Documentation Home Page HYPERSIM Home Page
Pour la documentation en FRANÇAIS, utilisez l'outil de traduction de votre navigateur Chrome, Edge ou Safari. Voir un exemple.

Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 2 Current »

Description

The current functionality is used to send raw data asynchronously from a HYPERSIM model to an OPAL-RT card. This is achieved by using the LoadIN ports of the FPGA.
As opposed to DataIN, the transfer of LoadIN data is relying on a slower communication link between the FPGA and the CPU and thus is not meant to be used at every step of the simulation. Doing otherwise will lead to decreased simulation performances.
The main purpose of this functionality is to send configuration-type data to the board. The transfer can be done during the initialization phase of a HYPERSIM model, during the execution of the simulation (with the help of a trigger mechanism) or both. Please check the Usage section below to find the implementation that suits your simulation's needs.

The number of supported LoadIN and LoadOUT ports depends on the RT-XSG revision used to generate the bitstream. If RT-XSG revision 3.5.4 or above is used, bitstreams can have up to 64 LoadIN input ports and 64 LoadOUT output ports. Otherwise, they will have 32 LoadIN input ports and 32 LoadOUT output ports. Each port has a maximum width of 250 32-bit data words. These ports are used to exchange data asynchronously between the HYPERSIM model and the FPGA chip of the card.

Note that knowledge of the bitstream logic that will use the data sent to the FPGA is necessary for a successful simulation.

Usage

Once the bitstream configuration file has been parsed, the ports capable of asynchronous raw data transfer are revealed to the user.
If the bitstream offers the possibility of transferring raw data asynchronously from the CPU to the FPGA then a new section named Load to board will appear as a subsection of the board's main configuration page. There, a dedicated configuration page will be created for each usable port, based on the information parsed from the bitstream configuration file. The title of each port's configuration page will contain the port number (count starting from 0) and the maximum number of DWORDS serviceable.

By clicking on a port's configuration page, the user will be able to:

  • choose the transfer mechanism best suited for the simulation
  • choose the data to send to the board (either a .mat file or custom data frames)
  • for custom data frames, define the frame format to send

The latter can be achieved by adding, removing and/or moving signals up or down in the data frame table using the buttons found above the table. The configurable parameters of each signal are detailed in the sections below.

Port configuration

Parameters

  • Send .mat configuration file
    When this check-box is selected, a new field will appear, where the user will have to input the path of the .mat file to be transferred to the FPGA.
    The transfer of data frames is disabled when this option is selected, therefore the Send configuration data real-time option will become grayed-out. 


  • .mat file path
    The path of the .mat file to transfer to the FPGA. 
    This parameter is only visible if the Send .mat configuration file check-box is enabled. 


  • Send configuration data real-time
    When this check-box is selected, the interface offers the user the possibility of transferring the data frame during the execution of the simulation, by the means of a trigger mechanism.
    Selecting this option will make the connectable points for the current port's data and trigger available in HYPERSIM. It will also cause the Also send configuration initial values field (explained below) to appear.
    For the trigger, only a '0' to '1' transition will enable the data transfer. Leaving the value of the trigger at '1' will not allow for data transfer.
    When only this option is active, the initial values of signals are disabled since LoadIN data is transferred at run-time. If they require to be transferred during the initialization phase of the model as well, then the check-box explained below needs to be activated as well. 


  • Also send configuration initial values
    By checking this box, the initial value parameter of the signals in the data frame is re-enabled and will be sent during the initialization phase of the simulation. Once the simulation is executing, data transfer will occur when the trigger's value will pass from '0' to '1' (rising-edge detection).
    This field is only visible if the Send configuration data real-time check-box is selected. 

Choosing a configuration

If the Send configuration data real-time check-box is selected then adding a signal to the data frame table will also make its connection point available for use in HYPERSIM.
In this scenario, if the table element has a vector size different from 1 (see section below for a description of the Vector size parameter), then its connectable item is an array.
If there is at least one signal to be sent real-time, the trigger connectable point will also be made available.
For sending data to the FPGA, connections must be made between the points in the model and the raw data output connectable points.

Given all of the above, the transfer scenarios for each port are described below:

  • Send data only during the initialization phase of the model:
    • None of the check-boxes have to be enabled
    • No connections need to be created
    • The initial value of each signal in the data frame table needs to be specified to its required value

  • Send data at run-time
    • Only the Send configuration data real-time check-box needs to be enabled
    • Connections need to be created for each signal (or vector of signals)
    • A connection also needs to be created for the port's trigger
    • Attention must be paid on how to the trigger is used. Note that starting the simulation with the trigger set to '1' will cause a rising-edge transition to be detected during the first step of the execution. This happens because the interface's connectable point passes from the blank value of '0' to the value received from the model which will be '1' in this case.

  • Send data during the initialization phase of the model and at run-time
    • Both the Send configuration data real-time and the Also send configuration initial values check-boxes need to be enabled
    • The initial value of each signal in the data frame table needs to be specified to its required value
    • Connections need to be created for each signal (or vector of signals)
    • A connection also needs to be created for the port's trigger
    • Attention must be paid on how to the trigger is used. Note that starting the simulation with the trigger set to '1' will cause a rising-edge transition to be detected during the first step of the execution. This happens because the interface's connectable point passes from the blank value of '0' to the value received from the model which will be '1' in this case.

