Transcript Xilinx Template (light) rev

Spartan-6 Memory Resources
Basic FPGA Architecture
Xilinx Training
Objectives
After completing this module you will be able to…
Fully utilize the Spartan-6 distributed and block memory
resources
Understand the features and limitations of the Spartan-6
dedicated memory controller block (MCB)
Use the Memory Interface Generator (MIG) to build your
custom memory controller and design an appropriate interface
to your off-chip memory component
Every Design Uses Memory
On-chip block RAM and ROM
Fast on-chip
memory
– Frequently used for…
• Finite State Machines
• FIFO
Spartan-6
• Large/local data storage
External memory for larger data storage
– Dedicated Memory Controller Block (MCB)
– Supports evolving controller standards
All memory solutions
– Must be fast and flexible
– Bitstream can pre-loaded with fixed data for a
ROM or RAM
External
memory
Memory Options
Distributed
RAM/SRL32
RAM / SRL 32
LOGIC
On-chip Block RAM
• Very granular, localized memory
• Minimal impact on logic routing
• Great for small FIFOs
Granularity
BRAM
• Efficient, on-chip blocks
• Ideal for mid-sized buffering
Fast Memory
Interfaces
Spartan-6
DRAM
DRAM • SDRAM
• DDR
• DDR3
• FCRAM
SRAM • RLDRAM
SRAM
• Sync SRAM
FLASH • DDR SRAM
• ZBT
• QDR
EEPROM FLASH
EEPROM
• Cost-effective bulk storage
• Various memory controller cores
• For large memory requirements
• Memory Controller Block
Capacity
Interfacing to External Memories
MCB supports many standards
– 1.5 V to 2.5 V
– Single or double data rate
– Different protocols
High performance
– With advanced ChipSync™ technology
• This is the Select I/O functionality that
enables the FPGA to be directly connected
to memory specific I/O standards
Spartan-6
Distributed RAM
Distributed LUT memory
– 64-bit blocks throughout the FPGA
available in 25% of the slices
– Single-port, dual-port, multi-port
– Can be used as 32-bit shift register
– Very fast (sub-nanosecond)
Ideal for small and fast memories
– Coefficient storage
– Small data buffers
– Small state machines
– Small FIFOs
– Shift registers
Slice3
Slice3
Logic
Logic RAM
Shift Register
Slice3
Slice3
Logic
Slice3
Logic
Logic RAM
Shift Register
Slice3
Logic RAM
Shift Register
Slice3
Slice3
Logic
Logic RAM
Shift Register
Slice3
Slice3
Logic
Slice3
Logic
R
A
M
R
A
M
Logic RAM
Shift Register
Slice3
Logic RAM
Shift Register
R
A
M
Distributed RAM Features
Distributed LUT memory
– Can be loaded by configuration
Synchronous (clocked) write operation
– But asynchronous (combinatorial) read
• Can make synchronous read when you
use the neighboring flip-flop
Slice3
Slice3
Logic
Logic RAM
Shift Register
Slice3
Slice3
Slice3
Logic RAM
Shift Register
Slice3
Slice3
Logic
Logic
Slice3
Logic
Logic RAM
Shift Register
Slice3
Logic RAM
Shift Register
Slice3
Logic
Logic
Slice3
Logic
R
A
M
R
A
M
Logic RAM
Shift Register
Slice3
Logic RAM
Shift Register
Easiest to build with the Core Generator
– Automatically builds the necessary input
decode and output multiplexer logic
Memories can be initialized
– Text I/O code in VHDL
– Coefficient file
– Or an initialization file placed in HDL code if inferring the memory
R
A
M
Block RAM Features
Multiple configuration options
– True dual-port, simple dual-port, single-port
18Kb Memory
Two independent ports address common data
– Individual address, clock, write enable, clock enable
– Independent widths for each port
Dual-Port
300-MHz operation when using data pipeline optionBRAM
