This block is listed in the following Xilinx® Blockset libraries: Index, DSP.
The Xilinx DSP48E2 block is an efficient building block for DSP applications that use UltraScale devices. DSP applications use many binary multipliers and accumulators that are best implemented in dedicated DSP resources. UltraScale devices have many dedicated low-power DSP slices, combining high speed with small size while retaining system design flexibility.
The DSP48E2 slice is effectively a superset of the DSP48E1 slice with these differences:
- Wider functionality
- More flexibility in the pre-adder
- Added fourth operand to ALU with WMUX
- Wide XOR of the X, Y, and Z multiplexers
- Additional unique features
Refer to the document titled UltraScale Architecture DSP Slice User Guide (UG579) for a detailed description of the DSP48E2 features.
Block Parameters
The block parameters dialog box can be invoked by double-clicking the icon in your Simulink® model.
- Basic tab
- Parameters specific to the Basic tab are as follows.
- 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. Theacin
port can only be connected to another DSP48 block. - B or BCIN input
- Specifies if the B input should be taken directly
from the b port or from the cascaded
bcin
port. Thebcin
port can only be connected to another DSP48 block.
- DSP48E2 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.
- Do not use multiplier
- When selected, the 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 selected, it instructs the 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 25-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 preadder.
- 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 above.
- 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 (48bit 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 (48bit hex value)
- Enter a 48-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 (48 bit hex value)
- Enter a 48-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
- This is a new feature in the DSP48E2 slice giving the ability to perform a 96-bit wide XOR function.
- XORSIMD Select Wide XOR SIMD
- The XORSIMD attribute is used to select the width of the XOR function. Select either XOR12 (the default), XOR24, XOR48, or XOR96.
- Optional Ports tab
- Parameters specific to the Optional Ports tab are as follows.
- Input Ports
-
- Consolidate control port
- When selected, combines the
opmode
,alumode
,carry_in
,carry_in_sel
, andinmode
ports into one 20-bit port. Bits 0 to 6 are the opmode, bits 7 to 10 are thealumode
port, bit 11 is thecarry_in
port, bits 12 to 14 are thecarry_in_sel
port, and bits 15-19 are the inmode bits. This option should be used when the Opmode block is used to generate a DSP48 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. Thepcin
port must be connected to thepcout
port of another DSP48 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 DSP48E 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 DSP48E 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 48-bit adder, thecarryout
port is 1-bit wide. When the mode of operation is set to two 24 bit adders, thecarryout
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, thecarryout
port is 4 bits wide with the bits corresponding to the addition of the 48 bit input split into 4 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 DSP48E 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 DSP48E 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. Theacout
port must be connected to theacin
port of another DSP48E block. - Provide bcout port
- When selected, the
bcout
output port is made available. Thebcout
port must be connected to thebcin
port of another DSP48E block. - Provide pcout port
- When selected, the
pcout
output port is made available. Thepcout
port must be connected to thepcin
port of another DSP48 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 DSP48E block and is used to support 96-bit accumulators/adders and subtracters which are built from two DSP48Es.
- 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 DSP48E block.
- Pipelining tab
-
Parameters specific to the Pipelining tab are as follows.
- 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 theacout
output port. A pipeline of length 0 removes the register from theacout
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
- Parameters specific to the Reset/Enable tab are as follows.
- 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'.
- Reset port for d and ad
- When selected, a port rst_a and rst_ad is made available. This resets the pipeline register for ports when set to '1'.
- Reset port for INMODE
- When selected, a port rst_inmode is made available. This resets the pipeline register for the inmode port 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 on this tab, the specified signal is inverted.
- Implementation tab
- Parameters specific to the Implementation tab are as follows.
- Use synthesizable model
- When selected, the DSP48E2 is implemented from an RTL description which might not map directly to the DSP48E2 hardware. This is useful if a design using the DSP48E2 block is targeted at device families that do not contain DSP48E2 hardware primitives.
Other parameters used by this block are explained in the topic Common Options in Block Parameter Dialog Boxes.