Different benchmark result with large file

[yongkun]

The benchmark result in README shows that size of compressed data is close to each other, but lz4 is much faster, takes about half of the time compared to snappy.

I feel that the file size is a little small (24KB) in this test.

So I tried large file, about 1.5GB, our real data file.

I am using python 2.7.6 on macbook air (1.7G i5).

Python 2.7.6 |Anaconda 1.6.1 (x86_64)| (default, Jan 10 2014, 11:23:15) 
[GCC 4.0.1 (Apple Inc. build 5493)] on darwin

My test results:

$ python bench.py 
Data Size:
  Input: 1597401203
  LZ4: 184286011 (0.12)
  Snappy: 238020116 (0.15)
  LZ4 / Snappy: 0.774246
Benchmark: 2 calls
  LZ4 Compression: 33.591156s
  Snappy Compression: 7.052323s
  LZ4 Decompression: 51.998798s
  Snappy Decompression : 29.753453s

My result shows that LZ4 has much better compression ratio, LZ4 data size after compression is 184MB, while snappy data size is 238MB, so lz4 is about 23% better in case of compression ratio.
But it takes about 4X+ and 1.7X+ time for compression and decompression respectively.
So it seems that LZ4 is quite slower than snappy with large data set.

Is it correct or maybe I need some tuning on block size?