Dual Port RAM System

Dual-port RAM that supports simultaneous read and write operations

• 

Libraries:
HDL Coder / HDL RAMs

Description

The Dual Port RAM System block models a dual-port RAM that supports simultaneous read and write operations through output ports for read and write data. To configure this block, open the block and set the Type of RAM parameter to dual port. By using this block, you can:

• Create parallel RAM banks by using vector data at the din, addr, and we ports.

• Specify an initial value for the RAM using the Initial output of RAM parameter.

• Perform asynchronous read in target hardware by selecting the Use asynchronous read feature in target hardware parameter.

• Perform cycle-accurate read operation by selecting the Model RAM with one cycle of delay parameter.

• Use the column-write method to write to specified bits in an addressed memory location.

Note

During a write, new data appears at the output of the write port wr_dout of the Dual Port RAM System block. If a read operation occurs simultaneously at the same address as a write operation, old data appears at the read output port rd_dout.

Limitations

• When you build the FPGA bitstream for the RAM, the global reset logic does not reset the RAM contents. To reset the RAM, implement the reset logic.

• The RAM address can be either fixed-point (fi) or integer, must be unsigned, and must be between 2 and 31 bits long.

Ports

Input

din — Data to write

scalar | vector

Data to write, specified as a scalar or a vector. You can write the data into the RAM memory location when the write enable signal at port we is true. This value can be an integer, or of type double, single or half. The value can also be a fi object and can be real or complex.

Bus Support:

You can use non-virtual bus and arrays of buses for HDL code generation.

Data Types: single | double | half | int8 | int16 | uint8 | uint16 | Boolean | fixed point

wr_addr — Address to write

scalar | vector

Address to write, specified as a scalar or a vector. Use this address to write to RAM when the signal at port we is true. This value can be either a fixed-point (fi) or an unsigned integer and must be between 2 and 31 bits long.

Data Types: uint8 | uint16 | fixed point

we — Write enable

scalar | Vector

Write enable, specified as a scalar or a vector. When the signal at port we is true, the block writes the data into the specified memory location. When the signal at port we is false, the block reads the value from the memory location specified by rd_addr port when the signal at port we is false.

Note

To use column-write method, the data type must be an integer or a fixed-point.

Data Types: uint8 | uint16 | uint32 | uint64 | Boolean | fixed point

rd_addr — Address to read

scalar | vector

Address to read, specified as a scalar or a vector. Use this address to read the data from the RAM. This value can be either a fixed-point (fi) or an unsigned integer and must be between 2 and 31 bits long.

Data Types: uint8 | uint16 | fixed point

Output

rd_dout — Read output data

scalar | vector

Read output data, returned as a scalar or vector. The block reads the old data from the memory location specified by the rd_addr port.

wr_dout — Write output data

scalar | vector

Write output data, returned as a scalar or vector. The block reads the new or old data from the memory location specified by the wr_addr port.

Parameters

Main

Type of RAM — RAM type

Dual port (default) | Simple dual port | Single port | True dual port | Simple tri port

Type of RAM, specified as either:

• Single port — Create a single port RAM with write data, address, and write enable data as inputs and read data as the output.

• Simple dual port — Create a simple dual port RAM with write data, write address, write enable, and read address as inputs, and data from the read address as the output.

• Dual port — Create a dual port RAM with write data, write address, write enable, and read address as inputs, and data from the read address and write address as the outputs.

• True dual port — Create a true dual port RAM with write data a and b, write and read addresses for a and b, and write enable a and b as inputs and data from write addresses a and b as the outputs.

• Simple tri port — Create a simple tri port RAM with write data, write address, write enable, and read addresses a and b as inputs and data from read addresses a and b as the outputs.

The code generator dynamically configures the input and output ports of the block based on the RAM type that you specify.

Programmatic Use

Block Parameter: RAMType

Type: character vector, string

Values: Dual port | Simple dual port | Single port | True dual port | Simple tri port

Default: Dual port

Use asynchronous read feature in target hardware — Option to enable asynchronous read feature in target hardware

off (default) | on

Enable this option to allow the hardware to execute a read instruction immediately, without waiting one cycle.

