Instruction prefetch throttling
An apparatus is provided for limiting the effective utilisation of an instruction fetch queue. The instruction fetch entries are used to control the prefetching of instructions from memory, such that those instructions are stored in an instruction cache prior to being required by execution circuitry while executing a program. By limiting the effective utilisation of the instruction fetch queue, fewer instructions will be prefetched and fewer instructions will be allocated to the instruction cache, thus causing fewer evictions from the instruction cache. In the event that the instruction fetch entries are for instructions that are unnecessary to the program, the pollution of the instruction cache with these unnecessary instructions can be mitigated.
1. An apparatus comprising:
an instruction fetch queue for storage of instruction fetch entries;
prefetching circuitry configured, for each of the instruction fetch entries, to fetch an instruction indicated by that instruction fetch entry and to cause the instruction to be allocated to an instruction cache; and
control circuitry configured to suppress allocation of a further instruction fetch entry to the instruction fetch queue in response to a number of the instruction fetch entries meeting or exceeding a limit value, wherein the limit value is less than a storage capacity of the instruction fetch queue.
2. The apparatus of claim 1 , wherein
in response to fetching the instruction indicated by the instruction fetch entry, the prefetching circuitry is configured to cause a deallocation of the instruction fetch entry; and
in response to the number of instruction fetch entries being less than the limit value after the deallocation, the control circuitry is configured to cause the further instruction fetch entry to be allocated to the instruction fetch queue.
3. The apparatus of claim 1 , wherein
the limit value is dynamically modifiable.
4. The apparatus of claim 3 , comprising:
event measurement circuitry configured to detect occurrences of one or more events; and
determination circuitry configured to update the limit value based on the events detected.
5. The apparatus of claim 4 , wherein:
the event measurement circuitry is configured to measure the number of occurrences of one or more events; and
the determination circuitry is configured to update the limit value based on the number of occurrences of the one or more events within a measurement period.
6. The apparatus of claim 5 , wherein
the measurement period is a predetermined number of allocations to the instruction fetch queue.
7. The apparatus of claim 4 , comprising:
prediction circuitry configured to generate the further instruction fetch entry,
wherein the one or more events are indicative of an increased risk of misprediction by the prediction circuitry.
8. The apparatus of claim 7 , wherein
the prediction circuitry is configured to generate a further instruction fetch entry after the further instruction fetch entry has been allocated to the instruction fetch queue.
9. The apparatus of claim 7 , wherein
the control circuitry is configured to suppress allocation of the further instruction fetch entry to the instruction fetch queue by causing the prediction circuitry to stall.
10. The apparatus of claim 7 , wherein
the one or more events comprises at least one of:
a miss in the instruction cache,
a miss in a branch target buffer, and
a miss in a branch prediction buffer.
11. The apparatus of claim 1 , comprising:
sequence tracking circuitry configured to determine a number of contiguous instruction fetch entries for instructions that are contiguous in program order, wherein
the control circuitry is configured to suppress allocation of the further instruction fetch entry to the instruction fetch queue in response to the number of contiguous instruction fetch entries meeting or exceeding the limit value.
12. The apparatus of claim 11 , wherein
in response to fetching an instruction indicated by a corresponding instruction fetch entry, the prefetching circuitry is configured to cause a deallocation of the corresponding instruction fetch entry; and
in response to the number of contiguous instruction fetch entries being less than the limit value after the deallocation, the control circuitry is configured to cause the further instruction fetch entry to be allocated to the instruction fetch queue.
13. The apparatus of claim 11 , wherein
the sequence tracking circuitry is configured to determine the number of contiguous instruction fetch entries based on a number of instruction fetch entries allocated to the instruction fetch queue since detecting a break in a predicted program flow.
14. The apparatus of claim 13 , wherein
the sequence tracking circuitry is configured to detect the break in the predicted program flow based on at least one of:
an instruction fetch entry corresponding to a predicted taken branch instruction; and
a pipeline flush.
15. A system comprising:
the apparatus of claim 1 , implemented in at least one packaged chip;
at least one system component; and
a board,
wherein the at least one packaged chip and the at least one system component are assembled on the board.
16. A chip-containing product comprising the system of claim 15 assembled on a further board with at least one other product component.
17. A method comprising:
storing instruction fetch entries in an instruction fetch queue;
for each of the instruction fetch entries, fetching an instruction indicated by that instruction fetch entry and causing the instruction to be allocated to an instruction cache; and
suppressing allocation of a further instruction fetch entry in response to a number of instruction fetch entries meeting or exceeding a limit value, wherein the limit value is less than a storage capacity of the instruction fetch queue.
18. A non-transitory computer-readable medium to store computer-readable code for fabrication of an apparatus comprising:
an instruction fetch queue for storage of instruction fetch entries;
prefetching circuitry configured, for each of the instruction fetch entries, to fetch an instruction indicated by that instruction fetch entry and to cause the instruction to be allocated to an instruction cache; and
control circuitry configured to suppress allocation of a further instruction fetch entry to the instruction fetch queue in response to a number of the instruction fetch entries meeting or exceeding a limit value, wherein the limit value is less than a storage capacity of the instruction fetch queue.