Home

Integrating the DSP563xx in Distributed Computing Environments

Contents

1. execution on the DSP side with the execution on the PC side has been omitted In this simple case the sync code is not even necessary since the time out value in the read data dsp function is much larger than the amount of time taken to do the actual calculations on the DSP lodpathname kfullpath kmsum lod S MOTOROLABIN NULL res scom load program amp did lodpathname if res ER OK kerror lib rtn Error d loading program s n res mdsp get error res kexit KEXIT FAILURE Receive the resulting signal from the DSP and write it into the output object using kpds put data part of the KHOROS Polymorphic Data Services res scom read data dsp amp did int inbufl max if res ER OK kerror lib rtn Error d reading from DSP s n res mdsp get error res kexit KEXIT FAILURE if kpds put data outobj KPDS VALUE ALL kaddr inbufl FALSE kerror lib rtn Error writing to output object s n clui info o file kexit KEXIT FAILURE main_library_call_end main_after_lib_call Finally close all the open kobjects shut down communications and do other cleanup tasks scom_shut_down amp did if kpds_set_attribute outobj KPDS HISTORY kpds history string kerror lib rtn Unable to set history on destination object kexit KEXIT FAILURE kpds close object inobj
2. 3 2 6 Transmit Routine The Monitor program also provides a low level Transmit routine which DSP algorithms can use to send back results Its use is simple and straightforward 1 A memory address space code is written to A2 register 4 for P memory 2 for X memory 1 for Y memory 2 The size of the buffer is written to the AO register 3 The starting address of the buffer is written to the RO register 4 A JSR is performed to the address stored at P 54 Again if none of the memory spaces is specified the 3 least significant bits of A2 are cleared and Y memory space is assumed Refer to the flowchart in Figure 3 on page 10 The Transmit routine is listed in Code Example 17 iegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Ne Ne Ne Ne Ne Ne Transmit _LoopT1 SavexY LoopT2 SaveY _LoopT3 EndT Transmit sends a buffer through the SCI to the host computer Freescale Semiconductor ING sentation on DSP56307 Code Example 17 Monitor Program Transmit Routine A2 lt 4 send buffer from P memory A2 lt 2 send buffer from X memory A2 lt 1 send buffer from Y memory AO lt number of 24 bit words to send RO start address of buffer JCLR 42 A2 SaveXY DO AO LoopT1 JCLR 41 X M SSR MOVEP P R0 X M STXL JCLR 41 X M SSR MOVEP P
3. Data communication between the workstation and the DSP is of major importance for the proper operation and satisfactory performance of a distributed DSP The system should be flexible and portable and provide high throughput error control and a simple communication protocol Workstation DSP Communication Protocol For More Information On This Product Go to www freescale com Freescale Semiconductor Inc This section focuses on the communication between the workstation and the DSP56307EVM board It describes the command protocol and a typical communication session and presents an overview of the communication library Most of the communication is based on the Client Server model in which the client is the workstation and the server is the DSP board The client sends various commands to the server which services them according to the command protocol described in this section The only exception to this is the result buffer sent by the digital signal processing algorithm back to the workstation This data transfer is initiated by the DSP board after completion of the processing All communication between the workstation and the DSP board in this application is carried out over an asynchronous serial link at 115200 bps This can be a major drawback in systems that require high throughput For better performance in these cases one can make use of Motorola s more powerful solutions which enable data transfers over parallel Ethernet and IS
4. to implement DSP algorithms This network uses signal processing glyphs provided by the KHOROS standard distribution package as well as specialized glyphs which are custom designed to run corresponding DSP algorithms on the DSP based board These specialized glyphs can be grouped in a custom toolbox which Cantata can use to build run and control the particular DSP algorithms needed in distributed applications NOTE In this document a software object refers to a program that performs data processing while a glyph is a graphical representation of a software object in Cantata For simplicity these terms are used interchangeably A basic custom toolbox called Freescale EVM was developed for this case study It contains two demonstration glyphs kmsum a simple adder of two input signals and kmf ft a simple 256 point FFT The two glyphs can be used to run the corresponding algorithms directly on the system s Freescale DSP board in a distributed manner over the Ethernet When the DSP network on the controlling computer is launched under Cantata it processes the corresponding data flow diagram When a local glyph is activated it launches the corresponding procedures on the controlling computer s own processor If the data flow on the DSP Network reaches one of the glyphs that implements a DSP algorithm on the Freescale EVM the KHOROS system on the controlling computer sends an execution command along with input parameters a
5. 2000 tten By Date cations int main int char argc argv main variable list char lib kobj kobj kobj kaddr inbufl kaddr inbuf2 u int32 t wl kmsum rtn NULL NULL NULL main ect inobjl ect inobj2 ect outobj NULL NULL w2 h d t e max sdid did err char lodpathname t res NULL main variable list end clui info commport string clui info commport flag Output is the sum of two inputs First Input data object TRUE if il specified Second Input data object TRUE if i2 specified Resulting output data object TRUE if o specified Communications Port for the EVM TRUE if commport specified The inputs and the output are represented by kobjects data structures identifying polymorphic data sets see Section 4 3 The KHOROS Environment Every I O operation on polymorphic data is done on these objects The only reason for having two separate input buffers is to make the source code easier to understand Used in determining the input s size Device ID for the communications library Return code yf Iniegra hg ihe D DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc XRHOHOSSORWES KHOROS init argc argv FREESCALE PRODUCT RELEASE DATE PRODUCT RELEASE NAME
6. PRODUCT RELEASE VERSION PRODUCT RELEASE MAJOR PRODUCT RELEASE MINOR SF REESCALE objects kroutine kmsum kexit handler kmsum free args NULL main get args call kclui init FREESCALE kmsum KGEN KROUTINE amp clui uis spec kmsum usage additions kmsum get args kmsum free args main get args call end main before lib call Open the input and output objects using kpds open input object and kpds open output object which are part of the KHOROS Polymorphic Data Services Note that error reporting is done via specialized methods because errors are most often reported graphically inside Cantata although they can also be reported to stderr when the kroutine is run in command line if inobjl kpds open input object clui info il file KOBJECT INVALID kerror lib rtn Cannot open input object s Wn clui info il file kexit KEXIT FAILURE if inobj2 kpds_open_input_object clui_info gt i2_file KOBJECT INVALID kerror lib rtn Cannot open input object s n clui info i2 file kexit KEXIT FAILURE if outobj kpds open output object clui info o file KOBJECT INVALID kerror lib rtn Cannot open output object s n clui info o file kexit KEXIT FAILURE Set the input data type to integer This is only the type in which data is read NOT the type of the ac
7. _Loopl Read address and block size NOP MOVE BO RO MOVE O AO MOVE BO R1 Save start address JCLR 2 A2 _LoadXY DO Bl Loop2 DO 3 Loop21 JCLR I2 X M SSR MOVEP X M SRxL B1 ASR 8 B B Loop21 NOP MOVEM BO P RO ADD B A _Loop2 Read block into P memory JMP SendCRC LoadXY JCLR 1 A2 _Loady DO Bl Loop3 DO 13 Loop31 JCLR 42 X M SSR MOVEP X M SRxL B1 ASR 8 B B Loop31 NOP MOVE BO X R0 ADD B A _Loop3 Read block into X memory JMP SendCRC LoadY DO Bl Loop4 DO 13 Loop41 JCLR 42 X M SSR MOVEP X M SRxL B1 ASR 8 B B _Loop41 NOP MOVE BO Y RO ADD B A _Loop4 Read block into Y memory JMP SendCRC The DSP56307 in a Distributed Environment For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 3 2 5 Run Command Procedure The Run Command procedure is implemented as follows The DSP receives the Run command followed by the start address of the program to run 2 The DSP sends back the command confirmation string Ok followed by the received address The address is pushed onto the upper half of the stack SSH 4 A value of C00200 is pushed to the lower half of the stack SSL Execution continues until the end of the Receive routine where a RTI instruction is issued The RTI instruction loads the PC register with the address at which execution is to resume from the SSH register which contains the value sent by the Run command
