U.S. patent application number 14/554849 was filed with the patent office on 2015-03-26 for high performance interconnect coherence protocol.
The applicant listed for this patent is Robert Beers, Robert G. Blankenship, Robert A. Maddox, Robert J. Safranek, Aaron T. Spink, Jeff Willey. Invention is credited to Robert Beers, Robert G. Blankenship, Robert A. Maddox, Robert J. Safranek, Aaron T. Spink, Jeff Willey.
Application Number | 20150089161 14/554849 |
Document ID | / |
Family ID | 50486430 |
Filed Date | 2015-03-26 |
United States Patent
Application |
20150089161 |
Kind Code |
A1 |
Beers; Robert ; et
al. |
March 26, 2015 |
HIGH PERFORMANCE INTERCONNECT COHERENCE PROTOCOL
Abstract
A coherence protocol message is sent corresponding to a
particular cache line. A potential conflict involving the
particular cache line is identified and a forward request is sent
to a home agent to identify the potential conflict. A forward
response can be received in response to the forward request from
the home agent and a response to the conflict can be
determined.
Inventors: |
Beers; Robert; (HILLBORO,
OR) ; Blankenship; Robert G.; (Tacoma, WA) ;
Safranek; Robert J.; (Portland, OR) ; Willey;
Jeff; (Timnath, CO) ; Maddox; Robert A.;
(Columbia, SC) ; Spink; Aaron T.; (San Francisco,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Beers; Robert
Blankenship; Robert G.
Safranek; Robert J.
Willey; Jeff
Maddox; Robert A.
Spink; Aaron T. |
HILLBORO
Tacoma
Portland
Timnath
Columbia
San Francisco |
OR
WA
OR
CO
SC
CA |
US
US
US
US
US
US |
|
|
Family ID: |
50486430 |
Appl. No.: |
14/554849 |
Filed: |
November 26, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13976957 |
Jun 27, 2013 |
|
|
|
PCT/US2013/032670 |
Mar 15, 2013 |
|
|
|
14554849 |
|
|
|
|
61717091 |
Oct 22, 2012 |
|
|
|
Current U.S.
Class: |
711/146 |
Current CPC
Class: |
G06F 12/0828 20130101;
G06F 12/0831 20130101; G06F 2212/621 20130101 |
Class at
Publication: |
711/146 |
International
Class: |
G06F 12/08 20060101
G06F012/08 |
Claims
1.-35. (canceled)
36. An apparatus comprising: protocol circuitry to: receive a snoop
request related to a particular address; determine that a
particular coherence agent has a conflicting request to the
particular address; and send a forward request to another coherence
agent to notify a conflict.
37. The apparatus of claim 36, wherein the protocol circuitry is to
send forward requests in response to detection of incoming snoop
requests to a same cache line as an outstanding read or
invalidation request.
38. The apparatus of claim 36, wherein the other coherence agent
comprises a home agent.
39. The apparatus of claim 36, wherein the conflicting request
comprises a particular request to the particular address by the
particular coherence agent that is still outstanding when the snoop
request is received.
40. The apparatus of claim 36, wherein the particular coherence
agent is further to receive a forward response from the other
coherence agent based on the forward request.
41. The apparatus of claim 40, wherein the particular coherence
agent is further to determine a response to the forward response
based at least in part on attributes of the snoop request.
42. The apparatus of claim 41, wherein the other coherence agent
comprises a second coherency agent, the snoop request corresponds
to another coherence protocol message by a third coherence agent
related to the particular address, and the attributes of the snoop
request include identification of the third coherence agent,
identification of a command included in the other coherence
protocol message, and a transaction identifier of the other
coherence protocol message.
43. The apparatus of claim 41, wherein the forward response is to
be sent after a completion is received, and the other coherence
agent comprises a home agent.
44. The apparatus of claim 41, wherein the response comprises
performing a writeback to memory prior to sending a snoop response
to the other agent.
45. The apparatus of claim 44, wherein a particular cache line
associated with the particular address is partially modified at the
particular coherence agent.
46. The apparatus of claim 40, wherein the protocol circuitry is
further to receive a completion after receiving the forward
response.
47. The apparatus of claim 40, wherein the protocol circuitry is
further to receive a completion before receiving the forward
response.
48. The apparatus of claim 36, wherein the protocol circuitry is
further to allocate a forward resource for the forward request.
49. The apparatus of claim 48, wherein the protocol circuitry is to
determine whether the particular coherence agent has an available
forward resource to respond to the snoop request.
50. The apparatus of claim 49, wherein the protocol circuitry is
further to block the snoop request when insufficient forward
resources are available.
51. The apparatus of claim 48, wherein the protocol circuitry is
further to reserve a forward resource for a forward response to the
forward request.
52. The apparatus of claim 36, wherein the forward request is to
comprise a command field, a destination node identifier field, a
home transaction identifier field, and a peer node identifier
field.
53. An apparatus comprising: a snoop receiver to receive a snoop
request that references a particular address; a conflict identifier
to determine that a coherence agent has an active request to an
address matching the particular address; a response generator to
generate a conflict forward request, wherein the conflict forward
request is to include a command field, a destination node
identifier field, a home transaction identifier field, and a peer
node identifier field; a transmitter to send a forward request to a
home agent to identify a conflict related to the snoop request; a
resource manager to reserve a forward resource for a response to
the forward request; and a response receiver to receive a forward
response from the home agent, wherein the forward response is to
identify a transaction identifier associated with the snoop
request.
54. An apparatus comprising: protocol circuitry to: receive a
conflict forward request from a coherence agent, wherein the
conflict forward request is associated with a snoop request;
generate a forward response to the conflict forward request,
wherein the forward response is to comprise a node identifier (NID)
of the coherence agent and a home agent transaction identifier
(TID) associated with a home agent associated with the snoop
request; and send the forward response to the coherence agent.
55. The apparatus of claim 54, wherein a home agent directory of
the home agent is not to be altered in response to the conflict
forward request.
56. The apparatus of claim 54, wherein the conflict forward request
is to include a node identifier of the coherence agent.
57. The apparatus of claim 56, wherein the conflict forward request
is to further include the home transaction identifier field.
58. A method comprising: receiving, at a coherence agent of a block
of a computing device, a snoop request related to a particular
address; determining that the coherence agent has a conflicting
request to the particular address; and send a forward request to
another coherence agent to identify a conflict.
59. The method of claim 58, further comprising receiving a forward
response to the foreward request from the other coherence
agent.
60. The method of claim 59, further comprising: determining a
response to the forward response; and sending a snoop response to
the other coherence agent associated with the forward response.
61. The method of claim 58, further comprising reserving a forward
resource for received responses to the forward request.
62. An apparatus comprising: a controller to interface between at
least a first processor to recognize a first instruction set and
second processor to recognize a second instruction set that is
different from the first instruction set, the controller comprising
interface logic to couple to a link comprising a plurality of
lanes; the interface logic to: receive a snoop request related to a
particular address; determine that a particular coherence agent has
a conflicting request to the particular address; and send a forward
request to another coherence agent to notify a conflict.
63. A system comprising: an interconnect fabric; a home agent; and
a coherence agent communicatively coupled to the home agent via the
interconnect fabric, the coherence agent to: receive a snoop
request related to a particular address; determine that the
coherence agent has a conflicting request to the particular
address; and send a forward request over the interconnect fabric to
the home agent to identify a conflict.
64. The system of claim 63, wherein the home agent is to: receive
another coherence protocol request from another coherence agent;
send the snoop request to the coherence agent, wherein the snoop
corresponds to the other coherence protocol request; receive the
forward request; send a forward response to the coherence agent in
response to the forward request; generate a complete for the
conflicting request; and generate a complete for the other
coherency protocol request.
65. The system of claim 63, wherein the home agent has a set of
resources and the set of resources are not pre-allocated to
coherence agents in the system.
Description
FIELD
[0001] The present disclosure relates in general to the field of
computer development, and more specifically, to software
development involving coordination of mutually-dependent
constrained systems.
BACKGROUND
[0002] Advances in semi-conductor processing and logic design have
permitted an increase in the amount of logic that may be present on
integrated circuit devices. As a corollary, computer system
configurations have evolved from a single or multiple integrated
circuits in a system to multiple cores, multiple hardware threads,
and multiple logical processors present on individual integrated
circuits, as well as other interfaces integrated within such
processors. A processor or integrated circuit typically comprises a
single physical processor die, where the processor die may include
any number of cores, hardware threads, logical processors,
interfaces, memory, controller hubs, etc.
[0003] As a result of the greater ability to fit more processing
power in smaller packages, smaller computing devices have increased
in popularity. Smartphones, tablets, ultrathin notebooks, and other
user equipment have grown exponentially. However, these smaller
devices are reliant on servers both for data storage and complex
processing that exceeds the form factor. Consequently, the demand
in the high-performance computing market (i.e. server space) has
also increased. For instance, in modern servers, there is typically
not only a single processor with multiple cores, but also multiple
physical processors (also referred to as multiple sockets) to
increase the computing power. But as the processing power grows
along with the number of devices in a computing system, the
communication between sockets and other devices becomes more
critical.
[0004] In fact, interconnects have grown from more traditional
multi-drop buses that primarily handled electrical communications
to full blown interconnect architectures that facilitate fast
communication. Unfortunately, as the demand for future processors
to consume at even higher-rates corresponding demand is placed on
the capabilities of existing interconnect architectures.
BRIEF DESCRIPTION OF THE DRAWINGS
[0005] FIG. 1 illustrates a simplified block diagram of a system
including a point-to-point interconnect to connect I/O devices in a
computer system in accordance with one embodiment;
[0006] FIG. 2 illustrates a simplified block diagram of a layered
protocol stack in accordance with one embodiment;
[0007] FIG. 3 illustrates an embodiment of a transaction
descriptor.
[0008] FIG. 4 illustrates an embodiment of a serial point-to-point
link.
[0009] FIG. 5 illustrates embodiments of potential High Performance
Interconnect (HPI) system configurations.
[0010] FIG. 6 illustrates an embodiment of a layered protocol stack
associated with HPI.
[0011] FIG. 7 illustrates a flow diagram of example coherence
protocol conflict management.
[0012] FIG. 8 illustrates a flow diagram of another example
coherence protocol conflict management.
[0013] FIG. 9 illustrates a flow diagram of another example
coherence protocol conflict management.
[0014] FIG. 10 illustrates an embodiment of a block diagram for a
computing system including a multicore processor.
[0015] Like reference numbers and designations in the various
drawings indicate like elements.
DETAILED DESCRIPTION
[0016] In the following description, numerous specific details are
set forth, such as examples of specific types of processors and
system configurations, specific hardware structures, specific
architectural and micro architectural details, specific register
configurations, specific instruction types, specific system
components, specific processor pipeline stages, specific
interconnect layers, specific packet/transaction configurations,
specific transaction names, specific protocol exchanges, specific
link widths, specific implementations, and operation etc. in order
to provide a thorough understanding of the present invention. It
may be apparent, however, to one skilled in the art that these
specific details need not necessarily be employed to practice the
subject matter of the present disclosure. In other instances, well
detailed description of known components or methods has been
avoided, such as specific and alternative processor architectures,
specific logic circuits/code for described algorithms, specific
firmware code, low-level interconnect operation, specific logic
configurations, specific manufacturing techniques and materials,
specific compiler implementations, specific expression of
algorithms in code, specific power down and gating techniques/logic
and other specific operational details of computer system in order
to avoid unnecessarily obscuring the present disclosure.
[0017] Although the following embodiments may be described with
reference to energy conservation, energy efficiency, processing
efficiency, and so on in specific integrated circuits, such as in
computing platforms or microprocessors, other embodiments are
applicable to other types of integrated circuits and logic devices.
