Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS20020013893 A1
Publication typeApplication
Application numberUS 09/918,123
Publication dateJan 31, 2002
Filing dateJul 30, 2001
Priority dateApr 22, 1998
Also published asCA2329423A1, CN1306639A, CN100449477C, EP1080404A1, US6321331, WO1999054809A1, WO1999054809A9
Publication number09918123, 918123, US 2002/0013893 A1, US 2002/013893 A1, US 20020013893 A1, US 20020013893A1, US 2002013893 A1, US 2002013893A1, US-A1-20020013893, US-A1-2002013893, US2002/0013893A1, US2002/013893A1, US20020013893 A1, US20020013893A1, US2002013893 A1, US2002013893A1
InventorsSubhash Roy, Paul Hembrook, Eugene Parrella, Richard Mariano
Original AssigneeTranswitch Corporation
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Real time debugger interface for embedded systems
US 20020013893 A1
Abstract
A debugging interface includes a pair of decoders and an event history buffer coupled to the sequencer of a processor. The first decoder is coupled to the program counter of the sequencer and the Instruction RAM of the processor. The second decoder is coupled to the cause register of the sequencer and the event history buffer is also coupled to the cause register. The first decoder provides a three bit real time output which is indicative of the processor activity on a cycle by cycle basis. The three bit output indicates seven different conditions: whether the last instruction executed by the processor was an inc, an exception, an exception with no event history buffer entry, or a branch taken, whether there has been no instruction executed since the last clock cycle, and whether a jump was an immediate jump or a jump to a register. The event history buffer is loaded with more detailed information about the instruction last executed when the first decoder indicates that the last instruction was an exception or a jump to a register, and when there is a change in state of an interrupt line or an internal processor exception. An exemplary implementation of the debugging interface is embodied on an ASIC chip having three processors. Each processor is provided with a first and second decoders and a single event history buffer for all processors is provided on the chip.
Images(4)
Previous page
Next page
Claims(3)
1. A method of debugging a processor, said method comprising:
a) providing information about processor activity in real time; and
b) associating the instructions executed by the processor with the information about processor activity.
2. A method according to claim 1, wherein:
the information about processor activity includes an indication of at least one of whether the last instruction executed was a jump, a jump based on the contents of a register, a branch taken, or an instruction which encountered an exception.
3. A method according to claim 1, further comprising:
c) providing information regarding the status of the processor when certain processor events occur, said certain processor events including at least one of a change in status of an interrupt line, an internal processor exception, and the execution of a jump instruction based on the contents of a register.
Description

[0001] This application is a continuation of U.S. Ser. No. 09/064,474, filed Apr. 22, 1998, which is hereby incorporated by reference herein in its entirety.

BACKGROUND OF THE INVENTION

[0002] 1. Field of the Invention

[0003] The invention relates to systems and methods for debugging software in real time. More particularly, the invention relates to systems and methods for the real time debugging of firmware in embedded systems, e.g. ASIC chips having one or more processors on a single chip.

[0004] 2. State of the Art

[0005] Software debugging may be accomplished in a number of ways, some of which are not performed in real time. A traditional debugging technique is to step through program instructions at a rate much slower than the rate at which the program is designed to run in real time. By stepping through the program instructions one-by-one, errors can be observed as they happen and the program code lines executed immediately prior to the error can be analyzed to find the cause of the error. This technique is not helpful, however, if the error in program execution is the result of timing errors or other types of errors which only occur when the program is running at real time speed. As used herein, the term “real time” means the rate at which a program must execute in order to process the incoming data rate which may be quite high.

[0006] A widely used technique for debugging a program which is running in real time is called “tracing”. Tracing involves recording the transactions performed by the computer as it executes the program code. The trace of activities performed by the computer during the time of a failure can be a useful guide in isolating possible causes of the failure.

[0007] Another useful debugging tool is to set breakpoints at selected places in the program. The breakpoints trap the flow of the software and provide insight into whether, when, and how certain portions of the software are entered and exited. An analysis of the flow of the software can provide information which is useful in isolating bugs.

