Xilinx Versal ACAP CPM4 User manual

Versal ACAP CPM Mode for
PCI Express
Product Guide
Vivado Design Suite
PG346 (v3.3) November 16, 2022
Xilinx is creating an environment where employees, customers, and
partners feel welcome and included. To that end, we’re removing non-
inclusive language from our products and related collateral. We’ve
launched an internal initiative to remove language that could exclude
people or reinforce historical biases, including terms embedded in our
software and IPs. You may still find examples of non-inclusive language
in our older products as we work to make these changes and align
with evolving industry standards. Follow this link for more
information.

Table of Contents
Chapter 1: Overview......................................................................................................4
Navigating Content by Design Process.................................................................................... 4
Introduction to the CPM4...........................................................................................................5
Introduction to the CPM5.........................................................................................................11
Use Modes..................................................................................................................................16
Licensing and Ordering............................................................................................................22
Chapter 2: Tandem Configuration.......................................................................23
Overview.....................................................................................................................................23
Enable the Tandem Configuration Solution...........................................................................24
Deliver Programming Images to Silicon................................................................................ 30
Tandem Configuration Performance......................................................................................31
Design Operation...................................................................................................................... 31
Loading Tandem PCIe for Stage 2...........................................................................................32
Known Issues and Limitations.................................................................................................36
Chapter 3: Product Specification......................................................................... 38
Performance.............................................................................................................................. 38
Minimum Device Requirements..............................................................................................39
Port Descriptions.......................................................................................................................40
Register Space......................................................................................................................... 122
Chapter 4: Designing with the Core................................................................. 124
Clocking.................................................................................................................................... 124
Resets........................................................................................................................................126
AXI4-Stream Interface Description....................................................................................... 126
Chapter 5: Design Flow Steps...............................................................................188
Customizing and Generating the CIPS IP Core for CPM4.................................................. 188
Customizing and Generating the CIPS IP Core for CPM5.................................................. 211
Appendix A: GT Selection and Pin Planning for CPM4............................237
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 2
Send Feedback

CPM4 GT Selection.................................................................................................................. 238
CPM4 Additional Considerations...........................................................................................240
CPM4 GTY Locations............................................................................................................... 242
Appendix B: GT Selection and Pin Planning for CPM5............................243
General Guidance for CPM5 ..................................................................................................244
Guidance for CPM5 in Specifically Identified Engineering Sample Devices.....................247
Guidance for CPM5 Migration from Specifically Identified Engineering Sample
Devices.................................................................................................................................251
CPM5 GTYP Locations............................................................................................................. 251
Appendix C: Debugging...........................................................................................253
Finding Help on Xilinx.com.................................................................................................... 253
PCIe Link Debug Enablement................................................................................................ 254
Appendix D: Using the High Speed Debug Port Over PCIe for
Design Debug........................................................................................................... 261
Overview...................................................................................................................................261
Implementing the HSDP-over-PCIe Example Design......................................................... 268
Appendix E: Limitations for CPM4 and CPM5..............................................273
Appendix F: Migrating............................................................................................. 275
Migrating to CPM4.................................................................................................................. 275
Migrating to CPM5.................................................................................................................. 279
Appendix G: Additional Resources and Legal Notices........................... 285
Xilinx Resources.......................................................................................................................285
Documentation Navigator and Design Hubs...................................................................... 285
References................................................................................................................................285
Revision History.......................................................................................................................286
Please Read: Important Legal Notices................................................................................. 287
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 3
Send Feedback