Similar techniques and teachings of embodiments described herein
may be applied to other types of circuits or semiconductor devices
that may also benefit from such features. For example, the
disclosed embodiments are not limited to server computer system,
desktop computer systems, laptops, Ultrabooks.TM., but may be also
used in other devices, such as handheld devices, smartphones,
tablets, other thin notebooks, systems on a chip (SOC) devices, and
embedded applications. Some examples of handheld devices include
cellular phones, Internet protocol devices, digital cameras,
personal digital assistants (PDAs), and handheld PCs. Here, similar
techniques for a high-performance interconnect may be applied to
increase performance (or even save power) in a low power
interconnect. Embedded applications typically include a
microcontroller, a digital signal processor (DSP), a system on a
chip, network computers (NetPC), set-top boxes, network hubs, wide
area network (WAN) switches, or any other system that can perform
the functions and operations taught below. Moreover, the
apparatus', methods, and systems described herein are not limited
to physical computing devices, but may also relate to software
optimizations for energy conservation and efficiency. As may become
readily apparent in the description below, the embodiments of
methods, apparatus', and systems described herein (whether in
reference to hardware, firmware, software, or a combination
thereof) may be considered vital to a "green technology" future
balanced with performance considerations.
[0018] As computing systems are advancing, the components therein
are becoming more complex. The interconnect architecture to couple
and communicate between the components has also increased in
complexity to ensure bandwidth demand is met for optimal component
operation. Furthermore, different market segments demand different
aspects of interconnect architectures to suit the respective
market. For example, servers require higher performance, while the
mobile ecosystem is sometimes able to sacrifice overall performance
for power savings. Yet, it is a singular purpose of most fabrics to
provide highest possible performance with maximum power saving.
Further, a variety of different interconnects can potentially
benefit from subject matter described herein. For instance, the
Peripheral Component Interconnect (PCI) Express (PCIe) interconnect
fabric architecture and QuickPath Interconnect (QPI) fabric
architecture, among other examples, can potentially be improved
according to one or more principles described herein, among other
examples.
[0019] FIG. 1 illustrates one embodiment of a fabric composed of
point-to-point Links that interconnect a set of components is
illustrated. System 100 includes processor 105 and system memory
110 coupled to controller hub 115. Processor 105 can include any
processing element, such as a microprocessor, a host processor, an
embedded processor, a co-processor, or other processor. Processor
105 is coupled to controller hub 115 through front-side bus (FSB)
106. In one embodiment, FSB 106 is a serial point-to-point
interconnect as described below. In another embodiment, link 106
includes a serial, differential interconnect architecture that is
compliant with different interconnect standard.
[0020] System memory 110 includes any memory device, such as random
access memory (RAM), non-volatile (NV) memory, or other memory
accessible by devices in system 100. System memory 110 is coupled
to controller hub 115 through memory interface 116. Examples of a
memory interface include a double-data rate (DDR) memory interface,
a dual-channel DDR memory interface, and a dynamic RAM (DRAM)
memory interface.
[0021] In one embodiment, controller hub 115 can include a root
hub, root complex, or root controller, such as in a PCIe
interconnection hierarchy. Examples of controller hub 115 include a
chipset, a memory controller hub (MCH), a northbridge, an
interconnect controller hub (ICH) a southbridge, and a root
controller/hub. Often the term chipset refers to two physically
separate controller hubs, e.g., a memory controller hub (MCH)
coupled to an interconnect controller hub (ICH). Note that current
systems often include the MCH integrated with processor 105, while
controller 115 is to communicate with I/O devices, in a similar
manner as described below. In some embodiments, peer-to-peer
routing is optionally supported through root complex 115.
[0022] Here, controller hub 115 is coupled to switch/bridge 120
through serial link 119. Input/output modules 117 and 121, which
may also be referred to as interfaces/ports 117 and 121, can
include/implement a layered protocol stack to provide communication
between controller hub 115 and switch 120. In one embodiment,
multiple devices are capable of being coupled to switch 120.
[0023] Switch/bridge 120 routes packets/messages from device 125
upstream, i.e. up a hierarchy towards a root complex, to controller
hub 115 and downstream, i.e. down a hierarchy away from a root
controller, from processor 105 or system memory 110 to device 125.
Switch 120, in one embodiment, is referred to as a logical assembly
of multiple virtual PCI-to-PCI bridge devices. Device 125 includes
any internal or external device or component to be coupled to an
electronic system, such as an I/O device, a Network Interface
Controller (NIC), an add-in card, an audio processor, a network
processor, a hard-drive, a storage device, a CD/DVD ROM, a monitor,
a printer, a mouse, a keyboard, a router, a portable storage
device, a Firewire device, a Universal Serial Bus (USB) device, a
scanner, and other input/output devices. Often in the PCIe
vernacular, such as device, is referred to as an endpoint. Although
not specifically shown, device 125 may include a bridge (e.g., a
PCIe to PCI/PCI-X bridge) to support legacy or other versions of
devices or interconnect fabrics supported by such devices.
[0024] Graphics accelerator 130 can also be coupled to controller
hub 115 through serial link 132. In one embodiment, graphics
accelerator 130 is coupled to an MCH, which is coupled to an ICH.
Switch 120, and accordingly I/O device 125, is then coupled to the
ICH. I/O modules 131 and 118 are also to implement a layered
protocol stack and associated logic to communicate between graphics
accelerator 130 and controller hub 115. Similar to the MCH
discussion above, a graphics controller or the graphics accelerator
130 itself may be integrated in processor 105.
[0025] Turning to FIG. 2 an embodiment of a layered protocol stack
is illustrated. Layered protocol stack 200 can includes any form of
a layered communication stack, such as a QPI stack, a PCIe stack, a
next generation high performance computing interconnect (HPI)
stack, or other layered stack. In one embodiment, protocol stack
200 can include transaction layer 205, link layer 210, and physical
layer 220. An interface, such as interfaces 117, 118, 121, 122,
126, and 131 in FIG. 1, may be represented as communication
protocol stack 200. Representation as a communication protocol
stack may also be referred to as a module or interface
implementing/including a protocol stack.
[0026] Packets can be used to communicate information between
components. Packets can be formed in the Transaction Layer 205 and
Data Link Layer 210 to carry the information from the transmitting
component to the receiving component. As the transmitted packets
flow through the other layers, they are extended with additional
information used to handle packets at those layers. At the
receiving side the reverse process occurs and packets get
transformed from their Physical Layer 220 representation to the
Data Link Layer 210 representation and finally (for Transaction
Layer Packets) to the form that can be processed by the Transaction
Layer 205 of the receiving device.
[0027] In one embodiment, transaction layer 205 can provide an
interface between a device's processing core and the interconnect
architecture, such as Data Link Layer 210 and Physical Layer 220.
In this regard, a primary responsibility of the transaction layer
205 can include the assembly and disassembly of packets (i.e.,
transaction layer packets, or TLPs). The translation layer 205 can
also manage credit-based flow control for TLPs. In some
implementations, split transactions can be utilized, i.e.,
transactions with request and response separated by time, allowing
a link to carry other traffic while the target device gathers data
for the response, among other examples.
[0028] Credit-based flow control can be used to realize virtual
channels and networks utilizing the interconnect fabric. In one
example, a device can advertise an initial amount of credits for
each of the receive buffers in Transaction Layer 205. An external
device at the opposite end of the link, such as controller hub 115
in FIG. 1, can count the number of credits consumed by each TLP. A
transaction may be transmitted if the transaction does not exceed a
credit limit. Upon receiving a response an amount of credit is
restored. One example of an advantage of such a credit scheme is
that the latency of credit return does not affect performance,
provided that the credit limit is not encountered, among other
potential advantages.
[0029] In one embodiment, four transaction address spaces can
include a configuration address space, a memory address space, an
input/output address space, and a message address space. Memory
space transactions include one or more of read requests and write
requests to transfer data to/from a memory-mapped location. In one
embodiment, memory space transactions are capable of using two
different address formats, e.g., a short address format, such as a
32-bit address, or a long address format, such as 64-bit address.
Configuration space transactions can be used to access
configuration space of various devices connected to the
interconnect. Transactions to the configuration space can include
read requests and write requests. Message space transactions (or,
simply messages) can also be defined to support in-band
communication between interconnect agents. Therefore, in one
example embodiment, transaction layer 205 can assemble packet
header/payload 206.
[0030] Quickly referring to FIG. 3, an example embodiment of a
transaction layer packet descriptor is illustrated. In one
embodiment, transaction descriptor 300 can be a mechanism for
carrying transaction information. In this regard, transaction
descriptor 300 supports identification of transactions in a system.
Other potential uses include tracking modifications of default
transaction ordering and association of transaction with channels.
For instance, transaction descriptor 300 can include global
identifier field 302, attributes field 304 and channel identifier
field 306. In the illustrated example, global identifier field 302
is depicted comprising local transaction identifier field 308 and
source identifier field 310. In one embodiment, global transaction
identifier 302 is unique for all outstanding requests.
[0031] According to one implementation, local transaction
identifier field 308 is a field generated by a requesting agent,
and can be unique for all outstanding requests that require a
completion for that requesting agent. Furthermore, in this example,
source identifier 310 uniquely identifies the requestor agent
within an interconnect hierarchy. Accordingly, together with source
ID 310, local transaction identifier 308 field provides global
identification of a transaction within a hierarchy domain.
[0032] Attributes field 304 specifies characteristics and
relationships of the transaction. In this regard, attributes field
304 is potentially used to provide additional information that
allows modification of the default handling of transactions. In one
embodiment, attributes field 304 includes priority field 312,
reserved field 314, ordering field 316, and no-snoop field 318.
Here, priority sub-field 312 may be modified by an initiator to
assign a priority to the transaction. Reserved attribute field 314
is left reserved for future, or vendor-defined usage. Possible
usage models using priority or security attributes may be
implemented using the reserved attribute field.
[0033] In this example, ordering attribute field 316 is used to
supply optional information conveying the type of ordering that may
modify default ordering rules. According to one example
implementation, an ordering attribute of "0" denotes default
ordering rules are to apply, wherein an ordering attribute of "1"
denotes relaxed ordering, wherein writes can pass writes in the
same direction, and read completions can pass writes in the same
direction. Snoop attribute field 318 is utilized to determine if
transactions are snooped. As shown, channel ID Field 306 identifies
a channel that a transaction is associated with.
[0034] Returning to the discussion of FIG. 2, a Link layer 210,
also referred to as data link layer 210, can act as an intermediate
stage between transaction layer 205 and the physical layer 220. In
one embodiment, a responsibility of the data link layer 210 is
providing a reliable mechanism for exchanging Transaction Layer
Packets (TLPs) between two components on a link. One side of the
Data Link Layer 210 accepts TLPs assembled by the Transaction Layer
205, applies packet sequence identifier 211, i.e. an identification
number or packet number, calculates and applies an error detection
code, i.e. CRC 212, and submits the modified TLPs to the Physical
Layer 220 for transmission across a physical to an external
device.
[0035] In one example, physical layer 220 includes logical sub
block 221 and electrical sub-block 222 to physically transmit a
packet to an external device. Here, logical sub-block 221 is
responsible for the "digital" functions of Physical Layer 221. In
this regard, the logical sub-block can include a transmit section
to prepare outgoing information for transmission by physical
sub-block 222, and a receiver section to identify and prepare
received information before passing it to the Link Layer 210.
[0036] Physical block 222 includes a transmitter and a receiver.
The transmitter is supplied by logical sub-block 221 with symbols,
which the transmitter serializes and transmits onto to an external
device. The receiver is supplied with serialized symbols from an
external device and transforms the received signals into a
bit-stream. The bit-stream is de-serialized and supplied to logical
sub-block 221. In one example embodiment, an 8b/10b transmission
code is employed, where ten-bit symbols are transmitted/received.
Here, special symbols are used to frame a packet with frames 223.
In addition, in one example, the receiver also provides a symbol
clock recovered from the incoming serial stream.
