US2021255947A1PendingUtilityA1

Guaranteed forward progress mechanism

Assignee: INTEL CORPPriority: Apr 17, 2017Filed: Dec 3, 2020Published: Aug 19, 2021
Est. expiryApr 17, 2037(~10.7 yrs left)· nominal 20-yr term from priority
G06F 12/0875G06F 12/00G06F 12/0811G06F 9/544G06T 1/60G06F 2212/455G06F 9/54
65
PatentIndex Score
0
Cited by
0
References
0
Claims

Abstract

An apparatus to facilitate guaranteed forward progress for graphics data is disclosed. The apparatus includes a plurality of ports to receive and transmit streams of graphics data, one or more buffers associated with each of the plurality of ports to store the graphics data and switching logic to virtually partition each of the one or more buffers to allocate a dedicated buffer to receive each of a plurality of independent streams of graphics data.

Claims

exact text as granted — not AI-modified
1 - 20 . (canceled) 
     
     
         21 . A system comprising:
 one or more processing nodes, each processing node including:
 a memory; 
 one or more processing cores to generate independent streams of graphics data to access memory; and 
 a plurality of network switches, each network switch including a plurality of ports to receive and transmit graphics data and one or more buffers to store graphics data; 
   wherein each network switch of the plurality of network switches is to manage the one or more buffers of the network switch to receive a plurality of independent streams of graphics data, including:   assigning credits for assessing resources of the one or more buffers; and   allowing access by each of the independent streams according to credits available for the independent stream.   
     
     
         22 . The system of  claim 21 , wherein assigning credits includes assigning dedicated credits for each of the plurality of the independent streams, the dedicated credits assigned to each independent stream being decremented as resources of the one or more buffers are used. 
     
     
         23 . The system of  claim 22 , wherein each network switch provides round robin arbitration for the plurality of independent streams, the round-robin arbitration being managed according to the assigned credits. 
     
     
         24 . The system of  claim 22 , wherein managing the one or more buffers of each network switch further includes:
 establishing a general buffer pool; and   assigning general purpose credits for use of the general buffer pool.   
     
     
         25 . The system of  claim 24 , wherein each of the plurality of independent streams is allowed to the general buffer pool utilizing the general purpose credits. 
     
     
         26 . The system of  claim 25 , wherein the general purpose credits are available to an independent stream upon the independent stream exhausting the dedicated credits assigned to the independent stream. 
     
     
         27 . The system of  claim 22 , wherein exhausting the dedicated credits assigned to an independent stream includes the independent stream being held up by a blocking event. 
     
     
         28 . The system of  claim 21 , wherein each of the plurality of network switches provides guaranteed forward progress for independent streams of graphics data utilizing the management of the one or more buffers. 
     
     
         29 . A method, comprising:
 generating independent streams of graphics data to access memory at one or more processing nodes, each processing node including a plurality of network switches, each network switch including a plurality of ports to receive and transmit graphics data and one or more buffers to store graphics data; and   managing the one or more buffers of each network switch to receive a plurality of independent streams of graphics data, including:
 assigning credits for assessing resources of the one or more buffers; and 
 allowing access by each of the independent streams according to credits available for the independent stream. 
   
     
     
         30 . The method of  claim 29 , wherein assigning credits includes assigning dedicated credits for each of the plurality of the independent streams, the dedicated credits assigned to each independent stream being decremented as resources of the one or more buffers are used. 
     
     
         31 . The method of  claim 30 , wherein each network switch provides round robin arbitration for the plurality of independent streams, the round-robin arbitration being managed according to the assigned credits. 
     
     
         32 . The method of  claim 30 , wherein managing the one or more buffers of each network switch further includes:
 establishing a general buffer pool; and   assigning general purpose credits for use of the general buffer pool.   
     
     
         33 . The method of  claim 32 , wherein each of the plurality of independent streams is allowed to the general buffer pool utilizing the general purpose credits. 
     
     
         34 . The method of  claim 33 , wherein the general purpose credits are available to an independent stream upon the independent stream exhausting the dedicated credits assigned to the independent stream. 
     
     
         35 . The method of  claim 30 , wherein exhausting the dedicated credits assigned to an independent stream includes the independent stream being held up by a blocking event. 
     
     
         36 . A fabric interconnect comprising:
 a plurality of network switches, each network switch including:
 a plurality of ports to receive and transmit streams of graphics data; and 
 one or more buffers associated with each of the plurality of ports to store the graphics data; 
   wherein each switch of the plurality of network switches is to manage the one or more buffers of the network switch to receive a plurality of independent streams of graphics data, including:   assigning credits for assessing resources of the one or more buffer, including assigning dedicated credits for each of the plurality of the independent streams; and   allowing access by each of the independent streams according to credits available for the independent stream, wherein the dedicated credits assigned to each independent stream are decremented as resources of the one or more buffers are used.   
     
     
         37 . The fabric interconnect of  claim 36 , wherein assigning credits includes assigning dedicated credits for each of the plurality of the independent streams, the dedicated credits assigned to each independent stream being decremented as resources of the one or more buffers are used. 
     
     
         38 . The fabric interconnect of  claim 37 , wherein each network switch provides round robin arbitration for the plurality of independent streams, the round-robin arbitration being managed according to the assigned credits. 
     
     
         39 . The fabric interconnect of  claim 37 , wherein managing the one or more buffers of each network switch further includes:
 establishing a general buffer pool; and   assigning general purpose credits for use of the general buffer pool.   
     
     
         40 . The fabric interconnect of  claim 39 , wherein each of the plurality of independent streams is allowed to the general buffer pool utilizing the general purpose credits, and wherein the general purpose credits are available to an independent stream upon the independent stream exhausting the dedicated credits assigned to the independent stream.

Join the waitlist — get patent alerts

Track US2021255947A1 — get alerts on status changes and closely related new filings.

We store only your email — no account needed. See our privacy policy.