8. 1 The PC sends a Reset command to the DSP board AA 80 2 The DSP board answers by sending the command confirmation string Ready after performing a soft reset 3 The PC downloads all the input data buffers as well as the DSP algorithm to the DSP board Each buffer is downloaded in the following sequence a The PC issues a Load command SAA 44 for loading into P memory AA 42 for loading into X memory AA 41 for loading into Y memory b The PC sends two data words to the DSP board containing the starting address where the data is to be loaded and the number of words to be sent c The PC sends the data words which the DSP board loads into memory while computing a checksum 4 The DSP board answers by sending the command confirmation string Ld followed by the checksum of the loaded buffer 5 The PC validates the checksum The PC sends a Run command AA 20 and the start address of the program 7 The DSP board acknowledges receiving the command by sending the string 0k followed by the start address supplied in the Run command Control is then transferred to the processing program which is responsible for sending the result buffers back to the PC 2 2 Communication Library Functions Overview The SCOM Serial COMmunication communication library is written in C and designed to run under Linux It provides a set of low level functions to perform the actual data transfer through the RS 232 serial port o
9. P R0 OVE B2 P RO OVE B1 P R0 4 OVEM BO P RO Save regs JCLR 2 X M_SSR OVEP X M SRxL A1 OVE 0 AO OVE tO A2 CMP FSAA A JNE End JCLR 2 X M_SSR OVEP X M SRxL A2 Read command byte OP JCLR 47 A2 LoadOrRun RESET DSP Reset OVEC 40 SP OVEC 0 SC OVEC C00300 SR OVEC 000309 OMR JMP SFF0000 Jump to bootstrap iegra hg fbi Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor ING sentation on DSP56307 3 2 4 Load Procedure As described in Section 2 the Load procedure downloads the number of 24 bit data words specified in the command protocol This is accomplished by simply moving the received bytes to the B1 register and then right shifting the B register After completing a cycle of three moves and three shifts BO holds the resulting 24 bit word which is moved to P X or Y memory according to the command protocol After each cycle the content of the B register is added to the A register At the end of the transfer AO holds the checksum of the received buffer which is then sent back to the workstation where it is validated The checksum is prefixed by the command confirmation string Ld The Load procedure is listed in Code Example 15 Code Example 15 Monitor Program Load Procedure LoadOrRun JCLR 46 A2 Run DO 46 Loopl JCLR 42 X M SSR MOVEP X M SRxL B2 ASR 8 B B
10. R0 X M STxM JCLR 41 X M SSR MOVEP P RO X M_STxH NOP save P memory JMP _Endt JCLR 1 A2 _SaveY DO A0 LoopT2 NOP NOP JCLR 41 X M SSR OVEP X R0O X M STXL OP NOP JCLR 41 X M SSR OVEP X R0 X M STxM NOP OP JCLR 41 X M SSR OVEP X RO X M STxH NOP oave X memory JMP EndT DO AO LoopT3 NOP NOP JCLR 1 X M_SSR OVEP Y RO X M_STxL NOP NOP JCLR 41 X M SSR OVEP Y R0 X M STxM NOP NOP JCLR 41 X M SSR OVEP Y RO X M_STxH NOP oave Y memory RTS Return from subroutine The DSP56307 in a Distributed Environment For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 4 Case Study As a case study a distributed digital signal processing system was developed using two personal computers interconnected with a TCP IP based Ethernet both running the KHOROS software package from Khoral Research Incorporated on the Linux platform The core of the system is the DSP56307 Evaluation Module which performs all the DSP specific operations initiated by the remote station which is the controlling computer This section describes the architecture of the case study system lists two DSP algorithms to be run in the system presents an overview of the KHOROS software and shows how the KHOROS software is used to implement one of the DSP algorithms in the system 4 1 General Architecture The general architecture of the proposed system is similar to the di
11. The RTI also loads the SR from SSL which contains the value C00200 This sets the core priority level to 3 and the interrupt priority level to 2 which enables the Monitor program to continue to receive and execute commands interrupting the running algorithm This ability is very useful for interrupting a lengthy operation to perform a less time consuming function The Run Command procedure is listed in Code Example 16 iegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor ING snentation on DSP56307 Code Example 16 Monitor Program Run Command Procedure _Run JCLR 5 A2 _End DO 13 Loop5 JCLR 42 X M SSR OVEP X M SRxL A2 ASR 8 A A _Loop5 Read address NOP LRA SRAdr RO OVE Al R1 OVE 4 A2 OVEM R1 P RO OVE 2 A0 JSR Transmit OVEC R1 SSH OVE SC00200 A0 OVEC AO SSL Set up return address JMP _End SendCRC LRA SLAdr RO MOVE 4 A2 NOP MOVEM AO P R0 MOVE 2 A0 JSR Transmit JMP End End LRA Regs RO NOP NOP OVE P RO R1 OVE P RO A2 OVE P RO Al OVE P RO AO0 OVE P RO B2 OVE P RO Bl OVE P RO BO OVE P RO MO OVE P ROReg RO Restore regs RTI Return from interrupt The DSP56307 in a Distributed Environment For More Information On This Product Go to www freescale com Freescale Semiconductor Inc
12. a be RR ERE UE KEE E e RIA RIC RE sever nen see EE 8 2 2 24 SCOM Write DD Lio du aias sre sphaera FRE oe Sepe rd 698 a ird 9 2S Helper Functions 24 2iccnpcbedereebaberibacscarecerdbavidegacerueaba 9 3 The DSP56307 in a Distributed Environment 9 3 1 General Description of the Monitor Program 1 0 2 0 00 cee eee eee eee 10 32 Monitor Program Implementation on DSP56307 2 2 0 0 0 0 00 cece eee eee 11 3 2 1 D finitons rm 12 3 2 2 Initialization 6 4 5 4 hed 3 pra RE ER EE AR E E A Ad REA RP dU ERES 13 3 2 3 Receive ROUUNG lt cere Sees he wh SER eee PRA FP ERE DAE Eee ERE 14 Abstract and Contents For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 3 2 4 Load Procedure isis ed exse eR RUE ELE UE REN eee REN M RES 15 3 2 5 Run Command Procedure 4c ci cuex derer erre Per Pea A Aer Grades 16 3 2 6 Transmit ROUDE sess ke E tassie he oa RE E EE PAG ERR ERR EET E E ome 18 4 Case Study auos sona hr ch oa ebd CCS dB Sc UI RC Ad rte do ise i E o RR 20 4 1 General Architect te mecsre aad en tdee sede a bent oee See see ees Ee dees Sel say 20 4 1 1 Controlling Computer 422 vus px TERES eeu een a S REN T wees ame 20 4 1 2 M DEESUIDOTE deo Ee nnter atan se ei ARTOEPEPLADTqQUN EV E Dex TP LPS RUE nes 21 4 1 3 DSP B sed Bond cote sosne vases tay Bose dagy a dead EAxL SUPER PEE 22 212 DSP ALSO usse ERR E x I Rex 9 ex ERR Rp xe acr tigen RP Rex RES 23 4 2 1 Ad
13. a data processing glyph or an input glyph e XVroutines interactive graphical programs Interaction with XVroutines is achieved by means of a standardized set of graphical user interface elements provided by KHOROS An example of a typical XVroutine is a visualization glyph e Libraries collections of routines that can be called from within other software objects These routines can be LKroutines public functions and private functions 11 Refer to the Advanced KHOROS Manuals for a more detailed description 4 3 5 Development Tools In addition to designing and simulating custom DSP networks KHOROS provides a set of very powerful software development tools which allows full access to the KHOROS core Programmers can conveniently design and implement their own software objects or libraries group them into custom toolboxes or extend the existing ones Two of the most important development tools are Craftsman and Composer Their purpose is to assist the programmer in every phase of the software development process by automating specific tasks Another very important tool is the Graphical User Interface Specification Editor or GUISE This tool allows the programmer to visually design the user interface component of a software object 4 3 6 Data Structure KHOROS uses a unified data structure model to implement all of its procedures thus providing comprehensive support for a very large set of real life applications Thi
14. buf size t count sid serial device id buf buffer of unsigned bytes count buffer size Error codes returned ER IO ER EXT ER OK 2 2 1 4 scom read buf Performs buffered input from the serial device until either count bytes are read or a time out occurs Code Example 5 The scom read buf Function include scom h err t scom read buf sdid sid u int8 t buf size t count sid serial device id buf buffer of unsigned bytes count buffer size Error codes returned ER TOUT ER OK Workstation DSP Communication Protocol For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 2 2 1 5 mdps get error Returns the associated string to err_no error code Code Example 6 The mdsp get error Function include mdspcom h char mdsp get error err t err no err no error code 2 2 2 High Level Functions 2 2 2 1 scsom reset dsp Resets the DSP board connected to the serial port designated by sid Code Example 7 The scom reset dsp Function include scom h err t scom_reset_dsp sdid sid sid serial device id Error codes returned ER IO ER TOUT ER EXT ER DSP ER OK 2 2 2 2 scsom run Sends a run command to the DSP The lower 24 bits of dsp_addr specify the starting address of the code to be run Code Example 8 The scom run Function include scom h err t scom run sdid sid u int32 t dsp addr sid se
