® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler 1.20 Features Provides an editor to enter the assembler instructions to configure the DFB block and an assembler that converts the assembly instructions to instruction words. Supports simulation of the assembly instructions. Supports a code optimization option that provides a mechanism to incorporate up to 128 very large instruction words inside the DFB Code RAM. Provides hardware signals such as DMA requests, DSI inputs and outputs, and interrupt lines. Supports semaphores to interact with the system software and the option to tie the semaphores to hardware signals. General Description The digital filter block (DFB) in PSoC 3 and PSoC 5 can be used as mini DSP processor and allows you to configure the DFB using assembly instructions. The component assembles the instructions entered in the editor and generates the corresponding hex code words, which can be loaded into the DFB. It also includes a simulator, which helps the user to simulate and debug the assembly instructions. The DFB consists of a programmable 24*24 multiplier/accumulator (MAC), an arithmetic logic unit (ALU), shifter, and various program and data memory to store instructions and data. The DFB runs on the bus clock, and can interface with both CPU and DMA. It can be used to offload the CPU and can speed up arithmetic calculations that involve intensive multiply accumulate operations. Typical operations you can use the DFB component to implement include: vector operations, matrix operations, filtering operations, and signal processing. See the Functional Description for details about the DFB. Cypress Semiconductor Corporation • 198 Champion Court • San Jose, CA 95134-1709 • 408-943-2600 Document Number: 001-84857 Rev. ** Revised December 4, 2012 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet Input/Output Connections This section describes the input and output connections for the DFB. An asterisk (*) in the list of I/Os means that the I/O may be hidden on the symbol under the conditions listed in the description of that I/O. in_1 – Input* Input terminal. Used to allow some DFB control and visibility to other resources on the chip, particularly the UDBs. This input displays when the Input 1 option is selected in the Configure dialog. in_2 – Input* Input terminal. Used to allow some DFB control and visibility to other resources on the chip, particularly the UDBs. This input displays when the Input 2 option is selected in the Configure dialog. out_1– Output* Output terminal. Allows the DFB signals to control other on-chip resources, particularly the UDBs. This output displays when the Output 1 option is selected in the Configure dialog. out_2– Output* Output terminal. Allows the DFB signals to control other on-chip resources, particularly the UDBs. This output displays when the Output 2 option is selected in the Configure dialog. dma_req_a – Output* DMA request output signal. It can be associated with the holding register or semaphore bits and is particularly useful to trigger DMA channels. This output displays when the DMA Request A Source option is selected in the Configure dialog. If the DMA request signal is configured to be associated with the output holding register, it generates a level-sensed signal to the DMA that is cleared when the register is read. If the DMA request signal is configured to be generated by a semaphore, it creates a single-cycle high pulse. dma_req_b – Output* DMA request output signal. It can be associated with the holding register or semaphore bits and is particularly useful to trigger DMA channels. This output displays when the DMA Request B Source option is selected in the Configure dialog. If the DMA request signal is configured to be associated with the output holding register, it generates a level-sensed signal to the DMA that is cleared when the register is read. Page 2 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler If the DMA request signal is configured to be generated by a semaphore, it creates a single-cycle high pulse. interrupt – Output Output of system interrupts. It can be associated with the holding registers and with semaphore bits. Component Parameters Drag a DFB onto your design and double-click it to open the Configure dialog. The Configure dialog contains several tabs with different parameters to set up the DFB component. Configure Tab Document Number: 001-84857 Rev. ** Page 3 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet Connections Enable input terminal 1 Enable input terminal 2 Enable output terminal 1 Enable output terminal 2 Output 1 Source Determines the internal signal that is mapped to the output global signal 1. Run bit – This is the same bit as the RUN bit in the DFB_CR register. Semaphore 0 Semaphore 1 Interrupt – This is the same signal as the primary DFB Interrupt output signal. Output 2 Source Determines the internal signal that is mapped to the output global signal 2. Semaphore 2 Datapath sign – This signal asserts any time the output of the ALU in the datapath unit is negative. It remains high for each cycle this condition is true. Datapath threshold crossed – This signal asserts any time the threshold of 0 is crossed in the ALU when one of the following instructions is executing: tdeca, tsuba, tsubb, taddabsa, or taddabsb. It remains high for each cycle this condition is true. Datapath ALU equals 0 – This signal asserts high when the output of the ALU in the Datapath unit equals 0 and one of the following ALU commands is executing: tdeca, tsuba, tsubb, taddabsa, or taddabsb. It remains high for each cycle this condition is true. Interrupt Generation Sources Configures the events on which interrupts will be generated: Semaphore 0 Semaphore 1 Semaphore 2 Page 4 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler Data in Holding Register A Data in Holding Register B DMA Request Mode Configures DMA Request sources. If DMA Request is not equal to None, a DMA request output terminal is created. DMA Request A Source: None Data in Holding Register A Semaphore 0 Semaphore 1 DMA Request B Source: None Data in Holding Register B Semaphore 0 Semaphore 1 This is all controlled using CSR configuration. When a DMA semaphore is programmed as a DMA_REQ, the HW converts any write of a ‘1’ to that semaphore to a single-cycle strobe. This clears the semaphore after one cycle. So, if a semaphore will be used as a source for a DMA request signal, it is cleared automatically. Code Tab The Code tab provides an editor to enter the assembler instructions, an assembly mechanism to verify and assemble code, and a simulator to simulate instructions. Document Number: 001-84857 Rev. ** Page 5 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet Assemble Assembles entered DFB assembler instructions. Status and error messages are displayed in the Output window of the customizer. Keyboard shortcut – [F6] Simulate Continuously The simulator runs until Bus Input Data (Bus1 Data and Bus2 Data) is exhausted. Bus1 data and Bus2 data are the input data for simulation that corresponds to data that will be streamed to DFB staging registers. Simulated contents of DFB sub-blocks after each instruction execution are displayed in the Output window. Keyboard shortcut – [F5] After simulation starts, use [Shift] F5] to stop it. Simulate Step by Step The simulator runs step by step until Bus Input Data is exhausted. At each step, the lines of code being executed are highlighted in the code editor. Simulation properties are displayed in the Simulation Properties panel. Keyboard shortcut – [F7] After simulation starts, use [F8] to execute the next simulation step. Page 6 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler Optimize Assembly States Enables the compactor feature. The compactor removes the 64-word program size barrier, allowing you to incorporate up to 128 instruction words inside the DFB Code RAM. If the Optimize Assembly States option is selected, the code compacting feature runs after successful code assembly. Code Editor Displays DFB assembler code with highlighting. It highlights comments, labels, commands, and values with different colors. Figure 1. Text Editor Toolbar Text Editor Open file: Opens a text file with DFB assembler code. Keyboard shortcut – [Ctrl] [O] Save file: Saves a text file with DFB Assembler code. Keyboard shortcut – [Ctrl] [S] Other text editor features: Includes cut, copy, paste, undo, redo, print, print preview, find/replace text, comment, uncomment, and other functionality. Output panel Displays log information for the assembly, compactor, and simulator processes. Lines with error text are highlighted red, lines that indicate successful operations are highlighted green. If you double-click on a line with error text, the customizer automatically activates the Code Editor and selects the line where the error is located. Use the keyboard shortcut [Alt] [C] to clear the log. Error List Displays a list of errors, warnings, and notes. If you double-click on an error, the customizer automatically activates Code Editor and selects the line where error is located. Bus1 Provides data for STAGEA input for the simulator. Enter it as a 24-bit value in hex, decimal, or binary format. For example, 99 is decimal, 0x63 is hex, and 0b1100011 is binary. Bus2 Provides data for STAGEB input for the simulator. Enter it as a 24-bit value in hex, decimal, or binary format. Document Number: 001-84857 Rev. ** Page 7 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet Bus data import Imports data to the Bus1/Bus2 text fields. Supports .txt and .data (an old data format that was used in the C-based implementation of the simulator) file formats. Bus data export Exports data from the Bus1/Bus2 text fields. Supports .txt and .data file formats. Simulator Properties Provides functionality to change input values and semaphores and view internal simulator values between simulations. GlobalInput1: Read/Write field GlobalInput2: Read/Write field Semaphore0: Read/Write field Semaphore1: Read/Write field Semaphore2: Read/Write field Cycle: Displays current cycle number RamA Index: Read-only field that represents the current index in RAM A. RamB Index: Read-only field that represents the current index in RAM B. Ram Selected: Read-only field that