[0008] Many state-of-the-art tracing and trapping methods are accomplished by a debug support circuit which is connected to the system bus, i.e. the bus which couples the CPU to memory. See, for example, U.S. Pat. No. 5,491,793 to Somasundaram et al. entitled “Debug Support in a Processor Chip.” Connecting a debug circuit to the system bus is convenient because addresses, instructions, and data can be accessed via the system bus. However, coupling the debug support circuit to the system bus increases the electrical load on the bus and interferes with the operation of the bus. Moreover, operation of the system bus may interfere with operation of the debug support circuit. In addition, the system bus may not provide all the information necessary for debugging a program running on a CPU which uses internal cache. These CPUs will not access the system bus if the information they need is available in cache. If an error occurs while the CPU is accessing internal cache, the debug support circuit will not be able to access the information it needs.

[0009] Another tracing and trapping method is disclosed in U.S. Pat. No. 5,833,310 to Whistel et al. entitled “On-Chip In-Circuit-Emulator Memory Mapping and Breakpoint Register Modules.” According to this method, an internal bus controller is coupled to the memory address bus and a match register. When a memory address written to the address bus matches an address in the match register, a memory mapping module maps a memory cycle to an external debug memory. The user can set specific bus event conditions for which memory is mapped by writing to a set of breakpoint registers. A disadvantage of this method is that it requires an additional set of I/O pins for the chip so that the external debug memory can be coupled to the chip. This may require a significant number of pins since the addresses to be mapped may be 32 or 64 bits wide.

[0010] Still another tracing and trapping method is disclosed in U.S. Pat. No. 5,513,346 to Satagopan et al. entitled “Error Condition Detector for Handling Interrupt in Integrated Circuits Having Multiple Processors.” According to this method, an interrupt processor controller intercepts all interrupts and routes them to the appropriate processor in a multiprocessor chip. The interrupt processor controller includes logic which determines when an interrupt will cause an error because a previously instigated interrupt has not been cleared. When such an error is detected, a bit is set in an error detect register, the bit corresponding to an interprocessor interrupt channel. The bits in the register are ORed and a single bit output indicates the occurrence of an error. The register may then be examined to determine the location of the interrupt error in the executing code. This method does not interfere with the system bus and does not require very many additional pins on the chip. However, the debugging information that it provides is limited.

[0011] The Motorola MPC-860 PowerQuicc™ includes a program development system interface port which provides a three bit output indicative of the state of the program execution as the program is being executed. The MPC-860 is a 40 mHz communications controller but the development system interface port is only operable at a rate of 4 mHz. Thus, the port can not be used for real time debugging. The specifications for the MPC-860 are found in the “MPC-860 POWERQUICC USER'S MANUAL”, Copyright 1996 - Motorola, Inc., Schaumberg, Ill., the complete disclosure of which is incorporated herein by reference.

[0012] ASIC design using one or more embedded processors poses additional debugging challenges. The prior art methods of trapping instructions at a given point in time implies that the system must be stopped to allow debugging of firmware. Once the system is stopped, however, real time events and their timing relationships are lost. If there is a firmware bug which is only identifiable in the presence of live traffic (during real time operations) it is necessary to obtain contextual information about the error before the firmware is changed.

SUMMARY OF THE INVENTION

[0013] It is therefore an object of the invention to provide a debugging interface for tracing instructions without loss of real time context and event interaction.

[0014] It is also an object of the invention to provide a debugging interface which does not interfere with the operation of a processor or system bus.

[0015] It is another object of the invention to provide a debugging interface which does not require many additional pins on a processor chip.

[0016] It is a further object of the invention to provide a debugging interface which provides access to a substantial amount of information about the executed instructions.

