Search International and National Patent Collections
Some content of this application is unavailable at the moment.
If this situation persists, please contact us atFeedback&Contact
1. (WO2017112374) METHOD AND APPARATUS FOR USER-LEVEL THREAD SYNCHRONIZATION WITH A MONITOR AND MWAIT ARCHITECTURE
Note: Text based on automatic Optical Character Recognition processes. Please use the PDF version for legal matters

THE CLAIMS

What is claimed is:

1. A proces sor comprising :

one or more model specific registers (MSRs) to be configured in a first execution state to specify support of a user-level thread synchronization architecture for the processor;

a plurality of execution cores having corresponding monitored address state storage to store a last monitored address for each of a plurality of execution threads that issues a MONITOR request; and

a cache memory to record MONITOR requests and associated states for addresses of memory storage locations, and

responsive to receipt of a MONITOR request for an address from an execution thread of the user- level, to store the address and record an associated state of the monitored address state storage for a fist execution core of the plurality of execution cores associated with the execution thread when said one or more MSRs is configured in the first execution state; and

further responsive to receipt of an MWAIT request from the execution thread for the address, to record an associated wait-to-trigger state of the monitored address state for said first execution core associated with the execution thread; and

wherein the first execution core is to transition the execution thread to an optimized sleep state responsive to the receipt of said MWAIT request from the execution thread when said one or more MSRs is configured in the first execution state.

2. The processor of claim 1, wherein the optimized sleep state is the lightest- weight sleep state of a plurality of C-states available to the first execution core.

3. The processor of claim 2, wherein the lightest-weight sleep state of the plurality of C-states available to the first execution core is a CI state.

4. The processor of claim 1, wherein the optimized sleep state is a sleep state selected by bits 7:4 of an EAX register from a plurality of C-states available to the first execution core.

5. The processor of claim 1, wherein the one or more MSRs are to be configured into the first execution state only by an operating system.

6. The processor of claim 1, wherein the first execution core is to trigger an invalid-opcode exception responsive to the receipt of said MWAIT request from the execution thread when said one or more MSRs is not configured in the first execution state.

7. A method comprising:

configuring one or more model specific registers (MSRs) to be in a first execution state to specify support of a user-level thread synchronization architecture for the processor;

stonng, in a monitored address state storage corresponding to a plurality of execution cores, a last monitored address for an execution thread of a plurality of execution threads when the execution thread is to issue a MONITOR request; and

in a cache memory to record MONITOR requests and associated states for addresses of memory storage locations, and

responsive to receipt of a MONITOR request for an address from an execution thread of the user-level, storing the address and recording an associated state of the monitored address state storage for a fist execution core of the plurality of execution cores associated with the execution thread when said one or more MSRs is configured in the first execution state; and

further responsive to receipt of an MWAIT request from the execution thread for the address, recording an associated wait-to-trigger state of the monitored address state for said first execution core associated with the execution thread; and

in the first execution core, transitioning the execution thread to an optimized sleep state responsive to the receipt of said MWAIT request from the execution thread when said one or more MSRs is configured in the first execution state.

The method of claim 4, wherein the optimized sleep state is the lightest-weight sleep state of a plurality of C-states available to the first execution core.

The method of claim 5, wherein the lightest- weight sleep state of the plurality of C-states available to the first execution core is a CI state.

A processor comprising:

one or more model specific registers (MSRs) to be configured in a first execution state to specify support of a user-level thread synchronization architecture for the processor;

a plurality of execution cores, each including a local storage to store a last monitored address for each of a plurality of execution threads when the execution thread is to issue a monitor request; and

one or more cache to store data corresponding to addresses of memory storage locations, a first cache of the one or more cache including state storage, wherein responsive to receipt of a request from an execution thread to monitor an address, the first cache is to store the address and to record request for the execution thread in the state storage and an associated state for a monitor state machine of the cache portion for the execution thread; then

responsive to receipt of a wait request from the execution thread for the address, the first cache is to record the wait request for the execution thread and an associated wait-to-trigger state for the monitor state machine;

wherein the execution thread is to transition to an optimized sleep state responsive to the receipt of said wait request from the execution thread when said one or more MSRs are configured in the first execution state.

11. The processor of claim 10, wherein the optimized sleep state is the lightest- weight sleep state of a plurality of C-states available to the first execution core.

12. The processor of claim 11, wherein the lightest- weight sleep state of the plurality of

C-states available to the first execution core is a CI state.

13. A processing system comprising:

a memory; and

a processor comprising:

one or more model specific registers (MSRs) to be configured in a first execution state to specify support of a user-level thread synchronization architecture for the processor;

a plurality of execution cores having corresponding monitored address state storage to store a last monitored address for each of a plurality of execution threads that issues a MONITOR request; and

a cache memory to record MONITOR requests and associated states for addresses of memory storage locations, and

responsive to receipt of a MONITOR request for an address from an execution thread of the user- level, to store the address and record an associated state of the monitored address state storage for a fist execution core of the plurality of execution cores associated with the execution thread when said one or more MSRs is configured in the first execution state; and

further responsive to receipt of an MWAIT request from the execution thread for the address, to record an associated wait-to-trigger state of the monitored address state for said first execution core associated with the execution thread; and

wherein the first execution core is to transition the execution thread to an optimized sleep state responsive to the receipt of said MWAIT request from the execution thread when said one or more MSRs is configured in the first execution state.

14. The processing system of claim 13, wherein the optimized sleep state is the lightest-weight sleep state of a plurality of C-states available to the first execution core.

15. The processing system of claim 14, wherein the lightest-weight sleep state of the plurality of C-states available to the first execution core is a CI state.

16. The processing system of claim 13, wherein the optimized sleep state is a sleep state selected by bits 7:4 of an EAX register from a plurality of C-states available to the first execution core.

17. The processing system of claim 13, wherein the one or more MSRs are to be configured into the first execution state only by an operating system.

18. The processing system of claim 13, wherein the first execution core is to trigger an invalid- opcode exception responsive to the receipt of said MWAIT request from the execution thread when said one or more MSRs is not configured in the first execution state.