Chapter 1
Overview
Navigating Content by Design Process
Xilinx® documentaon is organized around a set of standard design processes to help you nd
relevant content for your current development task. All Versal® ACAP design process Design
Hubs and the Design Flow Assistant materials can be found on the Xilinx.com website. This
document covers the following design processes:
•System and Soluon Planning: Idenfying the components, performance, I/O, and data
transfer requirements at a system level. Includes applicaon mapping for the soluon to PS,
PL, and AI Engine. Topics in this document that apply to this design process include:
•Introducon to the CPM4
•Introducon to the CPM5
•Use Modes
•Embedded Soware Development: Creang the soware plaorm from the hardware
plaorm and developing the applicaon code using the embedded CPU. Also covers XRT and
Graph APIs. The topic in this document that applies to this design process include:
•Register Space
•Host Soware Development: Developing the applicaon code, accelerator development,
including library, XRT, and Graph API use. The topic in this document that applies to this
design process include:
•Register Space
•Hardware, IP, and Plaorm Development: Creang the PL IP blocks for the hardware
plaorm, creang PL kernels, funconal simulaon, and evaluang the Vivado® ming,
resource use, and power closure. Also involves developing the hardware plaorm for system
integraon. Topics in this document that apply to this design process include:
•Chapter 5: Design Flow Steps
•Appendix A: GT Selecon and Pin Planning for CPM4
•Appendix B: GT Selecon and Pin Planning for CPM5
•PCIe Link Debug Enablement
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 4
Send Feedback

Introduction to the CPM4
The integrated block for PCIe® Rev. 4.0 with DMA and CCIX Rev. 1.0 (CPM4) consists of two
PCIe® controllers, DMA features, CCIX features, and network on chip (NoC) integraon. The
Versal® ACAP CPM Mode for PCI Express enables direct access to the two high-performance,
independently customizable PCIe controllers. The CPM4 uses up to 16 Versal device GTY
channels over the XPIPE. Applicaon designs can also interface to the CPM4 with so logic and
clocking resources in the programmable logic. All feature references are applicable to both
instances of CPM4 PCIe controllers, with the following excepons:
• CPM4 PCIe Controller 0 supports up to x16 operaon, and CPM4 PCIe Controller 1 supports
up to x8 operaon.
• CPM4 PCIe Controller 1 with up to x8 support is available only when CPM4 PCIe Controller 0
is congured with 8 lanes or fewer.
• The CPM4 DMA features are supported only with CPM4 PCIe Controller 0. For more
informaon about CPM4 DMA features, see the Versal ACAP CPM DMA and Bridge Mode for
PCI Express Product Guide (PG347).
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 5
Send Feedback

Figure 1: CPM4 Sub-Block for PCIe Function (CPM4 PCIE)
CPM4 PCIe Controller #1
CPM4 PCIe Controller #0
ARM(*) CoreSight I/F Module
AXI4-ST
Data
Link
Layer
Mux/Demux
Physical
Layer
Cfg
Reg
Space
Integrated PCIe RAM
Clock &
Reset
cfg
XPIPE Hard I/F
Up to 2 Quads
Gen1 (2.5 GT/s)
16b@125 MHz
Gen2 (5.0 GT/s)
16b@250 MHz
Gen3 (8.0 GT/s)
16b@500 MHz
Gen4 (16.0 GT/s)
32b@500 MHz
Transaction
Layer
(VC0,
CCIX VC1)
APB
Block
Program-
ming
Hard I/F
XPIPEIM
Init
Ctrl
AXI4-ST
Data
Link
Layer
Mux/Demux
CPM4 DMA
DMA Core
Physical
Layer
Cfg
Reg
Space
Integrated PCIe RAM
Clock &
Reset
cfg
512b
512b
512b
cfg
AXI4-ST TX
Integrated DMA RAM
AXI4-ST RX
512b
512b
cfg
XPIPE Hard I/F
Up to 4 Quads
Gen1 (2.5 GT/s)
16b@125 MHz
Gen2 (5.0 GT/s)
16b@250 MHz
Gen3 (8.0 GT/s)
16b@500 MHz
Gen4 (16.0 GT/s)
32b@500 MHz
Transaction
Layer
(VC0,
CCIX VC1)
PCIe Core Clock In
PCIe Reset In
Global Event Inputs
XPIPEIM
Misc Port
512b
512b
AXI4-
MM
Bridge
DMA
AXI4-MM
Switch
Init
Ctrl
XPIPE
Static
Switch
PS
Internal
Hard I/F
Programming Register Space
512b
512b
cfg
512b
128b
128b
128b
512b 512b
32b
attr_*0
dbg_0_0 dbg_0_1
attr_dma_*
AXI4-MM
Master1
AXI4-MM
Master0
AXI4-MM
Slave0
Enhanced
AXI4-ST + CFG
+
Fabric I/F
64/128/256/
512b
CCIX TL Hard
I/F #1
256b
500/625/
781.25 MHz
Enhanced
AXI4-ST +
CFG + Misc
DMA I/Os
Fabric I/F
64/128/256/
512b
To
On Chip NOC
Hard I/F
62.5/125
CCIX TL
Hard I/F #0
256b
500/625/
781.25 MHz
To On Chip NOC
Hard I/F
32b AXI4-MM
Lite (MCAP)
To On Chip
NOC Hard I/F
32b AXI4-MM
Lite (MCAP)
attr_*1
dbg_1_0 dbg_1_1
32b
RX
TX
RX
TX
X22665-072320
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 6
Send Feedback

