Method of organizing a programmable atomic unit instruction memory
Disclosed in some examples, are methods, systems, devices, and machine readable mediums that store instructions for programmable atomic transactions in a memory of the programmable atomic unit prior to execution of the programmable atomic transaction. The memory in some examples may be an instruction RAM. The memory in some examples may be partitioned into partitions of a fixed size that stores a same number of instructions. Each programmable atomic transaction may use one or more contiguously located instruction partitions. By loading the instructions ahead of time, the instructions are ready for execution when the transaction is requested.
1. An apparatus comprising:
a memory controller comprising a programmable atomic unit, the programmable atomic unit comprising:
a memory partitioned into multiple partitions;
a processor, the processor configured to:
store instructions of a programmable atomic operation in multiple contiguous partitions of the memory;
update a control structure to identify a number of the multiple contiguous partitions in which the instructions are stored;
receive, at the programmable atomic unit, a request from a requesting processor to execute the programmable atomic operation, the request specifying the programmable atomic operation by specifying a first partition on which the instructions are stored, the first partition one of the multiple contiguous partitions;
execute the instructions of the programmable atomic operation to modify a value of a memory controlled by the memory controller, the execution of the instructions starting at instructions stored in the first partition in the memory and, for each current instruction executed, verifying that the current instruction is within one of the multiple contiguous partitions; and
responsive to reaching a last instruction, send a response to the requesting processor.
2. The apparatus of claim 1 , wherein the processor is further configured to:
receive a message identifying the instructions of the programmable atomic operation; and
responsive to receipt of the message, store the instructions in the multiple contiguous partitions.
3. The apparatus of claim 2 , wherein the message is received from the requesting processor over a packet-based network.
4. The apparatus of claim 1 , wherein the processor is further configured to:
receive a second request to execute a second programmable atomic transaction, the request specifying a second partition;
determine that the second partition does not have any instructions based upon a value of a validity flag of the control structure corresponding to the second partition; and
send a second response to the requesting processor indicating that the second programmable atomic transaction failed.
5. The apparatus of claim 1 , wherein the control structure includes a maximum number of instructions to execute and wherein the processor is configured to execute the instructions by comparing a number of instructions executed within the maximum number of instructions.
6. The apparatus of claim 1 , wherein the response includes a status code.
7. A method comprising:
storing instructions of a programmable atomic operation in multiple contiguous partitions of a memory of a programmable atomic unit, the programmable atomic unit part of a memory controller;
updating a control structure to identify a number of the multiple contiguous partitions in which the instructions are stored;
receiving, at the programmable atomic unit, a request from a requesting processor to execute the programmable atomic operation, the request specifying the programmable atomic operation by specifying a first partition on which the instructions are stored, the first partition one of the multiple contiguous partitions;
executing the instructions of the programmable atomic operation to modify a value of a memory controlled by the memory controller, the execution of the instructions starting at instructions stored in the first partition in the memory and, for each current instruction executed, verifying that the current instruction is within one of the multiple contiguous partitions; and
responsive to reaching a last instruction, sending a response to the requesting processor.
8. The method of claim 7 , further comprising:
receiving a message identifying the instructions of the programmable atomic operation; and
responsive to receiving the message, storing the instructions in the multiple contiguous partitions.
9. The method of claim 8 , wherein the message is received from the requesting processor over a packet-based network.
10. The method of claim 7 , comprising:
receiving a second request to execute a second programmable atomic transaction, the request specifying a second partition;
determining that the second partition does not have any instructions based upon a value of a validity flag of the control structure corresponding to the second partition; and
sending a second response to the requesting processor indicating that the second programmable atomic transaction failed.
11. The method of claim 7 , wherein the control structure includes a maximum number of instructions to execute and wherein executing the instructions comprises comparing a number of instructions executed within the maximum number of instructions.
12. The method of claim 7 , wherein the response includes a status code.
13. The method of claim 7 , wherein the programmable atomic unit is part of a memory controller, and wherein the instructions of the programmable atomic transaction comprise instructions to modify a value of a memory location controlled by the memory controller.
14. A non-transitory, machine-readable medium storing instructions, which when executed by a machine, cause the machine to perform operations comprising:
storing instructions of a programmable atomic operation in multiple contiguous partitions of a memory of a programmable atomic unit, the programmable atomic unit part of a memory controller;
updating a control structure to identify a number of the multiple contiguous partitions in which the instructions are stored;
receiving, at the programmable atomic unit, a request from a requesting processor to execute the programmable atomic operation, the request specifying the programmable atomic operation by specifying a first partition on which the instructions are stored, the first partition one of the multiple contiguous partitions;
executing the instructions of the programmable atomic operation to modify a value of a memory controlled by the memory controller, the execution of the instructions starting at instructions stored in the first partition in the memory and, for each current instruction executed, verifying that the current instruction is within one of the multiple contiguous partitions; and
responsive to reaching a last instruction, sending a response to the requesting processor.
15. The non-transitory, machine-readable medium of claim 14 , wherein the operations further comprise:
receiving a message identifying the instructions of the programmable atomic operation; and
responsive to receiving the message, storing the instructions in the multiple contiguous partitions.
16. The non-transitory, machine-readable medium of claim 15 , wherein the message is received from the requesting processor over a packet-based network.
17. The non-transitory, machine-readable medium of claim 14 , wherein the operations further comprise:
receiving a second request to execute a second programmable atomic transaction, the request specifying a second partition;
determining that the second partition does not have any instructions based upon a value of a validity flag of the control structure corresponding to the second partition; and
sending a second response to the requesting processor indicating that the second programmable atomic transaction failed.
18. The non-transitory, machine-readable medium of claim 14 , wherein the control structure includes a maximum number of instructions to execute and wherein executing the instructions comprises comparing a number of instructions executed within the maximum number of instructions.