Publish the test suite/security metrics

[jan-wassenberg]

We'd like to publish the test suite, forked from smhasher and much faster thanks to data-parallel optimizations. This requires a portable thread pool and removal of other dependencies.

[jan-wassenberg]

Some progress: we have removed the dependencies and restructured the interface into a library. Open-source review is still pending, no estimate available yet.

[Bulat-Ziganshin]

i wonder why you cannot just publish the smahasher results as is. i'm interested because i think that even simple hashes like spooky or my own zzhash gathers a best possible results in this particular test, so i want to check your claim that highwayhash provides results better than any other hashes

[jan-wassenberg]

Sure, no problem, results are below. Just beware that it doesn't make much sense to compare results directly because this smhasher variant uses the PCG random generator and changes some of the test parameters (in particular, reps for AvalancheTest).
FYI we're planning on publishing this new parallel_smhasher by merging into github.com/aappleby/smhasher (to avoid the confusion of two similar but separate test suites), but I'm not sure when we'll get around to it.

Verification value 0x57EA5AB8 : Passed!
Running sanity check 1PASS
Running sanity check 2PASS

[[[ Keyset 'Zeroes' Tests ]]]

Keyset 'Zeroes' - 65536 keys
Testing distribution - Worst bias is the 13-bit window at bit 9 - 0.555%

[[[ Avalanche Tests ]]]

3-byte keys -> 64-bit, 9000000 reps worst bias is 0.162200%
4-byte keys -> 64-bit, 9000000 reps worst bias is 0.130178%
5-byte keys -> 64-bit, 9000000 reps worst bias is 0.134822%
6-byte keys -> 64-bit, 9000000 reps worst bias is 0.128756%
7-byte keys -> 64-bit, 9000000 reps worst bias is 0.119111%
8-byte keys -> 64-bit, 9000000 reps worst bias is 0.130133%
9-byte keys -> 64-bit, 9000000 reps worst bias is 0.134022%
10-byte keys -> 64-bit, 9000000 reps worst bias is 0.121000%
11-byte keys -> 64-bit, 9000000 reps worst bias is 0.132133%
12-byte keys -> 64-bit, 9000000 reps worst bias is 0.147022%
13-byte keys -> 64-bit, 9000000 reps worst bias is 0.131356%
14-byte keys -> 64-bit, 9000000 reps worst bias is 0.149067%
15-byte keys -> 64-bit, 9000000 reps worst bias is 0.125844%
16-byte keys -> 64-bit, 9000000 reps worst bias is 0.122489%
17-byte keys -> 64-bit, 9000000 reps worst bias is 0.138489%
18-byte keys -> 64-bit, 9000000 reps worst bias is 0.135956%
19-byte keys -> 64-bit, 9000000 reps worst bias is 0.139711%
20-byte keys -> 64-bit, 9000000 reps worst bias is 0.130533%
21-byte keys -> 64-bit, 9000000 reps worst bias is 0.137844%
22-byte keys -> 64-bit, 9000000 reps worst bias is 0.138133%
23-byte keys -> 64-bit, 9000000 reps worst bias is 0.148289%
24-byte keys -> 64-bit, 9000000 reps worst bias is 0.128600%
25-byte keys -> 64-bit, 9000000 reps worst bias is 0.130933%
26-byte keys -> 64-bit, 9000000 reps worst bias is 0.134578%
27-byte keys -> 64-bit, 9000000 reps worst bias is 0.131133%
28-byte keys -> 64-bit, 9000000 reps worst bias is 0.142044%
29-byte keys -> 64-bit, 9000000 reps worst bias is 0.138311%
30-byte keys -> 64-bit, 9000000 reps worst bias is 0.150844%
31-byte keys -> 64-bit, 9000000 reps worst bias is 0.137622%
32-byte keys -> 64-bit, 9000000 reps worst bias is 0.139111%
max bias: 0.162% average: 0.136%

[[[ Differential Tests ]]]

8303632 <=5-bit differentials in 64-bit keys -> 64 bit hashes.
1000 reps, 8303632000 total tests, expecting 0.00 random collisions
0 total collisions, of which 0 single collisions were ignored

11017632 <=4-bit differentials in 128-bit keys -> 64 bit hashes.
1000 reps, 11017632000 total tests, expecting 0.00 random collisions
0 total collisions, of which 0 single collisions were ignored