The CPM4 PCIe controllers are designed to the PCI Express Base Specicaon Revision 4.0 and
support the Gen4 data rate (16 GT/s per lane). They also support the Gen1 (2.5 GT/s per lane),
Gen2 (5 GT/s per lane) and Gen3 (8 GT/s per lane) data rates, and can interoperate with
components that are compliant with all versions of the PCI Express Base Specicaon.
The CPM4 PCIe controllers are available through the Vivado IP catalog in the Vivado Integrated
Design Environment (IDE). The combinaon of the CPM4 PCIe controllers, the GTY, and clocking
implement all layers of the PCI Express protocol, and the conguraon space and controller.
Protocol Layers
The layers of the protocol are the AXI4-Stream layer, the transacon layer, the data link layer and
the physical layer, and they are described in subsequent secons.
AXI4-Stream Layer
The AXI4-Stream layer implements Xilinx-specic requirements. In the transmit or outbound
direcon, the AXI4 layer interfaces the transacon layer with two AXI4-Stream interfaces. In the
receive or inbound direcon, the transacon layer output is forwarded to two AXI4-Stream
interfaces. Applicaon designs can aach to the AXI4-Stream interfaces, exchange informaon
with the Versal® ACAP CPM Mode for PCI Express encoded as a Xilinx-specic streaming
protocol implementaon, and run on top of the industry standard AXI4-Stream interface. The
CPM4 PCIe controllers support management of up to 256 (extended tag) or 768 (10 bit Tag)
outstanding customer iniated read requests, as part of the streaming protocol. The AXI4-Stream
layer supports:
•Recepon and transmission of address translaon services (ATS) invalid requests, ATS invalid
compleons, ATS page requests and ATS PRG response message TLPs, which enable ATS to
be implemented in the fabric logic.
• AXI4-Stream interface widths of 64 bits, 128 bits, 256 bits and 512 bits.
Transaction Layer
The transacon layer is the upper layer of the PCI Express architecture, and its primary funcon
is to accept, buer, and forward transacon layer packets (TLPs). TLPs communicate informaon
with the use of memory, I/O, conguraon, and message transacons. To maximize the eciency
of communicaon between devices, the transacon layer enforces PCI-compliant transacon
ordering rules and supports relaxed ordering (RO) of received transacons. The transacon layer
also manages TLP buer space through credit-based ow control. The transacon layer
implements built-in tag management for transmied non-posted transacons. It also implements
cut-through forwarding of transacons in the transmit (or outbound) direcon.
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 7
Send Feedback

