True Dual Port RAM System

True Dual Port RAM block based on the hdl.RAM system object with ability to provide initial value

Since R2023b

• 

Libraries:
HDL Coder / HDL RAMs

Description

The blocks are MATLAB System blocks that use the hdl.RAM System object™. You can specify the RAM type as Dual port, Simple dual port, Single port, True dual port, or Simple tri port. With the MATLAB System blocks, you can:

• Specify an initial value for the RAM. In the Block Parameters dialog box, enter a value for Initial output of RAM.

• Obtain faster simulation results when you use these blocks in your Simulink^® model.

• Create parallel RAM banks when you use vector data by leveraging the hdl.RAM System object functionality.

• Obtain higher performance and support for large data memories.

• Specify writable portions of bits in an addressed memory location by using column-write method. For more information, see Using the Column-Write Method to Selectively Write to Columns.

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

Ports

Input

din_a — Write data input

scalar | vector

Data to write into the RAM memory location when the signal at port we_a is true, specified as a scalar or a vector. This value can be double, single, half, integer, or a fixed-point (fi) object, and can be real or complex.

Bus Support:

You can use non-virtual bus and array of buses at the data port for HDL code generation.

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

addr_a — Write address

scalar | vector

Address to write, specified as a scalar or a vector. The RAM address to write the data from port din_a to. This value can be either fixed-point (fi) or integer, must be unsigned, and must be between 2 and 31 bits long. The values at port addr_a and addr_b cannot be the same address.

Data Types: uint8 | uint16 | fixed point

we_a — Write enable

scalar | vector

Write enable, specified as a scalar or a vector. When the signal at this port is true, the RAM writes the data from port din_a to the memory location specified by port addr_a. This value can be either a Boolean or an integer or a fixed-point (fi).

Note

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

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

din_b — Write data input

scalar | vector

Data to write into the RAM memory location when we_b is true, specified as a scalar or a vector. This value can be double, single, half, integer, or a fixed-point (fi) object, and can be real or complex.

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

addr_b — Write address

scalar | vector

Address to write or read, specified as a scalar or a vector. The RAM address where you write the data from port din_b into. This value can be either fixed-point (fi) or integer, must be unsigned, and must be between 2 and 31 bits long. The values at port addr_a and addr_b cannot be the same address.

Data Types: uint8 | uint16 | fixed point

we_b — Write enable

scalar | vector

Write enable, specified as a scalar or a vector. When the signal at this port is true, the RAM writes the data from port din_b to the memory location specified by port addr_b. This value can be either Boolean or integer or a fixed-point (fi).

Note

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

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

Output

dout_a — Write data output

scalar | vector

New or old output data that the RAM reads from the memory location specified by port addr_a.

dout_b — Write data output

scalar | vector

New or old output data that the RAM reads from the memory location specified by port addr_b.

Parameters

Main

Type of RAM — RAM type

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

Type of RAM, specified as either:

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

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

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

Behavior for write output — Write output behavior

New data (default) | Old data

Behavior for the write output, specified as either:

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

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

Initial output of RAM — Initial simulation output of RAM

0 (default) | scalar | vector

Initial simulation output of the System object block, specified as either:

• A scalar value

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

• An n-by-m matrix with 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 location in the RAM block, or vice-versa.

Advanced

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

on (default) | off

Since R2024b

If you specify:

• true

  • The RAM delays the input data by one cycle before the output can read it.

  • The RAM is cycle-accurate to the generated HDL code.

• false

  • The RAM reads and outputs the input data immediately, but adds one cycle of latency during HDL code generation.

  • You can leverage clock-rate pipelining when you specify an oversampling value or work with multirate models.

Programmatic Use

Block Parameter: ModelRAMDelay

Type: character vector, string

Values: "on" | "off"

Default: "on"

Data Types: Boolean

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.

• Both the write enable input types should be of same type for True Dual Port RAM System block

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 MATLAB System 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.

Version History

Introduced in R2023b

R2024b: Use clock-rate pipelining

You can use True 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 cycle of delay to model delay in your simulation. Model RAM with cycle of delay is enabled by default.

R2024b: Parameter name changes

These parameters are renamed:

Previous Name	Current Name
Specify the type of RAM	Type of RAM
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.