In-place resynthesis and remapping techniques for soft error mitigation in FPGA
In-place resynthesis for static memory (SRAM) based Field Programmable Gate Arrays (FPGAs) toward reducing sensitivity to single event upsets (SEUs). Resynthesis and remapping are described which have a low overheard and improve FPGA designs without the need of rerouting LUTs of the FPGA. These methods include in-place reconfiguration (IPR), in-place X-filling (IPF), and in-place inversion (IPV), which reconfigure LUT functions only, and can be applied to any FPGA architecture. In addition, for FPGAs with a decomposable LUT architecture (e.g., dual-output LUTs) an in-place decomposition (IPD) method is described for remapping a LUT function into multiple smaller functions leveraging the unused outputs of the LUT, and making use of built-in hard macros in programmable-logic blocks (PLBs) such as carry chain or adder. Methods are applied in-place to mapped circuits before or after routing without affecting placement, routing, and design closure.
1. An apparatus for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit for implementation on the FPGA circuit and routing of circuits through a synthesis process which arrives at a physical design;
performing a circuit analysis on said logic circuit;
performing in-place iterations of reconfiguring, don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
wherein said programming executable on said computer is configured to perform said in-place iterations by performing single event upset (SEU) fault analysis to obtain weight values for routing configuration memory (CRAM) bits, followed by performing in-place logic inversion which inverts functions of driving logic in response to reassigning look-up table (LUT) polarities, followed by adjusting all of the truth tables of its fanout LUTs and driven LUTs to preserve functionality, whereby total weight of configuration memory (CRAM) bits on routing multiplexers is minimized.
2. The apparatus as recited in claim 1 , wherein said programming executable on said computer is configured to perform said in-place iterations as multiple iterations of an in-place resynthesis (IPR) process in which a group of look-up-tables (LUTs) are selected as a sub-network and identical configuration bits corresponding to complementary inputs of said group of LUTs are maximized, whereby faults seen at a pair of complementary inputs have a lower probability of propagation toward increasing overall reliability of the circuit.
3. The apparatus as recited in claim 1 :
wherein said programming executable on said computer is configured to perform said in-place iterations as iterations of an in-place X-filling (IPF) process to a convergence in which states are determined for satisfiability don't cares (SDCs); and
wherein said programming executable on said computer is configured to perform said in-place X-filling (IPF) process in response to performing single event upset (SEU) fault analysis, followed by assigning satisfiability don't care (SDC) bits in response to criticality of LUT configuration bits by assigning said SDC bits to a logic value which maximizes correct logic output probability from a LUT toward minimizing fault impact from SEUs.
4. The apparatus as recited in claim 1 , wherein said FPGA utilizes unidirectional routing architecture having programmable interconnect points (PIPs).
5. The apparatus as recited in claim 4 , wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM).
6. An apparatus for increasing fault tolerance of a synthesized FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit for implementation on the FPGA circuit and mapping to a plurality of look-up tables (LUTs);
performing placement and routing;
performing single event upset (SEU) fault analysis to obtain weight values for each routing configuration memory (CRAM) bits; and
performing in-place logic inversion which inverts functions of driving logic in response to reassigning look-up table (LUT) polarities, followed by adjusting all of the truth tables of its fanout LUTs and driven LUTs to preserve functionality, whereby total weight of all configuration memory (CRAM) bits on routing multiplexers is minimized.
7. The apparatus as recited in claim 6 , wherein said programming executable on said computer is configured to perform driven logic adjustments to modify the logic functions of fan-out LUTs to preserve functionality affected by polarity inversion.
8. The apparatus as recited in claim 6 , wherein said programming executable on said computer is configured to perform said single event upset (SEU) analysis in response to a logic block SEU analysis, connection box SEU analysis, logic routing SEU analysis, and a switch box SEU analysis.
9. The apparatus as recited in claim 6 , wherein said FPGA utilizes unidirectional routing architecture having programmable interconnect points (PIPs).
10. The apparatus as recited in claim 9 :
wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM); and
wherein said bridging or driver errors arise in response to changing the driver of a net or bridging two nets with different drivers together, resulting in signal discrepancy on nets involved in driver switching or bridging due to the SEU.