CCIX Transaction Layer
The Cache Coherent Interconnect for Accelerators (CCIX) transacon layer requirements are
implemented by the oponal virtual channel 1 (VC1) in the design. Note that VC1 storage is in
addion to the PCI Express-compliant virtual channel 0 (VC0) storage. The CCIX transacon layer
interfaces with the CCIX protocol layer is implemented externally to the PCIe ports over the
CCIX transacon layer (ARM CXS) hard interface. For more informaon, see the Versal ACAP CPM
CCIX Architecture Manual (AM016).
Data Link Layer
The data link layer acts as an intermediate stage between the transacon layer and the physical
layer. Its primary responsibility is to provide a reliable mechanism for the exchange of informaon
between two components on a link. This includes data exchange (TLPs), error detecon and
recovery, inializaon services and the generaon and consumpon of data link layer packets
(DLLPs). DLLPs are used to transfer informaon between data link layers of two directly
connected components on the link. DLLPs convey informaon, such as power management, ow
control, and TLP acknowledgments. The data link layer supports 32 kilobyte replay buers and
the feature DLLP.
Physical Layer
The physical layer interfaces the data link layer with signaling technology for link data
interchange, and is subdivided into the logical sub-block and the electrical sub-block.
• The logical sub-block frames and de-frames TLPs and DLLPs. It also implements the link
training and status state machine (LTSSM), which handles link inializaon, training, and
maintenance. Scrambling and descrambling of data (for Gen1/Gen2/Gen3/Gen4 operaon) is
also performed in this sub-block.
• The electrical sub-block denes the input and output buer characteriscs that interface the
device to the PCIe link. The physical layer also supports lane reversal (for mul-lane designs)
and lane polarity inversion, as required by the PCI Express Base Specicaon 4.0 (hps://
www.pcisig.com/specicaons).
Data exchange with the other components on the link occurs over the serial lines of one or more
gigabit transceivers (GTs), which expose parallel interfaces at lower clock frequencies to the PCIe
controller. For Gen1, Gen2, Gen3 and Gen4 operaon, the physical layer is up-conguraon
capable in the downstream port mode only.
Standards
The CPM4 block adheres to the following standards:
•PCI Express Base Specicaon 4.0 Version 1.0, and Errata updates (available at hp://
pcisig.com/specicaons).
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 8
Send Feedback

• Cache Coherent Interconnect for Accelerators (CCIX) Transport Specicaon 1.0 (available at
hp://www.ccixconsorum.com).
Features
• Support for the following PCI Express architecture components:
○PCI Express Endpoint, Legacy Endpoint
○Root Port
○Switch Upstream and Downstream Ports
• 2.5 GT/s, 5.0 GT/s and 8.0 GT/s line rates with x1, x2, x4, x8, and x16 lane operaon.
• 16.0 GT/s line rate with x1, x2, x4, x8 lane operaon.
• CCIX support in PCI Express and EDR PHY Modes
○PCI Express support for Gen4x4, and Gen4x8
• Advanced Error Reporng (AER) and End-to-End CRC (ECRC)
• Two PCI Express virtual channels
○One PCI Express compliant virtual channel, eight trac classes
○One CCIX compliant virtual channel
• Support for mulple funcons and Single-Root IO Virtualizaon (SR-IOV)
○Up to 4 physical funcons
○Up to 252 virtual funcons
• PASID Prex capability supported
• Built-in lane reversal and receiver lane-lane de-skew
• 3 x 64-bit or 6 x 32-bit Base Address Registers (BARs) that are fully congurable
○Expansion ROM BAR supported
• All Interrupt types are supported:
○INTx
○32 mul-vector MSI capability
○MSI-X capability with up to 2048 vectors with oponal built-in vector tables
• Features that enable high-performance applicaons include:
○AXI4-Stream TLP Straddle on Requester Compleon Interface
○Address Translaon Services (ATS) and Page Request Interface (PRI) Messaging
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 9
Send Feedback