15. com Freescale Semiconductor Inc TEE Sonare Table 4 User Code Tags in Skeleton File Tags Denotes main variable list Program variables main variable list end main get args call Call to a special function which initializes the UIS the structure containing the main get args call end input parameters command line arguments of the program main before lib call Initialization code main before lib call end main library call Code that performs actual processing main library call end main after lib call Cleanup code main after lib call end When a Kroutine is created in Composer one input field and one output field are defined by default The kmsum Kroutine requires one additional input and a communications port to which the evaluation module is connected either locally or to a remote workstation For the case study these elements were added using the GUISE program and the GUI it generates When GUISE is started two dialog boxes appear One is the graphical user interface for the software object which is called a pane and the other is a form on which to make changes to the pane The form was used to add a second input for the second signal source as well as an additional String Selection control with a default value of dev ttyS1 the second serial device on a standard PC Figure 9 shows the pane for kmsum after th
16. enabled with a factor of 8 4 42 5 to yield a core frequency of 103 2192 MHz The SCI Receive Data interrupt handler is set by writing a JSR Receive at P 50 which is the address of the interrupt vector for the SCI Receive Data interrupt The entry point for the Transmit routine is written at address P 54 which is the address of the SCI Transmit Data interrupt vector The SCI Transmit Data interrupt is not used so it is disabled and its interrupt vector is used for storage The communication parameters are set to 10 bit asynchronous mode Transmit Interrupt disabled and Receive Interrupt enabled by writing B02 to the memory mapped SCI Control Register x FFFF9C SCI baud rate is set to 115200 by setting the SCCR Clock Divider to 13 which actually means a divisor of 14 After this set up is completed the Monitor program sends the acknowledgement string Ready to the workstation computer and the DSP goes into the WAIT state until a SCI receive interrupt occurs The initialization sequence is listed in Code Example 13 Init _Loop Code Example 13 Monitor Program tnitialization Sequence OVEC C00200 SR Set IPL to 2 OVEP X M IPR P A1 OR KSCO A OVEP A1 X M IPR P Set SCI IPL to 2 OVEP 45 460029 X M PCTL Set PLL factor to 8 4 OVE K550 RO OP OVE SOBFO80 Al OVEM Al1 P RO LRA Receive Al OVEM Al P RO Set up receive interrupt vector OVE 554 R0 LRA Trans
17. or other applications intended to support or sustain life or for any other application in which the failure of the Freescale Semiconductor product could create a situation where personal injury or death may occur Should Buyer purchase or use Freescale Semiconductor products for any such unintended or unauthorized application Buyer shall indemnify and hold Freescale Semiconductor and its officers employees subsidiaries affiliates and distributors harmless against all claims costs damages and expenses and reasonable attorney fees arising out of directly or indirectly any claim of personal injury or death associated with such unintended or unauthorized use even if such claim alleges that Freescale Semiconductor was negligent regarding the design or manufacture of the part For More Information On This Product Go to www freescale com Freescale Semiconductor Inc Abstract and Contents Distributed digital signal processing is the most suitable solution for many real life applications involving digital data acquisition and processing from multiple signal sources scattered over large areas The advantages of distributed computing over single processor or other multi processor architectures include high computing power at a low cost flexibility and scalability Several ways to implement distributed digital signal processing exist each with certain strengths and weaknesses Choosing the optimum implementation for a particular applicatio
18. the controlling computer the only limitation is the number of available communication interfaces on the workstation 1 2 Communication Digital signal processing systems usually require relatively high data throughputs especially when the data processing is performed in a distributed manner DSP based systems following the general architecture depicted in Figure 1 present two different kinds of data links 1 Workstation to workstation or Controlling computer to workstation data paths Data and command transactions between the workstations in the system are performed through the Ethernet using TCP IP protocols Featuring raw transfer rates at up to 10 Mbits per second for 10 Base T Ethernet or 100 Mbits per second for 100 Base T Ethernet or Fast Ethernet this type of data link should provide enough throughput for the majority of digital signal acquisition and processing applications including multimedia and digital image acquisition and processing 2 Workstation to DSP data links Most of the data processing required by a given application is performed by the DSP boards connected to workstations Thus the workstation DSP communication is of major importance for the overall system performance and can be a potential bottleneck for data transfers Although DSPs and DSP based boards generally feature a variety of data communication capabilities they typically do not provide complete glueless high performance communication interfaces to th
19. 0217 1 800 441 2447 or 303 675 2140 Fax 303 675 2150 LDCForFreescaleSemiconductor hibbertgroup com Information in this document is provided solely to enable system and software implementers to use Freescale Semiconductor products There are no express or implied copyright licenses granted hereunder to design or fabricate any integrated circuits or integrated circuits based on the information in this document Freescale Semiconductor reserves the right to make changes without further notice to any products herein Freescale Semiconductor makes no warranty representation or guarantee regarding the suitability of its products for any particular purpose nor does Freescale Semiconductor assume any liability arising out of the application or use of any product or circuit and specifically disclaims any and all liability including without limitation consequential or incidental damages Typical parameters which may be provided in Freescale Semiconductor data sheets and or specifications can and do vary in different applications and actual performance may vary over time All operating parameters including Typicals must be validated for each customer application by customer s technical experts Freescale Semiconductor does not convey any license under its patent rights nor the rights of others Freescale Semiconductor products are not designed intended or authorized for use as components in systems intended for surgical implant into the body
20. 1 kpds close object inobj2 kpds close object outobj if lodpathname kfree and NULL lodpathname main after lib call end Execution of the glyph ends here and data is available to the next node in the dataflow chain kexit KEXIT SUCCESS ingeratnae D DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor n Additional routines needed by the kmsum glyph Routine Name Purpose Input Output Written By Date Modifications kfprintf kstderr kmsum usage additions Prints usage additions in kmsum usage routine None None ghostwriter oname kmsum July 1 1997 void kmsum usage additions void usage additions usage additions end Routine Name Purpose Input Output Written By Date Modifications ARGSUSED void kmsum free args kexit status kaddr kmsum free args Frees CLUI struct allocated in kmsum get args None None ghostwriter oname kmsum July 1 1997 status client data do the wild and free thing if clui info NULL kfree and NULL clui_info gt il_file kfree and NULL clui_info gt i2_file kfree and NULL clui_info gt o_file kfree and NULL clui info commport string kfree and NULL clui info free handler additions free handler additions end tOutput is the sum of two input
21. A interfaces 2 1 The Command Protocol The command protocol used to control the behavior of the DSP board is tailored so that the various tasks it performs meet the exact requirements of a distributed digital signal processing system Every command consists of two bytes followed by zero to two 24 bit words depending on the particular command issued The first byte is always AA signaling the beginning of a new command The second byte is the actual command byte Its bit field encoding is shown in Figure 2 es a e EE gt x v Figure 2 Command Byte Bit Field Encoding Table 1 Command Byte Bit Field Descriptions Bit Meaning First Data Word Second Data Word Rs Issues a soft reset to the DSP board Ld Loads a data buffer in DSP memory in Address in P X or Y memory Buffer size number of 24 bit the specified address space P X or Y where data is to be stored words Rn Starts the execution of a program Address in P memory where program begins The two shaded bits in the command byte are reserved for later development and should be written with Os for future compatibility The command byte is followed by zero one or two additional 24 bit words depending on which command is issued In the case of the reset command no additional word follows the command byte For the load command the command byte is followed by two additional words The first one specifie