11. An apparatus for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit for implementation on the FPGA circuit and routing of circuits through a synthesis process which arrives at a physical design;
performing a circuit analysis on said logic circuit;
performing in-place iterations of don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
wherein said programming executable on said computer is configured to perform said in-place iterations by performing single event upset (SEU) fault analysis to obtain weight values for routing configuration memory (CRAM) bits, followed by performing in-place logic inversion which inverts functions of driving logic in response to reassigning look-up table (LUT) polarities, followed by adjusting all of the truth tables of its fanout LUTs and driven LUTs to preserve functionality, whereby total weight of configuration memory (CRAM) bits on routing multiplexers is minimized.
12. A method for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit within an FPGA circuit design application executing on a computer which routes FPGA circuits through a synthesis process and arrives at a physical design for said logic circuit;
performing a circuit analysis on said logic circuit;
performing in-place iterations of don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
performing driven logic adjustments to modify the logic functions of fan-out LUTs to preserve functionality affected by polarity inversion.
13. The method as recited in claim 12 , further configured for said FPGA utilizing a unidirectional routing architecture having programmable interconnect points (PIPs).
14. The method as recited in claim 13 , wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM).
15. An apparatus for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit for implementation on the FPGA circuit and routing of circuits through a synthesis process which arrives at a physical design;
performing a circuit analysis on said logic circuit;
performing in-place iterations of reconfiguring, don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
wherein said programming executable on said computer is configured to perform driven logic adjustments to modify the logic functions of fan-out LUTs to preserve functionality affected by polarity inversion.
16. The apparatus as recited in claim 15 , wherein said FPGA utilizes unidirectional routing architecture having programmable interconnect points (PIPs).
17. The apparatus as recited in claim 16 , wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM).
18. An apparatus for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit for implementation on the FPGA circuit and routing of circuits through a synthesis process which arrives at a physical design;
performing a circuit analysis on said logic circuit;
performing in-place iterations of reconfiguring, don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
wherein said programming executable on said computer is configured to perform said single event upset (SEU) analysis in response to a logic block SEU analysis, connection box SEU analysis, logic routing SEU analysis, and a switch box SEU analysis.
19. The apparatus as recited in claim 18 , wherein said FPGA utilizes unidirectional routing architecture having programmable interconnect points (PIPs).
20. The apparatus as recited in claim 19 , wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM).
21. An apparatus for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit for implementation on the FPGA circuit and routing of circuits through a synthesis process which arrives at a physical design;
performing a circuit analysis on said logic circuit;
performing in-place iterations of reconfiguring, don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream;
wherein said FPGA utilizes unidirectional routing architecture having programmable interconnect points (PIPs);
wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM); and
wherein said bridging or driver errors arise in response to changing the driver of a net or bridging two nets with different drivers together, resulting in signal discrepancy on nets involved in driver switching or bridging due to the SEU.
22. A method for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit within an FPGA circuit design application executing on a computer which routes FPGA circuits through a synthesis process and arrives at a physical design for said logic circuit;
performing a circuit analysis on said logic circuit;
performing in-place iterations of don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
performing said single event upset (SEU) analysis in response to a logic block SEU analysis, connection box SEU analysis, logic routing SEU analysis, and a switch box SEU analysis.
23. A method for increasing fault tolerance of an FPGA circuit, comprising:
a computer configured for designing an FPGA circuit; and
programming executable on said computer for:
describing a logic circuit within an FPGA circuit design application executing on a computer which routes FPGA circuits through a synthesis process and arrives at a physical design for said logic circuit;
performing a circuit analysis on said logic circuit;
performing in-place iterations of don't care X filling, and/or inversion of look-up table (LUT) bits toward increasing overall reliability of said logic circuit; and
updating said FPGA circuit in response to said in-place iterations;
wherein said in-place iterations are performed after placement and routing to preserve physical design while optimizing the logic circuit to mask faults originating upstream; and
wherein said FPGA is utilizing a unidirectional routing architecture having programmable interconnect points (PIPs);
wherein an FPGA utilizing unidirectional routing is subject to bridging or driver errors in response to a single-event upset (SEU) occurring on a routing bit of a configuration memory (CRAM); and
wherein said bridging or driver errors arise in response to changing the driver of a net or bridging two nets with different drivers together, resulting in signal discrepancy on nets involved in driver switching or bridging due to the SEU.