○Atomic Operaon Transacons Support
○Transacon Tag Scaling as Completer
○Flow Control Scaling
• Several ease of use and congurable features are supported:
○BAR and ID based ltering of received transacons
○Oponal ASPM support for endpoint port types only; ASPM is not supported for other
port types.
○Conguraon extend interface
○AXI4-Stream interfaces address align mode
○Debug and diagnoscs interface
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 10
Send Feedback

Introduction to the CPM5
Figure 2: CPM5 Sub-Block for PCIe Function (CPM5 PCIE)
CPM5 PCIe Controller #1
CPM5 PCIe Controller #0
ARM(*) CoreSight I/F Module
AXI4-ST
Data
Link
Layer
Mux/Demux
Physical
Layer
Cfg
Reg
Space
Integrated PCIe RAM
Clock &
Reset
cfg
XPIPE Hard I/F
Up to 2 Quads
Gen1 (2.5 GT/s)
16b@125 MHz
Gen2 (5.0 GT/s)
16b@250 MHz
Gen3 (8.0 GT/s)
16b@500 MHz
Gen4 (16.0 GT/s)
32b@500 MHz
Transaction
Layer
(VC0,
CCIX VC1)
APB
Block
Program-
ming
Hard I/F
XPIPEIM
Init
Ctrl
AXI4-ST
Data
Link
Layer
Mux/Demux
Physical
Layer
Cfg
Reg
Space
Integrated PCIe RAM
Clock &
Reset
cfg
512b
512b
cfg
XPIPE Hard I/F
Up to 4 Quads
Gen1 (2.5 GT/s)
16b@125 MHz
Gen2 (5.0 GT/s)
16b@250 MHz
Gen3 (8.0 GT/s)
16b@500 MHz
Gen4 (16.0 GT/s)
32b@500 MHz
Transaction
Layer
(VC0,
CCIX VC1)
PCIe Core Clock In
PCIe Reset In
Global Event Inputs
XPIPEIM
512b
Init
Ctrl
XPIPE
Static
Switch
PS
Internal
Hard I/F
Programming Register Space
512b
512b
cfg
512b
32b
attr_*0
dbg_0_0 dbg_0_1
attr_dma_*
Enhanced
AXI4-ST + CFG
+
Fabric I/F
CCIX TL Hard
I/F #1
256b
Enhanced
AXI4-ST +
CFG + Misc
DMA I/Os
Fabric I/F
64/128/256/
512b
To
On Chip NOC
Hard I/F
CCIX TL
Hard I/F #0
256b
To On Chip NOC
Hard I/F
32b AXI4-MM
Lite (MCAP)
To On Chip
NOC Hard I/F
32b AXI4-MM
Lite (MCAP)
attr_*1
dbg_1_0 dbg_1_1
32b
RX
TX
RX
TX
CPM5 DMA
DMA Core
512b
512b
512b
cfg
AXI4-ST TX AXI4-ST RX Misc Port
512b
AXI4-
MM
Bridge
DMA
AXI4-MM
Switch
128b
128b
128b
512b 512b
AXI4-MM
Master1
AXI4-MM
Master0
AXI4-MM
Slave0
CPM5 DMA
DMA Core
512b
512b
512b
cfg
AXI4-ST TX AXI4-ST RX Misc Port
512b
AXI4-
MM
Bridge
DMA
AXI4-MM
Switch
128b
128b
128b
512b 512b
AXI4-MM
Master1
AXI4-MM
Master0
AXI4-MM
Slave0
To
On Chip NOC
Or Fabric IF
X22665-072320
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 11
Send Feedback