2796416 <=3-bit differentials in 256-bit keys -> 64 bit hashes.
1000 reps, 2796416000 total tests, expecting 0.00 random collisions
0 total collisions, of which 0 single collisions were ignored

[[[ Differential Distribution Tests ]]]

Testing distribution - Worst bias is the 18-bit window at bit 43 - 0.064%
Testing distribution - Worst bias is the 18-bit window at bit 15 - 0.087%
Testing distribution - Worst bias is the 18-bit window at bit 12 - 0.096%
Testing distribution - Worst bias is the 18-bit window at bit 60 - 0.087%
Testing distribution - Worst bias is the 18-bit window at bit 55 - 0.090%
Testing distribution - Worst bias is the 18-bit window at bit 7 - 0.090%
Testing distribution - Worst bias is the 18-bit window at bit 42 - 0.093%
Testing distribution - Worst bias is the 18-bit window at bit 6 - 0.115%
Testing distribution - Worst bias is the 18-bit window at bit 1 - 0.087%
Testing distribution - Worst bias is the 18-bit window at bit 12 - 0.111%
Testing distribution - Worst bias is the 18-bit window at bit 26 - 0.103%
Testing distribution - Worst bias is the 18-bit window at bit 12 - 0.088%
Testing distribution - Worst bias is the 18-bit window at bit 22 - 0.079%
Testing distribution - Worst bias is the 17-bit window at bit 12 - 0.061%
Testing distribution - Worst bias is the 18-bit window at bit 53 - 0.093%
Testing distribution - Worst bias is the 18-bit window at bit 25 - 0.066%
Testing distribution - Worst bias is the 18-bit window at bit 40 - 0.070%
Testing distribution - Worst bias is the 18-bit window at bit 54 - 0.137%
Testing distribution - Worst bias is the 18-bit window at bit 29 - 0.089%
Testing distribution - Worst bias is the 18-bit window at bit 33 - 0.057%
Testing distribution - Worst bias is the 18-bit window at bit 63 - 0.089%
Testing distribution - Worst bias is the 18-bit window at bit 57 - 0.098%
Testing distribution - Worst bias is the 18-bit window at bit 36 - 0.077%
Testing distribution - Worst bias is the 18-bit window at bit 37 - 0.086%
Testing distribution - Worst bias is the 18-bit window at bit 7 - 0.034%
Testing distribution - Worst bias is the 18-bit window at bit 21 - 0.070%
Testing distribution - Worst bias is the 18-bit window at bit 60 - 0.078%
Testing distribution - Worst bias is the 18-bit window at bit 27 - 0.082%
Testing distribution - Worst bias is the 18-bit window at bit 58 - 0.103%
Testing distribution - Worst bias is the 18-bit window at bit 49 - 0.079%
Testing distribution - Worst bias is the 18-bit window at bit 55 - 0.096%
Testing distribution - Worst bias is the 18-bit window at bit 50 - 0.054%
Testing distribution - Worst bias is the 18-bit window at bit 63 - 0.072%
Testing distribution - Worst bias is the 17-bit window at bit 13 - 0.048%
Testing distribution - Worst bias is the 18-bit window at bit 4 - 0.064%
Testing distribution - Worst bias is the 18-bit window at bit 57 - 0.101%
Testing distribution - Worst bias is the 18-bit window at bit 47 - 0.092%
Testing distribution - Worst bias is the 18-bit window at bit 56 - 0.080%
Testing distribution - Worst bias is the 18-bit window at bit 14 - 0.101%
Testing distribution - Worst bias is the 18-bit window at bit 41 - 0.118%
Testing distribution - Worst bias is the 18-bit window at bit 24 - 0.066%
Testing distribution - Worst bias is the 18-bit window at bit 30 - 0.083%
Testing distribution - Worst bias is the 18-bit window at bit 35 - 0.091%
Testing distribution - Worst bias is the 18-bit window at bit 57 - 0.058%
Testing distribution - Worst bias is the 18-bit window at bit 12 - 0.090%
Testing distribution - Worst bias is the 17-bit window at bit 30 - 0.079%
Testing distribution - Worst bias is the 18-bit window at bit 28 - 0.078%
Testing distribution - Worst bias is the 18-bit window at bit 25 - 0.094%
Testing distribution - Worst bias is the 18-bit window at bit 5 - 0.079%
Testing distribution - Worst bias is the 18-bit window at bit 62 - 0.075%
Testing distribution - Worst bias is the 17-bit window at bit 7 - 0.056%
Testing distribution - Worst bias is the 18-bit window at bit 57 - 0.069%
Testing distribution - Worst bias is the 18-bit window at bit 29 - 0.073%
Testing distribution - Worst bias is the 18-bit window at bit 1 - 0.065%
Testing distribution - Worst bias is the 18-bit window at bit 4 - 0.065%
Testing distribution - Worst bias is the 18-bit window at bit 35 - 0.075%
Testing distribution - Worst bias is the 18-bit window at bit 40 - 0.078%
Testing distribution - Worst bias is the 18-bit window at bit 28 - 0.070%
Testing distribution - Worst bias is the 18-bit window at bit 4 - 0.077%
Testing distribution - Worst bias is the 18-bit window at bit 48 - 0.102%
Testing distribution - Worst bias is the 18-bit window at bit 39 - 0.067%
Testing distribution - Worst bias is the 18-bit window at bit 11 - 0.076%
Testing distribution - Worst bias is the 18-bit window at bit 63 - 0.069%
Testing distribution - Worst bias is the 18-bit window at bit 36 - 0.096%