[0017] In accord with these objects which will be discussed in detail below, the debugging interface of the present invention includes a first decoder coupled to the sequencer of a processor and to the Instruction RAM (IRAM) of the processor. The first decoder, according to the invention, provides a real time three bit output on a cycle by cycle basis which is indicative of the processor activity during the last clock cycle. According to a presently preferred embodiment, the three bit output indicates seven different conditions regarding processor activity. In particular, the three bit output indicates whether or not a new instruction has been executed since the last clock cycle, and if a new instruction has been executed, whether the last instruction executed by the processor was an immediate jump, a jump to register, or a branch taken. In addition, the three bit output will indicate whether execution of the instruction resulted in an exception. By recording this three bit output over time, and comparing it to the actual instructions listed in the program code, important debugging information is obtained about a program which was running in real time.

[0018] According to a preferred embodiment of the invention, a second decoder and an event history buffer are coupled to the cause register of the sequencer of the processor. In particular, the second decoder is coupled to the enable input of the history buffer and the cause register is coupled to the data input of the history buffer. The second decoder decodes the contents of the cause register and enables the history buffer whenever the contents of the cause register indicates an exception, a jump register instruction, or a change in the status of an interrupt line. Whenever the history buffer is enabled, information from the cause register and the program counter is loaded into the buffer. By recording the contents of the history buffer over time, and comparing the information to the actual program code, additional important debugging information is obtained about a program which was running in real time. According to this preferred embodiment of the invention, the seventh condition indicated by the three bit output of the first decoder is whether an exception was encountered without writing to the history buffer.

[0019] According to the presently preferred embodiment, each entry in the event history buffer is forty-four bits. Each forty-four bit entry in the history buffer includes the current sixteen bit time stamp, twenty three bits from certain fields of the cause register or program counter, one bit indicating whether the entry is related to a jump or an exception, two bits identifying the processor number (in a multiprocessor system), one bit identifying whether the history buffer has overflowed, and a time stamp rollover bit. The history buffer preferably has a depth of at least sixteen entries.

[0020] An exemplary implementation of the debugging interface is embodied on an ASIC chip having three processors. Each processor is provided with two decoders as described above and a single event history buffer is provided on the chip. Nine pins on the chip are used to provide access to the three bit outputs of each first decoder. Three pins on the chip provide serial access (data, clock, and enable) to the contents of the event history buffer. These twelve pins on the chip allow a diagnostic device to be coupled to the chip during real time operations without interfering with the operation of the chip. The outputs of the first decoders and the contents of the event history buffer can be recorded over time by the diagnostic device to provide a real time record of the processing events occurring in the chip during real time. This real time record taken together with knowledge of the program code being executed provides a true picture of the processors' execution sequence in real time and thereby expedite debugging of code.

[0021] Additional objects and advantages of the invention will become apparent to those skilled in the art upon reference to the detailed description taken in conjunction with the provided figures.

BRIEF DESCRIPTION OF THE DRAWINGS

[0022]FIG. 1 (represented as FIGS. 1A and 1B on two separate sheets) is a schematic block diagram of an exemplary implementation of a real time debugger interface according to the invention; and

[0023]FIG. 2 is a schematic block diagram of a debugging system coupled to a chip embodying a real time debugger interface according to the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0024] Referring now to FIG. 1, an exemplary ASIC chip 10 incorporating a debugger interface according to the invention includes three processors 12 a, 12 b, 12 c, sharing a common clock 16 via a clock bus 17. Each processor includes an instruction RAM (IRAM) 18 a, 18 b, 18 c, an arithmetic logic unit (ALU) 20 a, 20 b, 20 c, and a “sequencer” 22 a, 22 b, 22 c. Each sequencer includes a program counter 24 a, 24 b, 24 c and a cause register 26 a, 26 b, 26 c. Each program counter contains an index of the instructions in an associated IRAM and a pointer to the index as the instructions are executed by the processor. The cause registers store current information about interrupts, exceptions, and other processor functions.

[0025] According to one aspect of the invention, a first decoder 28 a, 28 b, 28 c is coupled to each IRAM 18 a, 18 b, 18 c, and to each sequencer 22 a, 22 b, 22 c, i.e., to each program counter and each cause register. Each first decoder has a three bit output 30 a, 30 b, 30 c which is available off the chip 10 via three pins (0, 1, 2) in real time.