Programmatic Use

Block Parameter: AsyncRead

Type: character vector, string

Values: 'on' | 'off'

Default: 'off'

Behavior for write output — Write output behavior

New data (default) | Old data

Select the behavior for write output as New data or Old data:

• New data — Send out new data at the address to the output.

• Old data — Send out old data at the address to the output.

Dependencies

To enable this parameter, clear the Use asynchronous read feature in target hardware parameter.

Programmatic Use

Block Parameter: WriteOutputValue

Type: character vector, string

Values: New data | Old data

Default: New data

Initial output of RAM — Initial simulation output of RAM

'0.0' (default) | Scalar | Vector

Specify the initial simulation output of the RAM by using one of these options:

• A scalar value.

• A vector with a one-to-one mapping between the initial value and the RAM words.

• An n-by-m matrix with a one-to-one mapping between the initial values and the RAM words in the RAM banks, where n represents the number of RAM banks and m represents the number of address locations in the RAM block, or vice-versa.

Programmatic Use

Block Parameter: RAMInitialValue

Type: integer, floating point

Values: 0.0 | Scalar | Vector

Default: 0.0

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Advanced

Model RAM with one cycle of delay — Model RAM with one cycle of delay

on (default) | off

Since R2024b

• When you select this parameter, the RAM delays the input data by one cycle before the output can read it. Use this option to ensure the block is cycle-accurate with the generated HDL code.

• When you clear this parameter, the RAM reads and outputs the input data immediately, but adds one cycle of latency during HDL code generation. Use this option to leverage clock-rate pipelining when you specify an oversampling value or work with multirate models.

Dependencies

To enable this parameter, clear the Use asynchronous read feature in target hardware parameter.

Programmatic Use

Block Parameter: ModelRAMDelay

Type: character vector, string

Values: 'on' | 'off'

Default: 'on'

Specify vector input access — Specify vector input access as serial or parallel

serial (default) | parallel

Since R2025a

• When you set this parameter to serial, the code generator infers parallel banks of RAM from vector inputs to RAM System block.

• When you set this parameter to parallel:

  • The code generator does not scale up RAM consumption with the size of the vector input when you generate HDL Code. Instead, it applies each operation of the input signals one at a time, starting with the first index. Enabling serial access allows you to leverage multicycle RAM access at a faster clock rate while modeling with a vector input at the data rate.

  • If the RAM System block exists in a clock-rate pipelining region, the serialization uses a clock-rate implementation instead of local multirate implementation.

Dependencies

To enable this parameter, clear the Model RAM with one cycle of delay parameter or enable the Use asynchronous read feature in target hardware parameter.

Programmatic Use

Block Parameter: VectorAccess

Type: character vector, string

Values: 'serial' | 'parallel'

Default: 'serial'

More About

Using the Column-Write Method to Selectively Write to Columns

You can use the column-write method to view the RAM as a collection of equally sized columns. During a write cycle, you can write into each of these columns separately. The data type and value of the write enable input, along with the data type of write data input, determine the size of each column and the columns in which the block writes in the addressed memory location.

In this context:

• DT is the data type of the write data input signal din.

• DW is the data width of the input data, which is equal to word length of the din value.

• DTWE is the data type of the write enable signal we. This signal determines which columns the block writes in the addressed memory location. The block writes the columns based on the position of the 1s in the binary representation of the value of we.

• NC is the number of columns which you can partition the RAM space to write the data, which is equal to word length of we value.

• WC is the width of each column, which is equal to DW divided by NC.

The table summarizes the relationship among the data types of the write data input, the data types of write enable input, the number of columns, and the width of each column.

DT	DW	DTWE	NC	WC in Bits
uint16	16	ufix4	4	4
uint32	32	ufix4	4	8
uint64	64	ufix4	4	16
uint32	32	uint8	8	4
uint64	64	uint8	8	8
int32	32	uint16	16	2

For example, if DT is uint16 and WE is ufix4, then DW is equal to 16, NC is equal to 4, and WC is equal to 4 bits. If the input to din is 980, its binary representation is 0000001111010100. The column-wise representation of din is c4 = 0000, c3 = 0011, c2 = 1101, and c1 = 0100, where c1 is the first column.

