0e7f9a6a6f
Allow zero parity shards ( #161 )
2021-03-08 16:13:24 +01:00
ab26eb4126
Add WithInversionCache and use pointer methods ( #160 )
...
There appears to be writes to value receivers.
Add `WithInversionCache(bool)` to disable cache.
Fixes #159
2021-01-13 10:21:28 +01:00
7c8682430c
tests: Set full data size as number of bytes ( #157 )
...
* Clean up deps.
* tests: Set full data size as number of bytes
Use total data size (data+parity) as benchmark sizes for more consistent benchmarks.
2020-12-18 09:09:17 +01:00
519603f6e1
Update packages ( #154 )
...
* Update packages
Update cpuid and clean up generated.
2020-12-09 22:56:01 +01:00
653e76aa26
Faster AVX2 encoding ( #153 )
...
* Remove 50% of bounds checks when copying.
* Use RIP only addressing, free one register.
```
benchmark old MB/s new MB/s speedup
BenchmarkGalois128K-32 57663.49 58005.87 1.01x
BenchmarkGalois1M-32 49479.31 49848.29 1.01x
BenchmarkGaloisXor128K-32 46310.69 46501.88 1.00x
BenchmarkGaloisXor1M-32 43804.86 43984.39 1.00x
BenchmarkEncode10x2x10000-32 25926.93 27457.75 1.06x
BenchmarkEncode100x20x10000-32 2635.82 2818.95 1.07x
BenchmarkEncode17x3x1M-32 63215.11 61576.76 0.97x
BenchmarkEncode10x4x16M-32 19551.54 19505.07 1.00x
BenchmarkEncode5x2x1M-32 79612.06 81985.14 1.03x
BenchmarkEncode10x2x1M-32 121478.29 127739.41 1.05x
BenchmarkEncode10x4x1M-32 70757.61 74423.67 1.05x
BenchmarkEncode50x20x1M-32 19811.96 20103.32 1.01x
BenchmarkEncode17x3x16M-32 27202.10 27825.34 1.02x
BenchmarkEncode_8x4x8M-32 19029.04 19701.31 1.04x
BenchmarkEncode_12x4x12M-32 22449.87 22480.51 1.00x
BenchmarkEncode_16x4x16M-32 24536.74 24672.24 1.01x
BenchmarkEncode_16x4x32M-32 24381.34 24981.99 1.02x
BenchmarkEncode_16x4x64M-32 24717.69 25086.94 1.01x
BenchmarkEncode_8x5x8M-32 16763.51 17154.04 1.02x
BenchmarkEncode_8x6x8M-32 15067.22 15205.87 1.01x
BenchmarkEncode_8x7x8M-32 13156.38 13589.40 1.03x
BenchmarkEncode_8x9x8M-32 11363.74 11523.70 1.01x
BenchmarkEncode_8x10x8M-32 10359.37 10474.91 1.01x
BenchmarkEncode_8x11x8M-32 9627.07 9463.24 0.98x
BenchmarkEncode_8x8x05M-32 30104.80 32634.89 1.08x
BenchmarkEncode_8x8x1M-32 36497.28 36425.88 1.00x
BenchmarkEncode_8x8x8M-32 12186.19 11602.41 0.95x
BenchmarkEncode_8x8x32M-32 11670.72 11413.71 0.98x
BenchmarkEncode_24x8x24M-32 21709.83 21652.50 1.00x
BenchmarkEncode_24x8x48M-32 22494.40 22280.59 0.99x
BenchmarkVerify10x2x10000-32 10567.56 10483.91 0.99x
BenchmarkVerify50x5x50000-32 28102.84 27923.63 0.99x
BenchmarkVerify10x2x1M-32 30298.33 30106.18 0.99x
BenchmarkVerify5x2x1M-32 16115.91 15847.03 0.98x
BenchmarkVerify10x4x1M-32 15382.13 14852.68 0.97x
BenchmarkVerify50x20x1M-32 8476.02 8466.24 1.00x