[[[ Keyset 'Combination Lowbits' Tests ]]]

Keyset 'Combination' - up to 8 blocks from a set of 8 - 19173960 keys
Testing distribution - Worst bias is the 20-bit window at bit 62 - 0.017%

[[[ Keyset 'Combination Highbits' Tests ]]]

Keyset 'Combination' - up to 8 blocks from a set of 8 - 19173960 keys
Testing distribution - Worst bias is the 20-bit window at bit 43 - 0.018%

[[[ Keyset 'Combination 0x8000000' Tests ]]]

Keyset 'Combination' - up to 20 blocks from a set of 2 - 2097150 keys
Testing distribution - Worst bias is the 18-bit window at bit 44 - 0.069%

[[[ Keyset 'Combination 0x0000001' Tests ]]]

Keyset 'Combination' - up to 20 blocks from a set of 2 - 2097150 keys
Testing distribution - Worst bias is the 18-bit window at bit 61 - 0.083%

[[[ Keyset 'Combination Hi-Lo' Tests ]]]

Keyset 'Combination' - up to 6 blocks from a set of 15 - 12204240 keys
Testing distribution - Worst bias is the 20-bit window at bit 52 - 0.031%

[[[ Keyset 'TwoBytes' Tests ]]]

Keyset 'TwoBytes' - up-to-4-byte keys, 652545 total keys
Testing distribution - Worst bias is the 16-bit window at bit 22 - 0.136%

Keyset 'TwoBytes' - up-to-8-byte keys, 5471025 total keys
Testing distribution - Worst bias is the 20-bit window at bit 25 - 0.052%

Keyset 'TwoBytes' - up-to-12-byte keys, 18616785 total keys
Testing distribution - Worst bias is the 20-bit window at bit 63 - 0.024%

Keyset 'TwoBytes' - up-to-16-byte keys, 44251425 total keys
Testing distribution - Worst bias is the 20-bit window at bit 58 - 0.008%

Keyset 'TwoBytes' - up-to-20-byte keys, 86536545 total keys
Testing distribution - Worst bias is the 20-bit window at bit 37 - 0.005%

[[[ Keyset 'Sparse' Tests ]]]

Keyset 'Sparse' - 32-bit keys with up to 6 bits set - 1149017 keys
Testing distribution - Worst bias is the 17-bit window at bit 11 - 0.126%

Keyset 'Sparse' - 40-bit keys with up to 6 bits set - 4598479 keys
Testing distribution - Worst bias is the 19-bit window at bit 25 - 0.035%

Keyset 'Sparse' - 48-bit keys with up to 5 bits set - 1925357 keys
Testing distribution - Worst bias is the 17-bit window at bit 22 - 0.067%

Keyset 'Sparse' - 56-bit keys with up to 5 bits set - 4216423 keys
Testing distribution - Worst bias is the 19-bit window at bit 10 - 0.038%

Keyset 'Sparse' - 64-bit keys with up to 5 bits set - 8303633 keys
Testing distribution - Worst bias is the 20-bit window at bit 19 - 0.039%

Keyset 'Sparse' - 96-bit keys with up to 4 bits set - 3469497 keys
Testing distribution - Worst bias is the 19-bit window at bit 13 - 0.083%

