A newer version of this document is available. Customers should click here to go to the newest version.
4.4.1. DCLSM Blind Window Control Register - DCLSM_BWCR
4.4.2. All Alarms’ Prior Alarms’ Fault Injection Register - ERRCTRL_ALL_ALARMS_PRIOR_AFI
4.4.3. INTREQ Configuration Register - ERRCTRL_INTREQ_CONF
4.4.4. Timeout Deadline and Status Register - ERRCTRL_TIMEOUT
4.4.5. Timeout Acknowledgment Register - ERRCTRL_TIMEOUT_ACK
4.4.6. Enable Key fRSmartComp Control Register - ERRCTRL_ENABLE_KEY
4.4.7. Root Fault Injection Control register - ERRCTRL_ROOT_INJ
4.4.8. Alarm Fault Injection Control register - ERRCTRL_ALARM_INJ
4.4.9. Event Mask Configuration register - ERRCTRL_MASKA and ERRCTRL_MASKB
4.4.10. Alarm Routing Configuration register - ERRCTRL_ROUTA and ERRCTRL_ROUTB
4.4.11. Error Controller PGO LOG Reset Control register - ERRCTRL_PGOLOGRST
4.4.12. PGO0 and PGO4 Configuration registers - ERRCTRL_PGO0 and ERRCTRL_PGO4
4.4.13. FN_MODEIN Control Register - ERRCTRL_FNMODEIN
4.4.14. FN_MODEOUT register - ERRCTRL_FNMODEOUT
4.4.15. All Alarms After Fault Injection - ERRCTRL_FNGIALARMS
4.4.16. Error Controller Context Register - ERRCTRL_FNGICTXT4
4.4.17. CMP Mismatch CONTEXT Registers - ERRCTRL_FNGICMPCTXT0 … ERRCTRL_FNGICMPCTXT3
4.4.18. STATISTICS registers: ERRCTRL_FNGISTAT0 and ERRCTRL_FNGISTAT4
4.4.19. State register - ERRCTRL_FNPERIPHGI4
2.1. Introduction
In the standard DCLS scheme, the outputs coming from the processors are compared by a comparator on a cycle-by-cycle basis. The two processors execute the same program flow. If the comparator detects a difference, it raises the alarm, signaling that a fault has been detected in one of the two processors without the opportunity to know in which of the two the fault has occurred.
Only one of the two processors (the Host) drives the system bus; the addresses and control information generated by the other processor (the Agent) does not affect the system.
Figure 3. Standard DCLS Architecture
Based on the standard DCLS architecture, the fRSmartComp is enhanced to handle an event of false positive faults along with timeout events. The following figure shows the two CPUs working in parallel with a timing skew between the two cores, implemented by means of flip-flop barriers. The timing skew provides time diversity, and hence a more robust implementation against common cause failures.
Figure 4. fRSmartComp DCLS Architecture with Time Diversity