The present invention relates to network ASICs. More particularly, the present invention relates to two modes of a configuration interface of a network ASIC.
A configuration interface of a chip needs to be extremely robust and satisfy several conditions as this is the interface through which all interfaces, such as PCIe, MDIO, I2C, etc., access an entire register or memory space.
Embodiments of the present invention are directed to a configuration interface of a network ASIC. The configuration interface allows for two modes of traversal of nodes. The nodes form one or more chains. Each chain is in a ring or a list topology. A master receives external access transactions. Once received by the master, an external access transaction traverses the chains to reach a target node. A target node either is an access to a memory space or is a module. A chain can include at least one decoder. A decoder includes logic that determines which of its leaves to send an external access transaction to. In contrast, if a module is not the target node, then the module passes an external access transaction to the next node coupled thereto; otherwise, if the module is the target node, the transmission of the external access transaction stops at the module.
In one aspect, a network device is provided. The network device includes memory, a controller receiving an external access transaction from a source, a plurality of slaves, wherein each of the plurality of slaves includes a set of registers, and a network ASIC including a configuration interface that allows for two modes of traversal of nodes for propagation of the external access transaction to a target node.
In some embodiments, the network device also includes a CPU, wherein the source is the CPU.
In some embodiments, the nodes form one or more chains. In some embodiments, each of the one or more chains is in a ring topology or a list topology.
In some embodiments, the chain includes at least one decoder. The at least one decoder includes logic that determines which of its leaves to send the external access transaction to. The nodes include the at least one decoder.
In some embodiments, the target node is an access to the memory. Alternatively, the target node is one of the plurality of slaves.
In some embodiments, the propagation of the external access transmission stops at one of the at least one slave when the one of the at least one slave is the target node. Alternatively, the propagation of the external access transmission is passed from one of the at least one slave when the one of the at least one slave is not the target node.
In some embodiments, the nodes are arranged in a tree structure.
In some embodiments, the nodes include the controller and the plurality of slaves.
In another aspect, a network device is provided. The network device includes a plurality of nodes forming one or more chains. The plurality of nodes includes a controller receiving an external access transaction from a source, a plurality of modules, wherein each module includes a set of registers, and at least one decoder, wherein each of the at least one decoder includes logic that determines which of the plurality of nodes to send the external access transaction to.
The network device also includes a network ASIC including a configuration interface that allows for two modes of traversal of the nodes for transmission of the external access transaction to a target node.
In some embodiments, the one or more chains include a ring of nodes. In some embodiments, the one or more chains include a list of nodes.
In some embodiments, the list of nodes includes the at least one decoder.
In some embodiments, the transmission of the external access transaction to one of the plurality of modules either stops at the one of the plurality of modules or is passed by the one of the plurality of module to the next node coupling therewith. The transmission is dependent on whether the one of the plurality of modules is the target node.
In some embodiments, the external access transaction includes a read instruction, a write instruction or a combination thereof.
In some embodiments, the set of registers is implemented as flip-flops.
In some embodiments, the network device also includes memory. In some embodiments, the plurality of nodes includes at least one access to the memory.
In yet another aspect, a method of implementing a network ASIC in a network device is provided. The network device includes a plurality of nodes arranged in a tree structure. The method includes receiving at a trunk of the tree an external access transaction from a source. In some embodiments, the trunk of the tree is a controller. In some embodiments, the source is a CPU.
The method also includes traversing the tree until the external access transaction reaches a target node. The method also includes, for each node in the tree that is traversed, when a corresponding node is of a first type, the corresponding node determining which link to send the external access transaction out on and sending the external access transaction out on that link, when the corresponding node is of a second type, the corresponding node processing the external access transaction accordingly, when the corresponding node is of a third type and the corresponding node is not the target node, the corresponding node sending the external access transaction out to the next node coupling therewith, and when the corresponding node is of the third type and the corresponding node is the target node, the corresponding node processing the external transaction accordingly. In some embodiments, the first type is decoder, the second type is an access to memory, and the third type is a module, wherein the module is a set of registers.
In some embodiments, the target node is either an access to memory or a module.
In some embodiments, the method further includes, after processing the external access transaction by the target node, the target node generating an acknowledgment, and sending the acknowledgment back to the trunk of the tree along a traversal route.
In some embodiments, the external access transaction includes a read instruction, a write instruction or a combination thereof.
The foregoing will be apparent from the following more particular description of example embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating embodiments of the present invention.
In the following description, numerous details are set forth for purposes of explanation. However, one of ordinary skill in the art will realize that the invention can be practiced without the use of these specific details. Thus, the present invention is not intended to be limited to the embodiments shown but is to be accorded the widest scope consistent with the principles and features described herein.
A network device includes a CPU (central processing unit) coupled with a network ASIC (application specific integrated circuit). The network ASIC is accessed by the CPU. The network device also includes memory, a master and one or more slaves. One or more of the memory, the master and the one or more slaves can be located on chip, such as on the network ASIC. The master is a controller and receives external access transactions, such as from the CPU. An external access transaction can include read and/or write instructions. A slave includes a set of registers. In some embodiments, the registers are implemented as flip-flops.
Embodiments of the present invention are directed to a configuration interface of a network ASIC. The configuration interface allows for two modes of traversal of nodes. The nodes form one or more chains. Each chain is in a ring or a list topology. A master receives external access transactions. Once received by the master, an external access transaction traverses the chains to reach a target node. A target node either is an access to a memory space or is a module. A chain can include at least one decoder. A decoder includes logic that determines which of its leaves to send an external access transaction to. In contrast, if a module is not the target node, then the module passes an external access transaction to the next node coupled thereto; otherwise, if the module is the target node, the transmission of the external access transaction stops at the module.