– All operations are synchronous; all outputs are latched
– Data output has an optional internal pipeline register
• Faster clock rate, but increased latency
Byte-write enable
– Enhances processor memory interfacing
Load block RAM during configuration
– Reset during operation clears the registers, not the data content
Do not violate address setup time while enabled
– Disable memory when address timing might be unpredictable
– This also saves power
Spartan-6 FPGA Block RAM Block
Each block RAM block can be used as
18 Kb
Block
RAM
• One 18 Kb Block RAM
9 Kb
Block RAM
or
9 Kb
Block RAM
• Two independent 9 Kb block RAMs
True Dual-Port Block RAM
True dual-port flexibility
– Can perform read and write operations
simultaneously and independently on port A and
port B
– Each port has its own clock, enable, and write enable
– Every write also performs a read operation
Addr A
Wdata A
36
Rdata A
18 Kb
Memory
Array
• Read before write, write before read, or no output
change
– Simultaneous read + write or write + write
to the same location can cause data corruption
Port A
36
Addr B
36
Wdata B
Port B
Rdata B
• Make sure that the address and control signals are stable
during operation
Block RAM configurations
– One block RAM: 16Kx1, 8Kx2, 4Kx4, 2Kx9, 1Kx18, 512x36
– Or two independent 9K block RAMs: 8Kx1, 4Kx2, 2Kx4, 1Kx9, 512x18
– Each port can have its own depth x width within that range
36
Simple Dual-Port Block RAM
One read port and one write port
– Natural structure for FIFOs
Addr A
Port A
36
Wdata A
Allows widest implementation
• Up to a 72-bit read and write in one cycle
– Doubles the memory bandwidth per block
“Parity bits” are not dedicated for parity
– All byte-wide data has an extra ninth storage bit
– Can be used for parity or for any other purpose
– Parity generation / checking would need LUT
logic
Rdata A
18 Kb
Memory
Array
– 72-bit data width on 18K block RAM
– 36-bit data width for 9K block RAM
36
Addr B
36
Wdata B
Port B
Rdata B
36
Block RAM Configurations
Each 9 K
18 K
True dual-port
8Kx1, 4Kx2, 2Kx4,
1Kx9, 512x18
16Kx1, 8Kx2,
4Kx4, 2Kx9,
1Kx18, 512x36
Two fully independent
read and write operations
Simple dual-port
8Kx1, 4Kx2, 2Kx4,
1Kx9, 512x18,
256x36
16Kx1, 8Kx2,
4Kx4, 2Kx9,
1Kx18, 512x36,
256x72
1 read & 1 write port
Read AND write in 1 cycle
Single-port
8Kx1, 4Kx2, 2Kx4,
1Kx9, 512x18,
256x36
16Kx1, 8Kx2,
4Kx4, 2Kx9,
1Kx18, 512x36,
256x72
1 read & 1 write port
Read OR write in 1 cycle
Byte-Wide Write Enable
Controls which byte is being written
– One write enable for each byte of data and its parity bit
– Useful when interfacing with processors
clk
address
a0
data
ABCD
we[3:0]
memory @a0
output
 Byte-write operation
during write-first mode
1001
FFFF
AFFD
AFFD
– Bytes not being written will show
an undefined value on the
output
– The output truly reflects the new
memory content
– Writing always implies a read
• Be careful of reading when
writing
Spartan-6 FPGA Memory Capacity
Spartan-6 Device
Distr. RAM (Kb)
Block RAM (Kb)
18-Kb Block RAM Blocks
XC6SLX4
32
144
8
XC6SLX9
90
576
32
XC6SLX16
136
576
32
XC6SLX25
229
936
52
XC6SLX45
401
2,088
116
XC6SLX100
930
4,824
268
XC6SLX150
1,355
4,824
268
XC6SLX25T
229
936
52
XC6SLX45T
401
2,088
116
XC6SLX100T
930
4,824
268
XC6SLX150T
1,355
4,824
268
Memory Controller Block (MCB)
Spartan-6 has a New dedicated memory
controller block
– Up to four controllers per device
– Saves between 500 and 2000 LUTs and
registers versus a soft implementation
MCB
Blk
DDR
DDR2
DDR3
LP DDR
Why a hard block?