Protocol Layers
The layers of the protocol are the AXI4-Stream layer, the transacon layer, the data link layer and
the physical layer, and they are described in subsequent secons.
AXI4-Stream Layer
The AXI4-Stream layer implements Xilinx-specic requirements. In the transmit or outbound
direcon, the AXI4 layer interfaces the transacon layer with two AXI4-Stream interfaces. In the
receive or inbound direcon, the transacon layer output is forwarded to two AXI4-Stream
interfaces. Applicaon designs can aach to the AXI4-Stream interfaces, exchange informaon
with the Versal® ACAP CPM Mode for PCI Express encoded as a Xilinx-specic streaming
protocol implementaon, and run on top of the industry standard AXI4-Stream interface. The
CPM4 PCIe controllers support management of up to 256 (extended tag) or 768 (10 bit Tag)
outstanding customer iniated read requests, as part of the streaming protocol. The AXI4-Stream
layer supports:
•Recepon and transmission of address translaon services (ATS) invalid requests, ATS invalid
compleons, ATS page requests and ATS PRG response message TLPs, which enable ATS to
be implemented in the fabric logic.
• AXI4-Stream interface widths of 64 bits, 128 bits, 256 bits, 512 bits, and 1024 bits.
Transaction Layer
The transacon layer is the upper layer of the PCI Express architecture, and its primary funcon
is to accept, buer, and forward transacon layer packets (TLPs). TLPs communicate informaon
with the use of memory, I/O, conguraon, and message transacons. To maximize the eciency
of communicaon between devices, the transacon layer enforces PCI-compliant transacon
ordering rules and supports relaxed ordering (RO) of received transacons. The transacon layer
also manages TLP buer space through credit-based ow control. The transacon layer
implements built-in tag management for transmied non-posted transacons. It also implements
cut-through forwarding of transacons in the transmit (or outbound) direcon.
CCIX Transaction Layer
The Cache Coherent Interconnect for Accelerators (CCIX) transacon layer requirements are
implemented by the oponal virtual channel 1 (VC1) in the design. Note that VC1 storage is in
addion to the PCI Express-compliant virtual channel 0 (VC0) storage. The CCIX transacon layer
interfaces with the CCIX protocol layer is implemented externally to the PCIe ports over the
CCIX transacon layer (ARM CXS) hard interface. For more informaon, see the Versal ACAP CPM
CCIX Architecture Manual (AM016).
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 12
Send Feedback

Data Link Layer
The data link layer acts as an intermediate stage between the transacon layer and the physical
layer. Its primary responsibility is to provide a reliable mechanism for the exchange of informaon
between two components on a link. This includes data exchange (TLPs), error detecon and
recovery, inializaon services and the generaon and consumpon of data link layer packets
(DLLPs). DLLPs are used to transfer informaon between data link layers of two directly
connected components on the link. DLLPs convey informaon, such as power management, ow
control, and TLP acknowledgments. The data link layer supports 32 kilobyte replay buers and
the feature DLLP.
Physical Layer
The physical layer interfaces the data link layer with signaling technology for link data
interchange, and is subdivided into the logical sub-block and the electrical sub-block.
• The logical sub-block frames and de-frames TLPs and DLLPs. It also implements the link
training and status state machine (LTSSM), which handles link inializaon, training, and
maintenance. Scrambling and descrambling of data (for Gen1/Gen2/Gen3/Gen4/Gen5
operaon) is also performed in this sub-block.
• The electrical sub-block denes the input and output buer characteriscs that interface the
device to the PCIe link. The physical layer also supports lane reversal (for mul-lane designs)
and lane polarity inversion, as required by the PCI Express Base Specicaon 5.0 (hps://
www.pcisig.com/specicaons).
Data exchange with the other components on the link occurs over the serial lines of one or more
gigabit transceivers (GTs), which expose parallel interfaces at lower clock frequencies to the PCIe
controller. For Gen1, Gen2, Gen3, Gen4, and Gen5 operaon, the physical layer is up-
conguraon capable in the downstream port mode only.
Standards
The CPM5 PCIe funconal mode adheres to the following standards:
• PCI Express Base Specicaon 5.0 Version 1.0, and Errata updates (available at hp://
pcisig.com/specicaons).
• Cache Coherent Interconnect for Accelerators (CCIX) Base Specicaon 1.1, and Errata/ECN
updates (available at hp://www.ccixconsorum.com).
• CCIX Transport Specicaon 1.0 (available at hp://www.ccixconsorum.com).
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 13
Send Feedback

