1 reply to this topic

[Baraclese]

Hi,
basically anytime you want to read or write to/from a data structure you would first have to de/compress it. The idea is to fit the data set into the cache and get a speedup because de/compression is still better than a cache miss. Has anybody tried that yet? I can envision a simple zero byte compression.. you would need an indexing table that holds the offsets to each individual compressed object, then the first bytes hold the size of the compressed data and then you can run it through the decompressor. I wonder if this is viable. Of course if reading of the data occurs more often than writing then one could do more elaborate compression which still offers fast decompression rates.

[iMalc]

You could use something like LZP (http://www.cbloom.co...c/index_lz.html) or LZO (http://www.oberhumer...opensource/lzo/) if you are interested in this.

I've only ever done compression specific to the data structure at hand though.