BenchmarkVerify10x4x16M-32 15101.03 15434.71 1.02x
BenchmarkReconstruct10x2x10000-32 26228.18 26960.19 1.03x
BenchmarkReconstruct50x5x50000-32 31091.42 30975.82 1.00x
BenchmarkReconstruct10x2x1M-32 58548.87 60281.92 1.03x
BenchmarkReconstruct5x2x1M-32 39499.23 41791.80 1.06x
BenchmarkReconstruct10x4x1M-32 41448.60 43053.15 1.04x
BenchmarkReconstruct50x20x1M-32 17185.99 17354.67 1.01x
BenchmarkReconstruct10x4x16M-32 18798.60 18847.43 1.00x
BenchmarkReconstructData10x2x10000-32 27208.48 27538.38 1.01x
BenchmarkReconstructData50x5x50000-32 32135.65 32078.91 1.00x
BenchmarkReconstructData10x2x1M-32 63180.19 67332.17 1.07x
BenchmarkReconstructData5x2x1M-32 47532.85 49932.17 1.05x
BenchmarkReconstructData10x4x1M-32 50059.14 52323.15 1.05x
BenchmarkReconstructData50x20x1M-32 26679.75 26714.11 1.00x
BenchmarkReconstructData10x4x16M-32 24854.99 24527.23 0.99x
BenchmarkReconstructP10x2x10000-32 115089.87 113229.75 0.98x
BenchmarkReconstructP10x5x20000-32 129838.75 132871.10 1.02x
BenchmarkParallel_8x8x64K-32 69951.43 69980.44 1.00x
BenchmarkParallel_8x8x05M-32 11752.94 11724.35 1.00x
BenchmarkParallel_20x10x05M-32 18553.93 18613.33 1.00x
BenchmarkParallel_8x8x1M-32 11639.19 11746.86 1.01x
BenchmarkParallel_8x8x8M-32 11799.36 11685.63 0.99x
BenchmarkParallel_8x8x32M-32 11510.94 11791.72 1.02x
BenchmarkParallel_8x3x1M-32 20268.92 20678.21 1.02x
BenchmarkParallel_8x4x1M-32 17616.05 17856.17 1.01x
BenchmarkParallel_8x5x1M-32 15590.87 15872.42 1.02x
BenchmarkStreamEncode10x2x10000-32 14917.08 15408.39 1.03x
BenchmarkStreamEncode100x20x10000-32 2014.81 2077.31 1.03x
BenchmarkStreamEncode17x3x1M-32 11839.37 12434.80 1.05x
BenchmarkStreamEncode10x4x16M-32 9151.14 9206.98 1.01x
BenchmarkStreamEncode5x2x1M-32 13598.55 13663.56 1.00x
BenchmarkStreamEncode10x2x1M-32 13192.91 13453.41 1.02x
BenchmarkStreamEncode10x4x1M-32 12109.90 12050.68 1.00x
BenchmarkStreamEncode50x20x1M-32 8640.73 8370.10 0.97x
BenchmarkStreamEncode17x3x16M-32 10473.17 10527.04 1.01x
BenchmarkStreamVerify10x2x10000-32 7032.23 7128.82 1.01x
BenchmarkStreamVerify50x5x50000-32 13023.46 13109.31 1.01x
BenchmarkStreamVerify10x2x1M-32 11941.63 11949.91 1.00x
BenchmarkStreamVerify5x2x1M-32 8029.93 8263.39 1.03x
BenchmarkStreamVerify10x4x1M-32 8137.82 8271.11 1.02x
BenchmarkStreamVerify50x20x1M-32 7378.87 7708.81 1.04x
BenchmarkStreamVerify10x4x16M-32 8973.18 8955.29 1.00x
```
2020-11-10 14:39:23 +01:00
7daa20bf74
Generate AVX2 code ( #141 )
...
Replaces AVX2 up to 10x8 configurations with specific generated functions.
If code size is a concern `-tags=nogen` can be used.
Biggest speedup when not memory constrained.