shows which ram (A or B) is currently executed Application Programming Interface Application Programming Interface (API) routines allow you to configure the component using software. The following table lists and describes the interface to each function. The subsequent sections discuss each function in more detail. By default, PSoC Creator assigns the instance name “DFB_1” to the first instance of a component in a given design. You can rename it to any unique value that follows the syntactic rules for identifiers. The instance name becomes the prefix of every global function name, variable, and constant symbol. For readability, the instance name used in the following table is “DFB.” Page 8 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Function Digital Filter Block (DFB) Assembler Description DFB_Start() Initializes and enables the DFB component using the DFB_Init() and DFB_Enable() functions. DFB_Stop() Turns off the run bit. If DMA control is used to feed the channels, allows arguments to turn off one of the TD channels. DFB_Pause() Pauses DFB and enables writing to the DFB RAM. DFB_Resume() Disables writing to the DFB RAM, clears any pending interrupts, disconnects the DFB RAM from the data bus, and runs the DFB. DFB_SetCoherency() Sets the coherency key to low/mid/high byte based on the coherencyKey parameter that is passed to the DFB. DFB_SetDalign() Allows 9- to 16-bit input and output samples to travel as 16-bit values on the AHB bus. DFB_LoadDataRAMA() Loads data to RAMA DFB memory. DFB_LoadDataRAMB() Loads data to RAMB DFB memory. DFB_LoadInputValue() Loads the input value into the selected channel. DFB_GetOutputValue() Gets the value from one of the DFB output holding registers. DFB_SetInterruptMode() Assigns the events that will trigger a DFB interrupt. DFB_GetInterruptSource() Looks at the DFB_SR register to see which interrupt sources have been triggered. DFB_ClearInterrupt() Clears the interrupt request. DFB_SetDMAMode() Assigns the events that will trigger a DMA request for the DFB. DFB_SetSemaphores() Sets semaphores specified with a 1. DFB_ClearSemaphores() Clears semaphores specified with a 1. DFB_GetSemaphores() Checks the current status of the DFB semaphores and returns that value. DFB_SetOutput1Source() Chooses which internal signals will be mapped to output 1. DFB_SetOutput2Source() Chooses which internal signals will be mapped to output 2. DFB_Sleep() Prepares the DFB component to go to sleep. DFB_Wakeup() Prepares DFB Component to wake up. DFB_Init() Initializes or restores the default DFB configuration provided with the customizer. DFB_Enable() Enables the DFB hardware block. Sets the DFB run bit. Powers on the DFB block. DFB_SaveConfig(void) Saves the user configuration of the DFB nonretention registers. This routine is called by DFB_Sleep() to save the component configuration before entering sleep. DFB_RestoreConfig() Restores the user configuration of the DFB nonretention registers. This routine is called by DFB_Wakeup() to restore the component configuration when exiting sleep. Document Number: 001-84857 Rev. ** Page 9 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet Global Variables Variable DFB_initVar Description Indicates whether the DFB has been initialized. The variable is initialized to 0 and set to 1 the first time DFB_Start() is called. This allows the component to restart without reinitialization after the first call to the DFB_Start() routine. If reinitialization of the component is required, then the DFB_Init() function can be called before the DFB_Start() or DFB_Enable() function. void DFB_Start(void) Description: This function initializes and enables the DFB component using the DFB_Init() and DFB_Enable() functions. Parameters: None Return Value: None Side Effects: None void DFB_Stop(void) Description: This function turns off the run bit. If DMA control is used to feed the channels, DFB_Stop() allows arguments to turn off one of the TD channels. Parameters: None Return Value: None Side Effects: Disables power to the DFB core. void DFB_Pause(void) Description: Parameters: This function pauses the DFB and enables writing to the DFB RAM. Turns off the run bit Connects the DFB RAM to the data bus, Clears the DFB run bit and passes the control of all DFB RAMs onto the bus None Return Value: None Side Effects: Page 10 of 42 None Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler void DFB_Resume(void) Description: This function disables writing to the DFB RAM, clears any pending interrupts, disconnects the DFB RAM from the data bus, and runs the DFB. It passes the control of all DFB RAM to the DFB and then sets the run bit. Parameters: None Return Value: None Side Effects: None void DFB_SetCoherency(uint8 coherencyKeyByte) Description: This function sets the coherency key to low, med, or high byte based on the coherencyKey parameter that is passed to the DFB. DFB_SetCoherency() allows you to select which of the three bytes of each of STAGEA, STAGEB, HOLDA, and HOLDB will be used as the key coherency byte. Coherency refers to the HW added to this block to protect against block malfunctions. This is needed in cases where register fields are wider than the bus access, which leaves intervals when fields are partially written or read (incoherent). The key coherency byte is the way the SW tells the HW which byte of the field will be written or read last when you want to update the field. When the key byte is written or read, the field is flagged coherent. If any other byte is written or read, the field is flagged incoherent. Parameters: uint8 coherencyKeyByte: Specifies bits in the DFB Coherency register. Value Description DFB_STGA_KEY_LOW The key coherency byte of the Staging A register is a low byte. DFB_STGA_KEY_MID The key coherency byte of the Staging A register is a med byte. DFB_STGA_KEY_HIGH The key coherency byte of the Staging A register is a high byte. DFB_STGB_KEY_LOW The key coherency byte of the Staging B register is a low byte. DFB_STGB_KEY_MID The key coherency byte of the Staging B register is a med byte. DFB_STGB_KEY_HIGH The key coherency byte of the Staging B register is a high byte. DFB_HOLDA_KEY_LOW The key coherency byte of the Holding A register is a low byte. DFB_HOLDA_KEY_MID The key coherency byte of the Holding A register is a med byte. DFB_HOLDA_KEY_HIGH The key coherency byte of the Holding A register is a high byte. DFB_HOLDB_KEY_LOW The key coherency byte of the Holding B register is a low byte. DFB_HOLDB_KEY_MID The key coherency byte of the Holding B register is a med byte. DFB_HOLDB_KEY_HIGH The key coherency byte of the Holding B register is a high byte. Return Value: None Side Effects: Coherency affects data loading using the DFB_LoadInputValue() function and data retrieval using the DFB_GetOutputValue() function. Note Default key byte configuration for Staging A and B, and Holding A and B registers is high byte. Document Number: 001-84857 Rev. ** Page 11 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet void DFB_SetDalign(uint8 dalignKeyByte) Description: This feature allows 9- to 16-bit input and output samples to travel as 16-bit values on the AHB bus. These bits, when set high, cause an 8-bit shift in the data to all access of the corresponding staging and holding registers. Because the DFB datapath is MSB aligned, it is convenient to the system SW to align values on bits 23:8 of the Staging and Holding register to bits 15:0 of the bus. A read of this register produces the last value written to the register. Parameters: uint8 dalignKeyByte: Specifies bits in the DFB Data Alignment register. Value Description DFB_STGA_DALIGN_LOW Writes normally DFB_STGA_DALIGN_HIGH Writes shifted left by eight bits DFB_STGB_DALIGN_LOW Writes normally DFB_STGB_DALIGN_HIGH Writes shifted left by eight bits DFB_HOLDA_DALIGN_LOW Reads normally DFB_HOLDA_DALIGN_HIGH Writes shifted left by eight bits DFB_HOLDB_DALIGN_LOW Reads normally DFB_HOLDB_DALIGN_HIGH Writes shifted left by eight bits Return Value: None Side Effects: None void DFB_LoadDataRAMA(int32 * ptr, uint32 * addr, uint8 size) Description: Parameters: This function loads data to the DFB RAM A memory. uint32 * ptr: Pointer on data source for load uint32 * addr: Start address for loading the data in DFB RAM A. uint8 size: Number of data words to load. Return Value: Value Side Effects: Page 12 of 42 Description DFB_SUCCESS Loading data is successful. DFB_NAME`_ADDRESS_OUT_OF_RANGE Error code: indicates that the address is out of range. DFB_DATA_OUT_OF_RANGE Error code: indicates a data overflow error. This function does not stop the DFB if it is already started. The recommended method is to call DFB_Init(), DFB_LoadDataRAMA(), then DFB_Enable(). Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler void DFB_LoadDataRAMB(uint32 * ptr, uint32 * addr, uint8 size) Description: This function loads data to DFB RAM B memory. Parameters: uint32 * ptr: Pointer on data source for load uint32 * addr: Start address for loading the data in DFB RAM B uint8 size: Number of data words to load Return Value: Side Effects: Value Description DFB_SUCCESS Loading data is successful DFB_NAME`_ADDRESS_OUT_OF_RANGE Error code: indicates that the address is out of range DFB_DATA_OUT_OF_RANGE Error code: indicates a data overflow error. This function does not stop the DFB if it is already started. The recommended method is to call DFB_Init(), DFB_LoadDataRAMB(), then DFB_Enable(). void DFB_LoadInputValue(uint8 channel, uint32 sample) Description: This function loads the input value into the selected channel. Parameters: channel: Use either DFB_CHANNEL_A (1) or DFB_CHANNEL_B (0) as arguments to the function. sample: 24-bit, right-justified input sample Return Value: None Side Effects: None Note: The write order is important. When the high byte is loaded, the DFB sets the input ready bit. Pay attention to byte order if coherency or data alignment is changed. int32 DFB_GetOutputValue(uint8 channel) Description: This function gets the value from one of the DFB Output Holding registers. Parameters: channel: Use either DFB_CHANNEL_A (1) or DFB_CHANNEL_B (0) as arguments to the function. Return Value: The current output value in the chosen channel’s holding register. This is a 24-bit number packed into the least-significant three