– Very common design component
– Benefits of a hard block
• Higher performance: 800 Mbps
• Lower cost: smaller than soft logic
• Lower power: compared to soft logic
– Easy to design
• Abstracts away complexity of memory interfacing
• CORE Generator™ tool / MIG wizard and EDK
support
Interface
Spartan-6
FPGA
DDR3 SDRAM
800 Mbps*
DDR2 SDRAM
800 Mbps*
DDR SDRAM
400 Mbps*
LP DDR
400 Mbps*
*For all speed grades, except -1L
MCB Features
Memory support
– DDR, DDR2, DDR3, LP DDR
standards
– Six 32-bit wide user ports
• Can be concatenated up to 128 bits
• Each port has 64-deep data FIFO and
4-deep command FIFO
Simple… but also programmable
– Controller options
• Set user interface, calibration,
addressing, and arbitration schemes
– Memory device options
• Control features and timing
parameters
Automatic calibration
– DQS centering
– DQ per-bit de-skew
– FPGA on-chip input termination
CMD FIFO 0
CMD FIFO 1
CMD FIFO 2
CMD FIFO 3
CMD FIFO 4
CMD FIFO 5
Arbiter Controller
32-bit
Bi-directional
32-bit
Bi-directional
32-bit
Uni-directional
32-bit
Uni-directional
32-bit
Uni-directional
32-bit
Uni-directional
Data Path
Dedicated Routing
Simple, multi-port user interface
Memory Controller Block
PHY
MCB Options
 Number of memory controllers
–
2 for medium-sized devices, 4 for the two largest devices
 I/O interface to a single external DRAM device: 4, 8, or 16 bits wide
 Internal “user interface” bus width is programmable: 32 to 128 bits
wide
bus LP DDR
DRAM size DRAM
width
x16
x8
128Mb
x4
x16
x8
256Mb
x4
x16
512Mb
x8
x4
x16
x8
1Gb
x4
x16
2Gb
x8
x4
4Gb
x16
DDR
DDR2
DDR3
LX4
M
C
B
3
M
C
LX16 B
1
M
C
B
4
M
C
B
3
LX100/T
M
M
C
C
B LX9 B
3
1
M
C
B
3
LX25/T
M
C
B
5
M
C
B
4
M
C
B
1
M
C
B
3
M
C
B
1
M
C
B
3
M
C
B
1
LX45/T
M
C
B
5
LX150/T
M
C
B
1
MCB Performance
Memory
Type
Max Theoretical Bandwidth
per Memory Controller Interface
Data Rate
Min
DDR
TBD *
DDR2
TBD *
DDR3
TBD *
LPDDR
TBD *
Max
400 Mbps
(200 MHz)
800 Mbps
(400 MHz)
800 Mbps
(400 MHz)
400 Mbps
(200 MHz)
4-bit
8-bit
16-bit
1.6 Gbps
3.2 Gbps
6.4 Gbps
3.2 Gbps
6.4 Gbps
12.8 Gbps
3.2 Gbps
6.4 Gbps
12.8 Gbps
1.6 Gbps
3.2 Gbps
6.4 Gbps
 Higher data rates than with soft-core implementations
– Data rates up to 800 Mbps (DDR2, DDR3)
– Maximum theoretical bandwidth up to 12.8 Gbps
– Max values are for all speed grades in standard voltage devices
Block Diagram
Spartan-6 FPGA
p0_cmd_full
p0_cmd_empty
Calibration Module (RTL)
Spartan-6 FPGA Memory Controller Block
CMD FIFO 0
CMD FIFO 1
CMD FIFO 2
CMD FIFO 3
CMD FIFO 4
CMD FIFO 5
Arbiter
p0_rd_clk
p0_rd_en
p0_rd_data
p0_rd_empty
p0_rd_full
p0_rd_overflow
p0_rd_count
p0_rd_error
p0_wr_clk
p0_wr_en
p0_wr_data
p0_wr_mask
p0_wr_empty
p0_wr_full
p0_wr_underrun
p0_wr_count
p0_wr_error


32-bit
Bi -directional
32-bit
Bi -directional
32-bit
Uni -directional
32-bit
Off-Chip Memory
Controller
Data Path
Dedicated Routing
p0_cmd_clk
p0_cmd_en
p0_cmd_bl
p0_cmd_instr
p0_cmd_addr
IP Wrapper
I/O Clock Network
User Interface
P
H
Y
I
O
B
mcbx_dram_clk
mcbx_dram_clk_n
mcbx_dram_cke
mcbx_dram_ras_n
mcbx_dram_cas_n
mcbx_dram_we_n
mcbx_dram_odt
mcbx_dram_ddr3_rst
mcbx_dram_ba
mcbx_dram_addr
mcbx_dram_dq
mcbx_dram_dqs
mcbx_dram_dqs_n
mcbx_dram_udm
mcbx_dram_ldm
Uni -directional
32-bit
Uni- directional
32-bit
Uni -directional
Simple user interface abstracts away complexity
MIG / EDK wrapper delivers complete interface solution
– Internal block assembly and signal connectivity is transparent to the user
Design Considerations
PLL creates two phases of MCB
system clock
User
Interface
User
Clks
Controller
Arbiter
Data Path
IOB
MCB Block
PHY Layer
– SYSCLK_2X & SYSCLK_2X_180
– Operate at 2X memory clock frequency
• Used for DDR I/O
• Divide by 2 in MCB creates memory
clock frequency for other logic (1X
clocks)
– Place MCBs on same side of device must
share system clocks = same data rate
Memory
Interface
1X Clks
:
2 :2
PLL
CLK
CLK