```
benchmark old MB/s new MB/s speedup
BenchmarkEncode_8x5x8M 5895.75 9648.18 1.64x
BenchmarkEncode_8x5x8M-4 16773.41 17220.67 1.03x
BenchmarkEncode_8x5x8M-16 18263.12 17176.28 0.94x
BenchmarkEncode_8x6x8M 5075.89 8548.39 1.68x
BenchmarkEncode_8x6x8M-4 14559.83 15370.95 1.06x
BenchmarkEncode_8x6x8M-16 16183.37 15291.98 0.94x
BenchmarkEncode_8x7x8M 4481.18 7015.60 1.57x
BenchmarkEncode_8x7x8M-4 12835.35 13695.90 1.07x
BenchmarkEncode_8x7x8M-16 14246.94 13737.36 0.96x
BenchmarkEncode_8x8x05M 5569.95 7947.70 1.43x
BenchmarkEncode_8x8x05M-4 17334.91 25271.37 1.46x
BenchmarkEncode_8x8x05M-16 29349.42 35043.36 1.19x
BenchmarkEncode_8x8x1M 4830.58 7891.32 1.63x
BenchmarkEncode_8x8x1M-4 17531.36 27371.42 1.56x
BenchmarkEncode_8x8x1M-16 29593.98 39241.09 1.33x
BenchmarkEncode_8x8x8M 3953.66 6584.26 1.67x
BenchmarkEncode_8x8x8M-4 11527.34 12331.23 1.07x
BenchmarkEncode_8x8x8M-16 12718.89 12173.08 0.96x
BenchmarkEncode_8x8x32M 3927.51 6195.91 1.58x
BenchmarkEncode_8x8x32M-4 11490.85 11424.39 0.99x
BenchmarkEncode_8x8x32M-16 12506.09 11888.55 0.95x
benchmark old MB/s new MB/s speedup
BenchmarkParallel_8x8x64K 5490.24 6959.57 1.27x
BenchmarkParallel_8x8x64K-4 21078.94 29557.51 1.40x
BenchmarkParallel_8x8x64K-16 57508.45 73672.54 1.28x
BenchmarkParallel_8x8x1M 4755.49 7667.84 1.61x
BenchmarkParallel_8x8x1M-4 11818.66 12013.49 1.02x
BenchmarkParallel_8x8x1M-16 12923.12 12109.42 0.94x
BenchmarkParallel_8x8x8M 3973.94 6525.85 1.64x
BenchmarkParallel_8x8x8M-4 11725.68 11312.46 0.96x
BenchmarkParallel_8x8x8M-16 12608.20 11484.98 0.91x
BenchmarkParallel_8x3x1M 14139.71 17993.04 1.27x
BenchmarkParallel_8x3x1M-4 21805.97 23053.92 1.06x
BenchmarkParallel_8x3x1M-16 24673.05 23596.71 0.96x
BenchmarkParallel_8x4x1M 10617.88 14474.54 1.36x
BenchmarkParallel_8x4x1M-4 18635.82 18965.65 1.02x
BenchmarkParallel_8x4x1M-16 21518.12 20171.47 0.94x
BenchmarkParallel_8x5x1M 8669.88 11833.96 1.36x
BenchmarkParallel_8x5x1M-4 16321.00 17500.30 1.07x
BenchmarkParallel_8x5x1M-16 17267.16 17191.04 1.00x
```
2020-05-20 12:48:34 +02:00
cf8495259a
Add pure XOR for 1 parity ( #138 )
...
WithFastOneParityMatrix will switch the matrix to a simple xor if there is only one parity shard.
The PAR1 matrix already has this property so it has little effect there.
2020-05-13 11:10:58 +02:00
2df03bd4d1
Ci test more archs ( #135 )
...
* ci: test more architectures
2020-05-09 10:35:17 +02:00
696c4018f8
bench: Fix reconstruct benchmarks ( #133 )
...
Always corrupt at least one shard and don't shuffle shards.
2020-05-06 15:42:49 +02:00
1b9e129671
Avx512 parallel81 ( #131 )
...
* AVX512 routine for 8x1 parallel processing (WIP)
* Testing and integration of Parallel81 assembly routine
2020-05-06 12:32:31 +02:00
cb7a0b5aef
Do fast by one multiplication ( #130 )
...
When multiplying by one we can use faster math.
2020-05-06 11:14:25 +02:00
65df535980
Make single goroutine encodes more efficient ( #122 )
...
Calculate the optimal per round size to keep data in cache when not using WithAutoGoroutines.
```
λ benchcmp before.txt after.txt
benchmark old ns/op new ns/op delta
BenchmarkParallel_8x8x05M-16 675225 321053 -52.45%
BenchmarkParallel_20x10x05M-16 3471988 600740 -82.70%
BenchmarkParallel_8x8x1M-16 3948606 728093 -81.56%
BenchmarkParallel_8x8x8M-16 47361588 5976467 -87.38%
BenchmarkParallel_8x8x32M-16 195044200 24365474 -87.51%
benchmark old MB/s new MB/s speedup
BenchmarkParallel_8x8x05M-16 6211.71 13064.22 2.10x
BenchmarkParallel_20x10x05M-16 3020.10 17454.73 5.78x
BenchmarkParallel_8x8x1M-16 2124.45 11521.34 5.42x
BenchmarkParallel_8x8x8M-16 1416.95 11228.85 7.92x
BenchmarkParallel_8x8x32M-16 1376.28 11017.04 8.00x
```
2020-05-03 19:37:22 +02:00
d2cfcb8065
Add commandline arg to disable asm for tests. ( #116 )
...
