APPARATUS AND METHOD FOR PROVIDING WORKLOAD DISTRIBUTION OF THREADS AMONG MULTIPLE COMPUTE UNITS

2022

Online Patent

Zugriff:

In some examples, thermal aware optimization logic determines a characteristic (e.g., a workload or type) of a wavefront (e.g., multiple threads). For example, the characteristic indicates whether the wavefront is compute intensive, memory intensive, mixed, and/or another type of wavefront. The thermal aware optimization logic determines temperature information for one or more compute units (CUs) in one or more processing cores. The temperature information includes predictive thermal information indicating expected temperatures corresponding to the one or more CUs and historical thermal information indicating current or past thermal temperatures of at least a portion of a graphics processing unit (GPU). The logic selects the one or more compute units to process the plurality of threads based on the determined characteristic and the temperature information. The logic provides instructions to the selected subset of the plurality of CUs to execute the wavefront.

Titel:	APPARATUS AND METHOD FOR PROVIDING WORKLOAD DISTRIBUTION OF THREADS AMONG MULTIPLE COMPUTE UNITS
Link:	Find this record on USPTO (Volltext)
Veröffentlichung:	2022
Medientyp:	Patent
Sonstiges:	Nachgewiesen in: USPTO Patent Applications Sprachen: English Document Number: 20220107849 Publication Date: April 7, 2022 Appl. No: 17/519290 Application Filed: November 04, 2021 Claim: 1. A method for providing workload distribution of a plurality of threads among a plurality of compute units (CUs), comprising: determining a characteristic of the plurality of threads; determining temperature information corresponding to at least a subset of the plurality of CUs, wherein the temperature information comprises: predictive thermal information indicating expected temperatures corresponding to at least the subset of the plurality of CUs; and historical thermal information indicating current or past thermal temperatures of at least a portion of a first processor; wherein the at least the portion of the first processor comprises at least one non-CU circuitry, wherein determining the temperature information comprises generating the predictive thermal information based on a location of the at least one non-CU circuitry, selecting the subset of the plurality of CUs to process the plurality of threads based on the determined characteristic and the temperature information; and providing instructions to the selected subset of the plurality of CUs to execute the plurality of threads. Claim: 2. The method of claim 1, wherein the determining the predictive thermal information comprises: determining a plurality of characteristics for a plurality of second threads currently being executed by the plurality of CUs; and generating the predictive thermal information based on the plurality of characteristics for the plurality of second threads currently being executed by the plurality of CUs. Claim: 3. The method of claim 1, further comprising: obtaining temperature readings from a plurality of temperature sensors; and generating the historical thermal information based on the temperature readings from the plurality of temperature sensors. Claim: 4. The method of claim 1, wherein the subset of the plurality of CUs is on the first processor, and wherein the method further comprises: determining a plurality of characteristics for a second plurality of threads currently being executed by a second plurality of CUs on a second processor, wherein the first processor and the second processor are stacked on top of each other; and generating the predictive thermal information based on the plurality of characteristics for the second plurality of threads currently being executed by the second plurality of CUs. Claim: 5. The method of claim 4, wherein the second processor comprises at least one non-CU circuitry, and wherein the determining the predictive thermal information comprises generating the predictive thermal information based on a location of the at least one non-CU circuitry. Claim: 6. The method of claim 5, wherein the second processor comprises a second plurality of temperature sensors, and wherein the method further comprises: obtaining temperature readings from the second plurality of temperature sensors; and generating the historical thermal information based on the temperature readings from the second plurality of temperature sensors. Claim: 7. The method of claim 1, wherein the determining the characteristic of the plurality of threads comprises: determining whether the plurality of threads is a memory intensive plurality of threads, a computational intensive plurality of threads, or a memory and computational intensive plurality of threads. Claim: 8. The method of claim 7, wherein the selecting the subset of the plurality of CUs is based on whether the plurality of threads is the memory intensive plurality of threads, the computational intensive plurality of threads, or the memory and computational intensive plurality of threads. Claim: 9. An apparatus for providing workload distribution of a plurality of threads among a plurality of CUs, comprising: a first processor, wherein the first processor comprises: a plurality of compute units (CU) configured to execute the plurality of threads; and thermal aware optimization logic, wherein the thermal aware optimization logic is configured to: determine a characteristic of the plurality of threads; determine temperature information corresponding to at least a subset of the plurality of CUs, wherein the temperature information comprises: predictive thermal information indicating expected temperatures corresponding to at least the subset of the plurality of CUs; and historical thermal information indicating current or past thermal temperatures of at least a portion of the first processor, wherein the portion of the first processor comprises at least one non-CU circuitry, and wherein the thermal aware optimization logic is configured to determine the predictive thermal information by generating the predictive thermal information based on a location of the at least one non-CU circuitry; select the subset of the plurality of CUs to process the plurality of threads based on the determined characteristic and the temperature information; and provide instructions to the selected subset of the plurality of CUs to