[0026] As mentioned above, the three bit output of each first decoder 28 provides an indication of the processor activity during the last clock cycle. Thus, the decoder 28 is arranged to indicate whether the program counter has moved its pointer to a new instruction. The decoder also decodes the instruction in the IRAM to provide information about the instruction, and decodes the contents of the cause register to provide an indication of an exception encountered during the execution of an instruction. According to a presently preferred embodiment, the first decoder 28 generates a three bit output which is interpreted as shown in Table 1, below.

TABLE 1
Output Mnemonic Description
000 NC No Change
001 INC Program Counter Increment
010 JI Program Counter Jump Immediate
011 JR Program Counter Jump Register
100 ECP Exception Encountered
101 PBT Program Counter Branch Taken
110 RSD Reserved
111 ENH Exception Encountered, No History Buffer
Entry Written

[0027] The output 000 indicates that there has been no change in the processor since the last clock cycle; i.e., the processor has not processed a new instruction and the program counter pointer has not changed. The output 001 indicates that the processor has processed the next instruction in the program; i.e., the program counter pointer has incremented to the next instruction in the index. The output 010 indicates that the last instruction processed by the processor was a “hard coded” jump to an instruction; i.e., the instruction in IRAM pointed to by the program counter includes code indicating that it is a jump instruction to an absolute address in the program. The output 011 indicates that the last instruction processed by the processor was a jump to an instruction based on the contents of a register; i.e., the instruction in IRAM pointed to by the program counter includes code indicating that it is a jump instruction to a location in the program determined by the value of a variable. The output 100 indicates that since the last clock cycle the processor has encountered an interrupt or an exception; i.e., the contents of the cause register contain code which indicates an interrupt or exception. The output 101 indicates that the last instruction processed by the processor was a pc branch taken; i.e., the instruction in IRAM pointed to by the program counter includes code indicating that it is a branch back to another instruction. The output 110 is not presently used, but is reserved for future use. The output 111 indicates that since the last clock cycle the processor has encountered an interrupt or an exception; and that no entry was made in the history buffer

[0028] The operation of the first decoder 28 and its output is illustrated with reference to a simple code listing which is shown below in Table 2.

TABLE 2
LINE NUMBER INSTRUCTION
10 Input A
20 B = 5
30 C = 2
40 D = B + C
50 If D = 7 then Goto 10
60 Goto A*10
70 B = 4
80 Goto 30
90 End

[0029] The listing in Table 2 has one “immediate” or “hard coded” jump instruction at line 80 and a conditional branch at line 50. It also has one jump instruction, line 60, based on the contents of a register, i.e. the value of A which is input at line 10. The three bit output of the first decoder during execution of the instructions shown in Table 2 is illustrated in Table 3 below where the values of variables A, B, C, and D are also shown.

TABLE 3
Current Next Three Bit
Line Line A B C D Mnemonic Output
10 20 ? ? ? ? INC 001
20 30 ? 5 ? ? INC 001
30 40 ? 5 2 ? INC 001
40 50 ? 5 2 7 INC 001
50 70 ? 5 2 7 PBT 101
70 80 ? 4 2 7 INC 101
80 30 ? 4 2 7 JI 010
30 40 ? 4 2 7 INC 001
40 50 ? 4 2 6 INC 001
50 60 ? 4 2 6 INC 001
60 ? ? 4 2 6 JR 011

[0030] When the first instruction (listed in line 10) is executed, the first decoder indicates that a program counter increment (INC) in the execution of the program has occurred and shows an output of “001”. As the program progresses from the instruction on line 10 through the instruction on line 40, the first decoder continues to indicate that a program counter increment (INC) in the execution of the program has occurred and continues to show an output of “001”. When the instruction on line 50 is executed, the first decoder indicates that a program counter branch taken (PBT) has occurred and shows an output of “101”. As seen in Tables 2 and 3, the program branches to line 70 because the conditional expression of line 50 is true based on the variable D=7. Upon execution of line 70, the first-decoder indicates that a program counter increment (INC) in the execution of the program has occurred and shows an output of “001”. When the instruction on line 80 is executed, the first decoder indicates that an immediate jump (JI) has occurred and shows an output of “010”. As seen in Tables 2 and 3, the program jumps to line 30. When the instructions on lines 30 and 40 are executed, the first decoder indicates that a program counter increment (INC) in the execution of the program has occurred and shows an output of “001”. When line 50 is executed (now for the second time) the first decoder indicates that a program counter increment (INC) in the execution of the program has occurred and shows an output of “001” because the condition (D=7) for the jump in line 50 is no longer valid. Line 60 is now executed and a jump to a location stored in a register occurs. The first decoder therefore indicates a jump to register (JR) by showing an output of “011”.