Keyset 'Sparse' - 256-bit keys with up to 3 bits set - 2796417 keys
Testing distribution - Worst bias is the 19-bit window at bit 44 - 0.069%

Keyset 'Sparse' - 2048-bit keys with up to 2 bits set - 2098177 keys
Testing distribution - Worst bias is the 18-bit window at bit 52 - 0.087%

[[[ Keyset 'Window' Tests ]]]

Keyset 'Windowed' - 128-bit key, 20-bit window - 128 tests, 1048576 keys per test
Testing distribution - Worst bias is the 17-bit window at bit 34 - 0.125%
Testing distribution - Worst bias is the 17-bit window at bit 61 - 0.100%
Testing distribution - Worst bias is the 17-bit window at bit 43 - 0.108%
Testing distribution - Worst bias is the 17-bit window at bit 18 - 0.155%
Testing distribution - Worst bias is the 17-bit window at bit 51 - 0.106%
Testing distribution - Worst bias is the 17-bit window at bit 39 - 0.106%
Testing distribution - Worst bias is the 17-bit window at bit 52 - 0.157%
Testing distribution - Worst bias is the 17-bit window at bit 52 - 0.096%
Testing distribution - Worst bias is the 17-bit window at bit 31 - 0.065%
Testing distribution - Worst bias is the 17-bit window at bit 23 - 0.141%
Testing distribution - Worst bias is the 17-bit window at bit 28 - 0.108%
Testing distribution - Worst bias is the 17-bit window at bit 22 - 0.080%
Testing distribution - Worst bias is the 16-bit window at bit 17 - 0.089%
Testing distribution - Worst bias is the 17-bit window at bit 22 - 0.108%
Testing distribution - Worst bias is the 17-bit window at bit 12 - 0.096%
Testing distribution - Worst bias is the 17-bit window at bit 11 - 0.118%
Testing distribution - Worst bias is the 17-bit window at bit 22 - 0.081%
Testing distribution - Worst bias is the 17-bit window at bit 55 - 0.123%
Testing distribution - Worst bias is the 17-bit window at bit 55 - 0.109%
Testing distribution - Worst bias is the 16-bit window at bit 40 - 0.100%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.096%
Testing distribution - Worst bias is the 17-bit window at bit 39 - 0.138%
Testing distribution - Worst bias is the 17-bit window at bit 42 - 0.076%
Testing distribution - Worst bias is the 17-bit window at bit 2 - 0.093%
Testing distribution - Worst bias is the 17-bit window at bit 28 - 0.096%
Testing distribution - Worst bias is the 17-bit window at bit 51 - 0.140%
Testing distribution - Worst bias is the 17-bit window at bit 1 - 0.115%
Testing distribution - Worst bias is the 17-bit window at bit 3 - 0.121%
Testing distribution - Worst bias is the 17-bit window at bit 23 - 0.127%
Testing distribution - Worst bias is the 17-bit window at bit 24 - 0.103%
Testing distribution - Worst bias is the 17-bit window at bit 58 - 0.118%
Testing distribution - Worst bias is the 17-bit window at bit 35 - 0.150%
Testing distribution - Worst bias is the 17-bit window at bit 34 - 0.124%
Testing distribution - Worst bias is the 17-bit window at bit 11 - 0.098%
Testing distribution - Worst bias is the 17-bit window at bit 55 - 0.097%
Testing distribution - Worst bias is the 17-bit window at bit 55 - 0.101%
Testing distribution - Worst bias is the 17-bit window at bit 39 - 0.109%
Testing distribution - Worst bias is the 16-bit window at bit 49 - 0.082%
Testing distribution - Worst bias is the 17-bit window at bit 16 - 0.115%
Testing distribution - Worst bias is the 17-bit window at bit 52 - 0.115%
Testing distribution - Worst bias is the 17-bit window at bit 57 - 0.098%
Testing distribution - Worst bias is the 17-bit window at bit 54 - 0.167%
Testing distribution - Worst bias is the 17-bit window at bit 12 - 0.147%
Testing distribution - Worst bias is the 17-bit window at bit 31 - 0.123%
Testing distribution - Worst bias is the 17-bit window at bit 50 - 0.092%
Testing distribution - Worst bias is the 17-bit window at bit 45 - 0.099%
Testing distribution - Worst bias is the 17-bit window at bit 30 - 0.101%
Testing distribution - Worst bias is the 17-bit window at bit 8 - 0.087%
Testing distribution - Worst bias is the 17-bit window at bit 58 - 0.131%
Testing distribution - Worst bias is the 17-bit window at bit 11 - 0.106%
Testing distribution - Worst bias is the 17-bit window at bit 8 - 0.120%
Testing distribution - Worst bias is the 17-bit window at bit 7 - 0.119%
Testing distribution - Worst bias is the 17-bit window at bit 54 - 0.116%
Testing distribution - Worst bias is the 17-bit window at bit 14 - 0.110%
Testing distribution - Worst bias is the 17-bit window at bit 1 - 0.054%
Testing distribution - Worst bias is the 17-bit window at bit 10 - 0.072%
Testing distribution - Worst bias is the 17-bit window at bit 8 - 0.092%
Testing distribution - Worst bias is the 17-bit window at bit 22 - 0.122%
Testing distribution - Worst bias is the 17-bit window at bit 27 - 0.128%
Testing distribution - Worst bias is the 17-bit window at bit 28 - 0.102%
Testing distribution - Worst bias is the 17-bit window at bit 30 - 0.096%