22. Freescale Semiconductor Integrating the DSP563xx in Distributed Computing Environments Application Note by Mihai V MICEA Mircea TRIFU and Adrian TRIFU AN2088 Rev 0 3 2001 S Freescale Semiconductor Inc 2004 All rights reserved T d oe z freescale For More Information On This semiconductor Go to www freescale Freescale Semiconductor Inc How to Reach Us Home Page www freescale com E mail support freescale com USA Europe or Locations Not Listed Freescale Semiconductor Technical Information Center CH370 1300 N Alma School Road Chandler Arizona 85224 1 800 521 6274 or 1 480 768 2130 support freescale com Europe Middle East and Africa Freescale Halbleiter Deutschland GmbH Technical Information Center Schatzbogen 7 81829 Muenchen Germany 44 1296 380 456 English 46 8 52200080 English 49 89 92103 559 German 33 1 69 35 48 48 French support freescale com Japan Freescale Semiconductor Japan Ltd Headquarters ARCO Tower 15F 1 8 1 Shimo Meguro Meguro ku Tokyo 153 0064 Japan 0120 191014 or 81 3 5437 9125 support japan freescale com Asia Pacific Freescale Semiconductor Hong Kong Ltd Technical Information Center 2 Dai King Street Tai Po Industrial Estate Tai Po N T Hong Kong 800 2666 8080 support asia freescale com For Literature Requests Only Freescale Semiconductor Literature Distribution Center P O Box 5405 Denver Colorado 8
23. as the core of a distributed DSP system as described in Section 1 This code provides a monitor like interface to provide processing autonomy to the DSP and to facilitate as much as possible its programming and command from higher level user applications The DSP56307 in a Distributed Environment For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 3 1 General Description of the Monitor Program The Monitor program is certainly one of the most important parts of the entire project Its purpose is to provide low level communication routines for the DSP algorithms to use as well as to implement the DSP side of the binary command protocol described in Section 2 It was designed to be automatically loaded at boot time from on board flash memory and to act as a command interpreter for commands sent by the workstation computer For details regarding the selection of the boot procedure required for stand alone operation refer to the DSP56307EVM User s Manual The Monitor program was written in the Atmel AT29LVO10A flash memory chip using the program flash supplied with the DSP56307EVM kit One of the major concerns when writing the code for the Monitor program was to make it as small as possible The goal was to make it less than 256 words the actual size of the code plus temporary storage is 251 words It is loaded at boot time at address P 3E00 which is the last 256 word segment of on chip program m
24. ce Hall 1996 6 Introduction to the KHOROS System Khoral Research Incorporated Advanced Khoros Manuals 1997 7 Toolbox Programming Manual Khoral Research Incorporated Advanced Khoros Manuals 1997 8 Programming Services I Foundation Services Khoral Research Incorporated Advanced Khoros Manuals 1997 9 Programming Services II Data Services Khoral Research Incorporated Advanced Khoros Man uals 1997 10 Programming Services III GUI and Visualization Khoral Research Incorporated Advanced Khoros Manuals 1997 11 Khoros Programming Tutorial Rafael Santos Internet Resources 1997 About the Authors Mihai V Micea is a lecturer at the Computer Software and Engineering Department at the Politehnica University of Timisoara and Executive Director of the DSP Applications Lab Timisoara DALT sponsored by Digital DNA from Freescale Adrian TRIFU and Mircea TRIFU are students at the Automation and Computer Science Faculty at the POLITEHNICA University of Timisoara and members of the research and development team at DALT Contacts e micha dsplabs utt ro Attp dsplabs utt ro dalt References For More Information On This Product Go to www freescale com Freescale Semiconductor Inc Iniegra hg ihe D DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com
25. ctor Inc 4 2 2 FFT Algorithm This algorithm calls the FFTR2CN macro O Motorola Inc to perform the actual FFT Code Example 19 FFT Algorithm Source Code km ft asm include sincos include bitrev include fftr2cn NoOfPoints equ 256 AddrOfInput equ 1 AddrOfCoefs equ 4096 AddrOfOutput equ 1024 org x 0 XLength ds 1 org y 0 YLength ds 1 sincos NoOfPoints AddrOfCoefs org P 100 Start fftr2cn NoOfPoints AddrOfInput AddrOfOutput AddrOfCoefs Send move SFFFFFF MO move 54 R0 move p RO R1 move AddrOfOutput RO move NoOfPoints A0 move 2 A2 jsr R1 move AddrOfOutput RO move 1 A2 jsr R1 rts End 4 3 The KHOROS Environment The KHOROS software package from Khoral Research is an advanced and complete digital signal processing and scientific software integration and development environment featuring advanced inter process and distributed computing support KHOROS originated as a research project at the University of New Mexico The first release of the KHOROS system KHOROS 1 0 Beta was made available via anonymous FTP in October 1990 6 The goal of the KHOROS software is to provide a complete application development environment that redefines the software engineering process to include all members of the project group from the application end user to the infrastructure programmer 6 4 3 1 Cantata The primary component of KHOROS is its visual design and simulation tool called Canta
26. der Also 2 2 5 4200s e rap SE d S PEE TEE d RE dap RE dS 23 4 2 2 EEG Ae OH uae Sau Re RE IHREN REN A ERE IN ER eA E qu E 24 4 3 The KHOROS EnviroDtBent 2ocedek e RE Re e RE KC ERA RES eadeew see sets 24 4 3 1 Cantata CPP bt oom ced ees eee TTC TIN 24 4 3 2 The Glypli 44 2 iaceo eC WAHR HR Ee mad Ao de ALEGRE 25 4 3 3 TOBIBOROR 3 504 wire He gS E RO E E Sapa AE E A E A E M ard 26 4 3 4 Software Objects siseses berd px a eR ai ipa EPa E E EE RIA EE E bee 26 4 3 5 Development Toolsi ou urbi b PR bat ae dta hoe RU AS ER a RSS 26 4 3 6 D t Structure TE wr 26 4 4 Using the RHOROS Soltwate 0 5 54d a Re ER RE REESE CRI E redone SE EPA 27 4 4 1 B rlding a ToolbOX a rne sabe dedere k eae a dt dct e tee A EE a 27 4 4 2 Adding a Glyph to a Toolbox sseleeeeeeeee e 27 4 4 2 1 Defining the Glyph Tasks sors as we SoqE S EE rep XY E eK RXENdXEF RM 28 4 4 2 2 Creating a Directory Structure ule esae prx e dr EP OO ERPR E Ex ERES 28 4 4 3 Defining the Glyph User Interface cc once idec dh Pese be RARE Qr b E UR nes 28 5 Conclusion 22s osdocr e REPE VIDA Ra Rug be cede Bab EOD Ee ET ted 34 6 Beferenc s co iisiebarecndicee ended RA ERE PERF EREERRRTARAE E ERE ELEC RS 35 Integrating the Dl in Distributed Computing Environments For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 1 Introduction Distributed digital signal processing is a top research and development topic backed up by an ever inc
27. e actual source code of the software object e kroutine kmsum uis contains the User Interface Specification UIS of the software object e kroutine kmsum help kroutine kmsum man vtkroutine kmsum html contain manual pages and help files for the software object The skeleton files for all of these directories are automatically generated by Craftsman 4 4 3 Defining the Glyph User Interface After a glyph has been added to a toolbox a User Interface Specification UIS must be constructed The UIS defines all the interaction between the user and the software object For the kmsum software object which is a non interactive Kroutine this interaction is limited to setting the command line parameters to the object code inputs output and flags before the routine is run For a command line type of interface the Ghostwriter tool is used to create the UIS If a graphical user interface is needed instead the GUISE application assists in the GUI design Both tools generate code inside the software object skeleton file that was created in Composer User written code in the skeleton file is enclosed in special comment fields called tags to prevent it from being altered The most commonly used tags are summarized in Table 4 The last three sets of tags are not required they are used to clarify the organization of the user code iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale
28. e host computer this task is left to the system designer The DSP56300 family of processors features several data communication interfaces suitable for a wide variety of system interconnections through its built in peripheral ports At the higher end of the transfer performance scale is the Host Interface Port a DSP to host interconnection offering transfer rates of up to 16 Mbits per second for the ISA bus and even higher rates for the PCI bus One potential disadvantage of this solution is that the DSP must be physically close to the host bus slot to minimize RFI which drastically reduces overall system flexibility At the lower end of the performance scale is the asynchronous Serial Communications Interface SCI which can transfer data at up to 115 2 kbits per second The SCI port on the DSP56307 Evaluation Module DSP56307EVM was chosen as the solution for this application This on board solution provides a data transfer which meets communication requirements for this application while maintaining system flexibility With the EVM configured to transfer data at the iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc ioni onitor Program maximum speed 115 2 kbits per second and the use of special handshaking the DSP can easily be controlled through user defined applications running on the workstation The implementation details of thi
