DSP58 - 2022.1 English

Vitis Model Composer User Guide (UG1483)

Document ID
UG1483
Release Date
2022-05-26
Version
2022.1 English

The Xilinx DSP58 block is an efficient building block for DSP applications that use Versal® devices. DSP applications use many binary multipliers and accumulators that are best implemented in dedicated DSP resources. Versal® devices have many dedicated low-power DSP slices, combining high speed with small size while retaining system design flexibility.

The DSP58 slice is effectively a super-set of the DSP48E2 slice with these differences. The DSP58 has the following.

  • Wider functionality
  • More flexibility in the pre-adder
  • New optional negate inport
  • More XOR operations
  • Additional unique features
Figure 1. DSP58

Block Parameters

The Block Parameters dialog box can be invoked by double-clicking the icon in your Simulink® model.

Basic tab
Input Configuration
A or ACIN input
Specifies if the A input should be taken directly from the a port or from the cascaded acin port. The acin port can only be connected to another DSP58 block.
B or BCIN input
Specifies if the B input should be taken directly from the b port or from the cascaded bcin port. The bcin port can only be connected to another DSP58 block.
Note: If the input to the block is NaN, you will see a behavioral simulation mismatch.
DSP58 Data-Path Configuration
SIMD Mode of Adder/Subtractor/Accumulator
This mode can be used to implement small add-subtract functions at high speed and lower power with less logic utilization. The adder and subtracter in the adder/subtracted/logic unit can also be split into two 24-bit units or four 12-bit units.
Mode of Multiplier
This option is disabled in the current release
Do not use multiplier
When this is selected, the DSP58 block is optimized in hardware for maximum performance without using the multiplier. If an instruction using the multiplier is encountered in simulation, an error is reported.
Use dynamic multiplier mode
When this is selected, it instructs the DSP58 block to use the dynamic multiplier mode. This indicates that the block is switching between A*B and A:B operations on the fly, and therefore needs to get the worst-case timing of the two paths.
Preadder Configuration
Use the 27-bit D data input to the pre-adder or alternative input to the multiplier. The pre-adder implements D + A as determined by the INMODE3 signal.
PREADDINSEL Select preadder input
Selects the input to be added with D in the pre-adder.
AMULTSEL Select A multiplexer output
Selects the input to the 27-bit A input of the multiplier. In the 7 series primitive, DSP48E1 the attribute is called USE_DPORT, but has been renamed due to new pre-adder flexibility enhancements (default AMULTSEL = A is equivalent to USE_DPORT=FALSE).
BMULTSEL Select B multiplexer output
Selects the input to the 18-bit B input of the multiplier.
Enable D Port
Automatically enabled when AD is selected.
Pattern Detection
Reset p register on pattern detection
If selected and the pattern is detected, reset the p register on the next cycle
AUTO RESET PRIORITY
When enabled by selecting the option above, select RESET (the default) or CEP (clock enabled for the P (output) resister).
Pattern Input
Pattern Input from c port
When selected, the pattern used in pattern detection is read from the c port.
Using Pattern Attribute (58bit hex value)
Value is used in pattern detection logic, which is best described as an equality check on the output of the adder/subtractor/logic unit.
Pattern attribute
A 58-bit value that is used in the pattern detector.
Mask Input
Mask input from c port
When selected, the mask used in pattern detection is read from the c port.
Using Mask Attribute (58 bit hex value)
A 58-bit value used to mask out certain bits during pattern detection.
MODE1
Selects rounding_mode 1 (C-bar left shifted by 1).
MODE2
Selects rounding_mode 2 (C-bar left shifted by 2).
Wide Xor tab
Parameters specific to the Wide Xor tab are as follows.
Use Wide XOR
Use this is feature to perfom a 116 bit XOR function.
XORSIMD Select Wide XOR SIMD
Optional Ports tab
Input Ports Use the XORSIMD attribute to select the width of the XOR function. Select either XOR12 (the default), XOR22, XOR24, XOR34, XOR58, or XOR116).
Consolidate control port
When selected, combines the opmode, alumode, carry_in, carry_in_sel, inmode, and negate ports into one 25-bit port. Bits 0 to 8 are the opmode, bits 9 to 12 are the alumode port, bit 13 is the carry_in port, and bits 14 to 16 are the carry_in_sel port, bits 17 to 21 are the inmode Use the XORSIMD attribute to select the width port, and bits 22 to 24 are the negate port. This option should be used when the Opmode block is used to generate a DSP58 instruction.