  • Upload a .mat file
    • Only the Send .mat configuration file check-box needs to be enabled
    • The file's path needs to be specified in the .mat file path field
    • No connections need to be created

Signal configuration

When not sending .mat files, each signal in the data frame has the following parameters:

  • Name
    The name of the signal (or vector of signals). The name is used if the data is to be transferred during execution as it is the name given to the interface's connectable point.


  • Type
    The type is selectable from a drop down list. The options are BitUnsignedSignedFloat and Double. They represent how the data will be encoded before being sent to the FPGA.
    If the Vector size is greater than 1, then this parameter will be applied to each element of the vector.
    Certain types force other parameters to change their value to a fixed one:

    • The Bit encoding forces the Size (bits) parameter to 1
    • Float forces the Size (bits) parameter to 32 and the Bit offset parameter to 0
    • Double forces the Size (bits) parameter to 64 and the Bit offset parameter to 0


  • Vector size
    Setting this parameter different from 1 will result in the interface creating an array of signals starting at the byte offset specified. Each element of the vector will have the type, size, initial, minimum and maximum values as specified in its table entry. In case the data is to be sent during execution, the name of the connectable created for the vector will be the one specified at the parameter Name.
    Using vectors can be beneficial when a data frame has contiguous signals that have the same parameters because it reduces the time necessary to configure the interface.
    Keeping the value at 1 will maintain the current signal as an individual element of the data frame.
    Current constraints are related to vector elements having to be aligned on byte-boundaries:

    • The Bit offset parameter cannot be different from 0
    • The Size (bits) parameter must be one of the standard data type sizes (i.e. 1, 8, 16, 32 or 64)
    • In the case of Type being set to Binary, the maximum vector size is 8

  • Byte offset
    This parameter represents the byte offset within the load port where the current table element will be written to.
    Note that the interface has the means to check for data overlap: to check that the configuration is valid, the user must click on the Apply button of the I/O Interfaces Configuration window.
    The current constraint is that on the OP4200 platform, the byte offset must be divisible by 4 when the signal is of type float and divisible by 8 when the signal is of type double. 

  • Bit offset
    This parameter represents the bit offset within the byte specified at the Byte offset parameter for the current table element.
    Note that the interface has the means to check for data overlap: to check that the configuration is valid, the user must click on the Apply button of the I/O Interfaces Configuration window.
    The Bit offset cannot be used (i.e. it is grayed out) for Float and Double signal types. It also can't be used in case the table element is a vector (i.e. Vector size is greater than 1). 

  • Size (bits)
    The Size (bits) parameter is where the size in bits of the current signal can be specified.
    The size is fixed to 1 for Bit types, to 32 for Float types and to 64 for Double types. The maximum size allowed by the interface is 64 (which is the size of the largest native data type).
    If the Vector size is greater than 1, then this parameter will be applied to each element of the vector.
    Note that the interface has the means to check for data overlap: to check that the configuration is valid, the user must click on the Apply button of the I/O Interfaces Configuration window.

  • Initial value
    This parameter is only relevant in a scenario where data is sent during the initialization phase of the model (either exclusively or combined with sending at run-time). 
    In other words, the value specified in this field will only be sent to the FPGA once, while the simulation is initializing.
    If the Vector size is greater than 1, then this parameter will be applied to each element of the vector.
    The value is constrained between the minimum and maximum limits (see items below for a description).

  • Min
    The minimum value that will be sent to the FPGA for this signal. If the Vector size is greater than 1, then this parameter will be applied to each element of the vector.
    Note that the value sent from the HYPERSIM model could be lower than the minimum. In this case, the interface will send the minimum configured.
    The value entered for this parameter is limited by the type and size of the signal. As an example, for an 8-bit signed signal, the minimum cannot be less than -128 (the smallest signed integer representable on 8 bits).
    Furthermore, the value cannot be greater than the Max value (see item below).
    For real-time transfers, the current limitation of this parameter is that it is not applicable for Float and Double types.

  • Max
    The maximum value that will be sent to the FPGA for this signal. If the Vector size is greater than 1, then this parameter will be applied to each element of the vector.
    Note that the value sent from the HYPERSIM model could be greater than the maximum. In this case, the interface will send the maximum configured.
    The value entered for this parameter is limited by the type and size of the signal. As an example, for an 8-bit signed signal, the maximum cannot be greater than 127 (the largest signed integer representable on 8 bits).
    Furthermore, the value cannot be lower than the Min value (see item above).
    For real-time transfers, the current limitation of this parameter is that it is not applicable for Float and Double types. 

Characteristics and limitations

In order to correctly configure the data frames for each port, the user must have knowledge of how the data sent to the ports of interest will be used by the logic implemented in the bitstream.
The current version of the asynchronous raw data output functionality of the OPAL-RT Board driver has the following limitations:

  • The bit offset must be 0 when the vector size is greater than 1
  • The size (in bits) must be standard (i.e. 1, 8, 16, 32 or 64) when the vector size is greater than 1
  • The maximum vector size is 8 for the binary data type
  • On the OP4200 platform, the byte offset must be divisible by 4 for all signals of the type float
  • On the OP4200 platform, the byte offset must be divisible by 8 for all signals of type double
  • The bit offset cannot be used for any signals of type float or double (i.e. these types of data must be byte-aligned)
  • When sending data real-time, the minimum and maximum parameters are not applicable for signals of type float
  • When sending data real-time, the minimum and maximum parameters are not applicable for signals of type double
  • No labels