* Add commandline test args
2020-04-22 15:38:21 +02:00
0abe9de20c
Update tests ( #115 )
...
Don't create new slices.
2020-02-21 11:30:44 -08:00
ec2eb9fb8c
Split: Reduce memory allocation ( #103 )
...
* [Split] Reduce memory allocation in Split function
2019-06-25 16:28:24 +02:00
f5e73dcfe2
Split blocks into size divisible by 16
...
Older systems (typically without AVX2) are more sensitive to misaligned load+stores.
Add parameter to automatically set the number of goroutines.
name old time/op new time/op delta
Encode10x2x10000-8 18.4µs ± 1% 16.1µs ± 1% -12.43% (p=0.000 n=9+9)
Encode100x20x10000-8 692µs ± 1% 608µs ± 1% -12.10% (p=0.000 n=10+10)
Encode17x3x1M-8 1.78ms ± 5% 1.49ms ± 1% -16.63% (p=0.000 n=10+10)
Encode10x4x16M-8 21.5ms ± 5% 19.6ms ± 4% -8.74% (p=0.000 n=10+9)
Encode5x2x1M-8 343µs ± 2% 267µs ± 2% -22.22% (p=0.000 n=9+10)
Encode10x2x1M-8 858µs ± 5% 701µs ± 5% -18.34% (p=0.000 n=10+10)
Encode10x4x1M-8 1.34ms ± 1% 1.16ms ± 1% -13.19% (p=0.000 n=9+9)
Encode50x20x1M-8 30.3ms ± 4% 25.0ms ± 2% -17.51% (p=0.000 n=10+8)
Encode17x3x16M-8 26.9ms ± 1% 24.5ms ± 4% -9.13% (p=0.000 n=8+10)
name old speed new speed delta
Encode10x2x10000-8 5.45GB/s ± 1% 6.22GB/s ± 1% +14.20% (p=0.000 n=9+9)
Encode100x20x10000-8 1.44GB/s ± 1% 1.64GB/s ± 1% +13.77% (p=0.000 n=10+10)
Encode17x3x1M-8 10.0GB/s ± 5% 12.0GB/s ± 1% +19.88% (p=0.000 n=10+10)
Encode10x4x16M-8 7.81GB/s ± 5% 8.56GB/s ± 5% +9.58% (p=0.000 n=10+9)
Encode5x2x1M-8 15.3GB/s ± 2% 19.6GB/s ± 2% +28.57% (p=0.000 n=9+10)
Encode10x2x1M-8 12.2GB/s ± 5% 15.0GB/s ± 5% +22.45% (p=0.000 n=10+10)
Encode10x4x1M-8 7.84GB/s ± 1% 9.03GB/s ± 1% +15.19% (p=0.000 n=9+9)
Encode50x20x1M-8 1.73GB/s ± 4% 2.09GB/s ± 4% +20.59% (p=0.000 n=10+9)
Encode17x3x16M-8 10.6GB/s ± 1% 11.7GB/s ± 4% +10.12% (p=0.000 n=8+10)
2017-11-18 22:00:55 +01:00
61c22eab55
Cauchy Matrix option ( #70 )
...
* Experimental Cauchy Matrix
Experimental support for Cauchy style matrix
http://web.eecs.utk.edu/~plank/plank/papers/CS-05-569.pdf
All matrices appear reversible.
* Remove Go 1.5 and 1.6 from CI tests.
* Fix comment.
* Increase max number of goroutines+docs.
2017-10-01 14:02:11 +02:00
ddcafc661e
Allow reconstructing into pre-allocated memory. ( #66 )
...
This changes the interface of Reconstruct and ReconstructData to accept
slices of zero length but sufficient capacity for shards to reconstruct,
and reslices them instead of allocating new memory.
2017-09-20 21:08:24 +02:00
d78bf472d8
add Update parity function ( #60 )
...
Add Update parity function
2017-08-20 11:42:39 +02:00
dc6af2dce5
Minor cleanup ( #61 )
...
* Remove some benchmarks
* Format tables a bit.
* Doc cleanup
2017-08-13 22:38:27 +02:00
0de37d7697
Add ReconstructData interface method ( #57 )
...
* Add ReconstructData interface method to allow reconstruction of any missing data shards
* Add support for just reconstructing data shards only to SteamEncoder.Reconstruct()
2017-07-20 12:15:46 +02:00
18d548df63
Add support for PAR1 ( #55 )
...