29. edure Transmit routine The DSP56307 in a Distributed Environment For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 3 2 1 Definitions The first portion of the Monitor program defines the code defining symbolic constants for the memory mapped registers It also declares string constants and reserves memory for data storage both of which are used in the command confirmation strings This code is listed in Code Example 12 Code Example 12 Monitor Program Equates and Defines IPR P equ SFFFFFE PCTL equ SFFFFFD PCRE equ SFFFF9F SCR equ SFFFF9C SCCR equ SFFFF9B SRxH equ SFFFF9A SRx equ SFFFF99 SRxL equ SFFFF98 STxH equ SFFFF 97 STX equ SFFFF96 STxL equ SFFFF95 SSR equ SFFFF 93 ORG P S3E00 JMP Init SReset DCB eR yda SRun DCB ko SRAdr DS 1i SLoad DCB dL SLAdr DS 1 ROReg DS al Regs DS 7 LastReg DS 1 iegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor ING sentation on DSP56307 3 2 2 Initialization After a soft reset the main program begins with an initialization sequence starting at the Init label This sequence consists of the following steps 1 The IPL in the SR register is set to 2 and the IPL of the SCI is set to 2 to ensure that the SCI Receive Data interrupt is not masked out The PLL is
30. emory in the default memory space configuration The Monitor program is composed of two main parts one part receives commands and data from the workstation and performs specific tasks The other part carries out all DSP to workstation data transfers Flow diagrams of these parts are shown in Figure 3 and Figure 4 Transmit Yes D Send P Y N memory es 2 Vv v Send X Send Y memory memory Ox Figure 3 Transmit Routine Flowchart iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor ING sentation on DSP56307 Receive v command byte Read Reset DSP v y No Reset Yes v Jump bootstrap Read block size and address Load Ce i Exit v Read block into P memory Yes emi No Vv v No v Read program address v Set up return address Read block into X memory Read block into Y memory O v Send checksum Figure 4 v RTI Receive Routine Flowchart 3 2 Monitor Program Implementation on DSP56307 This section presents the source code for the Monitor program including Symbolic constant definitions Initialization sequence Receive routine Load procedure Run Command proc
31. ese additions Output is the sum of two inputs Options Figure 9 Final Pane for kmsum The final source code of the adder glyph is listed in Code Example 20 Due to the length of the code only the main C file kmsum c is listed Case Ru For More Information On This Product Go to www freescale com KHORO fif def static c fendif Freescale Semiconductor Inc Code Example 20 Final Source Code of the Adder Glyph kmsum c S Id ined lint amp amp defined har rcsid KHOROS Id CODECENTER gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt XXXXXXXXXXXXXXXX X X X X gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt Main program for kmsum Private main Static Public gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt gt XXXXXXXXX X 4 4 include clui_inf Routi Wri Modifi kmsum h o struct clui info NULL ne Name main Purpose main program for kmsum Input clui info il file clui info il flag ui info i2 file ui info i2 flag cl int cl ui info o file ui info o flag cl int cl Output Returns Adrian Trifu May 10
32. f the workstation as well as some high level functions to implement the command protocol It was designed to handle multiple DSP boards simultaneously each identified by a special identifier called serial device id sdid Workstation DSP Communication Protocol For More Information On This Product Go to www freescale com Freescale Semiconductor Inc All functions have a parameter sid of type sdid The sdid structure is shown in Code Example 1 Code Example 1 The sdid Structure struct sdevid int fd struct termios oldtio h typedef struct sdevid sdid fd file descriptor associated with the serial device fil oldtio structure holding the previous settings of the serial device All functions return at least one error code The error codes are listed in Table 3 Table 3 SCOM Error Codes Error code Description ER OK The operation was successful ER UNKNOWN Returned by scom run when it does not receive the command confirmation string ER PARM At least one argument passed to a function is invalid ER TOUT A time out occurred before reading a specified number of bytes ER IO Returned by scom write buf when it fails to send the entire buffer ER SUM Received checksum is not the same as the computed one ER STX Syntax error is encountered in the LOD file ER DSP Attempt to reset the DSP board fails ER EXT Returned in all other cases ER EXT is a macro the actual value
33. from Khoral Research Incorporated running on the Linux platform All DSP specific operations were initiated on the controlling computer and run on a DSP56307 Evaluation Module attached to the workstation through the SCI A custom toolbox called Freescale EVM was created containing two specific DSP routines Glyphs kmsum for adding two input signals and kmf ft a 256 point Fast Fourier Transform both running on Freescale DSPs in a distributed manner The kmsum glyph was used illustrate the use of the KHOROS software The principles developed here can be applied to any DSP based application that can be implemented in a distributed manner such as digital signal acquisition and processing from sources scattered over large areas and integrating multiple measurement and instrumentation units into a larger digital signal analysis structure iegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc TEE Sonare 6 References 1 DSP56300 Family User s Manual Motorola Incorporated 1999 2 DSP56307 24 Bit Digital Signal Processor User s Manual Motorola Incorporated 1999 3 DSP56307EVM User s Manual Motorola Incorporated 1999 4 Freescale DSP Assembler Reference Manual Motorola Incorporated 1999 5 Digital Signal Processing Principles Algorithms and Applications A J G Proakis and D G Manolakis 3rd edition Prenti
34. h the DSP based board through its standard PC asynchronous serial interface which can exchange data at rates up to 115200 bps The primary reason for using this port is that it can connect to the SCI port on the Freescale EVM as described in Section 1 2 on page 2 Serial program and data communication with the DSP is implemented on the host workstation as a library of C callable functions on UNIX like platforms When execution of the DSP network on the controlling computer reaches one of the remote DSP glyphs it issues a specific command to the workstation and the routine that implements the corresponding glyph is launched The primary actions performed by this routine include the following e Receive additional input data if any e Load the input data and parameters into the DSP through the serial data link e Load the object code for the specific DSP routine into the DSP Wait for the results from the DSP e Send the results back to the controlling computer 4 1 3 DSP Based Board The Freescale DSP563xx Evaluation Modules EVM was selected as the DSP hardware core of the distributed system It uses an on board standard serial communication interface SCI port to communicate with the host workstation using the Monitor program described in Section 3 on page 9 The actual DSP routines to be run on the EVM resides on the host workstation in the lod loadable object code format They are loaded into the DSP along with additional in
35. he controlling computer e Anefficient mechanism for transacting commands and data between the system workstations The corresponding programming model emulates a client server architecture with the controlling computer as the client and the workstations servicing its various processing requests e Specialized server type programs implemented on the system workstations which can communicate with the client controlling computer receive processing commands and additional input data and send back the results The workstations can perform a specialized set of processing algorithms individually or a generic pack of processing routines can be implemented on each workstation The KHOROS Software Package from Khoral Research Incorporated meets all of these requirements and so was selected to provide the high level control software for the proposed system Section 4 presents a case study describing a distributed digital signal processing system with the general system architecture depicted in Figure 1 Two personal computers are interconnected with a TCP IP based Ethernet link running on a Linux platform one functioning as the controlling computer and the other as the system workstation The core of the system is the DSP56307 Evaluation Module which performs all the DSP specific operations initiated by the remote station controlling computer The KHOROS package is also described in detail in this section 2 Workstation DSP Communication Protocol