Testing distribution - Worst bias is the 17-bit window at bit 13 - 0.102%
Testing distribution - Worst bias is the 17-bit window at bit 13 - 0.105%
Testing distribution - Worst bias is the 17-bit window at bit 18 - 0.082%
Testing distribution - Worst bias is the 17-bit window at bit 57 - 0.162%
Testing distribution - Worst bias is the 17-bit window at bit 57 - 0.082%
Testing distribution - Worst bias is the 17-bit window at bit 27 - 0.080%
Testing distribution - Worst bias is the 17-bit window at bit 26 - 0.063%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.098%
Testing distribution - Worst bias is the 17-bit window at bit 34 - 0.106%
Testing distribution - Worst bias is the 17-bit window at bit 2 - 0.169%
Testing distribution - Worst bias is the 17-bit window at bit 39 - 0.116%
Testing distribution - Worst bias is the 17-bit window at bit 19 - 0.082%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.064%
Testing distribution - Worst bias is the 17-bit window at bit 2 - 0.108%
Testing distribution - Worst bias is the 17-bit window at bit 22 - 0.100%
Testing distribution - Worst bias is the 17-bit window at bit 58 - 0.162%
Testing distribution - Worst bias is the 17-bit window at bit 31 - 0.119%
Testing distribution - Worst bias is the 17-bit window at bit 49 - 0.100%
Testing distribution - Worst bias is the 17-bit window at bit 39 - 0.078%
Testing distribution - Worst bias is the 17-bit window at bit 26 - 0.091%
Testing distribution - Worst bias is the 17-bit window at bit 29 - 0.111%
Testing distribution - Worst bias is the 17-bit window at bit 7 - 0.149%
Testing distribution - Worst bias is the 17-bit window at bit 6 - 0.096%
Testing distribution - Worst bias is the 17-bit window at bit 46 - 0.094%
Testing distribution - Worst bias is the 17-bit window at bit 45 - 0.079%
Testing distribution - Worst bias is the 17-bit window at bit 41 - 0.100%
Testing distribution - Worst bias is the 17-bit window at bit 34 - 0.084%
Testing distribution - Worst bias is the 17-bit window at bit 36 - 0.083%
Testing distribution - Worst bias is the 17-bit window at bit 3 - 0.085%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.081%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.126%
Testing distribution - Worst bias is the 16-bit window at bit 37 - 0.100%
Testing distribution - Worst bias is the 17-bit window at bit 43 - 0.112%
Testing distribution - Worst bias is the 17-bit window at bit 12 - 0.165%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.080%
Testing distribution - Worst bias is the 17-bit window at bit 17 - 0.099%
Testing distribution - Worst bias is the 17-bit window at bit 7 - 0.082%
Testing distribution - Worst bias is the 17-bit window at bit 61 - 0.120%
Testing distribution - Worst bias is the 17-bit window at bit 2 - 0.194%
Testing distribution - Worst bias is the 17-bit window at bit 29 - 0.128%
Testing distribution - Worst bias is the 17-bit window at bit 11 - 0.121%
Testing distribution - Worst bias is the 17-bit window at bit 52 - 0.101%
Testing distribution - Worst bias is the 16-bit window at bit 51 - 0.109%
Testing distribution - Worst bias is the 17-bit window at bit 34 - 0.112%
Testing distribution - Worst bias is the 17-bit window at bit 51 - 0.121%
Testing distribution - Worst bias is the 17-bit window at bit 0 - 0.136%
Testing distribution - Worst bias is the 17-bit window at bit 50 - 0.112%
Testing distribution - Worst bias is the 17-bit window at bit 10 - 0.189%
Testing distribution - Worst bias is the 17-bit window at bit 10 - 0.150%
Testing distribution - Worst bias is the 17-bit window at bit 57 - 0.092%
Testing distribution - Worst bias is the 17-bit window at bit 63 - 0.148%
Testing distribution - Worst bias is the 17-bit window at bit 18 - 0.087%
Testing distribution - Worst bias is the 17-bit window at bit 2 - 0.102%
Testing distribution - Worst bias is the 17-bit window at bit 30 - 0.115%
Testing distribution - Worst bias is the 17-bit window at bit 47 - 0.118%
Testing distribution - Worst bias is the 17-bit window at bit 6 - 0.108%
Testing distribution - Worst bias is the 17-bit window at bit 63 - 0.068%
Testing distribution - Worst bias is the 17-bit window at bit 16 - 0.079%
Testing distribution - Worst bias is the 17-bit window at bit 58 - 0.161%
Testing distribution - Worst bias is the 17-bit window at bit 63 - 0.086%
Testing distribution - Worst bias is the 17-bit window at bit 26 - 0.110%
Testing distribution - Worst bias is the 17-bit window at bit 14 - 0.116%
Testing distribution - Worst bias is the 17-bit window at bit 56 - 0.110%
Testing distribution - Worst bias is the 17-bit window at bit 9 - 0.190%
Testing distribution - Worst bias is the 17-bit window at bit 9 - 0.148%
Testing distribution - Worst bias is the 17-bit window at bit 9 - 0.114%
Testing distribution - Worst bias is the 17-bit window at bit 46 - 0.080%
Testing distribution - Worst bias is the 17-bit window at bit 34 - 0.125%
OK