bytes of the output word, or 0xFF000000 for invalid channel numbers Side Effects: None Note: Because of the architecture of the DFB, any value read from the holding A or B registers will be MSB aligned unless shifted otherwise by the datapath shifter. Pay attention to byte order if coherency or data alignment is changed. Document Number: 001-84857 Rev. ** Page 13 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet void DFB_SetInterruptMode(uint8 events) Description: This function assigns the events that trigger a DFB interrupt. Parameters: events: Bits[0:5] of events represent the events that trigger DFB interrupts. Value Description DFB_HOLDA Interrupt is generated each time new valid data is written into output holding register A. DFB_HOLDB Interrupt is generated each time new valid data is written into output holding register B. DFB_SEMA0 Interrupt is generated each time a '1' is written into semaphore register bit 0. DFB_SEMA1 Interrupt is generated each time a '1' is written into semaphore register bit 1. DFB_SEMA2 Interrupt is generated each time a '1' is written into semaphore register bit 2. Return Value: None Side Effects: None Note: Do not configure semaphore 0 and semaphore 1 for both a DMA request and for an interrupt event. This is because, after one clock cycle, the system automatically clears any semaphore that triggered a DMA request. uint8 DFB_GetInterruptSource(void) Description: This function looks at the DFB_SR register to see which interrupt sources have been triggered. Parameters: None Return Value: uint8 value in which bits[0:5] of represent the events that triggered the DFB interrupt Value Side Effects: Page 14 of 42 Description DFB_HOLDA Holding register A is a source of the current interrupt. DFB_HOLDB Holding register B is a source of the current interrupt. DFB_SEMA0 Semaphore register bit 0 is a source of the current interrupt. DFB_SEMA1 Semaphore register bit 1 is a source of the current interrupt. DFB_SEMA2 Semaphore register bit 2 is a source of the current interrupt. None Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler void DFB_ClearInterrupt(uint8 interruptMask) Description: This function clears the interrupt request. Parameters: interruptMask: Mask of interrupts to clear Value Description DFB_HOLDA Clear interrupt from holding register A. (Reading holding register A also clears this bit.) DFB_HOLDB Clear interrupt from holding register B. (Reading holding register B also clears this bit.) DFB_SEMA0 Clear interrupt from semaphore register bit 0. DFB_SEMA1 Clear interrupt from semaphore register bit 1. DFB_SEMA2 Clear interrupt from semaphore register bit 2. Return Value: None Side Effects: Clearing semaphore interrupts also clears semaphore bits. void DFB_SetDMAMode(uint8 events) Description: This function assigns the events that trigger a DMA request for the DFB. Two different DMA requests can be triggered. Parameters: events: A set of four bits that configure what event, if any, triggers a DMA request for the DFB. Value Description DFB_DMAREQ1_DISABLED No request is generated DFB_DMAREQ1_HOLDA Output value ready in the holding register on channel A DFB_DMAREQ1_SEM0 Semaphore 0 DFB_DMAREQ1_SEM1 Semaphore 1 DFB_DMAREQ2_DISABLED No request is generated DFB_DMAREQ2_HOLDB Output value ready in the holding register on channel B DFB_DMAREQ2_SEM0 Semaphore 0 DFB_DMAREQ2_SEM1 Semaphore 1 Return Value: None Side Effects: None Note: Do not configure semaphore 0 and semaphore 1 as both a DMA request and an interrupt event. This is because, after one clock cycle, the system automatically clears any semaphore that triggered a DMA request. Document Number: 001-84857 Rev. ** Page 15 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet void DFB_SetSemaphores(uint8 mask) Description: This function sets semaphores specified with a 1. Parameters: mask: Mask specifying the bits to set Value Description DFB_SEMAPHORE0 Semaphore 0 DFB_SEMAPHORE1 Semaphore 1 DFB_SEMAPHORE2 Semaphore 2 Return Value: None Side Effects: None void DFB_ClearSemaphores(uint8 mask) Description: This function clears semaphores specified with a 1. Parameters: mask: Mask specifying the bits to clear. Value Description DFB_SEMAPHORE0 Semaphore 0 DFB_SEMAPHORE1 Semaphore 1 DFB_SEMAPHORE2 Semaphore 2 Return Value: None Side Effects: None uint8 DFB_GetSemaphores(void) Description: This function checks the current status of the DFB semaphores and returns that value. Parameters: None Return Value: uint8 value between 0 and 7 where bit 0 represents semaphore 0, and so on. Value Side Effects: Page 16 of 42 Description DFB_SEMAPHORE0 Semaphore 0 DFB_SEMAPHORE1 Semaphore 1 DFB_SEMAPHORE2 Semaphore 2 None Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler void DFB_SetOutput1Source(uint8 source) Description: This function allows you to choose which internal signals are mapped to output 1. Parameters: source: Internal signal that is mapped to output global signal 1. Signal Description DFB_RUN DFB run bit. This is the same bit as the run bit in the DFB_CR register. DFB_SEM0 Semaphore Bit 0. DFB_SEM1 Semaphore Bit 1. DFB_DFB_INTR DFB Interrupt. This is the same signal as the primary DFB interrupt output signal. Return Value: None Side Effects: None void DFB_SetOutput2Source(uint8 source) Description: This function allows you to choose which internal signals are mapped to output 2. Parameters: source: Internal signal that is mapped to output global signal 2. Signal Description DFB_SEM2 Semaphore bit 2. DFB_DPSIGN Datapath sign. This signal asserts any time the output of the ALU in the datapath unit is negative. It remains high for each cycle this condition is true. DFB_DPTHRASH Datapath threshold crossed. This signal asserts any time the threshold of 0 is crossed in the ALU and one of the following instructions is executing: tdeca, tsuba, tsubb, taddabsa, or taddabsb. It remains high for each cycle this condition is true. DFB_DPEQ Datapath ALU = 0. This signal asserts high when the output of the ALU in the datapath unit equals 0 and one of the following ALU commands is executing: tdeca, tsuba, tsubb, taddabsa, or taddabsb. It remains high for each cycle this condition is true. Return Value: None Side Effects: None Document Number: 001-84857 Rev. ** Page 17 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet void DFB_Sleep(void) Description: This is the preferred routine to prepare the component for sleep. The DFB_Sleep() routine saves the current component state. Then it calls the DFB_Stop() function and calls DFB_SaveConfig() to save the hardware configuration. Call the DFB_Sleep() function before calling the CyPmSleep() or the CyPmHibernate() function. See the PSoC Creator System Reference Guide for more information about power management functions. Parameters: None Return Value: None Side Effects: None void DFB_Wakeup(void) Description: This is the preferred routine to restore the component to the state when DFB_Sleep() was called. The DFB_Wakeup() function calls the DFB_RestoreConfig() function to restore the configuration. If the component was enabled before the DFB_Sleep() function was called, the DFB_Wakeup() function will also re-enable the component. Parameters: None Return Value: None Side Effects: Calling the DFB_Wakeup() function without first calling the DFB_Sleep() or DFB_SaveConfig() function may produce unexpected behavior. void DFB_Init(void) Description: Parameters: This function initializes or restores the default DFB component configuration provided with the customizer: Powers on the DFB (PM_ACT_CFG) and the RAM (DFB_RAM_EN) Moves CSA/CSB/FSM/DataA/DataB/Address calculation unit (ACU) data to the DFB RAM using an 8051/ARM core Changes RAM DIR to DFB Sets the interrupt mode Sets the DMA mode Sets the DSI outputs Clears all semaphore bits and pending interrupts None Return Value: None Side Effects: Page 18 of 42 All registers will be reset to their initial values. This reinitializes the component. This function turns off the run bit and enables power to the DFB block. Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler void DFB_Enable(void) Description: This function enables the DFB hardware block, sets the DFB run bit, and powers on the DFB block. Parameters: None Return Value: None Side Effects: None void DFB_SaveConfig(void) Description: This function saves the component configuration and nonretention registers. It also saves the current component parameter values, as defined in the Configure dialog or as modified by appropriate APIs. This function is called by the DFB_Sleep() function. Parameters: None Return Value: None Side Effects: None void DFB_RestoreConfig(void) Description: This function restores the component configuration and nonretention registers. It also restores the component parameter values to what they were before calling the DFB_Sleep() function. Parameters: None Return Value: None Side Effects: Calling this function without first calling the DFB_Sleep() or DFB_SaveConfig() function can produce unexpected behavior. Defines ClearInterruptSource(event) – Macro for clearing interrupts MISRA Compliance This section describes the MISRA-C:2004 compliance and deviations for the component. There are two types of deviations defined: project deviations – deviations that are applicable for all PSoC Creator components specific deviations – deviations that are applicable only for this component Document Number: 001-84857 Rev. ** Page 19 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet This section provides information on component-specific deviations. Project deviations are described in the MISRA Compliance section of the System Reference Guide along with information on the MISRA compliance verification environment. The DFB component does not have any specific deviations. Sample Firmware Source Code PSoC Creator provides many example projects that include schematics and example code in the Find Example Project dialog. For component-specific examples, open the dialog from the Component Catalog or an instance of the component in a schematic. For general examples, open the dialog from the Start Page or File menu. As needed, use the Filter Options in the dialog to narrow the list of projects available to select. Refer to the “Find Example Project” topic in the PSoC Creator Help for more