The table summarizes the results of using the column-write method for different input combinations.

Value of we	Binary Representation of we	Columns Selected for Writing in RAM	Data at Memory Location		dout
			Before Performing Write Operation	After Performing Write Operation
3	0011	c2, c1	c4 = 0000 c3 = 0000 c2 = 0000 c1 = 0000	c4 = 0000 c3 = 0000 c2 = 1101 c1 = 0100	212
4	0100	c3	c4 = 0000 c3 = 0000 c2 = 0000 c1 = 0000	c4 = 0000 c3 = 0011 c2 = 0000 c1 = 0000	768
6	0110	c3, c2	c4 = 0000 c3 = 0000 c2 = 0000 c1 = 0000	c4 = 0000 c3 = 0011 c2 = 1101 c1 = 0000	976
9	1001	c4, c1	c4 = 0000 c3 = 0000 c2 = 0000 c1 = 0000	c4 = 0000 c3 = 0000 c2 = 0000 c1 = 0100	4
9	1001	c4, c1	c4 = 1111 c3 = 1111 c2 = 1111 c1 = 1111	c4 = 0000 c3 = 1111 c2 = 1111 c1 = 0100	4084

Limitations

• Inputs with signed data types and with non-zero fraction lengths are not supported by the write enable input port.

• The word length of the write data input must a multiple of the word length of the write enable.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using Simulink® Coder™.

HDL Code Generation

Generate VHDL, Verilog and SystemVerilog code for FPGA and ASIC designs using HDL Coder™.

HDL Architecture

The block has a MATLABSystem architecture which indicates that the block implementation uses the hdl.RAM System object™.

HDL Block Properties

General
ConstrainedOutputPipeline	Number of registers to place at the outputs by moving existing delays within your design. Distributed pipelining does not redistribute these registers. The default is 0. For more details, see ConstrainedOutputPipeline.
InputPipeline	Number of input pipeline stages to insert in the generated code. Distributed pipelining and constrained output pipelining can move these registers. The default is 0. For more details, see InputPipeline.
OutputPipeline	Number of output pipeline stages to insert in the generated code. Distributed pipelining and constrained output pipelining can move these registers. The default is 0. For more details, see OutputPipeline.
RAMDirective	Specify whether to map RAM blocks in your design to the RAM blocks on the target FPGA. For UltraRAM mapping, Initial output of RAM must be set to 0. For more details, see RAMDirective.

Complex Data Support

This block supports code generation for complex signals.

Restrictions

The block does not support:

• Nested bus at the data port.

• Non-zero values at the Initial output of RAM when using non-virtual bus input to the data port.

Version History

Introduced in R2017b

R2024b: Use clock-rate pipelining

You can use Dual Port RAM System block inside a data rate feedback loop and use clock-rate pipelining optimization. This model design is helpful for applications that require programmable or tunable lookup tables without having to regenerate bitstreams.

Use the new parameter Model RAM with one cycle of delay to model delay in your simulation. Model RAM with one cycle of delay is enabled by default and is disabled when Use asynchronous read feature in target hardware is enabled.

R2024b: Parameter name changes

These parameters are renamed:

Previous Name	Current Name
Specify the type of RAM	Type of RAM
Enable asynchronous reads	Use asynchronous read feature in target hardware
Specify the output data for a write operation	Behavior for write output
Specify the RAM initial value	Initial output of RAM

R2024b: Write to RAM in columns using column-write operations in RAM System blocks

In this block, you can use the column-write method to selectively modify specific parts of the memory without altering the remaining parts at a specified memory address.

R2024b: Initialize RAM banks with unique initial values for vector data inputs

In this block, you can initialize RAM banks with unique initial values.

R2024a: Array of Buses Input Support

The block now supports HDL code generation for input data specified as an array of buses.

R2024a: Half-Precision and Boolean Data Support

The block now supports HDL code generation with input data of type Half or Boolean.