Provide c port
When selected, the c port is made available. Otherwise, the c port is tied to '0'.
Provide global reset port
When selected, the port rst is made available. This port is connected to all available reset ports based on the pipeline selections.
Provide global enable port
When selected, the optional en port is made available. This port is connected to all available enable ports based on the pipeline selections.
Cascadable Ports
Provide pcin port
When selected, the pcin port is exposed. The pcin port must be connected to the pcout port of another DSP58 block.
Provide carry cascade in port
When selected, the carry cascade in port is exposed. This port can only be connected to a carry cascade out port on another DSP58 block.
Provide multiplier sign cascade in port
When selected, the multiplier sign cascade in port (multsigncascin) is exposed. This port can only be connected to a multiplier sign cascade out port of another DSP58 block.
Output Ports
Provide carryout port
When selected, the carryout output port is made available. When the mode of operation for the adder/subtractor is set to one 58-bit adder, the carryout port is 1-bit wide. When the mode of operation is set to two 24 bit adders, the carryout port is 2 bits wide. The MSB corresponds to the second adder's carryout and the LSB corresponds to the first adder's carryout. When the mode of operation is set to four 12 bit adders, the carryout port is 4 bits wide with the bits corresponding to the addition of the 48 bit input split into four 12-bit sections.
Provide pattern detect port
When selected, the pattern detection output port is provided. When the pattern, either from the mask or the c register, is matched the pattern detection port is set to '1'.
Provide pattern bar detect port
When selected, the pattern bar detection (patternbdetect) output port is provided. When the inverse of the pattern, either from the mask or the c register, is matched the pattern bar detection port is set to '1'.
Provide overflow port
When selected, the overflow output port is provided. This port indicates when the operation in the DSP58 has overflowed beyond the bit P[N] where N is between 1 and 46. N is determined by the number of 1s in the mask whether set by the GUI mask field or the c port input.
Provide underflow port
When selected, the underflow output port is provided. This port indicates when the operation in the DSP58 has underflowed. Underflow occurs when the number goes below –P[N] where N is determined by the number of 1s in the mask whether set by the GUI mask field or the c port input.
Cascadable Ports
Provide ACOUT port
When selected, the acout output port is made available. The acout port must be connected to the acin port of another DSP58 block.
Provide BCOUT port
When selected, the bcout output port is made available. The bcout port must be connected to the bcin port of another DSP58 block.
Provide PCOUT port
when selected, the pcout output port is made available. The pcout port must be connected to the pcin port of another DSP58 block.
Provide multiplier sign cascade out port
When selected, the multiplier sign cascade out port (multsigncascout) is made available. This port can only be connected to the multiplier sign cascade in port of another DSP58 block and is used to support 96-bit accumulators/adders and subtracters which are built from two DSP58s.
Provide carry cascade out port
When selected, the carry cascade out port (carrycascout) is made available. This port can only be connected to the carry cascade in port of another DSP58 block.
Pipelining tab
Pipeline Options
Length of a/acin pipeline
Specifies the length of the pipeline on input register A. A pipeline of length 0 removes the register on the input.
Length of b/bCIN pipeline
Specifies the length of the pipeline for the b input whether it is read from b or bcin.
Length of acout pipeline
Specifies the length of the pipeline between the a/acin input and the acout output port. A pipeline of length 0 removes the register from the acout pipeline length. Must be less than or equal to the length of the a/acin pipeline.
Length of bcout pipeline
Specifies the length of the pipeline between the b/bcin input and the bcout output port. A pipeline of length 0 removes the register from the bcout pipeline length. Must be less than or equal to the length of the b/bcin pipeline.
