The present invention relates generally to integrated circuit memory access schemes, and more particularly to memory access in integrated circuits having multiple hardware accelerators.
As shown in
The HWA processor 140 has three co-processors 142, 144, and 146 configured to perform different data-processing algorithms, such as filtering, fast Fourier transform (FFT) processing, or sequencing, where each co-processor functions as a different hardware accelerator. The HWA processor 140 also has a local SRAM (static random access memory) 148 and controller (bus interface) 150. Similar to the HWA processors 110 and 120, as controlled by the controller 150, each hardware accelerator 142, 144, 146 of the HWA processor 140 accesses and processes data stored in the local memory 148 and stores the resulting processed data back into the local memory 148.
In addition to local and tightly coupled memories, the SOC 100 also has a system memory 160 with twelve banks 162 of independently addressable SRAM memory. The HWA processors 110, 120, and 140 are able to access (i.e., read data from and/or write data to) the memory banks 162 of the system memory 160 simultaneously in a non-blocking manner via a data crossbar switch (i.e., arbiter) 170 and system memory controller 180. In addition to being able to access the system memory 160 via the data crossbar switch 170, the HWA processor 140 is also able to access the system memory 160 via a dedicated bus 152 and system memory controller 180, bypassing the data crossbar switch 170. The bus 152 enables the HWA processor 140 to access system memory 160 faster than by using data the crossbar switch 170.
Assume, for example, an exemplary data-processing routine involving the sequence of a data-processing algorithm of hardware accelerator 142 of the HWA processor 140 followed by a data-processing algorithm of the hardware accelerator 144 of the HWA processor 140 followed by a data-processing algorithm of the hardware accelerator 112 of the HWA processor 110. The sequence of events for this exemplary data-processing routine may be as follows:
Because steps (5) and (7) involve the relatively slow data crossbar switch 170, this exemplary data-processing routine is slower than it would be if the SOC 100 were included a fast bus between the HWA processor 110 and the system memory controller 180 similar to the bus 152. Furthermore, the time that it takes for some data-processing routines to be performed using the data crossbar switch 170 may be too long for certain applications. However, designing the SOC 100 to have a fast bus between the HWA processor 110 and the system memory controller 180 would take up additional layout area.
The present invention is illustrated by way of example and is not limited by the accompanying figures, in which like references indicate similar elements. Elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the thicknesses of layers and regions may be exaggerated for clarity.
Detailed illustrative embodiments of the present invention are disclosed herein. However, specific structural and functional details disclosed herein are merely representative for purposes of describing example embodiments of the present invention. The present invention may be embodied in many alternate forms and should not be construed as limited to only the embodiments set forth herein. Further, the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments of the invention.
As used herein, the singular forms “a,” “an,” and “the,” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It further will be understood that the terms “comprises,” “comprising,” “includes,” and/or “including,” specify the presence of stated features, steps, or components, but do not preclude the presence or addition of one or more other features, steps, or components. It also should be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may in fact be executed substantially concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
In one embodiment, the present invention provides a method of processing data within an integrated circuit. A first set of data is processed using a first hardware processor to generate a second set of data. The second set of data is stored in a first memory. Addresses of the first memory are remapped with addresses of a second memory to swap the first and second memories. The second set of data in the first memory is accessed and processed by a second hardware processor using the remapped addresses of the first memory to generate a third set of data.
In another embodiment, the present invention is an integrated circuit comprising a first hardware processor, a first memory associated with the first hardware processor, a second hardware processor, and a second memory associated with the second hardware processor. The first hardware processor is configured to (i) process a first set of data to generate a second set of data and (ii) store the second set of data in the first memory. The second hardware processor is configured to (i) remap addresses of the first memory with addresses of the second memory to swap the first and second memories and (ii) access and process the second set of data in the first memory using the remapped addresses of the first memory to generate a third set of data.