[0037] As stated above, although transaction layer 205, link layer
210, and physical layer 220 are discussed in reference to a
specific embodiment of a protocol stack (such as a PCIe protocol
stack), a layered protocol stack is not so limited. In fact, any
layered protocol may be included/implemented and adopt features
discussed herein. As an example, a port/interface that is
represented as a layered protocol can include: (1) a first layer to
assemble packets, i.e. a transaction layer; a second layer to
sequence packets, i.e. a link layer; and a third layer to transmit
the packets, i.e. a physical layer. As a specific example, a high
performance interconnect layered protocol, as described herein, is
utilized.
[0038] Referring next to FIG. 4, an example embodiment of a serial
point to point fabric is illustrated. A serial point-to-point link
can include any transmission path for transmitting serial data. In
the embodiment shown, a link can include two, low-voltage,
differentially driven signal pairs: a transmit pair 406/411 and a
receive pair 412/407. Accordingly, device 405 includes transmission
logic 406 to transmit data to device 410 and receiving logic 407 to
receive data from device 410. In other words, two transmitting
paths, i.e. paths 416 and 417, and two receiving paths, i.e. paths
418 and 419, are included in some implementations of a link.
[0039] A transmission path refers to any path for transmitting
data, such as a transmission line, a copper line, an optical line,
a wireless communication channel, an infrared communication link,
or other communication path. A connection between two devices, such
as device 405 and device 410, is referred to as a link, such as
link 415. A link may support one lane--each lane representing a set
of differential signal pairs (one pair for transmission, one pair
for reception). To scale bandwidth, a link may aggregate multiple
lanes denoted by xN, where N is any supported link width, such as
1, 2, 4, 8, 12, 16, 32, 64, or wider.
[0040] A differential pair can refer to two transmission paths,
such as lines 416 and 417, to transmit differential signals. As an
example, when line 416 toggles from a low voltage level to a high
voltage level, i.e. a rising edge, line 417 drives from a high
logic level to a low logic level, i.e. a falling edge. Differential
signals potentially demonstrate better electrical characteristics,
such as better signal integrity, i.e. cross-coupling, voltage
overshoot/undershoot, ringing, among other example advantages. This
allows for a better timing window, which enables faster
transmission frequencies.
[0041] In one embodiment, a new High Performance Interconnect (HPI)
is provided. HPI can include a next-generation cache-coherent,
link-based interconnect. As one example, HPI may be utilized in
high performance computing platforms, such as workstations or
servers, including in systems where PCIe or another interconnect
protocol is typically used to connect processors, accelerators, I/O
devices, and the like. However, HPI is not so limited. Instead, HPI
may be utilized in any of the systems or platforms described
herein. Furthermore, the individual ideas developed may be applied
to other interconnects and platforms, such as PCIe, MIPI, QPI,
etc.
[0042] To support multiple devices, in one example implementation,
HPI can include an Instruction Set Architecture (ISA) agnostic
(i.e. HPI is able to be implemented in multiple different devices).
In another scenario, HPI may also be utilized to connect high
performance I/O devices, not just processors or accelerators. For
example, a high performance PCIe device may be coupled to HPI
through an appropriate translation bridge (i.e. HPI to PCIe).
Moreover, the HPI links may be utilized by many HPI based devices,
such as processors, in various ways (e.g. stars, rings, meshes,
etc.). FIG. 5 illustrates example implementations of multiple
potential multi-socket configurations. A two-socket configuration
505, as depicted, can include two HPI links; however, in other
implementations, one HPI link may be utilized. For larger
topologies, any configuration may be utilized as long as an
identifier (ID) is assignable and there is some form of virtual
path, among other additional or substitute features. As shown, in
one example, a four socket configuration 510 has an HPI link from
each processor to another. But in the eight socket implementation
shown in configuration 515, not every socket is directly connected
to each other through an HPI link. However, if a virtual path or
channel exists between the processors, the configuration is
supported. A range of supported processors includes 2-32 in a
native domain. Higher numbers of processors may be reached through
use of multiple domains or other interconnects between node
controllers, among other examples.
[0043] The HPI architecture includes a definition of a layered
protocol architecture, including in some examples, protocol layers
(coherent, non-coherent, and, optionally, other memory based
protocols), a routing layer, a link layer, and a physical layer
including associated I/O logic. Furthermore, HPI can further
include enhancements related to power managers (such as power
control units (PCUs)), design for test and debug (DFT), fault
handling, registers, security, among other examples. FIG. 6
illustrates an embodiment of an example HPI layered protocol stack.
In some implementations, at least some of the layers illustrated in
FIG. 6 may be optional. Each layer deals with its own level of
granularity or quantum of information (the protocol layer 605a,b
with packets 630, link layer 610a,b with flits 635, and physical
layer 605a,b with phits 640). Note that a packet, in some
embodiments, may include partial flits, a single flit, or multiple
flits based on the implementation.
[0044] As a first example, a width of a phit 640 includes a 1 to 1
mapping of link width to bits (e.g. 20 bit link width includes a
phit of 20 bits, etc.). Flits may have a greater size, such as 184,
192, or 200 bits. Note that if phit 640 is 20 bits wide and the
size of flit 635 is 184 bits then it takes a fractional number of
phits 640 to transmit one flit 635 (e.g. 9.2 phits at 20 bits to
transmit an 184 bit flit 635 or 9.6 at 20 bits to transmit a 192
bit flit, among other examples). Note that widths of the
fundamental link at the physical layer may vary. For example, the
number of lanes per direction may include 2, 4, 6, 8, 10, 12, 14,
16, 18, 20, 22, 24, etc. In one embodiment, link layer 610a,b is
capable of embedding multiple pieces of different transactions in a
single flit, and one or multiple headers (e.g. 1, 2, 3, 4) may be
embedded within the flit. In one example, HPI splits the headers
into corresponding slots to enable multiple messages in the flit
destined for different nodes.
[0045] Physical layer 605a,b, in one embodiment, can be responsible
for the fast transfer of information on the physical medium
(electrical or optical etc.). The physical link can be
point-to-point between two Link layer entities, such as layer 605a
and 605b. The Link layer 610a,b can abstract the Physical layer
605a,b from the upper layers and provides the capability to
reliably transfer data (as well as requests) and manage flow
control between two directly connected entities. The Link Layer can
also be responsible for virtualizing the physical channel into
multiple virtual channels and message classes. The Protocol layer
620a,b relies on the Link layer 610a,b to map protocol messages
into the appropriate message classes and virtual channels before
handing them to the Physical layer 605a,b for transfer across the
physical links. Link layer 610a,b may support multiple messages,
such as a request, snoop, response, writeback, non-coherent data,
among other examples.
[0046] The Physical layer 605a,b (or PHY) of HPI can be implemented
above the electrical layer (i.e. electrical conductors connecting
two components) and below the link layer 610a,b, as illustrated in
FIG. 6. The Physical layer and corresponding logic can reside on
each agent and connects the link layers on two agents (A and B)
separated from each other (e.g. on devices on either side of a
link). The local and remote electrical layers are connected by
physical media (e.g. wires, conductors, optical, etc.). The
Physical layer 605a,b, in one embodiment, has two major phases,
initialization and operation. During initialization, the connection
is opaque to the link layer and signaling may involve a combination
of timed states and handshake events. During operation, the
connection is transparent to the link layer and signaling is at a
speed, with all lanes operating together as a single link. During
the operation phase, the Physical layer transports flits from agent
A to agent B and from agent B to agent A. The connection is also
referred to as a link and abstracts some physical aspects including
media, width and speed from the link layers while exchanging flits
and control/status of current configuration (e.g. width) with the
link layer. The initialization phase includes minor phases e.g.
Polling, Configuration. The operation phase also includes minor
phases (e.g. link power management states).
[0047] In one embodiment, Link layer 610a,b can be implemented so
as to provide reliable data transfer between two protocol or
routing entities. The Link layer can abstract Physical layer 605a,b
from the Protocol layer 620a,b, and can be responsible for the flow
control between two protocol agents (A, B), and provide virtual
channel services to the Protocol layer (Message Classes) and
Routing layer (Virtual Networks). The interface between the
Protocol layer 620a,b and the Link Layer 610a,b can typically be at
the packet level. In one embodiment, the smallest transfer unit at
the Link Layer is referred to as a flit which a specified number of
bits, such as 192 bits or some other denomination. The Link Layer
610a,b relies on the Physical layer 605a,b to frame the Physical
layer's 605a,b unit of transfer (phit) into the Link Layer's 610a,b
unit of transfer (flit). In addition, the Link Layer 610a,b may be
logically broken into two parts, a sender and a receiver. A
sender/receiver pair on one entity may be connected to a
receiver/sender pair on another entity. Flow Control is often
performed on both a flit and a packet basis. Error detection and
correction is also potentially performed on a flit level basis.
[0048] In one embodiment, Routing layer 615a,b can provide a
flexible and distributed method to route HPI transactions from a
source to a destination. The scheme is flexible since routing
algorithms for multiple topologies may be specified through
programmable routing tables at each router (the programming in one
embodiment is performed by firmware, software, or a combination
thereof). The routing functionality may be distributed; the routing
may be done through a series of routing steps, with each routing
step being defined through a lookup of a table at either the
source, intermediate, or destination routers. The lookup at a
source may be used to inject a HPI packet into the HPI fabric. The
lookup at an intermediate router may be used to route an HPI packet
from an input port to an output port. The lookup at a destination
port may be used to target the destination HPI protocol agent. Note
that the Routing layer, in some implementations, can be thin since
the routing tables, and, hence the routing algorithms, are not
specifically defined by specification. This allows for flexibility
and a variety of usage models, including flexible platform
architectural topologies to be defined by the system
implementation. The Routing layer 615a,b relies on the Link layer
610a,b for providing the use of up to three (or more) virtual
networks (VNs)--in one example, two deadlock-free VNs, VN0 and VN1
with several message classes defined in each virtual network. A
shared adaptive virtual network (VNA) may be defined in the Link
layer, but this adaptive network may not be exposed directly in
routing concepts, since each message class and virtual network may
have dedicated resources and guaranteed forward progress, among
other features and examples.
[0049] In one embodiment, HPI can include a Coherence Protocol
layer 620a,b to support agents caching lines of data from memory.
An agent wishing to cache memory data may use the coherence
protocol to read the line of data to load into its cache. An agent
wishing to modify a line of data in its cache may use the coherence
protocol to acquire ownership of the line before modifying the
data. After modifying a line, an agent may follow protocol
requirements of keeping it in its cache until it either writes the
line back to memory or includes the line in a response to an
external request. Lastly, an agent may fulfill external requests to
invalidate a line in its cache. The protocol ensures coherency of
the data by dictating the rules all caching agents may follow. It
also provides the means for agents without caches to coherently
read and write memory data.
[0050] Two conditions may be enforced to support transactions
utilizing the HPI Coherence Protocol. First, the protocol can
maintain data consistency, as an example, on a per-address basis,
among data in agents' caches and between those data and the data in
memory. Informally, data consistency may refer to each valid line
of data in an agent's cache representing a most up-to-date value of
the data and data transmitted in a coherence protocol packet can
represent the most up-to-date value of the data at the time it was
sent. When no valid copy of the data exists in caches or in
transmission, the protocol may ensure the most up-to-date value of
the data resides in memory. Second, the protocol can provide
well-defined commitment points for requests. Commitment points for
reads may indicate when the data is usable; and for writes they may
indicate when the written data is globally observable and will be
loaded by subsequent reads. The protocol may support these
commitment points for both cacheable and uncacheable (UC) requests
in the coherent memory space.
[0051] The HPI Coherence Protocol also may ensure the forward
progress of coherence requests made by an agent to an address in
the coherent memory space. Certainly, transactions may eventually
be satisfied and retired for proper system operation. The HPI
Coherence Protocol, in some embodiments, may have no notion of
retry for resolving resource allocation conflicts. Thus, the
protocol itself may be defined to contain no circular resource
dependencies, and implementations may take care in their designs
not to introduce dependencies that can result in deadlocks.