information. Functional Description The Digital Filter Block is a 24-bit fixed-point, programmable limited-scope DSP with a 24*24 Multiply and Accumulate Unit (MAC), a multifunction Arithmetic Logic Unit (ALU), and data routing, shifting, holding, and rounding functions. Other important features of the DFB are: Two 24-bit-wide streaming data channels Two sets of control store RAMs, each of which can store up to 64 control words Two sets of data RAMs, each of which can store up to 128 24-bit-wide words Address calculation units (ACU) to calculate the data RAM address and two ACU RAMs, each of which can store up to 16 absolute data RAM addresses Two sets of 32*32 Finite State Machine RAM to store the control flow (branching) information One interrupt and two DMA request channels Three semaphore bits to interact with system software Data alignment and coherency protection support options for input and output registers The DFB supports up to two streaming data channels, where programming instructions, historic data and filter coefficients, and results are stored locally with new periodic data samples received from the AHB interface. In addition, the system SW can ‘load sample or coefficient data in or out of the DFB data RAMs, reprogram for different filter operation in ‘block mode’, or both. This allows for multichannel processing or deeper filters than are supported in local memory. The block provides SW-configurable interrupt and two-DMA-channel support. There are three semaphore bits for SW to interact with the DSP assembly program. Page 20 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler The DFB has two 24-bit input staging registers and two 24-bit output holding registers. These registers can be accessed by both the DFB and the AHB bus (CPU/DMA). Input data is generally streamed into staging registers using CPU or DMA and output is streamed out through DFB holding registers. The two sets of input /output registers make it well suited for stereo data processing applications (two channels in parallel). These inout registers support 32-bit, 16-bit, and 8-bit accesses and have coherency protection HW allowing them to be written or read in less than 32-bit accesses. Figure 2. DFB Application Diagram The typical use model is for data to be supplied to the DFB over the system bus from another onchip system data source such as an ADC. The data typically passes through main memory or is directly transferred from another chip resource through DMA. The DFB architecture supports parallel operation of the ACUs, ALU, MAC, and shifter. The operations allowed for each are encoded in bit fields. The basic assembly instruction set is given in terms of these functions. The instruction pipelining follows Figure 3 for the DFB processor. The diagram shows the locations of the pipeline registers so you can determine the instruction pipeline latency. While the ACU/DPRAM, MAC, and ALU/shifter all can operate in parallel, the instruction delay from one block to the next is one cycle of overhead. Suppose, for example, that you wanted to specify a new ACU address, multiply and accumulate based on that address, then see if the output exceeds a threshold. You would need to schedule the ACU address in the first cycle, the MAC in the second cycle, and the threshold in the third cycle after that. If the threshold flag from the data path is to be used in the controller for a branch, that branch cannot happen until the fourth cycle in the sequence. This is not usually an issue, as algorithms can be scheduled to avoid four-cycle delays in branches. Typically, an algorithm will see a one-cycle delay between the last command and the branch, because the steps leading up to the branch can be combined with previous statements in the algorithmic flow. The ACUs are positioned so there is no cycle delay between detection and branching. Document Number: 001-84857 Rev. ** Page 21 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet Figure 3. Data Flow/Pipelining Diagram Any of the semaphore bits can optionally be programmed to be associated with the system interrupt signal or either of the DMA_REQ outputs, leaving the DFB, either of the output DSI signals (Out_1, Out_2). Data in DFB and particularly in DATA RAM A/B memories are represented in two's complement format. The DFB operates on 24-bit signed arithmetic values. Valid values are from 0 to 16,777,215. The DFB component is oriented on filtering algorithms and has a range from –1 to 1. Value 1 (0.9999999) equals 0x7FFFFF (8388607), 0 equals 0x000000 (0), –1 equals 0x800000 (8388608), –0.0000001 equals 0xFFFFFF (16,777,215), 0.0000001 equals 0x000001 (1). The 24th bit is the sign. Simulator Output The output of the simulator contains information that can be used to debug the design. The data inputs coming into the DFB are simulated using the Bus1 (Stage register A) and Bus2 (Stage register B) windows located at the top right of the Code Tab. Semaphores, global inputs, and global outputs can be simulated by modifying the contents in the Simulation properties window. Note that the API calls made in the firmware do not affect the simulation properties. Therefore these behaviors need to be replicated in the simulator. These include populating the Data RAMs, writing to the Staging registers, reading from the Holding registers, and setting/unsetting the semaphores and global inputs. Page 22 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Signal Name Type Digital Filter Block (DFB) Assembler Description Cycle uint This is the program counter that counts the number of clock cycles. RamA uint The current RAM location in Control Store RAM A, also referred to as CStoreA. RamB uint The current RAM location in Control Store RAM B, also referred to as CStoreB. Ram sel string The Control Store RAM (either A or B) currently being executed. CFSM state uint Current state of the CFSM. This value is not the same as the RAM values of the CFSM content. The states are ordered as they appear in the assembler code. Therefore the first state to be called in the code is designated CFSM=1, and the second state to be called is designated CFSM=2 etc. Aaddr next uint The current address of Data RAM A Baddr next uint The current address of Data RAM B A2Mux Hex The output from Mux2 (A), which corresponds to either the output of Mux1 (A) or the current content in the Data Ram A. B2Mux Hex The output from Mux2 (B), which corresponds to either the output of Mux1 (B) or the current content in the Data Ram B. MacOut Hex The output of the Multiply and Accumulate Unit (MAC). AluOut Hex The output of the Arithmetic Logic Unit (ALU). ShiftOut Hex The output of the shifter located at the output of the ALU. DFB Compactor Using the optimizer feature (Optimize Assembly States tab), all 128 memory entries are available for program store in the DFB. The DFB supports code flow execution alternating from one 64entry code store to the other, creating zero overhead looping and branches. When the code stores are not identical, all 128 memory entries are available for programming. The compactor divides the program into states and puts the program states in one of the two control stores. The compactor also generates jump addresses between control stores. In a normal situation, jumps between program states go from one control store to another. A program cannot require a jump within the same control store. For example, you have a routine called FILTER, and it is partitioned to be in RAM A. You also have two routines, R1 and R2, both of which jump to FILTER. If R1 and R2 are both in RAM B, there is no problem, but if either is in RAM A, then the program cannot jump. This situation causes an error at assembly. To resolve this issue, the output panel provides Compactor information such as: the content of RAM A, RAM B and a description of control finite state machine (CFSM) content. The content of RAM A and RAM B contains information about program states. The CFSM content description contains information about jumps between program states. A code profiler was added to allow you to see where the lines of code were placed in the control store (cstore). Because the input assembly language is cycle- and line-oriented, there is a one- Document Number: 001-84857 Rev. ** Page 23 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet to-one correspondence between a line in the assembly language and an entry in the cstore. This allows you to gather meaningful information about how to optimize your code. When a DFB program has an issue with assembly state optimization, the following error appears after assembling: “ERROR: Unable to map to split RAMs. Found N states that can't be mapped. Please analyze results and verify code can be mapped.” When the state has indexed as -01, the file shows that the code is not in the correct RAM. You can deduce the problem by looking at the RAM and state information. Using the log file, you can construct the block diagram of the flow and identify the block allocation. To resolve your issue, you can place additional dummy states into your program. In the case of the problem mentioned earlier, it will cost a single dummy state and a single instruction. For example, if you have a problem with a jump to the nextState program state, you can jump to dummy state dummy1. From this state you can jump to nextState. See the following single instruction. dummy1: acu(hold,hold) dmux(sa,sa) alu(hold) mac(hold) jmp(eob, nextState) The following screenshot shows an example of this process. Page 24 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler Resources The DFB component uses the dedicated DFB hardware block in the silicon. API Memory Usage The component memory usage varies significantly, depending on the compiler, device, number of APIs used and component configuration. The following table provides the memory usage for all APIs available in the given component configuration. The measurements have been done with the associated compiler configured in Release mode with optimization set for Size. For a specific design the map file generated by the compiler can be analyzed to determine the memory usage. PSoC 3 (Keil_PK51) Configuration Default PSoC 5 (GCC) PSoC 5LP (GCC) Flash SRAM Flash SRAM Flash SRAM Bytes Bytes Bytes Bytes Bytes Bytes 1432 + size of DFB