For example, in the SOC 200, the memory addresses for the local TCM memory 214 of the HWA processor 210 can be swapped with the memory addresses for the system memory bank 262 within the controller 216. In this way, after such logical swapping, the hardware accelerator 212 can access the system memory bank 262 quickly without having to use the data crossbar switch 270. After the hardware accelerator 212 is done accessing system memory bank 262, the memory addresses can be logically swapped back the way they were initially. Alternatively, one or both of the memory blocks can be logically swapped with one or two other memory blocks in the SOC 200. Note that each memory bank 262 in the system memory 260 can be treated as a separate, independently swappable memory block. This means that two system memory banks can also be logically swapped with each other. Although not shown in
The ability to logically swap memory blocks in the SOC 200 enables the exemplary data-processing routine described in the Background section to be implemented using the following sequence of events:
Because steps (5) and (7) do not involve the relatively slow data crossbar switch 270, this exemplary data-processing routine is faster than the corresponding conventional data-processing routine described previously. Furthermore, there is no need for the SOC 200 to have a fast bus between the HWA processor 210 and the system memory controller 280 similar to the bus 252.
In certain implementations of the invention, the data (aka user data) that is processed by data-processing routines, such as that described above, has associated with it error-correction code (ECC) data and ECC address information that links particular sets of ECC data with particular sets of user data. When a set of user data is processed by a hardware accelerator, the corresponding ECC data and ECC address information are also accessed and used (aka consumed) to correct, if needed, the set of user data prior to its processing by the hardware accelerator. Furthermore, the resulting set of processed user data will have its own corresponding ECC data and ECC address information that will need to be stored in memory along with the set of processed user data.
For example, the third set of user data that is stored in system memory bank 262 in step (4) will have corresponding ECC data and ECC address information also stored in system memory bank 262. Similarly, the fourth set of user data stored in system memory bank 262 in step (5) will also have corresponding ECC data and ECC address information stored in system memory bank 262.
Note that, after step (4), the ECC address information for the third set of user data will be based on the memory addresses of system memory bank 262. Furthermore, it is also important for the ECC address information for the fourth set of user data also to be based on the memory addresses of system memory bank 262 after step (7).
According to one embodiment of the invention, after the memory addresses of hardware accelerator 212 and system memory bank 262 are logically swapped in step (5), when the third set of user data is accessed in step (6), the ECC address information is updated on the fly. This means that the ECC address information based on the memory addresses of system memory bank 262 are converted into appropriate ECC address information based on the memory addresses of TCM memory 214 as individual subsets of user data are read from system memory bank 262. Each subset of user data and corresponding ECC data of the third set are then processed to generate a corresponding resulting subset of processed user data and corresponding ECC data of the fourth set. As those individual subsets for the fourth set are written back into system memory bank 262, the memory addresses are then converted on the fly from being based on the memory addresses of TCM memory 214 back into appropriate memory address based on the memory addresses of system memory bank 262. When controller 250 writes data into system memory bank 262, the physical address gets routed to TCM memory 214, whose logical address is different. As such, the ECC data gets updated according to the different logical address so that, when another controller reads that data, it will see no ECC mismatch.
Unless stated otherwise, terms such as “first” and “second” are used to arbitrarily distinguish between the elements such terms describe. Thus, these terms are not necessarily intended to indicate temporal or other prioritization of such elements. Further, the use of introductory phrases such as “at least one” and “one or more” in the claims should not be construed to imply that the introduction of another claim element by the indefinite articles “a” or “an” limits any particular claim containing such introduced claim element to inventions containing only one such element, even when the same claim includes the introductory phrases “one or more” or “at least one” and indefinite articles such as “a” or “an.” The same holds true for the use of definite articles.
Although the invention is described herein with reference to specific embodiments, various modifications and changes can be made without departing from the scope of the present invention as set forth in the claims below. Accordingly, the specification and figures are to be regarded in an illustrative rather than a restrictive sense, and all such modifications are intended to be included within the scope of the present invention. Any benefits, advantages, or solutions to problems that are described herein with regard to specific embodiments are not intended to be construed as a critical, required, or essential feature or element of any or all the claims.
Number | Name | Date | Kind |
---|---|---|---|
5412788 | Collins | May 1995 | A |
5867642 | Chan | Feb 1999 | A |
6816986 | Fanning | Nov 2004 | B1 |
8412987 | Billing | Apr 2013 | B2 |
20080005516 | Meinschein | Jan 2008 | A1 |
20100262751 | Avudaiyappan | Oct 2010 | A1 |
20130339572 | Fanning | Dec 2013 | A1 |
20150095546 | Bennett | Apr 2015 | A1 |
Number | Date | Country | |
---|---|---|---|
20150286525 A1 | Oct 2015 | US |