36. i _ OA y DSP based L Board I ANS L E Networked g Computer iu DSP based Board Controlling Computer Ay Figure 1 Distributed DSP System Introduction For More Information On This Product Go to www freescale com Freescale Semiconductor Inc The primary challenge in such a distributed computing environment is to implement and manage multiple digital signal acquisitions and processing without increasing the workload of the host computer processor There are three main components of the system 1 A controlling computer which assists the user in building and running specific DSP based applications or algorithms and controls the flow of commands and data through the distributed system 2 Oneor more workstations interconnected to each other and to the controlling computer through the Ethernet The number of workstations and the relative distance are not subject to restrictions derived from this application Each workstation receives and interprets DSP related commands from the controlling computer passes the commands to the DSP based board connected to it and sends back results from the DSP to the controlling computer 3 A specialized DSP based board attached to each workstation which carries out the entire data acquisition process and implements most of the data processing algorithms The system architecture does not limit the number of DSP boards that can be connected to a particular workstation or even to
37. incorporate a DSP algorithm into a software object which KHOROS can use to implement the algorithm in a distributed processing network The general steps involved in creating a KHOROS software object include the following e Create a new toolbox or open an existing one using the Craftsman tool e Create the new software object using the Composer tool Design the user interface for the created software object using either Ghostwriter for a command line interface or GUISE for a graphical interface The adder algorithm is used to illustrate the process of building the glyphs and toolboxes KHOROS uses to implement a distributed processing system 4 4 1 Building a Toolbox Before a glyph can be created in KHOROS the user must first configure a toolbox in which it will reside The Craftsman tool is used either to modify a predefined toolbox or to create a new toolbox For the case study a new toolbox called MotorolaEVM was created to contain all the data processing glyphs that make use of the Freescale evaluation modules The creation of a toolbox is quite straightforward and is described in the KHOROS manuals 4 4 2 Adding a Glyph to a Toolbox Adding software objects to a toolbox is also straightforward thanks to Composer This tool features a C code editor plus predefined operators and data structures which can be used to create the software object If the user wishes to write custom code to create the object Composer can be configured
38. kes it possible to issue multiple commands in the same command byte this is not permitted in the protocol If more than one command bit is set the command corresponding to the most significant bit set executed For example if the Reset bit is set other command bits are ignored If no command bit is set the Run command is assumed Bits 0 2 of the command byte are ignored and for the Reset and the Run commands For the Load command these bits they are checked in the following order P X and Y if none of these bits is set Y is assumed Refer to the receive routine flowchart in Figure 4 on page 11 In response to the Reset command the routine issues a soft reset by writing the default boot time values to the SR SC SP and OMR registers and jumping to the bootstrap code at P SFF0000 The default setting for the MD MC MB and MA bits in the COM register is assumed to be 1001b meaning boot from byte wide memory The bootstrap code then reloads the Monitor program from the flash memory and runs the main program starting at Init label This code sets up of the DSP board and sends the Ready command confirmation string indicating that the Reset was successful The receive routine is listed in Code Example 14 Code Example 14 Monitor Program Receive Routine Receive OVEM RO P ROReg LRA Regs RO OVEM MO P LastReg OVE FFFFFF MO OVE R1 P R0 4 OVEM A2 P RO 4 OVEM Al1 P RO OVEM AO
39. mit Al OVEM Al1 P RO Set transmit entry point OVEP S0B02 X M_SCR Set up SCI communication parameters OVEP 13 X M_SCCR Set SCI baud rate OVEP 7 X M_PCRE Enable Rx Tx and SCLK pins for SCI OVE F4 A2 LRA SReset RO OVE 2 A0 JSR Transmit WAIT JMP _ Loop The DSP56307 in a Distributed Environment For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 3 2 3 Receive Routine When the SCI receives a byte it generates a SCI Receive Interrupt which wakes the DSP from the WAIT state Program execution resumes by servicing the pending interrupt The Receive Data interrupt handling routine saves all the registers used by the routine and restores them upon exiting the routine After saving the registers to memory the Monitor program reads the received byte from the lower 8 bits of the SCI Receive Data Register memory mapped at address X FFFF98 If this byte is equal to AA a new command is being issued by the workstation and the Monitor program waits to read more bytes according to the protocol If the received byte is not equal to AA the routine restores the saved registers and exits and the DSP resumes the WAIT state Once the AA byte is received the Receive Data Interrupt handling routine does not exit until the current command is fully processed The next received byte after the initial AA is the command byte illustrated in Figure 2 on page 4 Note that although the encoding scheme ma
40. n is often difficult depending largely on the requirements of the application This paper proposes a hardware and software structure for distributed digital signal processing which offers flexibility and scalability for many real life applications 1 IEFOGUCUON M 1 1 1 FN ITI Pr E aE E Sse E E E 1 1 2 Communication i ssceuaccs kac ERG ERG x GU CER E ERG Ge eed d ER eR RR 2 1 3 Monitor Progt M esos o appe daro FR S E REOR I RI a a a a a A 3 14 High Level Software i i siocasoiama Xe Ree A RAE RE RAE Rb TAS APP RTR RE E ERE 3 2 Workstation DSP Communication Protocol 3 2 1 The Command Protocol 5 ipa pa oic Ca Grac o ipio RO e ep p Rd d 4 2 2 Communication Library Functions Overview eeeeeeeee eee 5 2 2 Low Level Functions i aures die 9 dea crt ser P CR CHER RI HERRERA SCR IR po ORO ER 6 2 2 1 1 SCOM TI iun did eso eee ote See ee AR A sre Suid ees e dA bee dE d 6 2 2 1 2 seo sh t dOWIL Loue ed EAE raene earan E rean ELE RI E EE MR 7 2 2 1 3 SCO write Duf seaetenreeda tesi iatna a e ERE RP EEE Ea es 7 2 2 1 4 scom read buf is ond teaa aaa EE ep Ep due ER T 2 2 1 5 Maps get CUO e oaa sace CES porn SaaS Xa exo x pa Ia a E eae nds 8 2 2 2 High Level Functions 2 24 2 4224 cee e ERE REY ERGO EXE RGX GU Or bebe bee e Rn 8 2 2 2 SCSOM reset USD Li seda ri Pp ERE ETE ERE TOU PD qEARQIAPQUENU E RETE 8 2 222 BESOIN FUN ieee et cense cee RimEPEIRA SA ARE EA E dae ee A era doas 8 2 2 2 3 scom load DrOPEdH c
41. nd data to the corresponding workstation through the Ethernet and waits for the processing results to be returned from the DSP The received results are further used by the implemented DSP network The fact that some algorithms in a specific DSP network are executed by other components of the overall system is transparent to the user 4 1 2 Workstation The primary function of the workstation in the proposed distributed DSP system is to serve as a remote interface between the controlling computer s DSP network implemented under the Cantata application and the DSP based board where the actual signal processing algorithms are executed in distributed manner The system requirements for the workstation include e The LINUX operating system e An Ethernet interface for communicating with the controlling computer e The KHOROS package installed and configured to interpret the specific commands and input data issued by the controlling computer under Cantata and received on the DSP network e A custom designed glyph toolbox for Cantata implementing DSP algorithms to be run on the DSP based board in distributed manner This is the same toolbox implemented on the controlling computer Acommunication library implemented on the LINUX platform for direct interconnection with the corresponding DSP based board Case Stud For More Information on This Product Go to www freescale com Freescale Semiconductor Inc The workstation communicates wit
42. ponents Case Study For More Information On This Product Go to www freescale com Freescale Semiconductor Inc There are three basic types of glyphs used in a typical DSP network e Input glyphs input procedures and programs such as signal generators or data import nodes e Data processing glyphs procedures that implement a large variety of algorithms e Output glyphs used for visualization or storage of results 4 3 3 Toolboxes The various types of glyphs can be grouped together in a toolbox to clarify the way in which they are organized A toolbox can be regarded as a collections of related software objects as they are referred to in KHOROS terminology The purpose of a toolbox is to facilitate domain specific work while simultaneously enabling cross domain collaboration 6 Users can define their own toolboxes containing their own software objects in relation to or based on others if desired There is a one to one correspondence between software objects in a toolbox and their graphical representations as glyphs 4 3 4 Software Objects Software objects can be of several types including the following e Kroutines non interactive programs which read their data from a standard input process the data and write results to a standard output Every aspect of their execution is determined before running the program and once the execution has begun users have no means of intervening An example of a Kroutine could be