program* 2 1092 + size of DFB [1] program 5 1092 + size of DFB [1] program 5 DFB Data Memory (RAM) DATA A DATA B CS A CS B FSM 128x24 128x24 64x32 64x32 64x32 [2] ACU 16x14 The various DFB RAMS can be accessed by the DFB or the system (CPU/DMA) AHB bus, but not by both at once. In cases where bulk data must be moved into the DFB RAMs, pass the control of the DFB RAM to the system AHB bus (CPU/DMA), load the new data to the DFB RAMs, and pass control back to the DFB. The DFB_RAM_DIR register controls whether the DFB RAMs can be accessed by the DFB or the system bus (CPU/DMA). RAM Name Size Functions DATA A 128x24 Sample/Coeff Store DATA B 128x24 Sample/Coeff Store CS A 64x32 Control Store 1. The size of the DFB program is restricted by the DFB data memory size shown in the following table 2. For current DFB implementation, only half (32x32) of the FSM memory is available. Document Number: 001-84857 Rev. ** Page 25 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet CS B 64x32 Control Store FSM 64x32 Finite state machine RAM ACU 16x14 Address Store DC and AC Electrical Characteristics Specifications are valid for –40 °C £ TA £ 85 °C and TJ £ 100 °C, except where noted. Specifications are valid for 1.71 V to 5.5 V, except where noted. DC Specifications Parameter Description DFB operating current Conditions Min Typ Max Units 100 kHz (1.3 ksps) – 0.03 0.05 mA 500 kHz (6.7 ksps) – 0.16 0.27 mA 1 MHz (13.4 ksps) – 0.33 0.53 mA 10 MHz (134 ksps) – 3.3 5.3 mA 48 MHz (644 ksps) – 15.7 25.5 mA 67 MHz (900 ksps) – 21.8 35.6 mA 80 MHz (1.07 Msps) (only for PSoC5) – 26.1 42.5 mA Conditions Min Typ Max Units PSoC3 DC – 67.01 MHz PSoC5 DC – 80.01 MHz 64-tap FIR at FDFB AC Specifications Parameter FDFB Description DFB operating frequency DFB Assembler Directive Descriptions AREA The “area” directive takes an argument specifying the accessibility of a RAM segment. The selected RAM may or may not be enclosed in parenthesis: area(RAM_Name) or area RAM_Name Page 26 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler There are only three RAMs, data_a, data_b, and acu, that you would logically need to access and modify using the dw (define word) command. Control and CFSM RAM are accessible, but it does not make sense to manually code them when the assembler is designed to enter those opcodes for you. Comments Comments for the assembler are the same as C style line comments, with the stipulation that they occupy the entire line. // Designates the line to be a comment line. Everything is // ignored by the assembler ORG The “org” directive sets the current location counter (CLC) for the current RAM. Initially, each RAM’s CLC is set to zero. The value of a location must be an integer and must be constrained to the set of numbers that represent a valid memory location. Both of the following are acceptable formats for this instruction. org(location) or org location dw Define Word. After the instruction, the argument is the value that is place in the memory of the current section and CLC for that section. The CLC then increments. The customizer generates and displays an error message when the program tries to write a value that exceeds the maximum value of the RAM. Hexadecimal arguments beginning with the prefix “0x” are acceptable inputs for the “acu” area. This greatly clarifies the separation between values for side A and side B in the ACU RAM. Integer values representing 1-2 -23 to –1 in 24-bit, 2’s complement format (0 to 16,777,215) are acceptable inputs for data areas. In an ACU RAM, two 7-bit sides allow input data using 4-digit hexadecimal inputs. This makes it clear what value is placed in each side of the 14-bit wide RAM, as the first two digits are entered into side A and the second two into side B. The valid value range for each side is from 0x00 to 0x7F. Example: dw 0x123F // (Decimal 18 in ACU RAM side A, 63 in side B) Labels Labels are user-defined tags to refer to blocks of cstore code. Blocks of code are sections beginning with a valid label and ending with a jump instruction, which together become a state in the CFSM. With the exception of the keywords listed as instructions in this document, a label can be defined by a set of characters that start with a letter and are followed by a colon. Do not put declarations on the same line as a label. For DFB hardware commands to be executed they must be located inside of a label/jump block; the two of these together create a state. Two labels with no jump-terminated block separating them are not permitted. An example declaration is given. Document Number: 001-84857 Rev. ** Page 27 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet // MyLabel defines a new state for the cfsm. The location of // the state’s start in Cstore is attached MyLabel: VLIW Commands Each line of commands defines opcodes in a 32-bit very long instruction word (VLIW). The most inclusive form of command for a single instruction word is shown below in the required order. Required commands in each line are emphasized using bold text. The following section outlines each individual instruction. An instruction of the proper instruction sets should be entered in between the parentheses. ACU and DMUX instructions both require two instructions separated by a comma, the first for datapath side A, the second for side B. acu(,) addr() dmux(,) alu() mac() shift() write() jump() Each instruction contributes a short opcode to the VLIW, resulting in a 32-bit instruction control word for each line. ACU The address calculation unit (ACU) outputs the data RAM addresses required for the next instruction cycle. A single ACU is basically a counter with four registers in a default state of zero. reg – reg stores the current value that the ACU is operating on and outputs it on every cycle unless a command specifies otherwise. freg – freg can be loaded with the value that the data RAMs increment or decrement, when using the addf and subf commands. For example: load two into freg and you can increment through the data RAMs by two using ACU’s ‘addf’ instruction. mreg – mreg stores the maximum value before wraparound to the lreg value when modulus arithmetic is enabled. lreg – lreg stores the minimum value before wraparound to the mreg value when modulus arithmetic is enabled. Modulus arithmetic prevents the ACU from incrementing past the value of mreg and from decrementing below the value of lreg. The ACU will produce unexpected (although deterministic) results when modulus arithmetic is enabled and the current address is located outside of the lreg to mreg range. Good DFB programming practice requires that you ensure, using the “read” command or through careful inspection, that the value the ACU points to begins in a valid location. A 16-row-deep RAM accompanies the ACU to store values needed for storing absolute addresses of data RAM sections that are required during program execution. It also stores other values that the ACU might need access to, such as values for freg. During run time, the preferred way to place data into the ACU RAM is through system software intervention. Use of the ACU command is defined as follows: Page 28 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler acu(instruction_A, instruction_B) The two different instructions, instruction_A and instruction_B, represent members of the ACU instruction set that control the addresses for two data RAMs independently. The following table shows a comprehensive list of the ACU instruction set. Instruction Description hold Puts the registered output address value on the output unchanged. incr Increases the registered output address value (reg) by one and puts it on the output. decr Decreases the registered output address value (reg) by one and puts it on the output. read Gets the specified byte from the ACU RAM, loads it to the output address register, and puts the value on the output. (See the addr command.) write Puts the registered output address value into the specified ACU RAM row. (See the addr command.) loadf Loads freg with the value from the specified ACU RAM. The ACU output value remains the same as the previous cycle. (See the addr command for information about ACU RAM addressing.) loadl Loads lreg with the value from the specified ACU RAM. The ACU output value remains the same as the previous cycle. (See the addr command for information about ACU RAM addressing.) loadm Loads mreg with value of the specified ACU RAM. The ACU output value remains the same as the previous cycle. (See the addr command for information about ACU RAM addressing.) writel Puts the value of lreg on the output and then writes the output to the specified ACU RAM location. (See the addr command.) setmod Enables modulus arithmetic in the ACU. Modulus arithmetic is on by default. unsetmod Disables modulus arithmetic in the ACU. Modulus arithmetic is on by default. clear Sets the registered output value (reg) to zero. addf Increases the registered output address value (reg) by the value in register freg. When modulus arithmetic is disabled and the ACU output is outside the region defined between lreg and mreg, do not use this command. subf Decreases the registered output address value (reg) by the value in register freg. When modulus arithmetic is disabled and the ACU output is outside the region defined between lreg and mreg, do not use this command. writem Puts the value of mreg on the output and writes the output to the specified ACU RAM location. (See the addr command.) writef Puts the value of freg on the output and writes the output to the specified ACU RAM location. (See the addr command.) addr The addr command takes a value between zero and fifteen as an argument. It is used in several different ways. Be careful to write programs in such a way that commands within the same instruction do not require more than one access to the addr value. Multiple access of addr does not generate an error, but a warning, if you do not redefine the addr value. Document Number: 001-84857 Rev. ** Page 29 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet The addr command can act in five different ways. You cannot choose more than one at a time. To access a location in ACU RAM. You can specify a single ACU RAM location to be accessed by both the ACUs. (Both sides must access the same row of ACU