EINSim v2.0.0

This release provides a more extensible interface for specifying error injection models, test patterns, measurements to compute, and ECC schemes. Please see the CHANGES for full details.

These improvements break compatibility with v1.0.0, so this release does no longer includes the simulation dataset used in the original paper.

EINSim v1.0.0 + Full-Size Dataset

This release contains the full-size simulation dataset that we use in our DSN 2019 paper. It consists of a tarball containing the following files:

• BCH_0xFF_T1.xz (11.50 MiB)
• BCH_0xFF_T2.xz (13.03 MiB)
• BCH_0xFF_T3.xz (13.68 MiB)
• BCH_RANDOM_T1.xz (9.34 MiB)
• BCH_RANDOM_T2.xz (9.93 MiB)
• BCH_RANDOM_T3.xz (10.09 MiB)
• HSC_0xFF_T1.xz (12.35 MiB)
• HSC_RANDOM_T1.xz (9.36 MiB)
• REP_0xFF_T1.xz (12.02 MiB)
• REP_0xFF_T2.xz (13.85 MiB)
• REP_0xFF_T3.xz (16.09 MiB)
• REP_RANDOM_T1.xz (10.62 MiB)
• REP_RANDOM_T2.xz (13.00 MiB)
• REP_RANDOM_T3.xz (15.36 MiB)

These datasets comprise 980000 simulated models that are generated as a cross-product of the following parameters:

• RBERs from 0.0 to 0.2 in 10000 steps
• RANDOM and 0xFF data patterns
• Four different 'p' {0, 1, 2, 3} (aggregated)
• ECC schemes: Hamming (d = 3), BCH (d = 3, 5, 7), Repetition (d = 3, 5, 7)
• Data-word lengths: {4, 8, 16, 32, 64, 128, 256} (note: BCH simulations are output with non-truncated lengths)
• Burst length: 256
• Uniform-random error distribution
• Bursts that are entirely either true-cells or anti-cells with 50% probability