Features
The CPM5 block contains two high-performance PCI Express ports that are independently
customizable. Designs can interface with programmable logic and clocking resources using fabric
interconnect.
All feature references are applicable to both PCI Express ports in the CPM5 block, with the
following excepon:
• PCIe Port 0 supports up to x16 operaon
• PCIe Port 1 supports up to x8 operaon
Features of the CPM5 block are listed below:
• PCI Express Architecture Components Supported include:
○PCI Express Endpoint, Legacy Endpoint
○Root Port
○Switch Upstream and Downstream Ports
• x1, x2, x4, x8, or x16 link widths
• Gen1, Gen2, Gen3, Gen4, or Gen5 link speeds (Gen5 support limited to x8)
• CCIX supported in PCI Express and EDR PHY Modes.
○PCI Express Support: Gen4x4, Gen4x8, Gen4x16, Gen5x4, Gen5x8
○EDR Support: x4 and x8 link width capability
• AXI4-Stream Interfaces to Programmable Logic (PL)
○Congurable 64-bit/128-bit/256-bit/512-bit/1024-bit interface data path widths (newly
added 1024 bit support)
○Congurable 62.5 MHz/125 MHz/250 MHz interface data path speeds
○Four Independent Iniator/Target, Request/Compleon AXI4-Streams
• Parity protecon on internal logic data paths and external data path interfaces
• Advanced Error Reporng (AER) and End-to-End CRC (ECRC)
• Two PCI Express Virtual Channels:
○One PCI Express Compliant Virtual Channel, eight Trac Classes
○One CCIX Compliant Virtual Channel
• Supports mulple Funcons and Single-Root IO Virtualizaon
○Up to 16 Physical Funcons
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 14
Send Feedback

○Up to 4080 Virtual Funcons
• PASID Prex Capability Supported.
• Built-in lane reversal and receiver lane-lane de-skew
• 3 x 64-bit or 6 x 32-bit Base Address Registers (BARs) that are fully congurable
○Expansion ROM BAR supported
• All Interrupt types are supported
○INTx
○32 mul-vector MSI capability
○MSI-X capability with up to 32k vectors with built-in MSI-X vector tables
• Built-in Iniator Read Request/Compleon Tag Manager
○Congurable 256 (Extended Tag); OR
○Up to 768 (Scaled Tag) outstanding Iniator Read Request Transacons supported
• CCIX Conguraon Capabilies
○Transport and Protocol DVSECs
• Features that enable high performance applicaons
○AXI4 Streaming TLP Straddle on Requester Compleon Interface
○Up to 1024 RX Compleon Header Credits and 64 KB RX Compleon Payload Space
○Relaxed Transacon Ordering in the Receive Data Path
○Address Translaon Services (ATS) and Page Request Interface (PRI) Messaging
○Atomic Operaon Transacons Support
○Transacon Tag Scaling as Completer
○Flow Control Scaling
• PCI Express Component Measurement and Authencaon (CMA) capability
• Several ease of use and congurable features are supported:
○BAR and ID based ltering of received transacons
○Oponal ASPM support for endpoint port types only; ASPM is not supported for other
port types.
○Conguraon extend interface
○AXI4-Stream interfaces address align mode
○Debug and diagnoscs interface
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 15
Send Feedback

Use Modes
All design use modes support Endpoint, Legacy Endpoint, and Root Port conguraons.
PCI Express Endpoint Use Modes
Illustrative Example of Basic Bus Mastering Endpoint
By far the most common use of the Versal® ACAP CPM Mode for PCI Express is to construct a
bus mastering Endpoint using a CPM PCIe controller. This use model is applicable to most
applicaons that interface the Endpoint port on the ACAP (on an add-in card) to a root complex
or that switch downstream port through a PCI Express connector. The following gure shows a
block diagram of the bus mastering Endpoint use case.
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 16
Send Feedback