Additionally, the protocol may indicate where designs are able to
provide fair access to protocol resources.
[0052] Logically, the HPI Coherence Protocol, in one embodiment,
can include three items: coherence (or caching) agents, home
agents, and the HPI interconnect fabric connecting the agents.
Coherence agents and home agents can work together to achieve data
consistency by exchanging messages over the interconnect. The link
layer 610a,b and its related description can provide the details of
the interconnect fabric including how it adheres to the coherence
protocol's requirements, discussed herein. (It may be noted that
the division into coherence agents and home agents is for clarity.
A design may contain multiple agents of both types within a socket
or even combine agents behaviors into a single design unit, among
other examples.)
[0053] In one embodiment, home agents can be configured to guard
physical memory. Each home agent can be responsible for a region of
the coherent memory space. Regions may be non-overlapping, in that
a single address is guarded by one home agent, and together the
home agent regions in a system cover the coherent memory space. For
instance, each address can be guarded by at least one home agent.
Therefore, in one embodiment, each address in a HPI system's
coherent memory space can map to exactly one home agent.
[0054] Home agents in the HPI Coherence Protocol, in one
embodiment, can be responsible for servicing requests to the
coherent memory space. For read (Rd) requests, home agents may
generate snoops (Snp), process their responses, send a data
response, and send a completion response. For invalidation (Inv)
requests, home agents may generate necessary snoops, process their
responses, and send a completion response. For write requests, home
agents may commit the data to memory and send a completion
response.
[0055] Home agents may provide snoops in the HPI Coherence Protocol
and process snoop responses from coherence agents. Home agents can
also process forward requests, which are a special snoop response,
from coherence agents for conflict resolution. When a home agent
receives a forward request, it may send a forward response to the
coherence agent that generated the forward request (i.e., the agent
that detected a conflicting snoop request). Coherence agents can
use the ordering of these forward responses and completion
responses from the home agent to resolve conflicts.
[0056] A coherence agent may issue supported coherence protocol
requests. Requests may be issued to an address in the coherent
memory space. Data received for read requests (Rd) except RdCur may
be consistent. Data for RdCur requests may have been consistent
when the data packet was generated (although it may have become out
of date during delivery). Table 1 shows an exemplary,
non-exhaustive list of potential supported requests:
TABLE-US-00001 TABLE 1 Name Semantics Cache State RdCode Request a
cache line in F or S state. F or S RdData Request a cache line in
E, F, or S state. F or S RdMigr Request a cache line in M, E, F, or
S state. M and (F or S) RdInv Request a cache line in E state. If
line was previously E cached in M state, the line will be written
to memory before E data is delivered. RdInvOwn Request a cache line
in M or E state. M RdCur Request an uncacheable snapshot of a cache
line. InvItoE Request exclusive ownership of a cache line without M
or E receiving data. InvItoM Request exclusive ownership of a cache
line without M or E receiving data and with the intent of
performing a writeback soon afterward. InvXtoI Flush a cache line
from all caches. Requesting agent is to invalidate the line in its
cache before issuing this request. WbMtoI Write a cache line in M
state back to memory and M invalidate the line in the cache. WbMtoS
Write a cache line in M state back to memory and M and S transition
line to S state. WbMtoE Write a cache line in M state back to
memory and M and E transition line to E state. WbMtoIPtl Write a
cache line in M state back to memory, according M to a byte-enable
mask, and transition line to I state. WbMtoEPtl Write a cache line
in M state back to memory, according M and E to a byte-enable mask,
transition line to E state, and clear the line's mask in the cache.
EvctCln Notification to home agent that a cache line in E state was
E invalidated in the cache. WbPushMtoI Send a line in M state to
home agent and invalidate the M line in the cache; home agent may
either write the line back to memory or send it to a local cache
agent with M state. WbFlush Request that home flush writes to
implementation- specific addresses in its memory hierarchy. No data
is sent with the request.
[0057] HPI can support a Coherency protocol making use of
principles of the MESI protocol. Each cache line can be marked with
one or more supported states (e.g., coded in the cache line). A "M"
or "Modified" state can indicate that the cache line value has been
modified from that value which is in main memory. A line in the
M-state is only present in the particular and the corresponding
cache agent can be required to write the modified data back to
memory at some time in the future, for instance, before permitting
any other read of the (no longer valid) maing memory state. A
writeback can transition the line from the M-state to the E-state.
The "E" or "Exclusive" state can indicate that the cache line is
only present in the current cache but that its value matches that
in main memory. The cache line in E-state can transition to the
S-state at any time in response to a read request or may be changed
to the M-state by writing to the line. The "S" or "Shared" state
can indicates that the cache line may be stored in other caches of
the machine and has a value that matches that of the main memory.
The line may be discarded (changed to the I-state) at any time. The
"I" or "Invalid" state can indicate that a cache line is invalid or
unused. Other state can also supported in HPI, such as an "F" or
"Forward" shared state that indicates that the particular shared
line value is to be forwarded to other caches that are to also
share the line, among other examples.
[0058] Table 2 include exemplary information that can be included
in some Coherence protocol messages, including snoop, read, and
write requests, among other examples:
TABLE-US-00002 TABLE 2 Field Usage cmd Message command (or name or
opcode). addr Address of a coherent cache line. destNID Node ID
(NID) of destination (home or coherence) agent. reqNID NID of
requesting coherence agent. peerNID NID of coherence agent that
sent the (forward request) message. reqTID ID of the resource
allocated by the requesting agent for the transaction, also known
as RTID (or requesting transaction identifier). homeTID ID of the
resource allocated by the home agent to process the request, also
known as HTID (or home transaction identifier). data A cache line
of data. mask Byte mask to qualify the data.
[0059] Snoop messages may be generated by home agents and directed
toward coherence agents. A snoop (SNP) virtual channel can be used
for snoops and, in one embodiment, are the only messages that use
the SNP virtual channel. Snoops can include the requesting agent's
NID and the RTID it allocated for the request in case the snoop
results in data being sent directly to the requesting agent.
Snoops, in one embodiment, can also include the HTID allocated by
the home agent to process the request. The coherence agent
processing the snoop may include the HTID in the snoop response it
sends back to the home agent. Snoops may, in some instance, not
include the home agent's NID because it may be derived from the
included address, which the targeted coherence agent does when
sending its response. Fanout snoops (those with "SnpF" prefix) may
not include a destination NID because the Routing Layer is
responsible for generating the appropriate snoop messages to all
peers in the fanout region. An exemplary list of snoop channel
messages is listed Table 3:
TABLE-US-00003 TABLE 3 Command Semantics Fields SnpCode Snoop to
get data in F or S state. cmd, SnpData Snoop to get data in E, F,
or S state. addr, SnpMigr Snoop to get data in M, E, F, or S state.
destNID, SnpInv Snoop to invalidate the peer's cache, flushing any
M copy to reqNID, memory. reqTID, SnpInvOwn Snoop to get data in M
or E state. homeTID SnpCur Snoop to get an uncacheable snapshot of
a cache line. SnpFCode Snoop to get data in F or S state; Routing
layer to handle cmd, distribution to all fanout peers addr,
SnpFData Snoop to get data in E, F, or S state; Routing layer to
handle reqNID, distribution to all fanout peers reqTID, SnpFMigr
Snoop to get data in M, E, F, or S state; Routing layer to homeTID
handle distribution to all fanout peers SnpFInvOwn Snoop to get
data in M or E state: Routing layer to handle distribution to all
fanout peers. SnpFInv Snoop to invalidate the peer's cache,
flushing any M copy to memory; Routing layer to handle distribution
to all fanout peers. SnpCur Snoop to get an uncacheable snapshot of
a, cache line: Routing layer to handle distribution to all fanout
peers.
[0060] HPI may also support non snoop requests that they may issue
to an address, such as those implemented as non-coherent requests.
Examples of such requests can include a non-snoop read to request a
read-only line form memory, a non-snoop write to write a line to
memory, and a write a line to memory according to a mask, among
other potential examples.
[0061] In one example, four general types of response messages can
be defined in the HPI Coherence Protocol: data, completion, snoop,
and forward. Certain data messages can carry an additional
completion indication and certain snoop responses can carry data.
Response messages may use the RSP virtual channel, and the
communication fabric may maintain proper message delivery ordering
among ordered completion responses and forward responses.
[0062] Table 4 includes a listing of at least some potential
response messages supported by an example HPI Coherence
Protocol:
TABLE-US-00004 TABLE 4 Name Semantics Fields Data_M Data is M
state. cmd, Data_E Data is E state. destNID, Data_F Data is F
state. reqTID, Data_SI Depending upon request, data in S state or
uncacheable data "snapshot" data. Data_M Data is M state with an
ordered completion response. Data_E Data is E state with an ordered
completion response. Data_F Data is F state with an ordered
completion response. Data_SI Depending upon request, data in S
state or uncacheable "snapshot" data, with an ordered completion
response. CmpU Completion message with no ordering requirements.
cmd, CmpO Completion message to be ordered with forward destNID,
responses. reqTID RspI Cache is in I state. cmd, RspS Cache is in S
state. destNID, RspFwd Copy of cache line was sent to requesting
agent, cache homeTID state did not change. RspFwdI Copy of cache
line was sent to requesting agent, cache transitions to I state.
RspFwdS Copy of cache line was sent to requesting agent, cache
transitions to S state. RspIWb Modified line is being implicitly
written back to memory, cmd, cache was transitioned to I state.
destNID, RspSWb Modified line is being implicitly written back to
memory, homeTID, cache was transitioned to S state. data RspFwdIWb
Modified line is being implicitly written back to memory, copy of
cache line was sent to requesting agent, cache was transitioned to
I state. RspFwdSWb Modified line is being implicitly written back
to memory, copy of cache line was sent to requesting agent, cache
was transitioned to S state. RspCnflt Peer has an outstanding
request to same address, is cmd, requesting an ordered forward
response, and has destNID, allocated a resource for the forward.
homeTID, peerNID
[0063] In one example, data responses can target a requesting
coherence agent. A home agent may send any of the data responses. A
coherence agent may send only data responses not containing an
ordered completion indication. Additionally, coherence agents may
be limited to sending data responses only as a result of processing
a snoop request. Combined data and completion responses may always
be of the ordered-completion type and can be kept ordered with
forward responses by the communication fabric.
[0064] The HPI Coherence Protocol can uses the general unordered
completion message and a coherence-specific ordered completion
message. A home agent may send completion responses to coherent
requests and completion responses can be typically destined for a
coherence agent. The ordered completion response can be kept
ordered with forward responses by the communication fabric.
[0065] Snoop responses may be sent by coherence agents,
specifically in response to processing a snoop request, and target
the home agent handling the snoop request. The destNID is usually a
home agent (determined from the address in the snoop request) and
the included TID is for the home agent's resource allocated to
process the request. Snoop responses with "Wb" in the command are
for implicit writebacks of modified cache lines, and they carry the
cache line data. (Implicit writebacks can include those a coherence
agent makes due to another agent's request, whereas the other
requests are made explicitly by the coherence agent using its
request resources.)
[0066] Coherence agents can generate a forward request when a snoop
request conflicts with an outstanding request. Forward requests
target the home agent that generated the snoop, which is determined
from the address in the snoop request. Thus, the destNID is a home
agent. The forward request can also include the TID for the home
agent's resource allocated to process the original request and the
NID of the coherence agent generating the forward request
[0067] The HPI Coherence Protocol can support a single forward
response, FwdCnfltO. Home agents can send a forward response for
every forward request received and to the coherence agent in the
forward request's peerNID field. Forward responses carry the cache
line address so the coherence agent can match the message to the
forward resource it allocated. Forward response message can carry
the requesting agent's NID but, in some cases, not the requesting
agent's TID. If a coherence agent wants to support cache-to-cache
transfers for forward responses, it can save the requesting agent's
TID when processing the snoop and send a forward request. To
support conflict resolution, the communication fabric may maintain
ordering between the forward response and all ordered completions
sent before it to the same destination coherence agent.