RAM. Side A cannot read from row 1 while B reads from row 14.) To specify which input and output channels to write to or read from. Channel 1 is selected for odd addr values (when addr’s LSB is 1) and Channel 2 is selected for even addr values. See the dmux section for more information about input channels and input staging registers, and the write section for information on output staging registers. To provide a value to write semaphores and enable and disable semaphores as jump conditions. The addr value is automatically set by the 3-bit fields specified in the semaphore commands. (See the ALU section.) Enabling and disabling the saturation and rounding flags and clearing the saturation detection flag hides the explicit definition of the addr value from you and defines it automatically. (See the ALU section.) Enabling and disabling the global interrupts as jump conditions defines the addr value using the 2-bit field in the englobal command. (See the ALU section.). If the addr command is called as addr(1), both the side A and side B ACU will access ACU RAM row 1. If the command causes a bus read, bus data is read from staging register 1. If the command causes a bus write, data is written to output staging register 1. This is an example of the problem with having multiple accesses to the value of the addr command. When the addr value is defined once, all access to it must use the same value, or the customizer will return an error. The explicit definition of addr prevents the use of the semaphore, global jump condition enables, and saturation and rounding register commands. dmux Figure 4. Datapath Six bits in the control store RAM (cstore) output word control datapath muxing and specify the route data takes through the datapath. The MAC, ALU, and output shifter have three levels of Page 30 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler muxing for both side A and side B. Similar to the ACU, dmux takes an argument for side A and side B. dmux(instruction_A, instruction_B) When mux1 (see Figure 4) is set to allow access to the bus, it consumes whatever value is waiting in the input register. Unless the system is completely deterministic, code should wait for the controller to signal the availability of a value using a wait state that loops until the input channel ready jump condition is valid. The following table shows a comprehensive list of the dmux instruction set. Instruction Description ba Bus to ALU. mux1 passes the bus data to the data RAM. mux2 circumvents the data RAM and passes mux1 data directly to the MAC and mux3. mux3 circumvents the MAC and passes mux2 directly to the ALU input. The bus data comes from one of the two staging registers based on an addr declaration of either 0 or 1. sa Shifter to ALU. mux1 passes the shifter output data to the data RAM. mux2 circumvents the data RAM and passes mux1 data directly to the MAC and mux3. mux3 circumvents the MAC and passes mux2 directly to the ALU input. bra Bus to RAM, RAM to ALU. mux1 passes the bus data to the data RAM. mux2 passes data from the data RAM to the MAC and mux3. mux3 circumvents the MAC and passes mux2 directly to the ALU input. T The bus data comes from one of the two staging registers based on an addr declaration of either 0 or 1. sra Shifter to RAM, RAM to ALU. mux1 passes the shifter output data to the data RAM. mux2 passes data from the data RAM to the MAC and mux3. mux3 circumvents the MAC and passes mux2 directly to the ALU input. bm Bus to MAC, MAC to ALU. mux1 passes the bus data to the data RAM. mux2 circumvents the data RAM and passes mux1 data directly to the MAC and mux3. mux3 passes the MAC output to the ALU input. The bus data comes from one of the two staging registers based on an addr declaration of either 0 or 1. sm Shifter to MAC, MAC to ALU. mux1 passes the shifter output data to the data RAM. mux2 circumvents the data RAM and passes mux1 data directly to the MAC and mux3. mux3 passes the MAC output to the ALU input. brm Bus to RAM, to MAC. mux1 passes the bus data to the data RAM. mux2 passes the data from the data RAM to the MAC and mux3. mux3 passes the MAC output to the ALU input. The bus data comes from one of the two staging registers. Register choice is based on an addr declaration of either 0 or 1. srm Shifter to RAM to MAC. mux1 passes the shifter output data to the data RAM. mux2 passes the data from the data RAM to the MAC and mux3. mux3 passes the MAC output to the ALU input. ALU The ALU provides data control on the output end of the datapath. In addition to generic functions such as add and subtract, the ALU can set flags signaling that specific conditions for jumps Document Number: 001-84857 Rev. ** Page 31 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet between states have been met. The alu command includes five special instructions that require a bit field of three bits as input data. The ALU’s output feeds directly into the shift register. There are two format options for an “alu” command: alu(instruction) or alu(special_instruction, 3-bit_field) The following table shows a comprehensive list of the alu instruction set. Instruction Description set0 Sets the ALU output to zero. set1 Sets the ALU output to an integer value of one. That is, the LSB is one, all others are zero. seta Passes input A to the output. setb Passes input B to the output. nega Negates A and passes it to the output. negb Negates B and passes it to the output. passrama Passes RAM A’s current location value to the output. passramb Passes RAM B’s current location value to the output. add Evaluates and places ‘A + B’ on the ALU output. tdeca Evaluates and places ‘A – 1’ on the ALU output. A value of zero sets threshold detection. The design includes this command as a way to wait a set amount of time while a value counts down to zero. It is intended for use in a low-power wait mode. suba Evaluates and places ‘B – A’ on the ALU output. subb Evaluates and places ‘A – B’ on the ALU output. absa Evaluates and places ‘|A|’ on the ALU output. absb Evaluates and places ‘|B|’ on the ALU output. addabsa Evaluates and places ‘|A| + B’ on the ALU output. addabsb Evaluates and places ‘A + |B|’ on the ALU output. hold Maintains the ALU output value from the previous cycle. englobals Enables or disables the two global interrupts and the saturation detection flag as jump conditions for state changes. This is based on the 3-bit field following the command and a separating comma. The global interrupts are inputs to the DFB. The saturation detection flag is set when a wraparound would otherwise happen if the saturation logic holds the value at the maximum or minimum datapath value. This command sets the value for addr automatically, based on the entry to the bit field. englobals shares an ALU opcode with ensatrnd. The behavior of these commands is determined in hardware by the value of the addr opcode. The customizer generates errors if the program tries to define the addr command with different values. The following would set global interrupt 2 and saturation to be disabled as a jump condition and global interrupt 1 to be enabled. alu(englobals, 001) Page 32 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Instruction ensatrnd Digital Filter Block (DFB) Assembler Description Enables and disables saturation and rounding in the datapath by writing to the saturation and rounding registers. The value written to the saturation and rounding register it is taken from the last two bits of the 3-bit field following both the command and a separating comma. The first bit of the 3bit field strobes the saturation detection flag and clears it. This command uses and sets the value for addr automatically. Additionally this command shares an opcode with englobals; the behaviors of these two commands are determined in hardware by the value of the addr opcode. The customizer generates errors if the program tries to define the addr command with different values. The following ALU command would turn on rounding, turn off saturation and clear the saturation detection flag if it is set. alu(ensatrnd, 001) ensem Enables specified semaphores as jump conditions based on the 3-bit field following the command and a separating comma. This command uses and sets the value for addr automatically, based on the 3-bit field. Therefore, the customizer generates errors if the program tries to define the addr command with different values using this instruction. This instruction has a two cycle delay before the change becomes a valid jump condition. While using the semaphore as a jump condition, the condition “sem” can be used as a reminder to the programmer that a semaphore condition has been set. If the program no longer wants the semaphore as a jump condition, it must clear the enable flags with a call of the command with every bit in the field set to zero. alu(ensem, 000) setsem Sets the semaphores masked with ones in the 3-bit field to 1. Do not use setsem in the first instruction after reset or the semaphore will be repeatedly set. This command uses and sets the value for addr automatically. The customizer generates errors if the program tries to define the addr command with different values. The following sets semaphore 0 to true (1). alu(setsem, 100) clearsem Clears the semaphores masked with ones in the 3-bit field. (that is, sets them to zero). This command uses and sets the value for addr automatically. The customizer generates errors if the program tries to define the addr command with different values. The following sets the value of semaphore 0 to false (0). alu(clearsem, 100) tsuba Evaluates and places ‘B – A’ on the output. Sets threshold detection. tsubb Evaluates and places ‘A – B’ on the output. Sets threshold detection. taddabsa Evaluates and places ‘|A| + B’ on the output. Sets threshold detection. taddabsb Evaluates and places ‘A + |B|’ on the output. Sets threshold detection. sqlcmp Loads a value from mux3 of side A into the compare register to be used as a cutoff in squelch functions. sqlcnt Loads the lower 16 bits of mux3 on side A into the 16-bit count register. This value is decremented every time a squelch command is called, if the current value at the output does not meet the threshold set by the squelch compare register. It is reset to its original value every