1 new code com algorithm & decompressor in FPGA
2 deflate com&decom
LZ77
abstract
- a “refine and recycle” method applicable to LZ77-type decompressors & implementation in reconfigurable logic
- refine the write commands(for literal tokens) & read commands(for copy tokens) a set of commands that target a single bank of block ram, and rather than performing all the dependency calculations saves logic by recycling (read) commands that return with an invalid result.
- A single “Snappy” decompressor implemented in reconfigurable logic leveraging this method is capable of processing multiple literal or copy tokens per cycle and achieves up to 7.2GB/s, which can keep pace with an NVMe device.
6 results
6.1 setup
- target the Xilinx Virtex Ultrascale VU3P- 2 device on an AlphaData ADM-PCIE-9V3 board and integrated with the POWER9 CAPI 2.0 [27] interface
- compared an optimized software Snappy decompression implementation [1] compiled by gcc 7.3.0 with “O3” option and running on a POWER9 CPU in little endian mode with Ubuntu 18.04.1 LTS.
- benchmarks
1-3 “lineitem” table of the TPC-H benchmarks
Wiki [22] is an XML file dump from Wikipedia
Matrix[2] is a sparse matrix from the Matrix Market
high compression ratio file (Geo) that stores geographic information
6.2 integration
decompressor communicates with host memory through the CAPI 2.0 interface
6.3 resource utilization
6.4 end-to-end throughput performance
6.5 impacts of # of BCPs
6.6 comparison of decompression accelerators
By using 6 BCPs, output up to 31B per cycle at a clock frequency of 250MHz
around 14.5x and 3.7x faster then the prior work on ZLIB[18] and Snappy
more area-efficient, measured in MB/s per 1K LUTs and MB/s per BRAM (36kb), which is 1.4x more LUT efficient than the ZLIB implementation in [18] and 2.4x more BRAM efficient than the Snappy implementation in [25].
the Vitis Data Compression Library (VDCL) 25x and 4x faster than a single engine implementation and an 8-engine implementation
A FPGA-based Snappy decompressor-Filter
5 results
setup
[26] sample videos, \benchmark2,” http://www.sample-videos.com/
download-sample-text- le.php, accessed Dec 17, 2017.
[27] applied maths, \benchmark1,” http://www.applied-maths.com/download/
sample-data, accessed Dec 17, 2017.
HW of LZMA
CR = 57%