43. put data and parameters by the corresponding glyph routines on the host machine iegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc DSP Algorithms 4 2 DSP Algorithms Each of the two DSP algorithms in our case study the adder and the FFT starts with all input data already loaded into memory on the DSP board Each algorithm processes the data and calls the Transmit routine in the Monitor program to send the results back to the workstation 4 2 1 Adder Algorithm This algorithm starts with two data buffers in X and Y memory spaces starting at address 1 Address 0 in both X and Y memory hold its respective buffer size The algorithm performs an in place sum of the two signals using the larger buffer as a destination Code Example 18 Adder Algorithm Source Code kmsum asm org x 0 Length ds 1 x lt org y 0 K Length ds 1 org P 100 Start move 1 R0 move x XLength A move y YLength B cmp B A jgt XSum do A1 Loopl move x R0 A move y RO B add B A move Al y R0 _Loopl move 1 A2 move y YLength AO jmp _Send XSum do B1 Loop2 move x R0 A move y RO B add B A move Al x RO _Loop2 move 2 A2 move X XLength AO _ send move 54 R0 move p R0 R1 move 1 R0 jsr R1 rts End Case Stud For More Information on This Product Go to www freescale com Freescale Semicondu
44. reasing number of real life applications and IT companies involved Because it is still an advanced research topic and depends strongly on the application distributed DSP does not have a standard approach or a unified concept of design Distributed digital signal processing is the most suitable solution for most real life applications involving digital data acquisition and processing from multiple signal sources scattered over large areas The advantages of distributed computing over single processor or other multi processor architectures include high computing power at a low cost flexibility and scalability There are several ways to implement distributed digital signal processing each of which has its strong and weak points Choosing the optimum implementation for a particular application is often difficult depending largely on the requirements of the application This paper proposes a hardware and software structure for distributed digital signal processing which offers good flexibility and scalability for many real life applications 1 1 Architecture The general architecture of the proposed system uses an Ethernet based distributed computing environment running TCP IP protocols on a LINUX platform The digital signal processing hardware core of the system consists of one or more DSP based boards each connected to a host computer in the network as illustrated in Figure 1 Vy Networked Computer
45. returned is NO OF ERRS errno 2 2 1 Low Level Functions The communications library includes a set of functions which perform the low level operations associated with serial I O 2 2 1 1 scom init Initializes the serial port for e 8 bit mode e jstop bit no parity communication speed of 115200 bps Fills the sid structure with the appropriate device information Programs the serial device for buffered I O Sets a predefined time out value for the receiver iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Ing Functions Overview Code Example 2 The scom init Function include scom h err t scom init const char device sdid sid devic serial device file name dev ttySx sid serial device id Error codes returned ER PARM ER EXT ER OK 2 2 1 2 scom shut down Flushes the input and the output buffers of the serial device Closes the file descriptor associated with the serial device Code Example 3 The scom shut down Function include scom h err t scom shut down sdid sid sid serial device id Error codes returned ER PARM ER EXT ER OK 2 2 1 3 scom write buf Outputs the contents of a data buffer to the specified serial device Code Example 4 The scom write buf Function include scom h err t scom write buf sdid sid u int8 t
46. rial device id dsp addr address in DSP memory where to start the execution Error codes returned ER IO ER TOUT ER EXT ER UNKNOWN 2 2 2 3 scom load program Reads a LOD file containing a DSP executable code in hex format Loads all the program blocks into the DSP Issues a run command Validates check sums returned from the DSP board Code Example 9 The scom load program Function include scom h err t scom_load_program sdid sid const char pathn sid serial device id pathn name of the file containing the DSP algorithm to be loaded Error codes returned ER EXT ER STX ER PARM ER SUM ER IO ER TOUT ER OK iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Ingy Functions Overview 2 2 2 4 scom write dsp Downloads a buffer of 8 bit integers into DSP memory at the specified address in the P X or Y address space count must be a multiple of three each three byte group is concatenated to form a 24 bit word in DSP memory The first byte in the group is the least significant Code Example 10 The scom write dsp Function include scom h err_t scom_write_dsp sdid sid u_int8_t buf u_int32_t dsp_addr u_int32_t count u_int8_t pm sid serial device id buf buffer of unsigned bytes to load into the DSP memory dsp addr address where to load the buffer count
47. s Wn Gy the KHOROS Software This simple source code can be regarded as a general skeleton of any software object that uses Freescale EVMs to process data The basic steps to be taken are exactly the same no matter how complex the software object is For example a Freescale EVM based data acquisition glyph can be added which brings real world signals to KHOROS for further analysis and processing Case pudy For More Information On This Product Go to www freescale com Freescale Semiconductor Inc 5 Conclusion This paper provides a general framework for integrating Freescale DSP563xx based boards in distributed digital data acquisition and processing systems with special focus on the data communication between the DSP based board and the host workstation A specific communication protocol was developed using the asynchronous serial interface able to transact data up to 115200 bps To improve the autonomy of the Freescale DSP563xx based boards a special Monitor program was developed to transfer programs and data between the DSP and the workstation through the serial communication interface SCI as well as the corresponding communication library for the workstation A case study distributed digital signal processing system was designed using two personal computers one as a controlling computer and one as a workstation The computers were connected by a TCP IP based Ethernet Both computers used the KHOROS Software Package
48. s model called a polymorphic data model is based on the premise that data in DSP applications is generated either to model or acquired from real world phenomena and consequently is suitable for these purposes 9 Polymorphic data is composed of five so called segments including VALUE LOCATION TIME MASK and MAP iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc ie erionge Sonare The primary segment is the VALUE segment in which the actual data is stored as a time series of volumes in space 9 The VALUE segment has five dimensions three positional coordinates a time coordinate and an element which is the size number of components of one point in space time as illustrated in Figure 8 VALUE data can be given explicit positioning in space and time with the LOCATION and TIME segments value element VALUE Data vector each element has an implicit s Ketements position in the value data YO time width height Figure 8 Structure of the VALUE Segment The MASK and MAP segments are provided for convenience The MASK segment is used to mark the validity of each point of VALUE data The MAP segment is provided as an extension to the VALUE data VALUE data can be used to index into the MAP data 9 4 4 Using the KHOROS Software This section illustrates how the KHOROS software is used to
49. s system are presented in Section 2 1 3 Monitor Program One of the most important components of the project is a monitor emulation program developed on the DSP56307EVM to provide processing autonomy to the DSP and facilitate its programming and command from higher level user applications The Monitor program provides low level communication routines for DSP algorithms and implements the DSP side of the binary command protocol described in Section 2 It is automatically loaded at boot time from the DSP56307EVM on board flash memory and acts as a command interpreter for commands sent by the workstation The implementation of the Monitor program is described in detail in Section 3 1 4 High Level Software Another important issue related to the proper operation of a distributed DSP system is the implementation of high level software to provide the user with efficient control of the execution of DSP algorithms This software must include the following features e A proper user interface for controlling overall system operation It should provide an interactive and flexible mechanism for defining a particular DSP based application with maximum user control over its execution in a distributed processing system This software component resides in the controlling computer e Support for dividing a complex task into subcomponents to further distribute them for execution in the system This can be performed automatically or in a user controlled manner by t