[0068] In some systems, home agent resources are pre-allocated in
that "RTIDs" represent resources in the home agents and the caching
agents allocate RTIDs from system-configured pools when generating
new coherence requests. Such schemes can limit the number of active
requests any particular caching agent can have to a home agent to
the number of RTIDs it was given by the system, effectively slicing
up home resources statically among caching agents. Such schemes can
result inefficient allocation of resources and properly sizing a
home agent to support request throughput can become impractical for
large systems, among other potential issues. For instance, such
schemes can force RTID pool management upon the caching agents.
Additionally, in some systems, a caching agent may not reuse the
RTID until the home agent has completely processed the transaction.
Waiting until a home agent completes all processing, however, can
unnecessarily throttle caching agents. Additionally, certain flows
in the protocol can involve caching agents holding onto RTIDs
beyond the home agent release notification, further throttling
their performance, among other issues.
[0069] In one implementation, home agents can be allowed to
allocate their resources as requests arrive from cache agents. In
such instances, home agent resource management can be kept separate
from coherence agent logic. In some implementations, home resource
management and coherence agent logic can be at least partially
intermingled. In some instances, coherence agents can have more
outstanding requests to a home agent than the home agent can
simultaneously handle. For instance, HPI can allow requests to
queue up in the communication fabric. Further, to avoid deadlocks
caused by the home agent blocking incoming requests until resources
become available, the HPI Coherence protocol can be configured to
ensure that other messages can make progress around blocked
requests to ensure that active transactions reach completion.
[0070] In one example, resource management can be supported by
allowing an agent receiving a request to allocate resources to
process it, the agent sending the request allocating respective
resources for all responses to the request The HTID can represent
the resource that a home agent allocates for a given request
included in some protocol messages. The HTID (along with RNID/RTID)
in snoop requests and forward responses can be used to support
responses to a home agent as well as data forwarding to a
requesting agent, among other examples. Further, HPI can support
the ability of an agent to send an ordered complete (CmpO) early,
that is, before the home agent is finished processing the request,
when it is determined to be safe for a requesting agent to reuse
its RTID resource. General handling of snoops with similar
RNID/RTID can also be defined by the protocol.
[0071] In one illustrative example, when a particular request's
tracker state is busy, a directory state can be used to determine
when the home agent may send a response. For instance, an Invalid
directory state can allow a response to be sent, except for RdCur
requests which indicates there are no outstanding snoop responses.
An Unknown directory state can dictate that all peer agents have
been snooped and all their responses gathered before a response can
be sent. The Exclusive directory state can dictate that the owner
be snooped and all responses gathered before a response is sent, or
if the requesting agent is the owner then a response may
immediately be sent. The Shared directory state can specify that an
invalidating request (e.g., RdInv* or Inv*) has snooped all peer
agents and gathered all snoop responses. When a given request's
tracker state is writeback buffered (WbBuffered), the home agent
may send a data response. When the request's tracker state is
DataSent (indicating the home agent has already sent a data
response) or DataXfrd (indicating a peer transferred a copy of the
line), the home agent may send the completion response.
[0072] In instances such as those described above, a home agent may
send data and completion responses before all snoop responses have
been gathered. The HPI interface allows these "early" responses.
When sending early data and completions, the home agent may gather
all outstanding snoop responses before releasing the resource it
allocated for the request. The home agent can also continue
blocking further standard requests to the same address until all
snoop responses have been gathered, then releasing the resource. A
home agent sending a response message from a Busy or WbBuffered
state can use a sub-action table (e.g., included in a set of
protocol tables embodying the formal specification of the HPI
Coherence protocol) for which message to send and use a sub action
table for how to update the directory state, among other examples.
In some cases, an early completion can be performed without
pre-allocation by a home node.
[0073] In one embodiment, HPI Coherence protocol can omit the use
of either or both pre-allocated home resources and ordered request
channels. In such implementations, certain messages on the HPI RSP
communication channel can be ordered. For instance, specifically
"ordered completion" and "forward response" messages, can be
provided, that can be sent from the home agent to the coherence
agent. Home agents can send an ordered completion (CmpO or
Data_*_CmpO) for all coherent read and invalidation requests (as
well as other requests, such as a NonSnpRd requests, that are not
involved in cache-coherence conflicts).
[0074] Home agents can send forward responses (FwdCnfltO) to
coherence agents that send forward requests (RspCnflt) to indicate
a conflict. A coherence agent can generate a forward request
whenever it has an outstanding read or invalidation request and
detects an incoming snoop request to the same cache line as the
request. When the coherence agent receives the forward response, it
checks the current state of the outstanding request to determine
how to process the original snoop. The home agent can sent the
forward response to be ordered with a complete (e.g., CmpO or
Data_*_CmpO). The coherence agent can utilize information included
in the snoop to aid the coherence agent in processing a forward
response. For instance, a forward response may not include any
"type" information and no RTID. The nature of the forward response
can be derived from information obtained from the preceding
snoop(s). Further, a coherence agent may block outstanding snoop
requests when all of its "forward resources" are waiting for
forward responses. In some implementations, each coherence agent
can be designed to have at least one forward resource.
[0075] In some implementations, communication fabric requirements
can be upon the Routing Layer. In one embodiment, the HPI Coherence
protocol has one communication fabric requirement that is specific
to the Routing Layer. The coherence protocol can depend upon the
routing layer to convert a fanout snoop (SnpF* opcodes--Snoop (SNP)
Channel Messages") into the appropriate snoops for all of the
request's peers in the fanout set of Coherence Agents. The fanout
set is a configuration parameter of the Routing Layer that is
shared by the Protocol Layer. In this coherence protocol
specification it is described as a Home Agent configuration
parameter.
[0076] In some implementations above, the HPI Coherence Protocol
can utilizes four of the virtual channels: REQ, WB, SNP, and RSP.
The virtual channels can be used to unwind dependency cycles and
avoid deadlock. In one embodiment, every message can be delivered
without duplication on all virtual channels and an ordering
requirement upon the RSP virtual channel.
[0077] In some implementations, the communication fabric can be
configured to preserve an ordering among certain completion
messages and the FwdCnfltO message. The completion messages are the
CmpO message and any data message with CmpO attached (Data_*_CmpO).
Together, all of these messages are the "ordered completion
responses." The conceptual requirement between ordered completion
responses and the FwdCnfltO message is that a FwdCnfltO does not
"pass" an ordered completion. More specifically, if a home agent
sends an ordered completion response followed by a FwdCnfltO
message and both messages are destined for the same coherence
agent, then the communication fabric delivers the ordered
completion response before the FwdCnfltO, among other potential
examples.
[0078] It should be appreciated that while some examples of the
protocol flow are disclosed herein, the described examples are
merely intended to give an intuitive feel for the protocol and do
not necessarily cover all possible scenarios and behaviors the
protocol may exhibit.
[0079] A conflict may occur when requests to the same cache-line
address from more than one coherence agent occur around the same
time. As a specific example, a conflict can occur when a snoop for
a coherence agent's standard request arrives at a peer coherence
agent with an outstanding request to the same address. Because each
snoop may end up in a conflict, a single request can have multiple
conflicts. Resolving conflicts may be a coordinated effort among
the home agent, the coherence agents, and the communication fabric.
However, the primary responsibility lies with the coherence agents
detecting conflicting snoops.
[0080] In one embodiment, home agents, coherence agents, and
communication fabric can be configured to assist in successfully
resolving conflicts. For example, home agents may have outstanding
snoops for only one request per address at a time, such that, for a
given address, a home agent may have outstanding snoops for only
one request. This can serve to exclude the possibility of race
conditions involving two requests conflicting with each other. It
can also ensure that a coherence agent will not see another snoop
to the same address after it has detected a conflict but not yet
resolved it.
[0081] In another example, when a coherence agent processes a snoop
with an address matching an active standard request, it can
allocates a forward resource and sends a forward request to the
home agent. A coherence agent with an outstanding standard request
that receives a snoop to the same address can responds with a
RspCnflt snoop response. This response can be a forward request to
the home agent. Because the message is a request, before sending it
the coherence agent can allocate a resource to handle the response
that the home agent will send. (The coherence protocol allows
blocking conflicting snoops when the coherence agent has run out of
forward resources, in some instances.) The coherence agent may
store information about the conflicting snoop to use when
processing the forward response. After detecting a conflict and
until processing the forward response, a coherence agent may be
guaranteed to not see another snoop to the same address.
[0082] In some examples, when a home agent receives a forward
request, it does not record the snoop response. Instead, the home
agent can send a forward response to the conflicting coherence
agent. A forward request (RspCnflt), in one example, looks like a
snoop response but the home agent does not treat it as one. It does
not record the message as a snoop response, but instead sends a
forward response. Specifically, for every forward request
(RspCnflt) a home agent receives, it sends a forward response
(FwdCnfltO) to the requesting coherence agent.
[0083] The HPI Communication Fabric orders forward responses and
ordered completions between the home agent and the targeted
coherence agent. The fabric can thereby serve to differentiate an
early conflict from a late conflict at the conflicting coherence
agent. From a system-level perspective, an early conflict occurs
when a snoop encounters a request that the home agent has not yet
processed, and a late conflict occurs when a snoop encounters a
request that the home agent has already processed. From a home
agent's perspective, an early conflict is when a snoop for the
currently active request encounters a request that the home agent
has not yet received or started processing, and a late conflict is
when the snoop encounters a request it has already processed. In
other words, a late conflict is with a request to which the home
agent has already sent a completion response. Thus, when a home
agent receives a forward request for a late conflict, it will have
already sent the completion response to the conflicting agent's
outstanding request. By ordering the forward responses and ordered
completion responses from home agent to the coherence agent, the
coherence agent can determine whether the conflict was early or
late by the processing state of its conflicting request.
[0084] When a coherence agent receives a forward response, it uses
the state of its conflicting request to determine whether the
conflict was early or late and when to process the original snoop.
Because of the communication fabric's ordering requirement, the
state of the conflicting request indicates whether the conflict was
early or late. If the request state indicates the completion has
been received then it was a late conflict, otherwise it was an
early conflict. Alternatively, if the request state indicates the
request is still waiting for its response(s) then it was an early
conflict, otherwise it was a late conflict. The type of conflict
determines when to process the snoop: From a coherence agent's
perspective, an early conflict means the snoop is for a request
being processed before the agent's conflicting request, and a late
conflict means the snoop is for a request being processed after the
agent's conflicting request. Given that ordering, for an early
conflict, the coherence agent immediately processes the original
snoop; and for a late conflict, the coherence agent waits until the
conflicting request has received its data (for reads) and its
processor has had an opportunity to act upon the finished request
before processing the snoop. When the conflicting snoop is
processed, the coherence agent will generate a snoop response for
the home agent to finally record.
[0085] All conflicts with writeback requests can be late conflicts.
A late conflict from the coherence agent's perspective is when the
agent's request is processed before the snoop's request. By this
definition all conflicts with writeback requests can be treated as
late conflicts because the writeback is processed first. Otherwise,
data consistency and coherency could be violated if the home agent
were to process the request before the writeback commits to memory.
Because all conflicts with writebacks are deemed late conflicts,
coherence agents can be configured to block conflicting snoops
until an outstanding writeback request completes. Further,
writebacks can also block the processing of forwards. Blocking
forwards by an active writeback can also be implemented as a
protocol requirement for supporting uncacheable stores, among other
examples.