Figure 3: Basic PCI Express Bus Mastering Endpoint Use Case
Initiator Interface
TX
A
R
B
I
T
E
R
Mem Wr
CplD
RX DATA
TX DATA
Bus Mastering
(DMA) Logic
Bridge to User Application
Mem Rd
Mem Rd/Wr
Control & Status Registers
RX
CplD
PCIe
CPM PCIe Controller
CONTROL LOGIC
Mem Rd
Mem Wr
D
E
M
U
X
CplD
CplD
Completer Interfaces Requester Interfaces
X22666-071620
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 17
Send Feedback

PCI Express Two Function Endpoint
The following gure shows the architecture of a two-funcon Endpoint design. The CPM PCIe
Controller is congured to enable two built-in funcon conguraon spaces. This use case
enables the applicaon device driver to access and control two disnct applicaons
independently. The user logic implements the DMA, control registers and applicaons.
Figure 4: Illustrative Example of Two Function Endpoint Use Case
FUN0
Application
Function #0
Arbiter/Demux
CPM PCIe Controller
DMA
Control
Regs
TX
Buf
RX
Buf
FUN1
Control
Regs
TX
Buf
RX
Buf
Application
Function #1
X22667-071620
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 18
Send Feedback

PCI Express Endpoint with SR-IOV
The following gure shows the CPM PCIe Controller congured as a SR-IOV capable Endpoint,
interfacing with the user design. This use case addresses requirements for up to four physical and
252 virtual funcons, and minimizes the so logic requirement to implement an SR-IOV
Endpoint.
Figure 5: Illustrative Example of Endpoint with SR-IOV Use Case
VM1
VM0
VF0
CPM PCIe Controller
DMA
Control
Regs
TX
Buf
RX
Buf
VF1
Control
Regs
TX
Buf
RX
Buf
Application
PF Driver
Guest OS
Application
RX
Buf TX
Buf
VF Driver
PF
Host
PCIe Link
VMM
X22668-071620
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 19
Send Feedback

PCI Express Endpoint with AXI4 Memory Mapped Interface
This use case describes a PCI Express Endpoint funconal unit that implements AXI4 Memory
Mapped (AXI-MM) interfaces. This funconal block implements a so logic bridge between the
nave AXI4-Stream interface on the CPM PCIe controller and AXI4 Memory Mapped
interconnect.
PCI Express Endpoint Using Tandem PROM
This use case addresses the ability to congure the ACAP in two stages and bring up the PCI
Express protocol in less than 100 ms aer power to the ACAP is stable. This is accomplished
through a staged conguraon ow.
PCI Express Endpoint Using Tandem PCIe
This use case addresses the ability to inially load fully congurable PCI Express protocol
soluon from a small external ROM, so as to meet the 100 ms conguraon requirement. A PCIe
link is formed with a Root Complex or Switch component, which is subsequently used to
download the design that congures the rest of the ACAP. In this case the PCIe link is used by
the user applicaon. This is accomplished through a staged conguraon ow.
PCI Express Root Port Use Mode
Basic PCI Express Root Complex Use Case
The following gure shows a PCI Express Root Complex in the simplest form consisng of a PCI
Express Root Port to an AXI4 memory mapped bridge interfaced with the interconnect. The
interconnect consists of an Arm®-based processor system (PS) containing most of the crical
blocks such as CPU, memory controller and other important peripherals. One of the goals of this
use case is to minimize ACAP so logic requirements.
Chapter 1: Overview
PG346 (v3.3) November 16, 2022 www.xilinx.com
CPM Mode for PCI Express 20
Send Feedback
This manual suits for next models
1
Table of contents
Other Xilinx PCI Card manuals