430 likes | 594 Views
Functional Test Program Generation for Microprocessors. Prabhat Mishra. Outline. Introduction Test Program Generation: Taxonomy Test Program Generation Techniques Past Approaches Present Trends Future Directions Conclusion References. Introduction.
E N D
Functional Test Program Generation for Microprocessors Prabhat Mishra
Outline • Introduction • Test Program Generation: Taxonomy • Test Program Generation Techniques • Past Approaches • Present Trends • Future Directions • Conclusion • References
Introduction • Shrinking time-to-market, short product life .. • Need to reduce design cycle time for embedded microprocessors. • Functional verification is one of the major component of this design cycle time. • Existing validation approaches • Simulation based approaches using random tests • Formal techniques to verify certain properties
Test Generation: Taxonomy Test Generation Manufacturing Test Functional Test Verifies Functional Correctness Verifies stuck-at, bridging, delay.. faults
Test Generation: Taxonomy Test Generation Manufacturing Test Functional Test Fault Oriented Fault Independent Random Combined (RAPS, SMART) (Critical Path)
Test Generation: Taxonomy Test Generation Manufacturing Test Functional Test Without Fault Implicit (universal) Fault Functional Fault Fault Oriented Fault Independent Random Combined
Test Generation: Taxonomy Test Generation Manufacturing Test Functional Test Without Fault Implicit (universal) Fault Functional Fault Fault Oriented Fault Independent Random Combined BDD Heuristic
Test Generation: Taxonomy Test Generation Manufacturing Test Functional Test Without Fault Implicit (universal) Fault Functional Fault Fault Oriented Fault Independent Random Combined PseudoExhaustive Exhaustive BDD Heuristic
Test Generation: Taxonomy Test Generation Manufacturing Test Functional Test Without Fault Implicit (universal) Fault Functional Fault Fault Oriented Fault Independent Random Combined PseudoExhaustive Exhaustive BDD Heuristic
Test Generation: Taxonomy TestGeneration Manufacturing Test FunctionalTest Without Fault Implicit (universal) Fault Functional Fault Fault Oriented Fault Independent Random Combined PseudoExhaustive Exhaustive BDD Heuristic
Functional Test Program TestProgGen MOV R1, 011 MOV R2, 010 ADD R3, R1, R2 # R3 == 101 R3 == 101 ? Verifies the functionality of the processor using assembly programs
Test Program Generation Techniques • Past Approaches • Functional Fault Model based Test Generation • Present Trends • Micro-Architecture Coverage driven Test Gen. • Instruction-Level Self Test • Specification based Test Generation • Future Directions • Software based Self Test • Specification driven Validation
Test Program Generation Techniques • Past Approaches • Functional Fault Model based Test Generation • Nair et. al., IEEE Computers 1978 • Thatte et al. IEEE Computers,1980 • Brahme et al. IEEE Computers,1984 • Lin et al., DAC 1988 • Kannah et al., ATS 2000 • Present Trends • Micro-Architecture Coverage driven Test Gen. • Instruction-Level Self Test • Specification based Test Generation • Future Directions • Software based Self Test • Specification driven Validation
Testing Random-Access Memories • Presented a fault model which views faults in memories at a functional level instead of at a basic gate level. • Memory Fault Model • Memory cell array • One or more cells are stuck at 0 or 1. • One or more pairs of cells are coupled. • Decoder • Does not access the addressed cell • Access multiple cells, including the addressed cell • Read/Write Logic • Some output lines of the sense amplifier logic or write driver logic may be stuck-at-0 or stuck-at-1 Nair et. al., IEEE Computers 1978
