Low-density parity-check decoding with desaturation

1. A decoder, comprising:

a saturation monitor configured to:

generate a global de-saturation control signal that is determined based at least in part on a saturation threshold and a degree of saturation in a low-density parity-check (LDPC) decoding system that uses a fixed-point number representation;

plurality of variable nodes that is configured to:

in the event the global de-saturation control signal is a first value, output a more aggressively attenuated message to a plurality of check nodes compared to when the global de-saturation control signal is a second value; and

in the event the global de-saturation control signal is the second value, output a less aggressively attenuated message to the plurality of check nodes compared to when the global de-saturation control signal is the first value; and

the plurality of check nodes.

2. The decoder recited in claim 1 , wherein the degree of saturation is based at least in part on a number of messages, passed between at least one of the plurality of check nodes and at least one of the plurality of variable nodes, that are saturated to a maximum fixed-precision magnitude.

3. The decoder recited in claim 1 , wherein the degree of saturation is based at least in part on a number of saturated nodes in the LDPC decoding system.

4. The decoder recited in claim 1 , wherein:

the degree of saturation is based at least in part on a number of saturated nodes in the LDPC decoding system; and

a node is declared saturated if an associated minimum attenuated message after attenuation reaches a maximum fixed-precision magnitude.

5. The decoder recited in claim 1 , wherein outputting the more aggressively attenuated message to the plurality of check nodes includes:

outputting a first more aggressively attenuated message that is based at least in part on a first-most minimum variable node message; and

outputting a second more aggressively attenuated message that is based at least in part on a second-most minimum variable node message.

6. The decoder recited in claim 1 , wherein outputting the more aggressively attenuated message to the plurality of check nodes includes:

outputting a first more aggressively attenuated message (L min1 (j) ) that is based at least in part on a first-most minimum variable node message (l min1 (j) ), including by using L min1 (j) =└α ds ·l min1 (j) +β ds ┘, wherein α ds is associated with an attenuation factor and β ds is associated with an attenuation rounding; and

outputting a second more aggressively attenuated message (L min2 (j) ) that is based at least in part on a second-most minimum variable node message (l min2 (j) ) including by using L min2 (j) =└α ds ·l min2 (j) +β ds ┘.

7. The decoder recited in claim 1 , wherein:

a first attenuation factor (α) is used to generate the less aggressively attenuated message and a second attenuation factor (α ds ) is used to generate the more aggressively attenuated message, wherein α ds <α; and

a first attenuation rounding (β) is used to generate the less aggressively attenuated message and a second attenuation rounding (β ds ) is used to generate the more aggressively attenuated message, wherein └α ds +β ds ┘≤└α+β┘.

8. The decoder recited in claim 1 , wherein:

a first attenuation factor (α) is used to generate the less aggressively attenuated message and a second attenuation factor (α ds ) is used to generate the more aggressively attenuated message, wherein α ds <α;

a first attenuation rounding (β) is used to generate the less aggressively attenuated message and a second attenuation rounding (β ds ) is used to generate the more aggressively attenuated message, wherein └α ds +β ds ┘≤└α+β┘; and

at least one of α, α ds , β, or β ds is determined using density evolution.

9. A method, comprising:

generating a global de-saturation control signal that is determined based at least in part on a saturation threshold and a degree of saturation in a low-density parity-check (LDPC) decoding system that uses a fixed-point number representation;

in the event the global de-saturation control signal is a first value, outputting a more aggressively attenuated message to a plurality of check nodes compared to when the global de-saturation control signal is a second value; and

in the event the global de-saturation control signal is the second value, outputting a less aggressively attenuated message to the plurality of check nodes compared to when the global de-saturation control signal is the first value.

10. The method recited in claim 9 , wherein the degree of saturation is based at least in part on a number of messages, passed between at least one of the plurality of check nodes and at least one of the plurality of variable nodes, that are saturated to a maximum fixed-precision magnitude.

