manager for processing of the single write transaction.
21. A method for controlling storage of data, comprising:
receiving one or more write requests associated with data blocks;
receiving topological information associated with storage blocks configured in a plurality of parallel stripes of a storage system;
associating the data blocks with one or more storage blocks across the plurality of stripes as an association; and
writing the data blocks, in response to the association, to the one or more storage devices in a single write request.
22. The method of claim 21, further comprising: transmitting the association to a storage device manager.
23. The method of claim 21, further comprising: associating each data block with a single storage block.
24. The method of claim 21, further comprising: storing the data blocks in the association.
25. The method of claim 21, further comprising: storing the data blocks in a memory of the storage system.
26. The method of claim 21, further comprising: creating an array as the association.
27. The method of claim 21, further comprising: buffering a plurality of write requests into the single write request.
28. The method of claim 27, further comprising: buffering, in a buffer, the plurality of write requests into the single write request until a predetermined criteria is met.
29. The method of claim 28, further comprising: meeting the criteria when a buffer is full.
30. The method of claim 28, further comprising: meeting the criteria when the single write request is a predetermined logical length.
31. The method of claim 30, further comprising: receiving topological information of a plurality of storage devices of the storage system.
32. The method of claim 31, further comprising: using a RAID system as the plurality of storage devices.
33. The method of claim 21, further comprising: using a plurality of disks for the storage system.
34. The method of claim 21, further comprising: transmitting the association to the storage device manager.
35. The method of claim 21, further comprising: organizing the association as a combination of columns and rows.
36. The method of claim 35, further comprising: associating each column with a storage device.
37. The method of claim 36, further comprising: associating each row with a stripe of the storage device.
38. The method of claim 37, further comprising: arranging the association as a plurality of stripes on a plurality of storage devices.
39. A storage system, comprising:
a file system, the file system to receive one of more write requests associated with data blocks;
a storage device manager, the storage device manager to generate topological information of storage blocks configured in a plurality of parallel stripes of one or more storage devices, and to send the topological information to the file system; and
an association generated in the file system, the association to associate the data blocks with one or more storage blocks across the plurality of stripes of the one or more storage devices, the association to be sent to the storage device manager, the storage device manager to write the data blocks, in response to the association, to the one or more storage blocks as a single write request.
40. The storage system of claim 39, further comprising: an association that associates each data block with a single storage block.
41. The storage system of claim 39, further comprising: a memory to buffer the data blocks for the write request to the one or more storage devices.
42. The storage system of claim 39, further comprising: a memory to store the association containing the data blocks.
43. The storage system of claim 39, further comprising: one or more storage devices having storage blocks.
44. The storage system of claim 39, further comprising: an array as the association.
45. The storage system of claim 39, further comprising: a buffer in the file system to receive the one or more write requests.
46. The storage system of claim 39, further comprising: an arrangement of columns and rows in the association.
47. The storage system of claim 46, further comprising: each column representing a storage device of the one or more storage devices.
48. The storage system of claim 46, further comprising: each row representing a stripe of the one or more storage devices.
49. The storage system of claim 39, further comprising: a plurality of disks as the one or more storage devices.
50. The storage system of claim 39, further comprising: a RAID system as the plurality of storage devices.
51. A storage system, comprising:
means for receiving one or more write requests associated with data blocks;
means for receiving topological information associated with storage blocks configured in a plurality of parallel stripes of a storage system;
means for associating the data blocks with one or more storage blocks across the plurality of stripes as an association; and
means for writing the data blocks, in response to the association, to the one or more storage devices in a single write request.
52. A computer readable media, comprising: the computer readable media containing instructions for execution in a processor for the practice of the method of,
receiving one or more write requests associated with data blocks;
receiving topological information associated with storage blocks configured in a plurality of parallel stripes of a storage system;
associating the data blocks with one or more storage blocks across the plurality of stripes as an association; and
writing the data blocks, in response to the association, to the one or more storage devices in a single write request.