execute the plurality of threads. Claim: 10. The apparatus of claim 9, wherein the thermal aware optimization logic is configured to determine the predictive thermal information by: determining a plurality of characteristics for a plurality of second threads currently being executed by the plurality of CUs; and generating the predictive thermal information based on the plurality of characteristics for the plurality of second threads currently being executed by the plurality of CUs. Claim: 11. The apparatus of claim 9, wherein the first processor comprises a GPU and wherein a second processor further comprises a plurality of temperature sensors configured to provide temperature readings to the thermal aware optimization logic, and wherein the thermal aware optimization logic is configured to: obtain temperature readings from the plurality of temperature sensors; and generate the historical thermal information based on the temperature readings from the plurality of temperature sensors. Claim: 12. The apparatus of claim 9, further comprising: a third processor, wherein the second processor third processor are stacked on top of each other, wherein the third processor comprises a second plurality of CUs, and wherein the thermal aware optimization logic is configured to determine the predictive thermal information by: determining a plurality of characteristics for a second plurality of threads currently being executed by the second plurality of CUs; and generating the predictive thermal information based on the plurality of characteristics for the second plurality of threads currently being executed by the second plurality of CUs. Claim: 13. The apparatus of claim 12, wherein the third processor comprises at least one non-CU circuitry, and wherein the thermal aware optimization logic is configured to determine the predictive thermal information by generating the predictive thermal information based on a location of the at least one non-CU circuitry. Claim: 14. The apparatus of claim 13, wherein the third processor comprises a second plurality of temperature sensors configured to provide temperature readings to the thermal aware optimization logic, and wherein the thermal aware optimization logic is configured to: obtain temperature readings from the second plurality of temperature sensors; and generate the historical thermal information based on the temperature readings from the second plurality of temperature sensors. Claim: 15. The apparatus of claim 9, wherein the thermal aware optimization logic is configured to determine the characteristic of the plurality of threads by: determining whether the plurality of threads is a memory intensive plurality of threads, a computational intensive plurality of threads, or a memory and computational intensive plurality of threads. Claim: 16. The apparatus of claim 15, wherein the thermal aware optimization logic is configured to select the subset of the plurality of CUs based on whether the plurality of threads is the memory intensive plurality of threads, the computational intensive plurality of threads, or the memory and computational intensive plurality of threads. Claim: 17. An apparatus for providing workload distribution of a wavefront among a plurality of CUs, comprising: a first processor, wherein the first processor comprises: a plurality of compute units (CU) configured to execute a plurality of threads; and thermal aware optimization logic, wherein the thermal aware optimization logic is configured to: determine a characteristic of the plurality of threads; determine a plurality of characteristics for a plurality of second threads currently being executed by the plurality of CUs; and generate predictive thermal information based on the plurality of characteristics for the plurality of second threads currently being executed by the plurality of CUs; obtain temperature readings from a plurality of temperature sensors wherein at least one temperature sensor is configured to sense a temperature of non-CU circuitry, and generating the predictive thermal information is based on a temperature of the at least one non-CU circuitry; generate historical thermal information based on the temperature readings from the plurality of temperature sensors; select a subset of the plurality of CUs to process the plurality of threads based on the determined characteristic, the predictive thermal information, and the historical thermal information; and provide instructions to the selected subset of the plurality of CUs to execute the plurality of threads. Claim: 18. The apparatus of claim 17, further comprising: a second processor, wherein the first processor and the second processor are stacked on top of each other, wherein the second processor comprises a second plurality of CUs, a plurality of second temperature sensors, and wherein the thermal aware optimization logic is configured to generate the predictive thermal information by: determining a second plurality of characteristics for a third plurality of threads currently being executed by the second plurality of CUs; and generating the predictive thermal information based on the second plurality of characteristics for the third plurality of threads currently being executed by the second plurality of CUs, and wherein the thermal aware optimization logic is configured to generate the historical thermal information by: obtaining temperature readings from the plurality of second temperature sensors; and generating the historical thermal information based on the temperature readings from the plurality of second temperature sensors. Current International Class: 06; 06; 06; 04

Klicken Sie ein Format an und speichern Sie dann die Daten oder geben Sie eine Empfänger-Adresse ein und lassen Sie sich per Email zusenden.

BibTeX Citavi, JabRef, u.a.
(Literaturverwaltung)

PDF kein Volltext!
(Merkzettel, Notizen)

RIS Endnote, Citavi u.a.
(Literaturverwaltung)

MODS
(XML zur Weiterverarbeitung)

oder

Wählen Sie das für Sie passende Zitationsformat und kopieren Sie es dann in die Zwischenablage, lassen es sich per Mail zusenden oder speichern es als PDF-Datei.

Gewünschter Zitations-Stil:

oder

Bitte prüfen Sie, ob die Zitation formal korrekt ist, bevor Sie sie in einer Arbeit verwenden. Benutzen Sie gegebenenfalls den "Exportieren"-Dialog, wenn Sie ein Literaturverwaltungsprogramm verwenden und die Zitat-Angaben selbst formatieren wollen.