time the threshold is met. Document Number: 001-84857 Rev. ** Page 33 of 42 ® Digital Filter Block (DFB) Assembler Instruction sqa PSoC Creator™ Component Datasheet Description Takes the value on mux3 of side A and compares it to the value in the squelch compare register. If the current value is greater than the value of the compare register, the current value is passed to the output and the squelch count register is reset to its original value. If the current value is less than the compare register value, the command checks the squelch counter. If it is not at zero, the value is decremented and the current value on mux3 of side A is passed to the output. If the value of the count register is zero, zero is passed to the output. sqb Takes the value on mux3 of side B and compares it to the value in the squelch compare register. If the current value is greater than the value of the compare register, the current value is passed to the output and the squelch count register is reset to its original value. If the current value is less than the compare register value, the command checks the squelch counter. If it is not at zero, the value is decremented and the current value on mux3 of side B is passed to the output. If the value of the count register is zero, zero is passed to the output. MAC Multiply and Accumulate unit. Contains the hardware to multiply two fixed-point numbers and then add them to a previous value. ‘(A × B) + C’ There are four members in the MAC instruction set and they operate as follows: mac(instruction) The following table shows a comprehensive list of the MAC instruction set. Instruction Description loadalu Adds the previous ALU output (from the shifter) to the product and starts a new accumulation. clra Clears the accumulator and stores the current product. hold Holds the value in the accumulator from the previous cycle. No multiply. macc Multiply and Accumulate. Multiplies the values on mux2 of side A and side B. Adds the product to the current value of the accumulator. shift The shift command allows you to scale the ALU output. A valid shift command takes two arguments, direction and magnitude, and evaluates them to produce the correct opcode. Valid directions instructions are ‘right’, ‘left’, ‘r’, and ‘l’. A shift to the right allows magnitudes of 1, 2, 3, 4, and 8, while a shift to the left allows only the values 1 and 2. The output of the ALU is passed through the shifter and back out to the start of the datapath regardless of whether a shift occurs. shift(direction, magnitude) Page 34 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler The following table shows a comprehensive list of the shift instruction set. Instruction Description right, r Equivalent instructions specifying the direction to shift. left, l Equivalent instructions specifying the direction to shift. write A valid write command has zero to three arguments. A value is written for each argument. You can choose to write the value on mux1 of either side to the data RAM of the same side, or to write the current shifter output value to a staging register on the output. Check for the various pipeline delays in the DFB datapath before using the write instruction. There are system level implications if the first instruction following a reset contains a bus write command. The assembler prevents writes in the first instruction to prevent unintended problems. Each of the following is a valid write command. write(da, db, bus) write(da, db) write(db) The following table shows a comprehensive list of the write instruction set. Instruction Description da Writes the mux1A value to the specified data_a Ram location after the ACU instruction for the line, No delay in write execution. db Writes the mux1B value to the specified data_b Ram location after the ACU instruction for the line, No delay in write execution. abus Write shifter output to the bus’ holding register A. Holding register A is selected based on an addr definition of 1 (see the addr section). bbus Write shifter output to the bus’ holding register B. Holding register B is selected based on an addr definition of 0 (see the addr section). bus Write shifter output to the bus’ holding register. There are two output holding registers available. The register is selected based on an addr definition of either 1 or 0 (see the addr section). This instruction has been left for compatibility; do not use the bus instruction in new projects. Using the instruction generates a warning: “Potential addr() conflict attempting write(bus). Avoid this warning by using channel-specific bus write commands.” Jump Instructions A jump instruction allows the code to change its location to a different subroutine. Jump instructions have a general form of: JumpType(conditions, Target Routine) Document Number: 001-84857 Rev. ** Page 35 of 42 ® Digital Filter Block (DFB) Assembler PSoC Creator™ Component Datasheet where “conditions” is an ordered list of any of the enables available to restrict or allow jumps. The literals associated with these flags are described immediately following the description of the jump options. Be careful with jump instruction placement because some conditions require two cycles instead of one before the controller can use them as a valid jump condition. Additionally, a jump instruction cannot be used in the first instruction after a reset because the first state must be two instructions deep to set up pipelining. The Diagrams of Jump Instruction section contains a diagram of jump conditions. The following table shows a comprehensive list of the jump instruction set. Instruction jmp Description Jump” is similar to a standard “goto” instruction. If conditions are true, the code jumps to the target routine. If not, it falls through to next state numerically. When not in a loop, the fjlim (false jump limit) value is set to the max cstore location. jmp(eob, sign,…, Target State) jmpl “Jump Loop” sets the current block of code to be a loop. It is a two-way branch with the target routine as one branch when conditions are met, and the start of the current code block as the other branch when conditions are not met. When in a loop the following occurs: Bit 23 of the CFSM is set high. False jump address (FJADDR) is defined as the start of the current code block, which is the control store address of the label. (The label is not stored in the control store; the label references the first command of the block of code.) False jump limit (FJLIM) is set to the current CLC location, which is the end of the code block (Same for all jumps) The target routine’s label is used to provide the jump address (JADDR) and the CFSM RAM location of the next state on true conditions. Commands execute until the eob is detected. The program evaluates the condition and if it is false, set the program counter to FJADDR, restarting the block. If the condition is true, the program counter is set to JADDR and the state is updated to NextStateOnTrue. The format is identical to the jmp command. jmpl(conditions, Target State) jmpsl “Jump to Subroutine Loop” allows a jump to a subroutine block of code that is designated to be a loop. The effect jmpsl has on the current block of code is identical to the jmp command. If conditions are met, the code jumps to the specified subroutine, otherwise it falls through to the next state in code execution. However, jmpsl affects the subroutine’s state. Each time a subroutine is referenced, a copy of that subroutine is created as a state for the CFSM and properties for the new state are set. In the case of jmpsl, the created state is designated to be a loop and to have a return state of the next state in the code space. (The current state is the state that ends with the jmpsl command.) For more information about states and subroutines, see the jmpret entry in this table. jmpsl(conditions, Target Subroutine) jmpslr “Jump to Subroutine Loop with Return state” is identical to jmpsl in every way except that the return state is specified instead of defaulting to the state following the current state. jmpslr(conditions,…, Target Subroutine, Return State) Page 36 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Instruction Digital Filter Block (DFB) Assembler Description jmps and jmpsr These two are clones of jmpsl and jmpslr except that the created subroutine state is not a loop. Because of this, only the eob condition is specified in the jmpret statement. jmpret Subroutines are accessed only through one of the jump-to-subroutine commands. Subroutines differ from standard states because that their properties are determined by the state that calls them, instead of the jump condition at the end of the state. If a state calls a subroutine with the jump command “jmpslr(eob, sub1, anotherState), it defines the subroutine to be a loop and have a next state of “anotherState.” Subroutines are terminated with a jump instruction of the type jmpret (jump return), which provides the subroutine with its own set of jump conditions for loop termination (if the subroutine is called as a loop). Subroutines cannot call other subroutines because the jmpret command does not provide the necessary exit information to the sub-subroutine state. The jmpret command is used as a jmp command, but with no target state specified. jmpret(condition,condition,..) Jump Conditions Conditions enable or prevent state changes in code. The following conditions are, in essence, enable flags for hardware. When a condition is listed, it requires the signal complementing the enable to be true, in order for the jump to proceed. Note Datapath conditions have a two-cycle delay. That is, they must be true two cycles before the jump to recognize the condition as true. The following table shows a comprehensive list of jump conditions. Instruction Description eob End Of Block. A condition for a jump, which is always met because a jump instruction signifies the end of the block. eob only needs to be specified when an unconditional jump occurs. This is because of a software restriction, not a hardware restriction. dpsign A jump based on the MSB of the ALU output. Asserted when the ALU output is negative. Datapath conditions require a two-cycle delay to meet a jump condition. dpthresh Datapath Threshold. Asserted when the ALU detects a sign change. The