摘要:
An instruction processing pipeline having error detection and error recovery circuitry associated with one or more of the pipeline stages. If an error is detected within a signal value within that pipeline stage, then it can be repaired. Part of the error recovery may be to flush upstream program instructions from the instruction pipeline. When multi-threading, only those instructions from a thread including an instruction which has been lost as a consequence of the error recovery need be flushed from the instruction pipeline. The instruction pipeline may additionally/alternatively be provided with more than one main storage element associated with each signal value with these main storage elements used in an alternating fashion such that if a signal value has been erroneously captured and needs to be repaired, there is still available a main storage element to properly capture the signal value corresponding to the following program instruction.
摘要:
An integrated circuit 2 is provided with a plurality of pipeline stages 10. These pipeline stages 10 have speculative processing control circuitry 12 which permits speculative processing in downstream pipeline stages and triggers a first error recovery operation (partial pipeline flushing) if such speculative processing is determined to be based upon an error. The pipeline stage 10 further includes speculative error detecting circuitry 14 which generates a prediction nc regarding whether or not the processing circuitry 18 will produce an error. This prediction is used to trigger a second error recovery operation (partial pipeline stall). This second error recovery operation has a lower performance penalty than the first error recovery operation.
摘要:
An instruction processing pipeline 6 is provided. This has error detection and error recovery circuitry 20 associated with one or more of the pipeline stages. If an error is detected within a signal value within that pipeline stage, then it can be repaired. Part of the error recovery may be to flush upstream program instructions from the instruction pipeline 6. When multi-threading, only those instructions from a thread including an instruction which has been lost as a consequence of the error recovery need to be flushed from the instruction pipeline 6. Instruction can also be selected for flushing in dependence upon characteristics such as privileged level, number of dependent instructions etc. The instruction pipeline may additionally/alternatively be provided with more than one main storage element 26, 28 associated with each signal value with these main storage elements 26, 28 being used in an alternating fashion such that if a signal value has been erroneously captured and needs to be repaired, there is still available a main storage element 26, 28 to properly capture the signal value corresponding to the following program instruction. In this way flushes can be avoided.
摘要:
A data processing apparatus and method are provided that use monitoring circuitry to control operating parameters of the data processing apparatus. The data processing apparatus has functional circuitry for performing data processing, the functional circuitry including error correction circuitry configured to detect errors in operation of the functional circuitry and to repair those errors in operation. Tuneable monitoring circuitry monitors a characteristic indicative of changes in signal propagation delay within the functional circuitry and produces a control signal dependent on the monitored characteristic. In a continuous tuning mode operation, the tuneable monitoring circuitry modifies the dependency between the monitored characteristic and the control signal in dependence upon certain characteristics of the errors detected by the error correction circuitry. An operating parameter controller is then arranged, in the continuous mode of operation, to control one or more performance controlling operating parameters of the data processing apparatus in dependence upon the control signal. This enables efficient and robust control of those operating parameters in response to changes in environmental conditions.
摘要:
A data processing system 2 is used to perform processing operations to generate a result value. The processing circuitry which generates the result value has an error resistant portion 32 and an error prone portion 30. The probability of an error in operation of the error prone portion for a given set of operating parameters (clk, V) is greater than the probability of an error for that same set of operating parameters within the error resistant portion. Error detection circuitry 38 detects any errors arising in the error prone portion. Parameter control circuitry 40 responds to detected errors to adjust the set of operating parameters to maintain a non-zero error rate in the errors detected by the error detection circuitry. Errors within the one or more bits generated by the error prone portion are not corrected as the apparatus is tolerant to errors occurring within such bit values of the result value.
摘要:
A D-type flip-flop 2 includes tristate inverter circuitry 4, 6 passing a processing signal through to storage circuitry 8 from where the processing signal passes via a transmission gate 10 to slave storage circuitry 12. A transition detector 16 is coupled to the input node nm of the storage circuitry 8 and serves to generate an error signal if a transition is detected upon that input node during an error detecting period. Other forms of this technique may provide clock gating circuitry.
摘要:
A processor 2 is responsive to a stream of program instructions to issue program instructions under control of scheduling circuitry 6 to respective execution units 24 for execution. The execution units 24 can include error detecting circuitry 32 for detecting a change in an output signal which occurs after the output signal has latched and during an error detecting period following the latching of the output signal. The scheduling circuitry 6 is arranged so as to suppress issue of program instructions to an execution unit 24 having such error detecting circuitry 32 on consecutive processing cycles.
摘要:
An apparatus and method are provided for detecting an approaching error condition within a data processing apparatus. The data processing apparatus includes a second sequential storage structure which is arranged to latch the output signal generated by combinatorial circuitry dependent on a second clock signal. The second sequential storage structure has a main storage element to latch a value of the output signal for provision to subsequent combinatorial circuitry, and transition detection circuitry for detecting a change of the value of the output signal latched by the main storage element during a predetermined timing window, said change indicating an approaching error condition whilst the value stored in the main storage element is still correct. The second sequential storage structure can be operated in either a first mode of operation or a second mode of operation. In the first mode of operation, the predetermined timing window is a timing window ahead of a time at which the main storage element latches said value of the output signal, to thereby enable an approaching setup timing error due to a propagation delay within the combinatorial circuitry to be detected. In the second mode of operation, the predetermined timing window is a timing window after the time at which the main storage element latches said value of the output signal such that an approaching hold timing error due to an increase in skew between the first and second clock signals is detected. Such a technique provides a simple and efficient mechanism for detecting a variety of approaching error conditions whilst the second sequence storage structure continues to operate correctly.
摘要:
An integrated circuit includes processing pipeline stages formed of an input register, processing circuit and an output register. The output register employs speculative sampling and uses a subsequent speculation period during which any change in its input is detected and used to indicate a speculation error. In order to reduce the chances of a race condition giving rise to a false positive detection of a speculation error due to a too rapid signal propagation through the processing circuitry a transparent latch is disposed at the approximate midpoint, measured in terms of propagation delay, within the processing circuitry. This transparent latch is non-transmissive during the speculation period of the output register so as to prevent any new signal propagating from the input register during the speculation period from reaching the output register.
摘要:
An integrated circuit is provided with error detection circuitry and error repair circuitry. Error tolerance circuitry is responsive to a control parameter to selectively disable the error repair circuitry. The control parameter is dependent on the processing performed within the circuit. For example, the control parameter may be generated in dependence upon the program instruction being executed, the output signal value which is in error, the previous behavior of the circuit or in other ways.