Apparatus and method for accelerating operations in a processor which uses shared virtual memory

• US 10,467,012 B2
• Filed: 12/29/2016
• Issued: 11/05/2019
• Est. Priority Date: 03/30/2012
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Weissmann, Eliezer, Vaithianathan, Karthikeyan Karthik, Zach, Yoav, Ginzburg, Boris, Ronen, Ronny
• Primary Examiner(s): Warren, Tracy A
• Application Number:
  US15/394,458
• Publication Number:
  US 20170153984A1
• Time in Patent Office: 1,041 Days
• Field of Search: 711207, 711209, 711221, 710104, 710308, 710311, 712 11, 712 14, 712225, 712248
• CPC Class Codes: G06F 12/0292 using tables or multilevel address translation means G06F12/023 takes precedence; address translation in virtual memory systems G06F12/10
• CPC Class Codes: G06F 12/0811 with multilevel cache hierarchies
• CPC Class Codes: G06F 12/084 with a shared cache
• CPC Class Codes: G06F 12/0875 with dedicated cache, e.g. instruction or stack
• CPC Class Codes: G06F 12/1009 using page tables, e.g. page table structures
• CPC Class Codes: G06F 12/1027 using associative or pseudo-associative address translation means, e.g. translation look-aside buffer [TLB]
• CPC Class Codes: G06F 12/1045 associated with a data cache
• CPC Class Codes: G06F 12/1072 Decentralised address translation, e.g. in distributed shared memory systems
• CPC Class Codes: G06F 12/1081 for peripheral access to main memory, e.g. direct memory access [DMA]
• CPC Class Codes: G06F 12/1441 for a range
• CPC Class Codes: G06F 12/145 the protection being virtual, e.g. for virtual blocks or segments before a translation mechanism
• CPC Class Codes: G06F 2212/1024 Latency reduction
• CPC Class Codes: G06F 2212/283 Plural cache memories
• CPC Class Codes: G06F 2212/302 In image processor or graphics adapter
• CPC Class Codes: G06F 2212/452 Instruction code
• CPC Class Codes: G06F 2212/60 Details of cache memory
• CPC Class Codes: G06F 2212/62 Details of cache specific to multiprocessor cache arrangements
• CPC Class Codes: G06F 2212/65 Details of virtual memory and virtual address translation
• CPC Class Codes: G06F 2212/68 Details of translation look-aside buffer [TLB]
• CPC Class Codes: G06F 2212/683 Invalidation
• CPC Class Codes: G06F 2212/684 : TLB miss handling
• CPC Class Codes: G06F 3/0646 : Horizontal data movement in storage systems, i.e. moving data in between storage devices or systems
• CPC Class Codes: G06F 3/0662 : Virtualisation aspects
• CPC Class Codes: G06F 3/0668 : adopting a particular infrastructure
• CPC Class Codes: G06F 9/3851 : from multiple instruction streams, e.g. multistreaming
• CPC Class Codes: G06F 9/3881 : Arrangements for communication of instructions and data
• CPC Class Codes: G06F 9/3887 : controlled by a single instruction for multiple data lanes [SIMD]

Abstract

An apparatus and method are described for coupling a front end core to an accelerator component (e.g., such as a graphics accelerator). For example, an apparatus is described comprising: an accelerator comprising one or more execution units (EUs) to execute a specified set of instructions; and a front end core comprising a translation lookaside buffer (TLB) communicatively coupled to the accelerator and providing memory access services to the accelerator, the memory access services including performing TLB lookup operations to map virtual to physical addresses on behalf of the accelerator and in response to the accelerator requiring access to a system memory.

Claims

1. An apparatus comprising:

front end hardware logic to couple to an accelerator, the front end hardware logic to receive and schedule tasks for execution on the accelerator, the front end hardware logic comprising;

a translation lookaside buffer (TLB) to store virtual-to-physical address mappings for the accelerator, wherein the accelerator comprises a memory manager to submit one or more TLB lookup requests to the TLB of the front end hardware logic, and wherein the front end hardware logic is to perform all TLB lookup operations required by the accelerator, including mapping virtual to physical address on behalf of the accelerator in response to the accelerator requiring access to a system memory; and

page walker circuitry to provide page walk services to the accelerator to determine virtual-to-physical address mappings.

2. The apparatus as in claim 1 wherein the front end hardware logic further comprises:

address translation circuitry to perform a virtual-to-physical address translation for one or more of the accelerator execution circuits by querying the TLB containing the mapping of virtual-to-physical addresses.

3. The apparatus as in claim 2 wherein the address translation circuitry is to cause the page walker circuitry to access a page table from a memory hierarchy of one or more processor cores if the query to the TLB fails to locate a translation for a particular virtual address, the memory hierarchy comprising a system memory and a plurality of cache levels.

4. The apparatus as in claim 3 wherein the front end hardware logic is to access the memory hierarchy of the one or more processor cores using the address translation.

5. The apparatus as in claim 1 further comprising:

context save/restore circuitry to save and restore a context associated with a task scheduled for execution on the accelerator.

6. The apparatus as in claim 3 wherein the plurality of cache levels include a first cache integral to at least one processor core and a second cache to be shared by two or more of the processor cores.

7. The apparatus as in claim 1 wherein the front end hardware logic is to detect a shootdown operation for a restricted page, the front end hardware logic to flush one or more entries in the TLB responsive to detecting the shootdown operation.

9. The computer-readable medium as in claim 1 wherein the front end hardware logic further comprises:

address translation circuitry to perform a virtual-to-physical address translation for one or more of the accelerator execution circuits by querying the TLB containing the mapping of virtual-to-physical addresses.

10. The computer-readable medium as in claim 9 wherein the address translation circuitry to cause the page walker circuitry to access a page table from a memory hierarchy of one or more processor cores if the query to the TLB fails to locate a translation for a particular virtual address, the memory hierarchy comprising a system memory and a plurality of cache levels.

11. The computer-readable medium as in claim 10 wherein the front end hardware logic is to access the memory hierarchy of the one or more processor cores using the address translation.

13. The computer-readable medium as in claim 10 wherein the plurality of cache levels include a first cache integral to at least one processor core and a second cache to be shared by two or more of the processor cores.

8. A non-transitory computer-readable medium having stored thereon hardware description language code to implement front end hardware logic for an accelerator, the front end hardware logic to:

receive and schedule tasks for execution on the accelerator; and

perform memory management operations, including;

store virtual-to-physical address mappings for the accelerator in a translation lookaside buffer (TLB) of the front end hardware logic;

receive from a memory manager of the accelerator one or more TLB look up requests;

perform all TLB lookup operations required by the accelerator, including mapping virtual to physical address on behalf of the accelerator in response to the accelerator requiring access to a system memory; and

provide page walk services to the accelerator to determine virtual-to-physical address mappings.

12. The computer-readable medium as in claim 8 further comprising hardware description language code to implement:

context save/restore circuitry to save and restore a context associated with a task scheduled for execution on the accelerator.

14. The computer-readable medium as in claim 8 wherein the front end hardware logic is to detect a shootdown operation for a restricted page, the front end hardware logic to flush one or more entries in the TLB responsive to detecting the shootdown operation.