[0086] When a coherence agent receives a request to snoop its
cache, it can first check if the coherence protocol will allow it,
and then it may process the snoop and generate a response. One or
more state tables can be defined within a set of state tables that
defines the protocol specification. One or more state table can
specify when a coherence agent may process a snoop and whether it
will snoop the cache or instead generate a conflict forward
request. In one example, there are two conditions under which a
coherence agent processes a snoop. The first condition is when the
coherence agent has a REQ request (Rd* or Inv*) to the snoop
address and it has an available forward resource. In this case, the
coherence agent must generate a forward request (RspCnflt). The
second condition is when the coherence agent does not have a REQ,
Wb*, or EvctCln request to the snoop address. A state table can
define how a coherence agent is to process the snoop in accordance
with such respective conditions. In one example, under other
conditions, the coherence agent can block the snoop until either a
forward resource becomes available (first condition) or the
blocking Wb* or EvctCln receives its CmpU response (second
condition). Note that NonSnp* requests may not affect snoop
processing and a coherence agent can disregard NonSnp* entries when
determining how to process or block a snoop.
[0087] When generating a forward request, a coherence agent can
reserve a resource for the forward response. The HPI Coherence
protocol, in one example, may not require a minimum number of
forward response resources (beyond having at least one) and can
allow a coherence agent to block snoops when it has no forward
response resources available.
[0088] How a coherence agent processes a snoop in its cache can
depend upon the snoop type and current cache state. For a given
snoop type and cache state, however, there may be many allowed
responses. For example, a coherence agent with a full modified line
that receives a non-conflicting SnpMigr (or is processing a forward
response after a SnpMigr) may do any of the following: Downgrade to
S, send implicit writeback to Home and Data_F to requestor;
Downgrade to S, send implicit writeback to Home; Downgrade to I,
send Data_M to requestor; Downgrade to I, send implicit writeback
to Home and Data_E to requestor; Downgrade to I, send implicit
writeback to Home; among potentially other examples.
[0089] The HPI Coherence protocol allows a coherence agent to store
modified lines with partial masks in its cache. However, all rows
in for M copies can require a Full or Empty mask. The HPI Coherence
protocol, in one example, may restrict implicit writeback of
partial lines. A coherence agent wishing to evict a partial M line
due to a snoop request (or forward response) can first initiate an
explicit writeback and block the snoop (or forward) until the
explicit writeback is completed.
[0090] Saving information for forward responses: The HPI Coherence
Protocol, in one embodiment, allows a coherence agent to store
forward response information separate from the outgoing request
buffer (ORB). Separating the information allows the ORB to release
ORB resources and RTID when all responses are gathered, regardless
of the entry being involved in a conflict. State tables can be
utilized to specify what information to store for forward responses
and under what conditions.
[0091] Forward responses in the HPI Coherence protocol can contain
the address, the requesting agent's NID, and the home TID. It does
not contain the original snoop type or the RTID. A coherence agent
may store the forward type and the RTID if it wishes to use them
with the forward response, and it may use the address to match the
incoming forward response with the proper forward entry (and to
generate the home NID). Storing the forward type may be optional.
If no type is stored, the coherence agent can treat a forward
response as having FwdInv type. Likewise, storing the RTID can be
optional and may only occur when the coherence agent is to support
cache-to-cache transfers when processing forward responses.
[0092] As noted above, coherence agents can generate a forward
request when a snoop request conflicts with an outstanding request.
Forward requests target the home agent that generated the snoop,
which can be determined from the address in the snoop request.
Thus, the destNID can identify a home agent. The forward request
can also include the TID for the home agent's resource allocated to
process the original request and the NID of the coherence agent
generating the forward request.
[0093] In one embodiment, a coherence agent can block forwards for
writeback requests to maintain data consistency. Coherence agents
can also use a writeback request to commit uncacheable (UC) data
before processing a forward and can allow the coherence agent to
writeback partial cache lines instead of protocol supporting a
partial implicit writeback for forwards. Indeed, in one embodiment,
a coherence agent can be allowed to store modified lines with
partial masks in its cache (although M copies are to include a Full
or Empty mask).
[0094] In one example, early conflicts may be resolved by a forward
response encountering an outstanding standard request before it has
received any response. A corresponding protocol state table can
specify, in one example, that a forward response can be processed
as long as the standard request entry is still in ReqSent state.
Late conflicts can be resolved by a forward response arriving after
the outstanding request has received its completion response. When
this occurs either the request will have finished (already received
its data or was an Inv* request) or the entry is in its RcvdCmp
state. If the request is still waiting for its data, then the
coherence agent must block the forward until the data is received
(and used). If the conflicting Rd* or Inv* request has finished,
then the forward response may be processed as long as the coherence
agent has not initiated an explicit writeback of the cache line. It
can be permissible for a coherence agent to initiate an explicit
writeback while it has a forward response (or snoop request) to the
same address, thus allowing partial lines (e.g. Snoop Requests to
Partially Modified Lines") or uncacheable stores to be properly
committed to memory.
[0095] Turning to FIG. 7, a first example is illustrated of an
example conflict management scheme. A first cache (or coherence)
agent 705 can send a read request for a particular line of data to
home agent 710 resulting in a read of memory 715. Shortly after the
read request by cache agent 705, another cache agent 720 makes a
request for ownership (RFO) of the same line. However, the home
agent 710 has sent the Data_S_CmpO to the first cache agent 705
prior to receiving the RFO from cache agent 720. The RFO can result
in a snoop (SnpFO) being sent to the cache agent 705 (as well as
other cache agents), the snoop being received by the first cache
agent 705 prior to receiving the complete Data_S_CmpO. The cache
agent 705, upon receiving the snoop SnpO can identify a potential
conflict involving the line of memory requested in its original
read request and can notify the home agent 710 of the conflict by
responding to the SnpO with a forward responses conflict message
(RspCnflt). The home agent 710 can respond to the forward response
RspCnflt by sending a forward response (FwdCnfltO). The cache agent
705 can then receive the shared data complete Data_S_CmpO and
transition from an I state to S state. The forward response
FwdCnfltO can then be received by the cache agent 705 and cache
agent 705 can determine how to respond to the forward response
message FwdClfltO based on the snoop SnpFO that triggered the
sending of the forward response RspCnflt. In this example, the
cache agent 705 can consult a protocol state table, for instance,
to determine a response to the forward response message FwdClfltO.
In the particular example of FIG. 7, the cache agent 705 can
transition to an F-state and send the S-copy of the data it
received from the home agent 710 in the Data_S_CmpO message to the
second cache agent 720 in a Data_F message. The first cache agent
705 can also send a response message RspFwdS to the home agent 710
notifying the home agent 710 that the first cache agent has shared
its copy of the data with the second cache agent.
[0096] In another illustrative example, shown in the simplified
flow diagram of FIG. 8, the first cache agent 705 can send a
request for ownership (RFO) of a particular line of memory to the
home agent 710. Shortly thereafter, a second cache agent can send a
RdInvOwn message to the home agent 710 as a request for the same
line of memory in an M state. In connection with the RFO message
from the first cache agent 705, the home agent 710 can send a snoop
(SnpFO) to the second cache agent 720 which the second cache agent
720 can identify as a potential conflict involving the line of
memory subject to both the RFO and RdInvOwn requests. Accordingly,
the second cache agent 720 can send a forward request RspCnflt to
the home agent 720. The home agent 720 responds to the second cache
agent's 720 forward request with a forward response. The second
cache agent 720 determines a response to the forward response based
on information contained in the original snoop SnpFO. In this
example, the second cache agent 720 responds with a snoop response
RspI indicating that the second cache agent 720 is in an I-state.
The home agent 710 receives the snoop response RspI and determines
that it is appropriate to send the data complete exclusive
(Data_E_CmpO) to the first cache agent 705, which causes the first
cache agent to transition to an E state. With the complete sent,
the home agent 710 can then begin responding to the second cache
agent's RdInvOwn request, beginning with a snoop request SnpInvO of
the first cache agent 705. The first cache agent 705 can identify
that the snoop results in a request by the second cache agent 720
to obtain an exclusive M-state copy of the line. Consequently, the
first cache agent 705 transitions to the M state to send its copy
of the line as an M-state copy (with Data_M message) to the second
cache agent 720. Additionally, the first cache agent 705 also sends
a response message RspFwdI to indicate that the copy of the line
has been sent to the second cache agent 720 and that the first
cache agent has transitioned to an I-state (having given up
ownership of the copy to the second cache agent 720).
[0097] Turning next to the example of FIG. 9, another simplified
flowchart is shown. In this example, a cache agent 720 attempts to
request exclusive ownership of an uncacheable (UC) line without
receiving data (e.g., through a InvItoE message). A first cache
agent 705 send a competing message (RdInv) for the cache line in E
state. The HPI Coherence protocol can specify that if the requested
line was previously cached in M state, the line will be written to
memory before E data is delivered in response to the RdInv of the
first cache agent 705. The home agent 710 can send a complete
(CmpO) to the InvItoE request and send a snoop (SnpInv) to cache
agent 720 based on the RdInv request. If the cache agent 720
receives the snoop before the complete, the cache agent 720 can
identify that the snoop pertains to the same cache line as its
exclusive ownership request and indicate a conflict through a
forward requests RspCnflt. As in previous examples, the home agent
710 can be configured to respond to the forward request with a
forward response (FwdCnfltO). Multiple permissible responses may be
allowed to the forward response. For instance, the cache agent 720
can initiate an explicit writeback (e.g., WbMtoI) and block the
snoop (or forward) until the explicit writeback is completed (e.g.,
CmpU), as shown in the example of FIG. 9. The cache agent can then
complete the snoop response (RspI). The home agent 710 can then
process the RdInv request of the first cache agent 705 and return a
complete Data_E_CmpO, among other examples.
[0098] In examples, such as the example of FIG. 9, where a cache
agent receives a snoop when the agent has an outstanding read or
invalidation request to the same address and it has cached a
partial modified line (often referred to as a "buried-M"), the HPI
Coherence protocol, in one implementation, allows the agent to
either 1) perform an explicit writeback (partial) of the line while
blocking the snoop or 2) send a forward request (RspCnflt) to the
home agent. If (1) is chosen, the agent processes the snoop after
receiving the complete for the writeback. If (2) is chosen, it is
possible that the agent will receive forward response (FwdCnfltO)
while its outstanding read or invalidation request is still waiting
for responses and the agent still has a partial modified line. If
that is the case, the protocol allows the agent to block the
forward while performing an explicit writeback (partial) of the
line. During the writeback, the protocol guarantees the agent will
not receive responses for the outstanding read or invalidation
request. The mechanism described above (allowing coherence agents
to issue explicit writebacks and block snoops and forwards, even
when the agent has an outstanding read or invalidation request) is
also used to ensure partial or UC writes are posted to memory
before the writer acquires global observability.
[0099] Coherence agents use a two-step process for partial/UC
writes. First, they check if they have ownership of the cacheline
and issue an ownership (invalidation) request in the protocol if
they do not. Second, they perform the write. In the first step, if
they performed an ownership request, it is possible that the
request will conflict with other agents' requests for the line,
meaning the agent might receive a snoop while the ownership request
is outstanding. Per coherence protocol requirements, the agent will
issue a forward request for the conflicting snoop. While waiting
for the forward response, the agent may receive the ownership
request's completion, which grants ownership of the line to the
agent and allows the agent to initiate the writeback for the
partial/UC write. While this is occurring, the agent might receive
the forward response, which it is obligated to process also. The
coherence agent may not combine the two activities. The coherence
agent is to instead writeback the partial/UC write data separately
from processing the forward, and perform the writeback first. For
instance, a cache agent may use a writeback request to commit UC
data before processing forward and writeback partial cache lines,
among other examples and features.
[0100] HPI can be incorporated in any variety of computing devices
and systems, including mainframes, server systems, personal
computers, mobile computers (such as tablets, smartphones, personal
digital systems, etc.), smart appliances, gaming or entertainment
consoles and set top boxes, among other examples. For instance,
referring to FIG. 10, an embodiment of a block diagram for a
computing system including a multicore processor is depicted.