IN
CLKB
IBUFDS

FB
CLK SYSCLK_2X
OUT0
2X Clks
CLK
SYSCLK_2X_180
OUT1
BUFPLL_MCB
User interface clocks
– Port clocks for command, write, and read path
– Asynchronous to system clocks
– FIFOs handle clock domain transfer
IO Clock Network
2nd MCB Block
On the same side of device
(Only in larger parts)
Clock Example:
DDR2 800 Mbps
2X clk = 800 MHz
1X clk = 400 MHz
Design Considerations
I/O pins for MCBs are predefined
– But are general-purpose I/O when a particular MCB is not used
Package selection determines access to MCBs
– Higher pin count packages have more MCB blocks bonded out
Migration across devices within same package
– Up or down one device density in most cases
– Applies only within a device family (LX or LXT, for example)
The left and lower left MCB has the best migration path
– MCB pins shared less with other functions compared to right side
Design Considerations
MCB blocks can be connected in parallel to create wider
interfaces
 This will require extra CLB logic
MCB blocks interface to a SINGLE memory device (x4, x8,
or x16)
 There is No support for two x8 MC interfacing to a x16 memory
Even for LPDDR, use an external VREF supply in the bank
 Supports soft calibration module input termination tuning
Use a PLL nearest the center of the device to drive the
BUFPLL_MCB
Two MCB Design Flows
For ISE® tool design flow
– Memory Interface Generator (MIG) wizard within the CORE Generator tool
– Best for non-embedded applications
– Simple GUI-driven tool for configuring MCB block
– Supports all MCB memory standards (DDR3, DDR2, DDR, LPDDR)
For EDK design flow
– Multi Port Memory Controller (MPMC)
– Best for Embedded applications
– MCB block is underlying hardware implementation of the MPMC peripheral
– Simple GUI-driven tool within EDK / XPS
– Supports all MCB memory standards (DDR3, DDR2, DDR, LPDDR)
Memory Interface Generator (MIG)
Easy to customize your memory controller and interface design
– Memory architecture, data rate, bus width
– CAS latency, burst length
– I/O bank assignments
– Generates RTL source code and UCF from hardware-verified IP
– Delivered as part of the ISE software (CORE Generator utility)
– MCB supports DDR, DDR2, DDR3, and LPDDR
Soft Memory Controller supports other additional memory
interfaces
MIG Design Flow
Open CORE Generator™
Run MIG, choose your memory
parameters and generate rtl and ucf files
Integrate MIG .ucf constraints to
overall design constraints file
Import RTL and build options into
ISE project
Customize MIG design
Integrate/customize MIG memory RTL
testbench
Perform functional simulation
Optional RTL customization
Synthesize design
Place and route design
Timing simulation
Verify in hardware
MIG
 Enables you to customize your
memory controller
– Some options are specific to the
memory controller standard
(such as DDR2 and DDR3)
 Options for the physical layer and
the FPGA controller
– Debug signals
– IOB options for power or speed
 MIG bank selection options
– Displays pins required
– Restricted I/O columns are
disabled
MIG Output Files
UCF file folder
– Pinout and clocking constraints
– Batch file (ise_flow.bat) with
recommended build options
RTL file folder
– Functional modules (physical layer, user
interface, controller, testbench)
– Unencrypted for ease of customization
Simulation file folder
– HDL simulation files including memory
device models Synthesis files folder
Summary
Distributed LUT
RAM
Distributed LUT RAM
– Fast, localized memories
– Great for small FIFOs
Block RAM
– Bigger on-chip memories
High-Performance
Block RAM
– Great for mid-sized buffering
FPGA
Dedicated Memory Controller (MCB)
– Fast connection to popular standard RAMs
– Memory controller cores
– Ideal for large memory requirements
Memories can be built with the Core
Generator or Memory Interface Generator
(MIG)
External Memory
Interfacing
Where Can I Learn More?