50. s the address in P X or Y memory of the DSP where the data should be stored while the second word is the size of the buffer expressed as a number of 24 bit words The load command is followed by a number of data words equal to the buffer size The run command byte is followed by one additional word containing the address in P memory where program execution begins Each 24 bit word is sent as a series of three bytes least significant byte first The choice of an unencoded command byte was made to reduce the amount of code needed to distinguish between the commands on the DSP Refer to Chapter 4 to see how this is actually done iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Ing Functions Overview Every command received by the DSP board is acknowledged by a command confirmation string These strings are shown in Table 2 Table 2 Command Confirmation Strings Command String Subsequent Word Reset Ready Load Ld Buffer checksum Run Ok Run address The command confirmation string for the load command is followed by a 24 bit checksum computed as a modulo 27 sum of all the data words in the buffer The command confirmation string for the run command is followed by the address supplied in the run command A typical communication session between the workstation and the DSP board might proceed as follows
51. size of buffer pm a flag specifying the address space P X or Y memory Should contain one of the following values PMEM XMEM or YMEM Error codes returned ER EXT ER PARM ER SUM ER IO ER TOUT ER OK 2 2 2 5 Helper Functions The following two functions are referred to as helper functions because they perform data conversion and are used in conjunction with specific DSP algorithms scom write data dspis used to load a buffer of 24 bit integers to the DSP data memory It does so by ignoring the most significant 8 bits of every 32 bit integer scom read data dsp is used to read a buffer of 24 bit words from the DSP It stores each word in a 32 bit signed integer extending its sign bit to the 8 most significant bits of the 32 bit integer Code Example 11 The Helper Functions include scom h err t scom write data dsp sdid sid int buf u int32 t count u int8 t pm err t scom read data dsp sdid sid int buf u int32 t count sid serial device id buf buffer of 32 bit integers of data count size of buffer pm a flag specifying the address space XMEM or YMEM Error codes returned scom write data dsp ER EXT ER PARM ER SUM ER IO ER TOUT ER OK scom read data dsp ER TOUT ER OK 3 The DSP56307 in a Distributed Environment This section focuses on the implementation of the code running on the DSP56307 Evaluation Module DSP56307EVM
52. stributed computing generic architecture presented in Section 1 and is illustrated in Figure 5 DSP Network Controlling computer DSPS6307EVM Workstation Figure 5 Architecture of the Proposed System The system has three main components a controlling computer a workstation and a DSP56307 Evaluation Module 4 1 1 Controlling Computer The controlling computer is a special kind of workstation that controls overall system operation It s functions include e Assisting the user in implementing the desired DSP network e Controlling the distribution of tasks among the various processors e Gathering and presenting the final results The system requirements for the controlling computer include e The Linux operating system e An Ethernet interface running TCP IP protocol iiegra hg fie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc g f eneral Architecture e The KHOROS package installed and configured e The Cantata component of the KHOROS package installed and configured e A custom designed glyph toolbox created in Cantata which implements and distributes the DSP algorithms to be run on the DSP based board The Cantata application is a highly interactive graphic environment which uses a building block called a glyph see Section 4 3 2 on page 25 On the controlling computer Cantata is used to build a DSP Network
53. t relevant to our discussion res scom init clui info commport string amp did if res ER OK kerror lib rtn Error in scom init Invalid tty specified Wn kexit KEXIT FAILURE main_before_lib_call_end Case pudy For More Information On This Product Go to www freescale com Freescale Semiconductor Inc main library call Read the input data into the two defined buffers if inbufl kpds get data inobjl KPDS VALUE ALL NULL NULL kerror lib rtn Error reading input object s n clui info il file kexit KEXIT FAILURE if inbuf2 kpds get data inobj2 KPDS VALUE ALL NULL NULL kerror lib rtn Error reading input object s n clui info i2 file kexit KEXIT FAILURE Send the data from the input buffers to the the DSP X and Y memory spaces respectively res scom write data dsp amp did int inbufl u int32 t wl EM if res ER OK kerror lib rtn Error d writing X memory s n res mdsp get error res kexit KEXIT FAILURE res scom_write_data_dsp amp did int inbuf2 u_int32_t w2 YMEM if res ER OK kerror lib rtn Error d writing Y memory s n res mdsp get error res kexit KEXIT FAILURE Load the object code of the actual DSP routine to be currently executed on the DSP56307 For simplicity the code to synchronize the
54. ta a data flow visual language integrated in a powerful visual programming environment It is extremely useful for developing and testing DSP algorithms in a very intuitive manner Cantata enables a user to visually design a graphical data flow structure called a DSP network assign nodes of the DSP network to the various computers on the network and simulate the network while taking care of the various synchronization issues involved A screen capture of a DSP network is shown in Figure 6 iegra he bie Er DSP563xx in Distributed Computing Environments More Information On This Product Go to www freescale com Freescale Semiconductor Inc KHOROS Environment Figure 6 Example of a Cantata Screen Layout 4 3 2 The Glyph The building block of the DSP network is the glyph A glyph is a visual representation of a process which runs on either a local or remote machine and performs a specific task such as the generation of a signal processing of a previously generated set of signals visualization of signals etc These fundamental construction blocks can be linked together in various ways to form the desired DSP network A glyph can best be viewed as a black box with inputs outputs and other specific controls as illustrated in Figure 7 Pena Accegd Freescale Semiconductor Inc EN output Connection Input Connection Output Data Connaction Input Data Connect ion Figure 7 Glyph Layout and Com
55. to generate just a Case Stud For More Information on This Product Go to www freescale com Freescale Semiconductor Inc skeleton C code file to which the user can add the C statements needed to completely define all the object functions The code written by the user inside the skeleton file is placed between a corresponding pair of special tags In this case study two glyphs were added to the MotorolaEVM toolbox kmsum and kmf ft The following paragraphs describe implementing kmsum to illustrate the process 4 4 2 1 Defining the Glyph Tasks The kmsum Kroutine must perform the following tasks e Open its two inputs and one output Initialize communication with the attached DSP evaluation board local or remote e Send data from the two inputs to the X and Y data memory on the DSP board Download the adder software routine to the DSP board e Run the adder software e Receive the results from the DSP board Write these results to the output e Close the inputs and output and perform any other required cleanup tasks 4 4 2 2 Creating a Directory Structure The KHOROS software automatically creates the directory structure it requires In our case study the following directories were created within the freescale objects directory e kroutine the directory in which all Kroutines source code of the current toolbox are stored e kroutine kmsum the root of the Kroutine in this toolbox e kroutine kmsum src contains th
56. tual data All type conversions are made automatically by the read write functions kpds set attribute inobjl KPDS VALUE DATA TYPE KINT kpds set attribute inobj2 KPDS VALUE DATA TYPE KINT To determine how much data must be processed taking into account that the two inputs can have different lengths the greater of the two sizes is taken as the size of the output kpds get attribute inobjl KPDS VALUE SIZE amp wl amp h amp d amp t amp e kpds get attribute inobj2 KPDS VALUE SIZE amp w2 amp h amp d amp t amp e max wl gt w2 wl w2 Finally create the value segment in the output object and set its dimensions according to the maximum value determined earlier The input and output objects are now ready to be used kpds create value outobj kpds set attribute outobj KPDS VALUE SIZE max 1 1 1 1 kpds set attribute outobj KPDS VALUE DATA TYPE KINT Before moving on to the next pair of tags the the serial communication must be initialized with the evaluation board This is done with a call to scom_init part of the communications library which is described in Chapter 2 For simplicity it is assumed that the size of the inputs does not exceed the size of the X and Y memories Otherwise the data would have to be tested and eventually sent in slices However this is no

Download Pdf Manuals

Related Search

Related Contents

2001 IMPREZA SERVICE MANUAL QUICK REFERENCE INDEX  User`s Manual - BCM Advanced Research  PIKE 1064E/1068E User Guide  Le manuel d`utilisation du jeu G.Nius  LPFG user`s manual - Algorithmic Botany  取扱説明書 (5.75 MB/PDF)    

Copyright © All rights reserved. Failed to retrieve file