Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
powturbo
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
14 ms
·
91.
▲
by
powturbo
10y ago
Java Critical Natives Interface. Access TurboRLE incl. SIMD! from Java as fast as calling from C
92.
▲
Show HN: TurboRLE: Bringing Turbo Run Length Encoding Incl. SIMD to Java
(github.com)
2 points
by
powturbo
10y ago
|
1 comments
93.
▲
by
powturbo
10y ago
Almost all current implementations (except TurboANX) are in general not better than huffman coding. See: https://sites.google.com/site/powturbo/entropy-coder Better test this yourself on your own data with TurboBe
94.
▲
by
powturbo
11y ago
hi Nathan, glad you want to benchmark TurboRLE on Haswell and Skylake. Well, it is very simple: 1- download the lkatest TurboRLE from github 2- cd TurboRLE 3- make 4- ./trle file The enwik9bwt must be generated. The two other files can
95.
▲
Show HN: TurboRLE – Turbo Run Length Encoding using SIMD
(github.com)
23 points
by
powturbo
11y ago
|
11 comments
96.
▲
by
powturbo
11y ago
You can make your own tests and your data with TurboBench: https://github.com/powturbo/TurboBench
97.
▲
Show HN: TurboBench Compressor Benchmark + nice Plots
(github.com)
1 points
by
powturbo
11y ago
|
0 comments
98.
▲
by
powturbo
11y ago
- The only benchmark program including LzTurbo - 100% in-memory benchmark, no I/O overhead - Include (>50) allmost all popular, latest or fastest compressors in one compiled package - No other compressor benchmark includes more cod
99.
▲
Show HN: TurboBench – Compressor Benchmark
(github.com)
4 points
by
powturbo
11y ago
|
1 comments
100.
▲
by
powturbo
11y ago
Internet Scenario Benchmark (browser plugin simulation): html8 : 100MB random html pages from a 2GB Alexa Top sites corpus. number of pages = 1178 average length = 84886 bytes. The pages (length + content) are concatenated into a single ht
101.
▲
by
powturbo
11y ago
Benchmarking the fastest, most efficient or popular entropy coders - Asymmetric Numeral Systems incl. FSE & TurboANX - Asymmetric Binary Systems - Arithmetic Coding / Range Coder w/ bitwise/bytewise range coders - Fastest
102.
▲
Show HN: Entropy Coder Benchmark
(sites.google.com)
1 points
by
powturbo
11y ago
|
1 comments
103.
▲
by
powturbo
11y ago
libfor benchmarked: https://github.com/powturbo/TurboPFor
104.
▲
by
powturbo
11y ago
- Direct Access w/o decompression - Fastest Variable Byte implementation - Novel Variable Simple faster than simple16, better than simple8-b - Scalar Bit Packing decoding as fast as SIMD-Packing - Bit Packing incl. Direct Access/U
105.
▲
Show HN: TurboPFor – Bringing Fastest Integer Compression Incl. SIMD to Java
(github.com)
14 points
by
powturbo
11y ago
|
1 comments
106.
▲
TurboPFor: Bringing Fastest Integer Compression Incl. SIMD to Java
(github.com)
2 points
by
powturbo
11y ago
|
0 comments
107.
▲
by
powturbo
11y ago
- Direct Access w/o decompression - Fastest Variable Byte implementation - Novel Variable Simple faster than simple16, better than simple8-b - Scalar Bit Packing decoding as fast as SIMD-Packing - Bit Packing incl. Di
108.
▲
TurboPFor:Fastest Integer Compression+Inverted Index. PForDelta,Bit Packing,
(github.com)
1 points
by
powturbo
11y ago
|
1 comments
109.
▲
by
powturbo
12y ago
Efficient and fast Run Length Encoding library - 100% C (C++ compatible headers), without inline assembly - Efficient compression - Maximal 1 byte overhead - No modification of the raw data, preserving compressibility - Order preserving
110.
▲
TurboRLE: Turbo Run Length Encoding
(github.com)
3 points
by
powturbo
12y ago
|
1 comments
111.
▲
by
powturbo
12y ago
TurboPFor - DIRECT ACCESS Integer Compression w/o decompressing entire blocks. Instant access to individual compressed array elements X-Times faster.
112.
▲
TurboPFor – DIRECT ACCESS Integer Compression
(github.com)
4 points
by
powturbo
12y ago
|
2 comments
113.
▲
TurboPFOR: Direct Access “Integer Compression” w/o Decompression
(github.com)
1 points
by
powturbo
12y ago
|
0 comments
114.
▲
by
powturbo
12y ago
* Fastest Integer Compression - 100% C, without inline assembly - Fastest "Variable Byte" implementation - Novel "Variable Simple" faster than simple16 and more compact than simple64 - Scalar "Binary Packing" w
115.
▲
TurboPFor: Fastest Integer Compression
(github.com)
2 points
by
powturbo
12y ago
|
1 comments