Processor 1000 includes any processor or processing device, such as
a microprocessor, an embedded processor, a digital signal processor
(DSP), a network processor, a handheld processor, an application
processor, a co-processor, a system on a chip (SOC), or other
device to execute code. Processor 1000, in one embodiment, includes
at least two cores--core 1001 and 1002, which may include
asymmetric cores or symmetric cores (the illustrated embodiment).
However, processor 1000 may include any number of processing
elements that may be symmetric or asymmetric.
[0101] In one embodiment, a processing element refers to hardware
or logic to support a software thread. Examples of hardware
processing elements include: a thread unit, a thread slot, a
thread, a process unit, a context, a context unit, a logical
processor, a hardware thread, a core, and/or any other element,
which is capable of holding a state for a processor, such as an
execution state or architectural state. In other words, a
processing element, in one embodiment, refers to any hardware
capable of being independently associated with code, such as a
software thread, operating system, application, or other code. A
physical processor (or processor socket) typically refers to an
integrated circuit, which potentially includes any number of other
processing elements, such as cores or hardware threads.
[0102] A core often refers to logic located on an integrated
circuit capable of maintaining an independent architectural state,
wherein each independently maintained architectural state is
associated with at least some dedicated execution resources. In
contrast to cores, a hardware thread typically refers to any logic
located on an integrated circuit capable of maintaining an
independent architectural state, wherein the independently
maintained architectural states share access to execution
resources. As can be seen, when certain resources are shared and
others are dedicated to an architectural state, the line between
the nomenclature of a hardware thread and core overlaps. Yet often,
a core and a hardware thread are viewed by an operating system as
individual logical processors, where the operating system is able
to individually schedule operations on each logical processor.
[0103] Physical processor 1000, as illustrated in FIG. 10, includes
two cores--core 1001 and 1002. Here, core 1001 and 1002 are
considered symmetric cores, i.e. cores with the same
configurations, functional units, and/or logic. In another
embodiment, core 1001 includes an out-of-order processor core,
while core 1002 includes an in-order processor core. However, cores
1001 and 1002 may be individually selected from any type of core,
such as a native core, a software managed core, a core adapted to
execute a native Instruction Set Architecture (ISA), a core adapted
to execute a translated Instruction Set Architecture (ISA), a
co-designed core, or other known core. In a heterogeneous core
environment (i.e. asymmetric cores), some form of translation, such
a binary translation, may be utilized to schedule or execute code
on one or both cores. Yet to further the discussion, the functional
units illustrated in core 1001 are described in further detail
below, as the units in core 1002 operate in a similar manner in the
depicted embodiment.
[0104] As depicted, core 1001 includes two hardware threads 1001a
and 1001b, which may also be referred to as hardware thread slots
1001a and 1001b. Therefore, software entities, such as an operating
system, in one embodiment potentially view processor 1000 as four
separate processors, i.e., four logical processors or processing
elements capable of executing four software threads concurrently.
As alluded to above, a first thread is associated with architecture
state registers 1001a, a second thread is associated with
architecture state registers 1001b, a third thread may be
associated with architecture state registers 1002a, and a fourth
thread may be associated with architecture state registers 1002b.
Here, each of the architecture state registers (1001a, 1001b,
1002a, and 1002b) may be referred to as processing elements, thread
slots, or thread units, as described above. As illustrated,
architecture state registers 1001a are replicated in architecture
state registers 1001b, so individual architecture states/contexts
are capable of being stored for logical processor 1001a and logical
processor 1001b. In core 1001, other smaller resources, such as
instruction pointers and renaming logic in allocator and renamer
block 1030 may also be replicated for threads 1001a and 1001b. Some
resources, such as re-order buffers in reorder/retirement unit
1035, ILTB 1020, load/store buffers, and queues may be shared
through partitioning. Other resources, such as general purpose
internal registers, page-table base register(s), low-level
data-cache and data-TLB 1051, execution unit(s) 1040, and portions
of out-of-order unit 1035 are potentially fully shared.
[0105] Processor 1000 often includes other resources, which may be
fully shared, shared through partitioning, or dedicated by/to
processing elements. In FIG. 10, an embodiment of a purely
exemplary processor with illustrative logical units/resources of a
processor is illustrated. Note that a processor may include, or
omit, any of these functional units, as well as include any other
known functional units, logic, or firmware not depicted. As
illustrated, core 1001 includes a simplified, representative
out-of-order (OOO) processor core. But an in-order processor may be
utilized in different embodiments. The OOO core includes a branch
target buffer 1020 to predict branches to be executed/taken and an
instruction-translation buffer (I-TLB) 1020 to store address
translation entries for instructions.
[0106] Core 1001 further includes decode module 1025 coupled to
fetch unit 1020 to decode fetched elements. Fetch logic, in one
embodiment, includes individual sequencers associated with thread
slots 1001a, 1001b, respectively. Usually core 1001 is associated
with a first ISA, which defines/specifies instructions executable
on processor 1000. Often machine code instructions that are part of
the first ISA include a portion of the instruction (referred to as
an opcode), which references/specifies an instruction or operation
to be performed. Decode logic 1025 includes circuitry that
recognizes these instructions from their opcodes and passes the
decoded instructions on in the pipeline for processing as defined
by the first ISA. For example, as discussed in more detail below
decoders 1025, in one embodiment, include logic designed or adapted
to recognize specific instructions, such as transactional
instruction. As a result of the recognition by decoders 1025, the
architecture or core 1001 takes specific, predefined actions to
perform tasks associated with the appropriate instruction. It is
important to note that any of the tasks, blocks, operations, and
methods described herein may be performed in response to a single
or multiple instructions; some of which may be new or old
instructions. Note decoders 1026, in one embodiment, recognize the
same ISA (or a subset thereof). Alternatively, in a heterogeneous
core environment, decoders 1026 recognize a second ISA (either a
subset of the first ISA or a distinct ISA).
[0107] In one example, allocator and renamer block 1030 includes an
allocator to reserve resources, such as register files to store
instruction processing results. However, threads 1001a and 1001b
are potentially capable of out-of-order execution, where allocator
and renamer block 1030 also reserves other resources, such as
reorder buffers to track instruction results. Unit 1030 may also
include a register renamer to rename program/instruction reference
registers to other registers internal to processor 1000.
Reorder/retirement unit 1035 includes components, such as the
reorder buffers mentioned above, load buffers, and store buffers,
to support out-of-order execution and later in-order retirement of
instructions executed out-of-order.
[0108] Scheduler and execution unit(s) block 1040, in one
embodiment, includes a scheduler unit to schedule
instructions/operation on execution units. For example, a floating
point instruction is scheduled on a port of an execution unit that
has an available floating point execution unit. Register files
associated with the execution units are also included to store
information instruction processing results. Exemplary execution
units include a floating point execution unit, an integer execution
unit, a jump execution unit, a load execution unit, a store
execution unit, and other known execution units.
[0109] Lower level data cache and data translation buffer (D-TLB)
1050 are coupled to execution unit(s) 1040. The data cache is to
store recently used/operated on elements, such as data operands,
which are potentially held in memory coherency states. The D-TLB is
to store recent virtual/linear to physical address translations. As
a specific example, a processor may include a page table structure
to break physical memory into a plurality of virtual pages.
[0110] Here, cores 1001 and 1002 share access to higher-level or
further-out cache, such as a second level cache associated with
on-chip interface 1010. Note that higher-level or further-out
refers to cache levels increasing or getting further way from the
execution unit(s). In one embodiment, higher-level cache is a
last-level data cache--last cache in the memory hierarchy on
processor 1000--such as a second or third level data cache.
However, higher level cache is not so limited, as it may be
associated with or include an instruction cache. A trace cache--a
type of instruction cache--instead may be coupled after decoder
1025 to store recently decoded traces. Here, an instruction
potentially refers to a macro-instruction (i.e. a general
instruction recognized by the decoders), which may decode into a
number of micro-instructions (micro-operations).
[0111] In the depicted configuration, processor 1000 also includes
on-chip interface module 1010. Historically, a memory controller,
which is described in more detail below, has been included in a
computing system external to processor 1000. In this scenario,
on-chip interface 101 is to communicate with devices external to
processor 1000, such as system memory 1075, a chipset (often
including a memory controller hub to connect to memory 1075 and an
I/O controller hub to connect peripheral devices), a memory
controller hub, a northbridge, or other integrated circuit. And in
this scenario, bus 1005 may include any known interconnect, such as
multi-drop bus, a point-to-point interconnect, a serial
interconnect, a parallel bus, a coherent (e.g. cache coherent) bus,
a layered protocol architecture, a differential bus, and a GTL
bus.
[0112] Memory 1075 may be dedicated to processor 1000 or shared
with other devices in a system. Common examples of types of memory
1075 include DRAM, SRAM, non-volatile memory (NV memory), and other
known storage devices. Note that device 1080 may include a graphic
accelerator, processor or card coupled to a memory controller hub,
data storage coupled to an I/O controller hub, a wireless
transceiver, a flash device, an audio controller, a network
controller, or other known device.
[0113] Recently however, as more logic and devices are being
integrated on a single die, such as SOC, each of these devices may
be incorporated on processor 1000. For example in one embodiment, a
memory controller hub is on the same package and/or die with
processor 1000. Here, a portion of the core (an on-core portion)
1010 includes one or more controller(s) for interfacing with other
devices such as memory 1075 or a graphics device 1080. The
configuration including an interconnect and controllers for
interfacing with such devices is often referred to as an on-core
(or un-core configuration). As an example, on-chip interface 1010
includes a ring interconnect for on-chip communication and a
high-speed serial point-to-point link 1005 for off-chip
communication. Yet, in the SOC environment, even more devices, such
as the network interface, co-processors, memory 1075, graphics
processor 1080, and any other known computer devices/interface may
be integrated on a single die or integrated circuit to provide
small form factor with high functionality and low power
consumption.
[0114] In one embodiment, processor 1000 is capable of executing a
compiler, optimization, and/or translator code 1077 to compile,
translate, and/or optimize application code 1076 to support the
apparatus and methods described herein or to interface therewith. A
compiler often includes a program or set of programs to translate
source text/code into target text/code. Usually, compilation of
program/application code with a compiler is done in multiple phases
and passes to transform hi-level programming language code into
low-level machine or assembly language code. Yet, single pass
compilers may still be utilized for simple compilation. A compiler
may utilize any known compilation techniques and perform any known
compiler operations, such as lexical analysis, preprocessing,
parsing, semantic analysis, code generation, code transformation,
and code optimization.
[0115] Larger compilers often include multiple phases, but most
often these phases are included within two general phases: (1) a
front-end, i.e. generally where syntactic processing, semantic
processing, and some transformation/optimization may take place,
and (2) a back-end, i.e. generally where analysis, transformations,
optimizations, and code generation takes place. Some compilers
refer to a middle, which illustrates the blurring of delineation
between a front-end and back end of a compiler. As a result,
reference to insertion, association, generation, or other operation
of a compiler may take place in any of the aforementioned phases or
passes, as well as any other known phases or passes of a compiler.
As an illustrative example, a compiler potentially inserts
operations, calls, functions, etc. in one or more phases of
compilation, such as insertion of calls/operations in a front-end
phase of compilation and then transformation of the
calls/operations into lower-level code during a transformation
phase. Note that during dynamic compilation, compiler code or
dynamic optimization code may insert such operations/calls, as well
as optimize the code for execution during runtime. As a specific
illustrative example, binary code (already compiled code) may be
dynamically optimized during runtime. Here, the program code may
include the dynamic optimization code, the binary code, or a
combination thereof.
[0116] Similar to a compiler, a translator, such as a binary
translator, translates code either statically or dynamically to
optimize and/or translate code. Therefore, reference to execution
of code, application code, program code, or other software
environment may refer to: (1) execution of a compiler program(s),
optimization code optimizer, or translator either dynamically or
statically, to compile program code, to maintain software
structures, to perform other operations, to optimize code, or to
translate code; (2) execution of main program code including
operations/calls, such as application code that has been
optimized/compiled; (3) execution of other program code, such as
libraries, associated with the main program code to maintain
software structures, to perform other software related operations,
or to optimize code; or (4) a combination thereof.