ALU asserts dpthresh only when the program uses ALU threshold detection operands (tsuba, tsubb, taddabsa, …). Datapath conditions require a two-cycle delay to meet a jump condition. dpeq Datapath Equity. Asserted when the ALU hardware detects an output value of zero. The ALU asserts dpeq only when the program uses ALU threshold detection operands (tsuba, tsubb, taddabsa, …). Datapath conditions require a two-cycle delay to meet a jump condition. acuaeq ACU A Equals. Asserted when ACU A detects a wraparound condition. This can be either zero or the maximum data RAM location if modulus arithmetic is disabled or the max and min modulo counter limit when it is enabled. Datapath conditions require a one-cycle delay to meet a jump condition. acubeq ACU B Equals. Asserted when ACU B detects a wraparound condition. This can be either zero or the modulo counter limit. Datapath conditions require a one-cycle delay to meet a jump condition. Document Number: 001-84857 Rev. ** Page 37 of 42 ® Digital Filter Block (DFB) Assembler Instruction PSoC Creator™ Component Datasheet Description in1 Channel 1 Input Register Value Ready signal. When it is asserted, a new input cycle is available for consumption. It remains asserted until cleared by a bus read. Datapath conditions require a onecycle delay to meet a jump condition. in2 Channel 2 Input Register Value Ready signal. When it is asserted, a new input cycle is available for consumption. It remains asserted until cleared by a bus read. Datapath conditions require a onecycle delay to meet a jump condition. sem The sem condition has no effect on the opcodes. It improves code clarity and reminds the programmer that a semaphore is currently a jump condition. (See the ALU section for details.) globals The globals condition has no effect on the opcodes. It improves code clarity reminds the programmer that a global input is currently a jump condition. (See the ALU section for details.) sat The “sat” condition has no effect on the opcodes. It improves code clarity and reminds the programmer that a saturation event is currently required to enable a jump. (See the ALU section for details.) DMUX Commands Code Name Function Mux 1 Function Mux 2 Function Mux 3 0 ba Bus register Bus register Bus register 1 sa Previous shifter output Previous shifter output Previous shifter output 2 bra Bus register Current RAM value Current RAM value 3 sra Previous shifter output Current RAM value Current RAM value 4 bm Bus register Bus register MAC accumulator 5 sm Previous shifter output Previous shifter output MAC accumulator 6 brm Bus register Current RAM value MAC accumulator 7 srm Previous shifter output Current RAM value MAC accumulator MAC Commands Code Name Function 0 loadalu Adds the ALU value to the product and starts a new accumulation. 1 clra Clears the accumulator. Loads it with the current product. 2 hold Holds the accumulator, no multiply (no power in mult). 3 macc Standard operation – multiply and accumulate with the previous values Page 38 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler ACU Commands Code Name Function 0 hold Puts reg on the output. 1 incr Puts reg + 1 on the output, writes to reg. 2 decr Puts reg – 1 on the output, writes to reg. 3 read Loads reg from ACU RAM, puts the value on the output. 4 write Puts reg into the specified ACU RAM row. 5 loadf Loads freg from ACU RAM, puts reg on the output. 6 loadl Loads lreg from ACU RAM, puts reg on the output. 7 loadm Loads mreg from ACU RAM, puts reg on the output 8 writel Puts lreg on the output, writes to ACU RAM. 9 setmod Sets arithmetic to modulo mreg. 10 unsetmod Sets arithmetic to wraparound. 11 clear Sets reg to 0, put 0 on the output. 12 addf Adds reg to freg, puts the result on the output, stores it in reg. 13 subf Subtracts freg from reg, puts the result on the output, stores it in reg. 14 writem Puts mreg on the output, writes to the ACU RAM. 15 writef Puts freg on the output, writes to the ACU RAM. ALU Commands Code Name Function 0 set0 Sets the ALU output to 0. 1 set1 Sets the ALU output to 1. 2 seta Passes A to the ALU output. 3 setb Passes B to the ALU output. 4 nega Sets the ALU output to –A. 5 negb Sets the ALU output to –B. 6 passrama Passes the RAM A output directly to the ALU output. 7 passramb Passes the RAM B output directly to the ALU output. 8 add Adds A and B and puts the result on the ALU output. 9 tdeca Puts A – 1 on the ALU output, sets threshold detection. Document Number: 001-84857 Rev. ** Page 39 of 42 ® Digital Filter Block (DFB) Assembler Code PSoC Creator™ Component Datasheet Name Function 10 suba Puts B – A on the ALU output. 11 subb Puts A – B on the ALU output. 12 absa Puts |A| on the ALU output. 13 absb Puts |B| on the ALU output. 14 addabsa Puts |A| + B on the ALU output. 15 addabsb Puts A + |B| on the ALU output. 16 hold Holds the ALU output from the previous cycle. 17 englobals Enables global and saturation jump conditions using a three-bit field to specify which events are active jump conditions. 17 ensatrnd Writes to the saturation and rounding enable registers using a threebit field to enable and disable them. 18 ensem Enables semaphores as jump conditions using a three-bit field to specify which are active. 19 setsem Sets the semaphores high using the three-bit mask. 20 clearsem Sets the semaphores low using mask, addr[2:0]. 21 tsuba Puts B – A on the ALU output, sets threshold detection. 22 tsubb Puts A – B on the ALU output, sets threshold detection. 23 taddabsa Put |A| + B on the ALU output, set threshold detection 24 taddabsb Puts A + |B| on the ALU output, sets threshold detection 25 sqlcmp Loads the squelch comparison register with a value from side A, passes side B. 26 sqlcnt Loads the squelch count register with a value from side A, passes side B. 27 sqa Squelch side A. If the value is above the threshold, passes it. If the value is below the threshold and the squelch count register is zero, passes zero. 28 sqb Squelch side B. If the value is above the threshold, pass it. If the value is below the threshold and the squelch count register is zero, passes zero. 29-31 undefined Undefined opcodes Page 40 of 42 Document Number: 001-84857 Rev. ** ® PSoC Creator™ Component Datasheet Digital Filter Block (DFB) Assembler Diagrams of Jump Instruction jmps jmp True State 0 jmp(cond, State 2) State 0 True False Cond State 2 jmps(cond, Sub 1) False Cond Cond Cond State 1 State 2 State 1 jmpslr jmpsl True State 0 True State 0 jmpsl(cond, Sub 1) Cond jmpslr(cond, Sub 1, State 2) Cond False False Cond Cond Sub 1 State 1 False Return Cond True Return State 1 Sub 1 Return Cond True State 2 Cond jmpl Return Cond jmpsr State 0 jmpsr(cond, Sub 1, State 2) State 0 jmpl(cond, State 1) True Cond False False Cond State 1 Document Number: 001-84857 Rev. ** Sub1 True Cond False Cond State1 State2 Page 41 of 42 Digital Filter Block (DFB) Assembler ® PSoC Creator™ Component Datasheet Component Changes This section lists the major changes in the component from the previous version. Version 1.20 1.10 Description of Changes Reason for Changes / Impact Added Simulator Output information to the datasheet Changes to meet MISRA compliance MISRA compliant with some global deviations Updated DFB LoadDataRAMx() API source To improve the speed of data transfer Updated DC and AC Electrical Characteristics section. Added PSoC5LP device support. © Cypress Semiconductor Corporation, 2012. The information contained herein is subject to change without notice. Cypress Semiconductor Corporation assumes no responsibility for the use of any circuitry other than circuitry embodied in a Cypress product. Nor does it convey or imply any license under patent or other rights. Cypress products are not warranted nor intended to be used for medical, life support, life saving, critical control or safety applications, unless pursuant to an express written agreement with Cypress. Furthermore, Cypress does not authorize its products for use as critical components in life-support systems where a malfunction or failure may reasonably be expected to result in significant injury to the user. The inclusion of Cypress products in lifesupport systems application implies that the manufacturer assumes all risk of such use and in doing so indemnifies Cypress against all charges. PSoC® and CapSense® are registered trademarks, and SmartSense™, PSoC Creator™, and Programmable System-on-Chip™ are trademarks of Cypress Semiconductor Corp. All other trademarks or registered trademarks referenced herein are property of the respective corporations. Any Source Code (software and/or firmware) is owned by Cypress Semiconductor Corporation (Cypress) and is protected by and subject to worldwide patent protection (United States and foreign), United States copyright laws and international treaty provisions. Cypress hereby grants to licensee a personal, non-exclusive, non-transferable license to copy, use, modify, create derivative works of, and compile the Cypress Source Code and derivative works for the sole purpose of creating custom software and or firmware in support of licensee product to be used only in conjunction with a Cypress integrated circuit as specified in the applicable agreement. Any reproduction, modification, translation, compilation, or representation of this Source Code except as specified above is prohibited without the express written permission of Cypress. Disclaimer: CYPRESS MAKES NO WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, WITH REGARD TO THIS MATERIAL, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. Cypress reserves the right to make changes without further notice to the materials described herein. Cypress does not assume any liability arising out of the application or use of any product or circuit described herein. Cypress does not authorize its products for use as critical components in lifesupport systems where a malfunction or failure may reasonably be expected to result in significant injury to the user. The inclusion of Cypress’ product in a life-support systems application implies that the manufacturer assumes all risk of such use and in doing so indemnifies Cypress against all charges. Use may be limited by and subject to the applicable Cypress software license agreement. Page 42 of 42 Document Number: 001-84857 Rev. **