The nodes also include at least one decoder (105, generally). Each decoder 105 forms a part of a branch of the tree. Two decoders 105a, 105b are shown in
The nodes also include a plurality of modules (110, generally) as the leaves of the tree. In
In some embodiments, the leaves of the tree can also include at least one access to a memory space (115, generally). In
It should be noted that nodes coupled downstream with a decoder 105 are also referred to as leaves. For example, the leaves of the decoder 105a are the module 110a, the decoder 105b, the memory access block 115a and the module 110g. For another example, the leaves of the decoder 105b are the module 110f and the memory access block 115b.
In
Still referring to
For example, assume that the target node is the memory access block 115b. The external access transaction will traverse the list that includes links L1, L3 and L12 to the memory access block 115b. The following describes the corresponding sequence of steps. At first, the master 100 sends the external access transaction on link L1 to the decoder 105a. Once received, the decoder 105a determines that the outgoing link is L3 in order for the external access transaction to reach the memory access block 115b and sends the external access transaction out on link L3 to the decoder 105b. Once received, the decoder 105b determines that the outgoing link is L12 in order for the external access transaction to reach the memory access block 115b and sends the external access transaction out on L12 to the memory access block 115b. The external access transaction has now reached the target node. After the memory access block 115b (the target node) processes the external access transaction accordingly, the memory access block 115a generates and sends an acknowledgment to the decoder 105b. After the decoder 105b receives the acknowledgment from the memory access block 115a, the decoder 105b forwards the acknowledgment to the decoder 105a. Similarly, after the decoder 105a receives the acknowledgment from the decoder 105b, the decoder 105a forwards the acknowledgment to the master 100.
For another example, assume that the target node is the module 110c. The external access transaction traverses the ring that includes links L1, L2, L6 and L7 to the module 110c. The following describes the corresponding sequence of steps. At first, the master 100 sends the external access transaction on link L1 to the decoder 105a. Once received, the decoder 105a determines that the outgoing link is L2 in order for the external access transaction to reach the module 110c and sends the external access transaction out on link L2 to the module 110a. Once received, the module 110a sends the external access transaction out on the only link (e.g., link L6) to the next node (e.g., the module 110b) because the module 110a is not the target node. Once received, the module 110b sends the external access transaction out on the only link (e.g., link L7) to the next node (e.g., the module 110c) because the module 110b is not the target node. The external access transaction has now reached the target node. After the module 110c (the target node) processes the external access transaction, the module 110c generates and sends an acknowledgment to the module 110b. After the module 110b receives the acknowledgment from the module 110c, the module 110b forwards the acknowledgment to the module 110a. Similarly, after the module 110a receives the acknowledgment from the module 110b, the module 110a forwards the acknowledgment to the decoder 105a. Similarly, after the decoder 105a receives the acknowledgment from the decoder 105b, the decoder 105a forwards the acknowledgment to the master 100.
At a step 255, when the corresponding node is of a first type, the corresponding node determines which link to send the external access transaction out on and sends the external access transaction out on that link. In some embodiments, the first type is a decoder. In some embodiments, the decoder has hardware logic to make such a determination. At a step 260, when the corresponding node is of a second type, the corresponding node processes the external access transaction accordingly. The corresponding node that is of the second type is the target node. In some embodiments, the second type is an access to memory. At a step 265, when the corresponding node is of a third type and the corresponding node is not the target node, the corresponding node sends the external access transaction out to the next node coupled downstream therewith. In some embodiments, the third type is a module. A module includes a set of registers. At a step 270, when the corresponding node is of the third type and the corresponding node is the target node, the corresponding node processes the external transaction accordingly. After the external access transaction is processed by the target node, an acknowledgment is sent back along a traversal route. The target node typically generates the acknowledgment that is sent back to the controller via the traversal route.
The tradeoffs between the list approach and ring approach include cost and latency. For example, the benefits of the ring approach are that coupling the nodes in the ring uses the least number of wires and has minimal routing overhead, thereby reducing costs. However, these benefits are at the expense of latency. Operation latency can be very high with large number of nodes in the ring. Furthermore, flow control lock-up by a single node in the ring can potentially block the entire ring. In contrast, the benefit of the list approach is that there is low operation latency. However, this benefit is at the expense of routing and wire cost.
The module 300 also includes an input labeled as ack_i and an output labeled as ack_o. An acknowledgment either generated or received by the module 300 is sent out from the ack_o output to the preceding node coupled upstream with the module 300. An acknowledgment sent from the node coupled downstream with the module 300 is received at the ack_i input. This input is not enabled in a particular module, such as the module 110f and the module 110g of
The decoder 400 also includes an input labeled as ack_i and an output labeled as ack_o. An acknowledgment sent from the node coupled downstream with the decoder 400 is received at the ack_i input and is sent out through the ack_o output to the preceding node coupled upstream with the decoder 400.
One of ordinary skill in the art will realize other uses and advantages also exist. While the invention has been described with reference to numerous specific details, one of ordinary skill in the art will recognize that the invention can be embodied in other specific forms without departing from the spirit of the invention. Thus, one of ordinary skill in the art will understand that the invention is not to be limited by the foregoing illustrative details, but rather is to be defined by the appended claims.
Number | Name | Date | Kind |
---|---|---|---|
7796627 | Hurley et al. | Sep 2010 | B2 |
9241304 | Dawson et al. | Jan 2016 | B2 |
9253043 | Adolphson | Feb 2016 | B2 |
20140126578 | Yoon | May 2014 | A1 |
20150146626 | Sinha | May 2015 | A1 |
Number | Date | Country | |
---|---|---|---|
20160117282 A1 | Apr 2016 | US |