[0031] Referring once again to FIG. 1, according to another aspect of the invention, each cause register 26 a, 26 b, 26 c is coupled to the data input D of an event history buffer 14 and a second decoder 32 a, 32 b, 32 c is coupled to each cause register and to the enable input E of the history buffer 14. The clock 16 provides the common clock signal to the clock input C of the history buffer 14 via the clock bus 17, and a timestamp register 19 is also coupled to the clock bus 17. The contents of the history buffer 14 are made available off chip by three pins for the data, clock, and enable (D, C, E) of the history buffer 14. According to this aspect of the invention, when certain conditions are detected by one of the second decoders 32, the history buffer is enabled via the appropriate decoder, and information from the cause register, the timestamp register, and the program counter is stored in the history buffer. More particularly, the second decoder 32 enables the history buffer whenever the first decoder contains code which indicates that the processor is processing an instruction to jump to a location stored in a register, whenever the first decoder contains code indicating an exception was encountered, and whenever the first decoder contains code indicating a change in state of an interrupt line.

[0032] According to a presently preferred embodiment, when the history buffer is enabled, it captures forty-four bits of information from the cause register or program counter, and the timestamp register. The forty-four bits of information are preferably organized as illustrated in Table 4 below.

TABLE 4
43 42 41 40-18 17 16 15-0
Mode Proc Cause/PC HOVRF TR Time Stamp

[0033] The first bit, bit location 43, is a mode identifier indicating whether the entry being stored has program counter information or cause register information. A two bit processor identification number is stored in binary form at bit locations 42, 41. This number is used to indicate which processor's information is being stored (in the case of a multiprocessor system). The next twenty-three bits at bit locations 40 through 18 are used to store cause register information or program counter information depending on the mode as explained above. If program counter information is being stored, the contents of the program counter are stored at bit locations 40 through 18. If cause register information is being stored, bit location 40 is used to indicate whether the exception occurred while the processor was executing an instruction in the branch delay slot. (This applies to pipelined processors such as RISC processors.) Bit locations 39 through 35 are used to store processor related exception conditions. Bit locations 34 through 18 are used to store an indication of all pending interrupts (external, software, co-processor. The HOVRF field at bit location 17 is used to indicate whether the internal event history buffer has overflowed. The TR bit 16 is used to indicate a timestamp rollover and bits 15 through 0 are used to store a sixteen bit timestamp. According to the presently preferred embodiment, the forty-four bits captured in the history buffer 14 are serially output on data pin D over forty-four clock cycles (bit serial output).

[0034] As mentioned above, the event history buffer records information when an event (either an unmasked exception or a PC jump register instruction) has occurred. According to a presently preferred embodiment, this requires an additional mask register per cause register and a free running timestamp counter. The event masks are provided by a JTAG test register load instruction in the static debug interface. When the cause register bits corresponding to an exception are unmasked or a PC jump register instruction is encountered, an entry is made in the history buffer.

[0035] Those skilled in the art will appreciate that the outputs of the first decoder 28 and the contents of the history buffer 14 provide a relatively complete indication of each processor's execution sequence in real time, particularly when viewed in light of the actual program code which is being executed. Therefore, according to the invention, a debugging system may be coupled to the first decoders and history buffer as illustrated in FIG. 2.