11. The method recited in claim 9 , wherein the degree of saturation is based at least in part on a number of saturated nodes in the LDPC decoding system.

12. The method recited in claim 9 , wherein:

the degree of saturation is based at least in part on a number of saturated nodes in the LDPC decoding system; and

a node is declared saturated if an associated minimum attenuated message after attenuation reaches a maximum fixed-precision magnitude.

13. The method recited in claim 9 , wherein outputting the more aggressively attenuated message to the plurality of check nodes includes:

outputting a first more aggressively attenuated message that is based at least in part on a first-most minimum variable node message; and

outputting a second more aggressively attenuated message that is based at least in part on a second-most minimum variable node message.

14. The method recited in claim 9 , wherein outputting the more aggressively attenuated message to the plurality of check nodes includes:

outputting a first more aggressively attenuated message (L min1 (j) ) that is based at least in part on a first-most minimum variable node message (l min1 (j) ), including by using L min1 (j) =└α ds ·l min1 (j) +β ds ┘, wherein α ds is associated with an attenuation factor and β ds is associated with an attenuation rounding; and

outputting a second more aggressively attenuated message (L min2 (j) ) that is based at least in part on a second-most minimum variable node message (l min2 (j) ) including by using L min2 (j) =└α ds ·l min2 (j) +β ds ┘.

15. The method recited in claim 9 , wherein:

a first attenuation factor (α) is used to generate the less aggressively attenuated message and a second attenuation factor (α ds ) is used to generate the more aggressively attenuated message, wherein α ds <α; and

a first attenuation rounding (β) is used to generate the less aggressively attenuated message and a second attenuation rounding (β ds ) is used to generate the more aggressively attenuated message, wherein └α ds +β ds ┘≤α+β┘.

16. The method recited in claim 9 , wherein:

a first attenuation factor (α) is used to generate the less aggressively attenuated message and a second attenuation factor (α ds ) is used to generate the more aggressively attenuated message, wherein α ds <α;

a first attenuation rounding (β) is used to generate the less aggressively attenuated message and a second attenuation rounding (β ds ) is used to generate the more aggressively attenuated message, wherein └α ds +β ds ┘≤└α+β┘; and

at least one of α, α ds , β, or β ds is determined using density evolution.

17. A computer program product embodied in a non-transitory computer readable medium and comprising computer instructions for:

generating a global de-saturation control signal that is determined based at least in part on a saturation threshold and a degree of saturation in a low-density parity-check (LDPC) decoding system that uses a fixed-point number representation;

in the event the global de-saturation control signal is a first value, outputting a more aggressively attenuated message to a plurality of check nodes compared to when the global de-saturation control signal is a second value; and

in the event the global de-saturation control signal is the second value, outputting a less aggressively attenuated message to the plurality of check nodes compared to when the global de-saturation control signal is the first value.

18. The computer program product recited in claim 17 , wherein the degree of saturation is based at least in part on a number of messages, passed between at least one of the plurality of check nodes and at least one of the plurality of variable nodes, that are saturated to a maximum fixed-precision magnitude.

19. The computer program product recited in claim 17 , wherein outputting the more aggressively attenuated message to the plurality of check nodes includes:

outputting a first more aggressively attenuated message that is based at least in part on a first-most minimum variable node message; and

outputting a second more aggressively attenuated message that is based at least in part on a second-most minimum variable node message.

20. The computer program product recited in claim 17 , wherein outputting the more aggressively attenuated message to the plurality of check nodes includes:

outputting a first more aggressively attenuated message (L min2 (j) ) that is based at least in part on a first-most minimum variable node message (l min1 (j) ), including by using L min1 (j) =└α ds ·l min1 (j) +β ds ┘, wherein is associated with an attenuation factor and β ds is associated with an attenuation rounding; and

outputting a second more aggressively attenuated message (L min2 (j) ) that is based at least in part on a second-most minimum variable node message (l min2 (j) ), including by using L min2 (j) =└α ds ·l min2 (j) +β ds ┘.