Translation bandwidth optimized prefetching strategy through multiple translation lookaside buffers

Patent metadata

Field	Value
Publication number	US-11645208-B2
Application number	US-202117215296-A
Country	US
Kind code	B2
Filing date	Mar 29, 2021
Priority date	Mar 29, 2021
Publication date	May 9, 2023
Grant date	May 9, 2023

Abstract

A computer system includes a processor and a prefetch engine. The processor is configured to generate a demand access stream. The prefetch engine is configured to generate a first prefetch request and a second prefetch request based on the demand access stream, to output the first prefetch request to a first translation lookaside buffer (TLB), and to output the second prefetch request to a second TLB that is different from the first TLB. The processor performs a first TLB lookup in the first TLB based on one of the demand access stream or the first prefetch request, and performs a second TLB lookup in the second TLB based on the second prefetch request.

First claim

What is claimed is:

1. A computer-implemented method for performing prefetching in a computer processing system, the method comprising:
generating, by a processor, a demand access stream;
generating, by a prefetch engine, a first prefetch request and a second prefetch request based on the demand access stream;
outputting, from the prefetch engine, the first prefetch request to a first translation lookaside buffer (TLB) and the second prefetch request to a second TLB that is different from the first TLB; and
performing a first TLB lookup in the first TLB based on one of the demand access stream or the first prefetch request, and performing a second TLB lookup in the second TLB based on the second prefetch request, wherein the first TLB lookup and the second TLB lookup are performed simultaneously,
wherein generating the first and second prefetch requests further comprises:
monitoring, by the prefetch engine, the demand access stream;
determining a demand load associated with the first TLB in response to monitoring the demand access stream;
generating the first prefetch request in response to the demand load being less than or equal to a demand threshold; and
generating the second prefetch request in response to the demand load exceeding the demand threshold.

2. The computer-implemented method of claim 1 , wherein performing the first TLB lookup further comprises:
outputting a first virtual address (VA) included in the demand access stream from the processor;
outputting a second VA included in the first prefetch request from the prefetch engine;
selectively delivering one of the first VA or the second VA to the first TLB; and
performing the first TLB lookup based on the first VA or the second VA delivered to the first TLB.

3. The computer-implemented method of claim 2 , wherein performing the second TLB lookup further comprises:
outputting a third VA included in the second prefetch request to the second TLB; and
performing the second TLB lookup based on the third VA.

4. The computer-implemented method of claim 3 , outputting the first and second prefetch requests further comprises:
outputting the second VA to a first prefetch buffer;
outputting the third VA to a second prefetch buffer that is separate from the first prefetch buffer;
outputting the second VA from the first prefetch buffer to the first TLB; and
outputting the third VA from the second prefetch buffer to the second TLB independently from outputting the second VA from the first prefetch buffer.

5. The computer-implemented method of claim 2 , further comprising:
in response to a TLB hit associated with the first TLB lookup, translating the first VA or the second VA into a real address (RA);
in response to a TLB miss associated with the first TLB lookup, outputting the first VA or the second VA to the second TLB;
performing a third TLB lookup in the second TLB based on the first VA or the second VA;
in response to a TLB hit associated with the third TLB lookup, determining a RA corresponding to the first VA or the second VA, and returning the RA corresponding to the first VA or the second VA to the first TLB; and
translating, using the first TLB, the first VA or the second VA using the RA obtained from the second TLB.

6. A computer system comprising:
a processor configured to generate a demand access stream;
a prefetch engine configured to generate a first prefetch request and a second prefetch request based on the demand access stream, to output the first prefetch request to a first translation lookaside buffer (TLB), and to output the second prefetch request to a second TLB that is different from the first TLB,
wherein the processor performs a first TLB lookup in the first TLB based on one of the demand access stream or the first prefetch request, and performs a second TLB lookup in the second TLB based on the second prefetch request, wherein the first TLB lookup and the second TLB lookup are performed simultaneously,
wherein the prefetch engine generates the first and second prefetch requests in response to monitoring the demand access stream, determining a demand load associated with the first TLB in response to monitoring the demand access stream, generating the first prefetch request in response to the demand load being less than or equal to a demand threshold, generating the second prefetch request in response to the demand load exceeding the demand threshold.

7. The computer system of claim 6 , wherein the processor performs the first TLB lookup in response to outputting a first virtual address (VA) included in the demand access stream from the processor, outputting a second VA included in the first prefetch request from the prefetch engine, selectively delivering one of the first VA or the second VA to the first TLB, and performing the first TLB lookup based on the first VA or the second VA delivered to the first TLB.

8. The computer system of claim 7 , wherein the processor performs the second TLB lookup in response to outputting a third VA included in the second prefetch request to the second TLB, and performing the second TLB lookup based on the third VA.

9. The computer system of claim 8 , outputting the processor performs the first and second prefetch requests in response to outputting the second VA to a first prefetch buffer, outputting the third VA to a second prefetch buffer that is separate from the first prefetch buffer, outputting the second VA from the first prefetch buffer to the first TLB, and outputting the third VA from the second prefetch buffer to the second TLB independently from outputting the second VA from the first prefetch buffer.

10. The computer system of claim 7 , wherein in response to a TLB hit associated with the first TLB lookup, the processor translates the first VA or the second VA into a real address (RA), wherein in response to a TLB miss associated with the first TLB lookup, the processor outputs the first VA or the second VA to the second TLB.

11. The computer system of claim 10 , wherein the processor performs a third TLB lookup in the second TLB based on the first VA or the second VA, and
in response to a TLB hit associated with the third TLB lookup, the processor determines a RA corresponding to the first VA or the second VA, returns the RA corresponding to the first VA or the second VA to the first TLB, and uses the first TLB to translate the first VA or the second VA using the RA obtained from the L2 TLB.

12. A computer program product to control a computer processing system to perform prefetching, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by an electronic computer processor to control the computer processing system to perform operations comprising:
generate a demand access stream;
generate, by a prefetch engine, a first prefetch request and a second prefetch request based on the demand access stream;
output, from the prefetch engine, the first prefetch request to a first translation lookaside buffer (TLB) and the second prefetch request to a second TLB that is different from the first TLB;
perform a first TLB lookup in the first TLB based on one of the demand access stream or the first prefetch request; and
perform a second TLB lookup in the second TLB based on the second prefetch request,
wherein the first TLB lookup and the second TLB lookup are performed simultaneously,
wherein generating the first and second prefetch requests further comprises:
monitoring, by the prefetch engine, the demand access stream;
determining a demand load associated with the first TLB in response to monitoring the demand acce

Assignees

• IBM

Inventors

• Campbell David
• Lloyd Bryan
• Rohrbaugh Iii George W
• Britto Vivek
• Karve Mohit

Classifications

• G06F2212/654

  Look-ahead translation · CPC title

• G06F12/0862 Primary

  with prefetch · CPC title

• G06F2212/681

  Multi-level TLB, e.g. microTLB and main TLB · CPC title

• G06F2212/684

  TLB miss handling · CPC title

• G06F2212/602

  Details relating to cache prefetching · CPC title