Instruction decode cluster offlining
An embodiment of an integrated circuit may comprise a core and an instruction decoder communicatively coupled to the core to decode one or more instructions for execution by the core, where the instruction decoder includes two or more decode clusters in a parallel arrangement, and circuitry to offline a decode cluster of the two or more decode clusters. Other embodiments are disclosed and claimed.
1. An integrated circuit, comprising: a core an execution unit to execute instructions; and
an instruction decoder communicatively coupled to the execution unit, wherein the instruction decoder comprises:
multiple decode clusters coupled, in parallel with each other, to receive and decode portions of one or more fetch streams, wherein the portions are to be received, each by a respective decode cluster of the multiple decode clusters, based on an order of the multiple decode clusters relative to each other; and
circuitry coupled to the multiple decode clusters, the circuitry to:
perform a transition of a first decode cluster of the multiple decode clusters from an online state with respect to the one or more fetch streams, to an offline state with respect to the one or more fetch streams, wherein:
while in the online state, the first decode cluster is to be included in the order of the multiple decode clusters relative to each other; and
while in the offline state, the first decode cluster is to be bypassed in the order of the multiple decode clusters relative to each other;
generate a control signal based on one or more bypasses of the first decode cluster; and
provide the control signal to operate a multiplexer circuit which is to be coupled between the multiple decode clusters and one or more instruction queues.
2. The integrated circuit of claim 1 , wherein the circuitry is further to:
power down the first decode cluster while the first decode cluster is in the offline state.
3. The integrated circuit of claim 1 , wherein the circuitry is further to:
utilize the first decode cluster for a prefetch activity while the first decode cluster is in the offline state.
4. The integrated circuit of claim 3 , wherein the circuitry is to:
utilize the first decode cluster for the prefetch activity in parallel with instruction decode activity on one or more online decode clusters of the multiple decode clusters.
5. The integrated circuit of claim 4 , wherein the prefetch activity includes one or more of an instruction translation look-aside buffer (ITLB) page walk, an ITLB fill, an instruction cache request, and instruction cache fill, an instruction end byte detection, and a branch prediction update.
6. A method at a processor, the method comprising:
receiving portions of one or more fetch streams with multiple decode clusters which are coupled in parallel with each other, wherein the portions are received, each by a respective decode cluster of the multiple decode clusters, based on an order of the multiple decode clusters relative to each other;
decoding the portions with the multiple decode clusters;
during the decoding, performing a transition of a first decode cluster of the multiple decode clusters from an online state with respect to the one or more fetch streams, to an offline state with respect to the one or more fetch streams, wherein:
while in the online state, the first decode cluster is included in the order of the multiple decode clusters relative to each other; and
while in the offline state, the first decode cluster is bypassed in the order of the multiple decode clusters relative to each other;
generating a control signal based on one or more bypasses of the first decode cluster; and
operating a multiplexer circuit with the control signal, wherein the multiplexer circuit is coupled between the multiple decode clusters and one or more instruction queues.
7. The method of claim 6 , further comprising:
powering down the first decode cluster while the first decode cluster is in the offline state.
8. The method of claim 6 , further comprising:
utilizing the first decode cluster for a prefetch activity while the first decode cluster is in the offline state.
9. The method of claim 8 , further comprising:
utilizing the first decode cluster for the prefetch activity in parallel with instruction decode activity on one or more online decode clusters of the multiple decode clusters.
10. The method of claim 9 , wherein the prefetch activity includes one or more of an instruction translation look-aside buffer (ITLB) page walk, an ITLB fill, an instruction cache request, and instruction cache fill, an instruction end byte detection, and a branch prediction update.
11. An apparatus, comprising:
a back end unit to execute one or more decoded instructions; and
a front end unit communicatively coupled to the back end unit to decode one or more instructions, wherein the front end unit comprises:
multiple decode clusters coupled, in parallel with each other, to receive and decode portions of one or more fetch streams, wherein the portions are to be received, each by a respective decode cluster of the multiple decode clusters, based on an order of the multiple decode clusters relative to each other; and
circuitry coupled to the multiple decode clusters, the circuitry to:
perform a transition of a first decode cluster of the multiple decode clusters from an online state with respect to the one or more fetch streams, to an offline state with respect to the one or more fetch streams, wherein:
while in the online state, the first decode cluster is to be included in the order of the multiple decode clusters relative to each other; and
while in the offline state, the first decode cluster is to be bypassed in the order of the multiple decode clusters relative to each other;
generate a control signal based on one or more bypasses of the first decode cluster; and
provide the control signal to operate a multiplexer circuit which is to be coupled between the multiple decode clusters and one or more instruction queues.
12. The apparatus of claim 11 , wherein the circuitry is further to:
power down the first decode cluster while the first decode cluster is in the offline state.
13. The apparatus of claim 11 , wherein the circuitry is further to:
utilize the first decode cluster for a prefetch activity while the first decode cluster is in the offline state.
14. The apparatus of claim 13 , wherein the circuitry is to:
utilize the first decode cluster for the prefetch activity in parallel with instruction decode activity on non-offlined decode clusters of the multiple decode clusters.
15. The apparatus of claim 14 , wherein the prefetch activity includes one or more of an instruction translation look-aside buffer (ITLB) page walk, an ITLB fill, an instruction cache request, and instruction cache fill, an instruction end byte detection, and a branch prediction update.