3 ms·
The algorithm used (EPH) seems bit curious. The paper says "The EPH algorithm was implemented in the C language and is available at http://cmph.sf.net" http://c
by zokier 9y ago
The algorithm used (EPH) seems bit curious. The paper says "The EPH algorithm was implemented in the C language
and is available at http://cmph.sf.net" http://cmph.sf.net", but that page has no mention of EPH and I even checked archive.org. I wonder why they ended up never actually releasing a version of cmph with that algorithm. Two years later they seem to have come up with another algorithm, CHD, which was actually released in cmph. Interestingly enough the CHD paper has no comparisons to EPH either.
- rurban 9y agoYes, interesting. Never heard of EPH before. cmph contains only: BDZ, BDZ_PH, BMZ, BMZ8, BRZ, CHD, CHD_PH, CHM and FCH. EPH seems to be better than BDZ, CHM and FCH, but only for really huge tables, like >100.000 entries. For smaller hash tables a simple perfect hash or even an optimized memcmp switch table is still faster. https://github.com/rurban/Perfect-Hash#benchmarks https://github.com/rurban/Perfect-Hash#benchmarks cmph is usually 2-3x slower than a trivial PH, and much slower for small tables.