User Guides
– Spartan-6 FPGA User Guide
• Describes the complete FPGA architecture, including distributed memory,
block memory and the MCB
– Sparfan-6 FPGA Memory Controller User Guide
• Detailed description of all MCB functionality
Xilinx Education Services courses
– www.xilinx.com/training
• Xilinx tools and architecture courses
• Hardware description language courses
• Basic FPGA architecture, Basic HDL Coding Techniques, and other Free
videos!
Trademark Information
Xilinx is disclosing this Document and Intellectual Property (hereinafter “the Design”) to you for use in the development of designs to operate on,
or interface with Xilinx FPGAs. Except as stated herein, none of the Design may be copied, reproduced, distributed, republished, downloaded,
displayed, posted, or transmitted in any form or by any means including, but not limited to, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written consent of Xilinx. Any unauthorized use of the Design may violate copyright laws, trademark laws, the laws of
privacy and publicity, and communications regulations and statutes.
Xilinx does not assume any liability arising out of the application or use of the Design; nor does Xilinx convey any license under its patents,
copyrights, or any rights of others. You are responsible for obtaining any rights you may require for your use or implementation of the Design.
Xilinx reserves the right to make changes, at any time, to the Design as deemed desirable in the sole discretion of Xilinx. Xilinx assumes no
obligation to correct any errors contained herein or to advise you of any correction if such be made. Xilinx will not assume any liability for the
accuracy or correctness of any engineering or technical support or assistance provided to you in connection with the Design.
THE DESIGN IS PROVIDED “AS IS" WITH ALL FAULTS, AND THE ENTIRE RISK AS TO ITS FUNCTION AND IMPLEMENTATION IS WITH
YOU. YOU ACKNOWLEDGE AND AGREE THAT YOU HAVE NOT RELIED ON ANY ORAL OR WRITTEN INFORMATION OR ADVICE,
WHETHER GIVEN BY XILINX, OR ITS AGENTS OR EMPLOYEES. XILINX MAKES NO OTHER WARRANTIES, WHETHER EXPRESS,
IMPLIED, OR STATUTORY, REGARDING THE DESIGN, INCLUDING ANY WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE, TITLE, AND NONINFRINGEMENT OF THIRD-PARTY RIGHTS.
IN NO EVENT WILL XILINX BE LIABLE FOR ANY CONSEQUENTIAL, INDIRECT, EXEMPLARY, SPECIAL, OR INCIDENTAL DAMAGES,
INCLUDING ANY LOST DATA AND LOST PROFITS, ARISING FROM OR RELATING TO YOUR USE OF THE DESIGN, EVEN IF YOU HAVE
BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. THE TOTAL CUMULATIVE LIABILITY OF XILINX IN CONNECTION WITH
YOUR USE OF THE DESIGN, WHETHER IN CONTRACT OR TORT OR OTHERWISE, WILL IN NO EVENT EXCEED THE AMOUNT OF
FEES PAID BY YOU TO XILINX HEREUNDER FOR USE OF THE DESIGN. YOU ACKNOWLEDGE THAT THE FEES, IF ANY, REFLECT
THE ALLOCATION OF RISK SET FORTH IN THIS AGREEMENT AND THAT XILINX WOULD NOT MAKE AVAILABLE THE DESIGN TO YOU
WITHOUT THESE LIMITATIONS OF LIABILITY.
The Design is not designed or intended for use in the development of on-line control equipment in hazardous environments requiring fail-safe
controls, such as in the operation of nuclear facilities, aircraft navigation or communications systems, air traffic control, life support, or weapons
systems (“High-Risk Applications”). Xilinx specifically disclaims any express or implied warranties of fitness for such High-Risk Applications. You
represent that use of the Design in such High-Risk Applications is fully at your risk.
© 2012 Xilinx, Inc. All rights reserved. XILINX, the Xilinx logo, and other designated brands included herein are trademarks of Xilinx, Inc. All
other trademarks are the property of their respective owners.