Testing of Microprocessors • Graph model of the architecture • Node = Register groups • Edge = Instructions • Functional fault models • Register decoding • Instruction decoding • Data Storage • Data Transfer • Data Manipulation • Test generation procedure • Register decoding Thatte et. al., IEEE Computers 1980
Graph Model of the Microprocessor Registers R1: accumulator R2: general purpose register R3: scratch-pad register In I2 I1 I3 I7 Instructions I1: Load R1 I2: Load R2 I3: Add R1 = R1 + R2 I4: Mov R2 = R1 I5: Store R1 I6: Store R2 I7: Mov R3 = R1 I8: Mov R1 = R3 I4 R2 R1 R3 I3 I8 I6 I5 Out
Graph Model of the Microprocessor In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out Registers R1: accumulator R2: general purpose register R3: scratch-pad register Instructions I1: Load R1 I2: Load R2 I3: Add R1 = R1 + R2 I4: Mov R2 = R1 I5: Store R1 I6: Store R2 I7: Mov R3 = R1 I8: Mov R1 = R3 Write (R3) = {I1, I7} Read(R3) = {I8, I5}
Microprocessor Fault Models • Register Decoding • fd(Ri) =Ri • Write (Ri) and Read (Ri) • Instruction Decoding • No instruction executed – f (Ij / ) • Different instruction is executed – f(Ij / Ik ) • F(Ij / Ij + Ik ) • Data Storage • Any cell of a register can be stuck-at-0 or stuck-at-1 • Data Transfer • A line in a transfer path can be stuck at 0 or 1 • Two lines in a transfer path can be coupled • Data Manipulation
Test Generation for Register Decoding • Initialize Q with all registers so that Ri lies ahead of Rj iff l(Ri) <= l(Rj) • A = register at the front of the Q • Repeat • Write each register Ri of set A with data ONE and register Rj at the front of Q with data ZERO • Read each register Ri of A • Read Rj • A = A U Rj • Q = Q - Rj • Until Q is empty f(Ri) , for every Ri A f(Ri) f(Rj) = for every Ri, Rj A f(Ri) = 1
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3}
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3} R1 ZERO R1 ZERO ? LOAD R1, ZERO STORE R1 # Out == ZERO ?
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3} A = {R1} Q = {R2, R3} R1 ZERO R1 ZERO ?
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3} A = {R1} Q = {R2, R3} R1 ZERO R1 ZERO ? R1 ONE, R2 ZERO R1 ONE ?, R2 ZERO ?
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3} A = {R1} Q = {R2, R3} A = {R1, R2} Q = {R3} R1 ZERO R1 ZERO ? R1 ONE, R2 ZERO R1 ONE ?, R2 ZERO ?
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3} A = {R1} Q = {R2, R3} A = {R1, R2} Q = {R3} R1 ZERO R1 ZERO ? R1 ONE, R2 ZERO R1 ONE ?, R2 ZERO ? (R1, R2) ONE, R3 ZERO (R1, R2) ONE ?, R2 ZERO ?
Register Decoding Algorithm In I2 I1 I3 I7 I4 R2 R1 R3 I3 I8 I6 I5 Out A = {} Q = {R1, R2, R3} A = {R1} Q = {R2, R3} A = {R1, R2} Q = {R3} A = {R1, R2, R3} Q = {} R1 ZERO R1 ZERO ? R1 ONE, R2 ZERO R1 ONE ?, R2 ZERO ? (R1, R2) ONE, R3 ZERO (R1, R2) ONE ?, R2 ZERO ?
Extended Fault Model • Representation for Instructions • An instruction is composed of a sequence of microinstructions. • Each microinstruction is made up of a set of microorders which are executed in parallel • SWAP, CLR, NOT, MOV, EXG, ADD, MUL, DIV • Simple ADD has 4 microorders • Fault model for the Instruction Sequencing • One or more microorders can be inactive • Microorders which are normally inactive become active • F(I) = I + + - - Brahme et. al., IEEE Computers 1984
Further Improvements • Chen-Shang Lin and Hong-Fa Ho, DAC 88 • Signalflow model of the processor • Functional fault models are derived from Turing machine model • Developed O-algorithm to eliminate redundant tests and obtain improved fault coverage • Rajesh Kannah, C.P. Ravikumar, ATS 2000 • Reduce the test application time. • Use structural information to eliminate certain functional tests using the notion of fault-grading.