[[[ Keyset 'Cyclic' Tests ]]]

Keyset 'Cyclic' - 8 cycles of 8 bytes - 10000000 keys
Testing distribution - Worst bias is the 20-bit window at bit 23 - 0.028%

Keyset 'Cyclic' - 8 cycles of 9 bytes - 10000000 keys
Testing distribution - Worst bias is the 20-bit window at bit 34 - 0.033%

Keyset 'Cyclic' - 8 cycles of 10 bytes - 10000000 keys
Testing distribution - Worst bias is the 20-bit window at bit 7 - 0.043%

Keyset 'Cyclic' - 8 cycles of 11 bytes - 10000000 keys
Testing distribution - Worst bias is the 20-bit window at bit 63 - 0.036%

Keyset 'Cyclic' - 8 cycles of 12 bytes - 10000000 keys
Testing distribution - Worst bias is the 20-bit window at bit 43 - 0.035%

[[[ Keyset 'Text' Tests ]]]

Keyset 'Text' - keys of form "Foo[XXXX]Bar" - 14776336 keys
Testing distribution - Worst bias is the 20-bit window at bit 50 - 0.026%

Keyset 'Text' - keys of form "FooBar[XXXX]" - 14776336 keys
Testing distribution - Worst bias is the 20-bit window at bit 33 - 0.021%

Keyset 'Text' - keys of form "[XXXX]FooBar" - 14776336 keys
Testing distribution - Worst bias is the 20-bit window at bit 51 - 0.021%

[[[ Keyset 'Seed' Tests ]]]

Keyset 'Seed' - 1000000 keys
Testing distribution - Worst bias is the 17-bit window at bit 55 - 0.087%

[jan-wassenberg]

Here's a preprint with statistical analysis and preliminary cryptanalysis: https://arxiv.org/abs/1612.06257

We greatly reduced the variance of bias estimates by using 900M iterations instead of the 9M above.
Interesting result: the avalanche effect of SipHash13 and HighwayHash is comparable to that of the Blake2bp cryptographic hash.

We welcome any comments you might have!

[pombredanne]

I guess this binding will have to be named "The Snake" http://www.thesnake421.com/ aka. the snake highway.