IP Library Granted Patent US 10,656,696
Granted Patent B1
US 10,656,696 · App. 15/907,719 · Granted May 19, 2020

Reducing chiplet wakeup latency

Inventors: Benjamin Tsien (Fremont, CA); Michael J. Tresidder (Spicewood, TX); Ivan Yanfeng Wang (Richmond Hill, CA); Kevin M. Lepak (Austin, TX); Ann Ling (Folsom, CA); Richard M. Born (Fort Collins, CA); John P. Petry (San Diego, CA); Bryan P. Broussard (Austin, TX); Eric Christopher Morton (Austin, TX)
Assignees: Advanced Micro Devices, Inc.; ATI Technologies ULC
G06F1/3206G06F1/3253G06F1/3287
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 10,656,696
App. No.
15/907,719
Granted
May 19, 2020
Kind
B1
Abstract

Systems, apparatuses, and methods for reducing chiplet interrupt latency are disclosed. A system includes one or more processing nodes, one or more memory devices, a communication fabric coupled to the processing unit(s) and memory device(s) via link interfaces, and a power management unit. The power management unit manages the power states of the various components and the link interfaces of the system. If the power management unit detects a request to wake up a given component, and the link interface to the given component is powered down, then the power management unit sends an out-of-band signal to wake up the given component in parallel with powering up the link interface. Also, when multiple link interfaces need to be powered up, the power management unit powers up the multiple link interfaces in an order which complies with voltage regulator load-step requirements while minimizing the latency of pending operations.

Claims (42)

1. A system comprising:

one or more processing nodes, wherein each processing node of the one or more processing nodes comprises one or more processor cores;

a communication fabric coupled to the one or more processing nodes via one or more link interfaces; and

a power management unit;

wherein the power management unit is configured to:

detect a request to wake up a processor core of a processing node;

determine that a link interface that connects the communication fabric to the processing node is in a non-operational state; and

send an out-of-band signal over the link interface to wake up the processor core prior to the link interface returning to an operational state.

2. The system as recited in claim 1 , wherein the power management unit is further configured to send a signal to power up the link interface in parallel with sending the out-of-band signal over the link interface to wake up the processor core.

3. The system as recited in claim 1 , wherein:

the processing node is on a first semiconductor die;

the communication fabric is on a second semiconductor die; and

the second semiconductor die is different from the first semiconductor die.

4. The system as recited in claim 1 , wherein the out-of-band signal to wake up the processor core is sent on the link interface prior to training being performed on the link interface.

5. The system as recited in claim 1 , wherein the power management unit is further configured to:

detect a condition for powering up multiple link interfaces; and

determine an order for powering up the multiple link interfaces based on complying with voltage regulator load-step requirements and minimizing a latency of pending operations.

6. The system as recited in claim 5 , wherein:

the power management unit is further configured to maintain a table with entries for a plurality of components; and

each entry specifies an order of powering up link interfaces to minimize the latency of pending operations for a corresponding component.

7. A method comprising:

detecting, by a power management unit comprising circuitry, a request to wake up a processor core of a processing node;

determining, by the power management unit, that a link interface is in a non-operational state, wherein the link interface connects a communication fabric to the processing node; and

sending, by the power management unit, an out-of-band signal over the link interface to wake up the processor core prior to the link interface returning to an operational state.

8. The method as recited in claim 7 , further comprising sending a signal to power up the link interface in parallel with sending the out-of-band signal over the link interface to wake up the processor core.

9. The method as recited in claim 7 , wherein the processing node is on a first semiconductor die, wherein the communication fabric is on a second semiconductor die, and wherein the second semiconductor die is different from the first semiconductor die.

10. The method as recited in claim 7 , wherein the out-of-band signal to wake up the processor core is sent on the link interface prior to training being performed on the link interface.

11. The method as recited in claim 7 , further comprising:

detecting a condition for powering up multiple link interfaces; and

determining an order for powering up the multiple link interfaces based on complying with voltage regulator load-step requirements and minimizing a latency of pending operations.

12. The method as recited in claim 11 , further comprising maintaining a table with entries for a plurality of components, wherein each entry specifies an order of powering up link interfaces to minimize the latency of pending operations for a corresponding component.

13. An apparatus comprising:

a processing node comprising a plurality of processor cores; and

a control unit;

wherein the control unit is configured to:

detect a request to wake up a processor core of the processing node;

determine that a link interface is in a non-operational state, wherein the link interface connects the control unit to the processing node; and

send an out-of-band signal over the link interface to wake up the processor core prior to the link interface returning to an operational state.

14. The apparatus as recited in claim 13 , wherein the control unit is further configured to send a signal to power up the link interface in parallel with sending the out-of-band signal over the link interface to wake up the processor core.

15. The apparatus as recited in claim 13 , wherein the processing node is on a first semiconductor die, wherein the control unit is on a second semiconductor die, and wherein the second semiconductor die is different from the first semiconductor die.

16. The apparatus as recited in claim 13 , wherein the out-of-band signal to wake up the processor core is sent on the link interface prior to training being performed on the link interface.

17. The apparatus as recited in claim 13 , wherein the control unit is further configured to determine an order for powering up multiple link interfaces based on complying with voltage regulator load-step requirements and minimizing a latency of pending operations.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Mar 21, 2018
From: TSIEN, BENJAMIN; TRESIDDER, MICHAEL J.; LEPAK, KEVIN M.; LING, ANN; BORN, RICHARD M.; PETRY, JOHN P.; BROUSSARD, BRYAN P.; MORTON, ERIC CHRISTOPHER
To: ADVANCED MICRO DEVICES, INC.
Reel/Frame 045304/0763 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Feb 28, 2018
From: WANG, IVAN YANFENG
To: ATI TECHNOLOGIES ULC
Reel/Frame 045063/0460 →