Test Program Generation Techniques • Past Approaches • Functional Fault Model based Test Generation • Present Trends • Micro-Architecture Coverage driven Test Generation • Iwashita et al., ICCAD 1994 • Ho et al., ISCA 1995 • Ur et al., DAC 1999 • Campenhout et al., DAC 1999 • Specification driven Test Generation • Geist et al., DAC 1999 (Specman Elite from Verisity) • Mishra et al., HLDVT 2002 • Instruction-Level Self Test • Lai et al., DAC 2001 • Chen et al., DAC 2003 • Future Directions • Software based Self Test • Specification driven Validation
Test Generation for Pipelined Processors Iwashita et. al., ICCAD 1994
FSM Traversal based Test Generation Ho et. al., ISCA 1995
Micro Architecture Coverage Driven … Ur et. al., DAC 1999
High-Level Test Generation Campenhout et. al., DAC 1999 DPTRACE: Path Selection DPRELAX: Value Selection CTRLJUST: Signal Justification
Test Program Generation Techniques • Past Approaches • Functional Fault Model based Test Generation • Present Trends • Micro-Architecture Coverage driven Test Generation • Iwashita et al., ICCAD 1994 • Ho et al., ISCA 1995 • Ur et al., DAC 1999 • Campenhout et al., DAC 1999 • Instruction-Level Self Test • Lai et al., DAC 2001 • Chen et al., DAC 2003 • Specification driven Test Generation • Geist et al., DAC 1999 (Specman Elite from Verisity) • Mishra et al., HLDVT 2002 • Future Directions • Software based Self Test • Specification driven Validation
Specification driven Test Gen. using Model Checker Architecture Specification ADL Specification Properties (SMV) Processor Model Simulator Generation Not Enough Properties SMV Counterexamples Coverage Report Simulator Automatic Test Programs Manual Mishra et. al., HLDVT 2002
Test Program Generation • Specification of Properties • Stall Decode Unit of the DLX architecture stall: assert G(ID._stall = 0); • Test Program • The Decode unit (ID) will be stalled in cycle 4. Fetch CycleOpcode DestSrc1Src2 1 NOP 2 ADD R3 R1 R2 3 ADD R4 R3 R2
Test Program Generation Techniques • Past Approaches • Functional Fault Model based Test Generation • Present Trends • Micro-Architecture Coverage driven Test Generation • Iwashita et al., ICCAD 1994 • Ho et al., ISCA 1995 • Ur et al., DAC 1999 • Campenhout et al., DAC 1999 • Specification driven Test Generation • Geist et al., DAC 1999 (Specman Elite from Verisity) • Mishra et al., HLDVT 2002 • Instruction-Level Self Test • Lai et al., DAC 2001 • Chen et al., DAC 2003 • Future Directions • Software based Self Test • Specification driven Validation
Instruction-Level DFT Lai et. al., DAC 2001
Software-Based Self-Test Chen et. al., DAC 2003
Conclusion • The boundary is blurring ….. • Confluence of Functional Verification and Manufacturing testing • Software-Based Self-Test • Top-down and Bottom-up Validation • Specification driven Test Generation • Micro-Architecture Coverage driven Techniques • Combined Simulation and Formal Techniques
References • Test Generation for Microprocessors: S. Thatte et al., IEEE Computers, June 1980 • Functional Testing of Microprocessors: D. Brahme et al., IEEE Computers, 1984 • Automatic Test Program Generation for Pipelined Processors: H. Iwashita et al., ICCAD 1994 • Architecture Validation for Processors: R. Ho et al., ISCA 1995 • Automatic Functional Test Program Generation for Microprocessors: C. Lin et al., DAC 1998 • Micro Architecture Coverage Directed Generation of Test Programs: S. Ur et al., DAC 1999 • High-Level Test Generation for design Verification of Pipelined Microprocessors: D. Campenhout et al., DAC 1999 • Functional Testing of Microprocessors with Graded Fault Coverage: R. Kannah et al., ATS 2000 • Instruction Level DFT for Testing Processor and IP Cores in System-on-a-Chip: W. Lai et al., DAC 2001 • Automatic Functional Test Program Generation for Pipelined Processors using Model Checking, by P. Mishra et al., HLDVT 2002. • A Scalable Software-Based Self-Test Methodology for Programmable Processors: L. Chen et al., DAC 2003
Thank you ! Thank you !