PAR1 is a file format which uses a Reed-Solomon code similar
to the current one, except it uses a different (flawed) coding
matrix.
Add support for it via a WithPAR1Matrix option, so that this code
can be used to encode/decode PAR1 files. Also add the option to
existing tests, and add a test demonstrating the flaw in PAR1's
coding matrix.
Also fix an mistakenly inverted test in testOpts().
Incidentally, PAR1 is obsoleted by PAR2, which uses GF(2^16)
and tries to fix the flaw in the coding matrix; however, PAR2's
coding matrix is still flawed! The real solution is to build the
coding matrix like in this repository.
PAR1 spec:
http://parchive.sourceforge.net/docs/specifications/parity-volume-spec-1.0/article-spec.html
Paper describing the (flawed) Reed-Solomon code used by PAR1:
http://web.eecs.utk.edu/~plank/plank/papers/CS-96-332.html
2017-06-20 20:24:57 +02:00
87c4e5ae75
Allow 256 total shards ( #54 )
...
* Allow 256 total shards
2017-06-19 11:26:52 +02:00
5abf0ee302
Add options ( #46 )
...
* Add options
Make constants changeable as options.
The API remains backwards compatible.
* Update documentation.
* Fix line endings
* fmt
* fmt
* Use functions for parameters.
Much neater.
2017-02-19 11:13:22 +01:00
c54154da9e
Add Inverse Matrix caching in a Thread-Safe Lookup Tree ( #36 )
...
* Add matrix inversion caching
* Benchmark and Parallel Benchmark tests for Reconstruct
2016-09-12 21:31:07 +02:00
b1c8b4b073
Make Join return an error if a reconstruction is required first
...
If one or more required data shards are still nil and we can't correctly join
them before a reconstruction, return ErrReconstructRequired.
2016-08-05 19:23:08 +02:00
ba30981088
Add checks for data and parity to not exceed 255 shards in total.
...
Fixes #16
2016-06-03 01:31:01 -07:00
9f0bea8a29
Tests: backport go1.6 rand.Read for speedup tests
2016-04-07 18:34:47 +08:00
976a24f33b
Move examples to separate file/package
...
This makes the reedsolomon package prefix show up in the documentation examples.
+ StreamEncoder example.
2015-11-03 12:12:42 +01:00
86bd0f239b
seed RNG in TestSplitJoin
2015-08-08 18:20:40 -04:00
458f451fc2
add codeSomeShardsP test
2015-08-08 13:52:00 -04:00
bb7bd0036a
fully test Split/Join functions
2015-08-08 13:51:11 -04:00
64b705bbf6
fully test Reconstruct function
...
Well, I can't figure out how to trigger the Invert error.
It may not be possible; need more domain knowledge to be sure.
2015-08-08 13:50:18 -04:00
f81ea8daaf
fully test Verify function
2015-08-08 13:50:18 -04:00
0238782585
fully test Encode function
2015-08-08 13:50:18 -04:00
10fbe96890
use slice literal
2015-08-06 22:56:32 -04:00
640ab74d9d
fully test the New function
2015-08-06 22:47:11 -04:00
d31049df42
Add another example that shows that sets can be xor'ed and still remain valid.
2015-06-23 14:35:16 +02:00
8ebf356efb
The number of data shards must be below 257. Check that and update documentation.
2015-06-23 13:39:57 +02:00
6861078d3b
Add more information to example.
2015-06-22 15:52:10 +02:00
0cb21eccc5
Rename example function.
2015-06-22 15:48:52 +02:00
7794948a5b
Add split/merge example.
2015-06-22 15:44:22 +02:00
619e2b7d65
Add benchmark with 17 data shards and 3 parity shards with 16MB each, and correct comments.
2015-06-21 17:07:17 +02:00
ab50161bb9
Update benchmarks.
2015-06-20 20:51:26 +02:00
36a0e57744
Begin docs.
2015-06-20 13:10:51 +02:00
d54843ee41
Add Encoder example (and test)
2015-06-20 11:29:26 +02:00
c5de03551c
Minor adjustments for golint.
2015-06-20 10:11:33 +02:00
cf70107291
Add verification test that also tests failure.
2015-06-19 19:20:44 +02:00
e3aca6cd9d
Shorten the variable names and make an encoder interface, so it isn't possible to create it without calling New.
2015-06-19 18:54:58 +02:00
67f8d8b8c7
Add another benchmark.
2015-06-19 18:25:48 +02:00
Shawn Zivontsis