[0117] While the present invention has been described with respect
to a limited number of embodiments, those skilled in the art will
appreciate numerous modifications and variations therefrom. It is
intended that the appended claims cover all such modifications and
variations as fall within the true spirit and scope of this present
invention.
[0118] A design may go through various stages, from creation to
simulation to fabrication. Data representing a design may represent
the design in a number of manners. First, as is useful in
simulations, the hardware may be represented using a hardware
description language or another functional description language.
Additionally, a circuit level model with logic and/or transistor
gates may be produced at some stages of the design process.
Furthermore, most designs, at some stage, reach a level of data
representing the physical placement of various devices in the
hardware model. In the case where conventional semiconductor
fabrication techniques are used, the data representing the hardware
model may be the data specifying the presence or absence of various
features on different mask layers for masks used to produce the
integrated circuit. In any representation of the design, the data
may be stored in any form of a machine readable medium. A memory or
a magnetic or optical storage such as a disc may be the machine
readable medium to store information transmitted via optical or
electrical wave modulated or otherwise generated to transmit such
information. When an electrical carrier wave indicating or carrying
the code or design is transmitted, to the extent that copying,
buffering, or re-transmission of the electrical signal is
performed, a new copy is made. Thus, a communication provider or a
network provider may store on a tangible, machine-readable medium,
at least temporarily, an article, such as information encoded into
a carrier wave, embodying techniques of embodiments of the present
invention.
[0119] A module as used herein refers to any combination of
hardware, software, and/or firmware. As an example, a module
includes hardware, such as a micro-controller, associated with a
non-transitory medium to store code adapted to be executed by the
micro-controller. Therefore, reference to a module, in one
embodiment, refers to the hardware, which is specifically
configured to recognize and/or execute the code to be held on a
non-transitory medium. Furthermore, in another embodiment, use of a
module refers to the non-transitory medium including the code,
which is specifically adapted to be executed by the microcontroller
to perform predetermined operations. And as can be inferred, in yet
another embodiment, the term module (in this example) may refer to
the combination of the microcontroller and the non-transitory
medium. Often module boundaries that are illustrated as separate
commonly vary and potentially overlap. For example, a first and a
second module may share hardware, software, firmware, or a
combination thereof, while potentially retaining some independent
hardware, software, or firmware. In one embodiment, use of the term
logic includes hardware, such as transistors, registers, or other
hardware, such as programmable logic devices.
[0120] Use of the phrase `configured to,` in one embodiment, refers
to arranging, putting together, manufacturing, offering to sell,
importing and/or designing an apparatus, hardware, logic, or
element to perform a designated or determined task. In this
example, an apparatus or element thereof that is not operating is
still `configured to` perform a designated task if it is designed,
coupled, and/or interconnected to perform said designated task. As
a purely illustrative example, a logic gate may provide a 0 or a 1
during operation. But a logic gate `configured to` provide an
enable signal to a clock does not include every potential logic
gate that may provide a 1 or 0. Instead, the logic gate is one
coupled in some manner that during operation the 1 or 0 output is
to enable the clock. Note once again that use of the term
`configured to` does not require operation, but instead focus on
the latent state of an apparatus, hardware, and/or element, where
in the latent state the apparatus, hardware, and/or element is
designed to perform a particular task when the apparatus, hardware,
and/or element is operating.
[0121] Furthermore, use of the phrases `to,` `capable of/to,` and
or `operable to,` in one embodiment, refers to some apparatus,
logic, hardware, and/or element designed in such a way to enable
use of the apparatus, logic, hardware, and/or element in a
specified manner. Note as above that use of to, capable to, or
operable to, in one embodiment, refers to the latent state of an
apparatus, logic, hardware, and/or element, where the apparatus,
logic, hardware, and/or element is not operating but is designed in
such a manner to enable use of an apparatus in a specified
manner.
[0122] A value, as used herein, includes any known representation
of a number, a state, a logical state, or a binary logical state.
Often, the use of logic levels, logic values, or logical values is
also referred to as 1's and 0's, which simply represents binary
logic states. For example, a 1 refers to a high logic level and 0
refers to a low logic level. In one embodiment, a storage cell,
such as a transistor or flash cell, may be capable of holding a
single logical value or multiple logical values. However, other
representations of values in computer systems have been used. For
example the decimal number ten may also be represented as a binary
value of 1010 and a hexadecimal letter A. Therefore, a value
includes any representation of information capable of being held in
a computer system.
[0123] Moreover, states may be represented by values or portions of
values. As an example, a first value, such as a logical one, may
represent a default or initial state, while a second value, such as
a logical zero, may represent a non-default state. In addition, the
terms reset and set, in one embodiment, refer to a default and an
updated value or state, respectively. For example, a default value
potentially includes a high logical value, i.e. reset, while an
updated value potentially includes a low logical value, i.e. set.
Note that any combination of values may be utilized to represent
any number of states.
[0124] The embodiments of methods, hardware, software, firmware or
code set forth above may be implemented via instructions or code
stored on a machine-accessible, machine readable, computer
accessible, or computer readable medium which are executable by a
processing element. A non-transitory machine-accessible/readable
medium includes any mechanism that provides (i.e., stores and/or
transmits) information in a form readable by a machine, such as a
computer or electronic system. For example, a non-transitory
machine-accessible medium includes random-access memory (RAM), such
as static RAM (SRAM) or dynamic RAM (DRAM); ROM; magnetic or
optical storage medium; flash memory devices; electrical storage
devices; optical storage devices; acoustical storage devices; other
form of storage devices for holding information received from
transitory (propagated) signals (e.g., carrier waves, infrared
signals, digital signals); etc, which are to be distinguished from
the non-transitory mediums that may receive information there
from.
[0125] Instructions used to program logic to perform embodiments of
the invention may be stored within a memory in the system, such as
DRAM, cache, flash memory, or other storage. Furthermore, the
instructions can be distributed via a network or by way of other
computer readable media. Thus a machine-readable medium may include
any mechanism for storing or transmitting information in a form
readable by a machine (e.g., a computer), but is not limited to,
floppy diskettes, optical disks, Compact Disc, Read-Only Memory
(CD-ROMs), and magneto-optical disks, Read-Only Memory (ROMs),
Random Access Memory (RAM), Erasable Programmable Read-Only Memory
(EPROM), Electrically Erasable Programmable Read-Only Memory
(EEPROM), magnetic or optical cards, flash memory, or a tangible,
machine-readable storage used in the transmission of information
over the Internet via electrical, optical, acoustical or other
forms of propagated signals (e.g., carrier waves, infrared signals,
digital signals, etc.). Accordingly, the computer-readable medium
includes any type of tangible machine-readable medium suitable for
storing or transmitting electronic instructions or information in a
form readable by a machine (e.g., a computer).
[0126] The following examples pertain to embodiments in accordance
with this Specification. One or more embodiments may provide an
apparatus, a system, a machine readable storage, a machine readable
medium, and a method to send a coherence protocol message
corresponding to a particular cache line, identify a potential
conflict involving the particular cache line, and send a forward
request to a home agent to identify the potential conflict.
[0127] One or more examples can further provide receiving a snoop
corresponding to the particular cache line.
[0128] One or more examples can further provide identifying that
the snoop is received while a request is outstanding, and the
potential conflict is identified based on identifying that the
snoop is received while the request is outstanding.
[0129] One or more examples can further provide receiving a forward
response from the home agent based on the forward request.
[0130] One or more examples can further provide determining a
response to the forward response based at least in part on
attributes of the snoop.
[0131] In at least one example, the snoop corresponds to another
coherence protocol message by another agent corresponding to the
particular cache line, and the attributes of the snoop include
identification of the other agent, identification of a command
included in the other coherence protocol message, and a transaction
identifier of the other coherence protocol message.
[0132] In at least one example, the response to the forward
response includes a snoop response and the protocol layer logic is
further to send the snoop response to the home agent after
receiving a completion to the coherence protocol message.
[0133] In at least one example, the response comprises performing a
writeback to memory prior to sending a snoop response to the home
agent.
[0134] In at least one example, the particular cache line is
partially modified at the agent.
[0135] One or more examples can further provide receiving a
completion after receiving the forward response,
[0136] One or more examples can further provide receiving a
completion before receiving the forward response.
[0137] One or more examples can further provide allocating a
resource for responses to the request.
[0138] One or more examples can further provide allocating a
forward resource for a forward responses to the forward
request.
[0139] One or more embodiments may provide an apparatus, a system,
a machine readable storage, a machine readable medium, and a method
to receive a first coherence protocol request from a first cache
agent, send a snoop request to a second cache agent, wherein the
snoop request corresponds to the first coherence protocol request,
receive a forward request from the second cache agent corresponding
to the snoop request, wherein the forward request identifies a
potential conflict with the first coherence protocol request, and
send a forward response to the second cache agent in response to
the forward request.
[0140] One or more examples can further provide receiving another
coherence protocol request from the second cache agent, wherein the
first coherence protocol request and the other coherence protocol
request each pertain to a common cache line.
[0141] In at least one example, the other coherence protocol
request is received by the agent before the first coherence
protocol request, and the protocol layer logic is further to
process the other coherence protocol request and return a complete
message to the other coherence protocol request.
[0142] One or more examples can further provide receiving a
response from the second agent to the forward response and generate
a complete to the first coherence protocol request upon receiving
the response to the forward response.
[0143] In at least one example, the agent comprises a home
agent.
[0144] One or more examples can further provide system with an
interconnect fabric, a home agent to service requests to a coherent
memory space, and a cache agent communicatively coupled to the home
agent via the interconnect fabric. The interconnect fabric can
ensure ordering of the response to the snoop and the complete for
the other coherency protocol request. The home agent can have a set
of resources and the set of resources are not pre-allocated to
cache agents in the system.
[0145] One or more examples can further provide an agent including
a layered protocol stack including a protocol layer, wherein the
protocol layer is to initiate an allocation of resources, without
home agent intervention, to process a first request in response to
the agent receiving the first request, and to initiate an
allocation of resources, without home agent intervention, to
process responses to a second request in response to the agent
transmitting the second request.
[0146] In at least one example, allocation of resources includes on
of: HTID, RNID, RTID, or a combination thereof.
[0147] In at least one example, allocation of resources includes on
of: resources to process snoop requests and forward responses.
[0148] One or more examples can further provide an agent including
a layered protocol stack including a protocol layer, wherein the
protocol layer is to imitate use of an ordered response channel to
perform conflict resolution.
[0149] One or more examples can further provide a coherence agent
including a layered protocol stack including a protocol layer,
wherein the protocol layer is to block a forward for a writeback
request to maintain data consistency.
[0150] In at least one example, the protocol layer is to initiate a
writeback request to commit unchacheable data before processing the
forward
[0151] In at least one example, the protocol layer is further to
support explicit writeback of partial cache lines.
[0152] Reference throughout this specification to "one embodiment"
or "an embodiment" means that a particular feature, structure, or
characteristic described in connection with the embodiment is
included in at least one embodiment of the present invention. Thus,
the appearances of the phrases "in one embodiment" or "in an
embodiment" in various places throughout this specification are not
necessarily all referring to the same embodiment. Furthermore, the
particular features, structures, or characteristics may be combined
in any suitable manner in one or more embodiments.
[0153] In the foregoing specification, a detailed description has
been given with reference to specific exemplary embodiments. It
will, however, be evident that various modifications and changes
may be made thereto without departing from the broader spirit and
scope of the invention as set forth in the appended claims. The
specification and drawings are, accordingly, to be regarded in an
illustrative sense rather than a restrictive sense. Furthermore,
the foregoing use of embodiment and other exemplarily language does
not necessarily refer to the same embodiment or the same example,
but may refer to different and distinct embodiments, as well as
potentially the same embodiment.
* * * * *