[0036] Turning now to FIG. 2, the outputs 30 a, 30 b, 30 c of the first decoders and the D, C, E terminals of the history buffer are coupled to a debugging computer 44 which preferably has a copy of the program code stored therein. The three-bit outputs 30 a, 30 b, 30 c of the first decoders and the D, C, E terminals of the history buffer are preferably coupled to an interface buffer 40 which is coupled by a serial, parallel, or network connection 42 to the debugging computer 44. The interface buffer 40 is a rate decoupling buffer. In a present embodiment of the invention, the debugger interface is provided on a 100 MHz three processor system. In that system, the data rate for reading the event history buffer is approximately 1 gigabit/sec. Current PCs cannot keep up with that data rate. Therefore, the buffer 40 is provided to prevent the loss of event history data.

[0037] As the program is running on the ASIC 10, the debugging computer 44 collects information from the first decoders and the history buffer. The information collected by the computer 44 is associated with each line of code being executed by the ASIC by stepping through the copy of the code which is stored in the computer 44. When a bug is encountered, the complete history of instruction execution leading up to the failure can be reviewed with the computer 44. The debugging system is non-invasive and permits debugging of programs operating in real time.

[0038] There have been described and illustrated herein embodiments of a real time debugger interface for embedded systems. While particular embodiments of the invention have been described, it is not intended that the invention be limited thereto, as it is intended that the invention be as broad in scope as the art will allow and that the specification be read likewise. Thus, while particular encoding schemes have been disclosed with reference to the first decoder output and the history buffer contents, it will be appreciated that other encoding schemes could be utilized provided that they achieve substantially the same results as described herein. Also, while the invention has been illustrated with reference to a three-processor ASIC chip, it will be recognized that the invention may be applied in other types of chips having greater or fewer processors. Moreover, while particular configurations have been disclosed in reference to the indications provided by the first decoders, it will be appreciated that other configurations could be used as well, provided that they achieve substantially the same results as described herein. It will therefore be appreciated by those skilled in the art that yet other modifications could be made to the provided invention without deviating from its spirit and scope as so claimed.

Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US7225098Nov 21, 2002May 29, 2007Stmicroelectronics S.A.Monitoring device with optimized buffer
US7673121Nov 14, 2002Mar 2, 2010Stmicroelectronics S.A.Circuit for monitoring a microprocessor and analysis tool and inputs/outputs thereof
US8533678 *Jul 13, 2007Sep 10, 2013Digi International Inc.Embedded device program debug control
US8566501 *Sep 14, 2011Oct 22, 2013Etron Technology, Inc.Circuit for simultaneously analyzing performance and bugs and method thereof
US20090019270 *Jul 13, 2007Jan 15, 2009Digi International Inc.Embedded device program debug control
US20120079161 *Sep 14, 2011Mar 29, 2012Hsuan-Ching ChaoCircuit for simultaneously analyzing performance and bugs and method thereof
EP1783649A1 *Oct 10, 2005May 9, 2007Nagracard S.A.Secure microprocessor with jump verification
WO2004046929A1 *Nov 14, 2002Jun 3, 2004Jehan-Philippe BarbieroTransmission of a digital message between a microprocessor monitoring circuit and an analysis tool
WO2007042478A1 *Oct 6, 2006Apr 19, 2007Nagracard SaSecure microprocessor with jump verification
Classifications
U.S. Classification712/227, 714/E11.212, 714/E11.214
International ClassificationG06F11/28, G06F11/36
Cooperative ClassificationG06F11/3636, G06F11/3656
European ClassificationG06F11/36B7H, G06F11/36B5
Legal Events
DateCodeEventDescription
Jul 28, 2008ASAssignment
Owner name: TR TECHNOLOGIES FOUNDATION LLC, DELAWARE
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:TRANSWITCH CORPORATION;REEL/FRAME:021291/0963
Effective date: 20071226
Jul 30, 2001ASAssignment
Owner name: TRANSWITCH CORPORATION, CONNECTICUT
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ROY, SUBHASH C.;HEMBROOK, PAUL;PARRELLA, EUGENE L.;AND OTHERS;REEL/FRAME:012041/0817;SIGNING DATES FROM 20010724 TO 20010726