Automatic hardware ZLW insertion for IPU image streams
A system for automatic hardware ZLW insertion for IPU image streams is described herein. The system includes a memory and a processor. The memory is to store imaging data. The processor is coupled to the memory. The processor is to receive an image stream request and determine a data transfer type. The processor is also to insert a zero length write (ZLW) instruction ahead of the image stream request in response to the image stream request beginning on a different page when compared to the current page in a page history. Additionally, the processor is to insert a ZLW instruction ahead of the image stream request in response to the image stream request crossing a page boundary.
1. A system, comprising:
a memory configured to store imaging data; and
a processor coupled to the memory, the processor to:
receive an image stream request;
determine a data transfer type;
in response to the image stream request beginning on a different page when compared to the current page in a page history, insert a zero length write (ZLW) instruction ahead of the image stream request based on the data transfer type; and
in response to the image stream request crossing a page boundary, insert a ZLW instruction ahead of the image stream request based on the data transfer type.
2. The system of claim 1 , wherein the ZLW instruction is to cause a page address translation to be pre-fetched.
3. The system of claim 1 , wherein the ZLW instruction is to cause a page address translation to be pre-fetched, wherein the ZLW instruction is inserted by hardware.
4. The system of claim 1 , wherein the ZLW instruction is issued N pages ahead of the image stream request.
5. The system of claim 1 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on an image stream throughput.
6. The system of claim 1 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on a look-aside buffer configuration.
7. The system of claim 1 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on an IOMMU replacement policy.
8. The system of claim 1 , wherein the data transfer type is a 1D linear burst.
9. The system of claim 1 , wherein the data transfer type is a 2D block burst.
10. The system of claim 1 , wherein the ZLW instruction is a write request that does not carry data.
11. A method, comprising:
receiving an image stream request;
determining a data transfer type;
inserting a zero length write (ZLW) instruction ahead of the image stream request in response to the image stream request beginning on a different page when compared to the current page in a page history based on the data transfer type; and
inserting a ZLW instruction ahead of the image stream request in response to the image stream request crossing a page boundary based on the data transfer type.
12. The method of claim 11 , wherein the ZLW instruction is to cause a page address translation to be pre-fetched.
13. The method of claim 11 , wherein the ZLW instruction is to cause a page address translation to be pre-fetched, wherein the ZLW instruction is inserted by hardware.
14. The method of claim 11 , wherein the ZLW instruction is issued N pages ahead of the image stream request.
15. The method of claim 11 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on an image stream throughput.
16. An apparatus, comprising:
a controller to receive an image stream request;
a ZLW insertion unit to determine a data transfer type, and in response to the image stream request beginning on a different page when compared to the current page in a page history, insert a zero length write (ZLW) instruction ahead of the image stream request based on the data transfer type, and in response to the image stream request crossing a page boundary, insert a ZLW instruction ahead of the image stream request based on the data transfer type.
17. The apparatus of claim 16 , wherein the ZLW instruction is to cause a page address translation to be pre-fetched.
18. The apparatus of claim 16 , wherein the ZLW instruction is to cause a page address translation to be pre-fetched, wherein the ZLW instruction is inserted by hardware.
19. The apparatus of claim 16 , wherein the ZLW instruction is issued N pages ahead of the image stream request.
20. The apparatus of claim 16 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on an image stream throughput.
21. At least one non-transitory machine readable medium comprising a plurality of instructions that, in response to being executed on a computing device, cause the computing device to: receive an image stream request; determine a data transfer type; insert a zero length write (ZLW) instruction ahead of the image stream request in response to the image stream request beginning on a different page when compared to the current page in a page history based on the data transfer type; and insert a ZLW instruction ahead of the image stream request in response to the image stream request crossing a page boundary based on the data transfer type.
22. The machine readable medium of claim 21 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on a look-aside buffer configuration.
23. The machine readable medium of claim 21 , wherein the ZLW instruction is issued N pages ahead of the image stream request, wherein the value of N is based on an IOMMU replacement policy.
24. The machine readable medium of claim 21 , wherein the data transfer type is a 1D linear burst.
25. The machine readable medium of claim 21 , wherein the data transfer type is a 2D block burst.