Pipeline c
Indicates whether the input from the c port should be registered.
Pipeline p
Indicates whether the outputs p and pcout should be registered.
Pipeline multiplier
indicates whether the internal multiplier should register its output.
Pipeline opmode
Indicates whether the opmode port should be registered.
Pipeline alumode
Indicates whether the alumode port should be registered.
Pipeline carry in
Indicates whether the carry in port should be registered.
Pipeline carry in select
Indicates whether the carry in select port should be registered.
Pipeline preadder input register d
Indicates to add a pipeline register to the d input.
Pipeline preadder output register ad
Indicates to add a pipeline register to the ad output.
Pipeline INMODE register
Indicates to add a pipeline register to the INMODE input.
Reset/Enable Ports tab
Provide Reset Ports
Reset port for a/acin
When selected, a port rst_a is made available. This resets the pipeline register for port a when set to '1'.
Reset port for b/bcin
When selected, a port rst_b is made available. This resets the pipeline register for port b when set to '1'.
Reset port for c
When selected, a port rst_c is made available. This resets the pipeline register for port c when set to '1'.
Reset port for multiplier
when selected, a port rst_m is made available. This resets the pipeline register for the internal multiplier when set to '1'.
Reset port for P
When selected, a port rst_p is made available. This resets the output register when set to '1'.
Reset port for carry in
When selected, a port rst_carryin is made available. This resets the pipeline register for carry in when set to '1'.
Reset port for alumode
When selected, a port rst_alumode is made available. This resets the pipeline register for the alumode port when set to '1'.
Reset port for controls (opmode and carry_in_sel)
When selected, a port rst_ctrl is made available. This resets the pipeline register for the opmode register (if available) and the carry_in_sel register (if available) when set to '1'.
Provide Enable Ports
Enable port for first a/acin register
When selected, an enable port ce_a1 for the first a pipeline register is made available.
Enable port for second a/acin register
When selected, an enable port ce_a2 for the second a pipeline register is made available.
Enable port for first b/bcin register
When selected, an enable port ce_b1 for the first b pipeline register is made available.
Enable port for second b/bcin register
When selected, an enable port ce_b2 for the second b pipeline register is made available.
Enable port for c
When selected, an enable port ce_c for the port C register is made available.
Enable port for multiplier
When selected, an enable port ce_m for the multiplier register is made available.
Enable port for p
When selected, an enable port ce_p for the port P output register is made available.
Enable port for carry in
When selected, an enable port ce_carry_in for the carry in register is made available.
Enable port for alumode
When selected, an enable port ce_alumode for the alumode register is made available.
Enable port for multiplier carry in
When selected, an enable port mult_carry_in for the multiplier register is made available.
Enable port for controls (opmode and carry_in_sel)
When selected, the enable port ce_ctrl is made available. The port ce_ctrl controls the opmode and carry in select registers.
Enable port for d
When selected, an enable port is added input register d.
Enable port for ad
When selected, an enable port is add for the preadder output register ad.
Enable port for INMODE
When selected, an enable port is added for the INMODE register.
Inversion Options tab
When a checkbox is selected under this tab, the specified signal is inverted.
Implementation tab
Use synthesizable model
When selected, the DSP58 is implemented from an RTL description which might not map directly to the DSP58 hardware. This is useful if a design using the DSP58 block is targeted at device families that do not contain DSP58 hardware primitives.
Mode of operation for the adder/subtractor
This mode can be used to implement small add-subtract functions at high speed and lower power with less logic utilization. The adder and subtractor in the adder/subtracted/logic unit can also be split into two 24-bit fields or four12-bit fields. This is achieved by setting the mode of operation to "Two 24-bit adders" or "Four 12-bit adders".

Other parameters used by this block are explained in the topic Common Options in Block Parameter Dialog Boxes.