I ran Drup's benchmark on stream-like datastructure, which is provided in LILiS source code [1]. Interestingly, it features cfstream and core's sequence:
--- Lsystem Von_koch for 7 iterations ---
Stream+: LazyList+: Enum+: Seq+: Core+: CFStream+: Sequence+: Rate Stream+ LazyList+ CFStream+ Enum+ Core+ Seq+ Sequence+
Stream+ 3.58+-0.00/s -- -30% -78% -90% -93% -94% -95%
LazyList+ 5.14+-0.00/s 44% -- -68% -86% -90% -91% -92%
CFStream+ 15.9+- 0.0/s 344% 209% -- -57% -69% -72% -76%
Enum+ 37.0+- 0.1/s 933% 619% 132% -- -28% -34% -44%
Core+ 51.2+- 0.0/s 1331% 896% 222% 38% -- -8% -23%
Seq+ 56.0+- 0.1/s 1463% 988% 252% 51% 9% -- -16%
Sequence+ 66.2+- 0.0/s 1751% 1188% 316% 79% 29% 18% --
--- Lsystem dragon for 15 iterations ---
Rate Stream+ LazyList+ CFStream+ Enum+ Core+ Seq+ Sequence+
Stream+ 2.15+-0.00/s -- -29% -78% -90% -92% -94% -95%
LazyList+ 3.04+-0.00/s 41% -- -70% -85% -89% -91% -93%
CFStream+ 9.98+-0.00/s 364% 228% -- -52% -64% -71% -77%
Enum+ 20.7+- 0.0/s 861% 580% 107% -- -25% -40% -52%
Core+ 27.5+- 0.0/s 1180% 805% 176% 33% -- -21% -36%
Seq+ 34.7+- 0.0/s 1513% 1041% 248% 68% 26% -- -20%
Sequence+ 43.3+- 0.1/s 1912% 1323% 334% 109% 57% 25% --
--- Lsystem fern for 9 iterations ---
Rate Stream+ LazyList+ CFStream+ Enum+ Core+ Seq+ Sequence+
Stream+ 0.437+-0.001/s -- -34% -78% -90% -93% -93% -95%
LazyList+ 0.659+-0.001/s 51% -- -68% -86% -90% -90% -92%
CFStream+ 2.03+- 0.00/s 365% 208% -- -56% -68% -69% -75%
Enum+ 4.57+- 0.00/s 946% 593% 125% -- -29% -29% -44%
Core+ 6.42+- 0.00/s 1369% 874% 216% 40% -- -1% -21%
Seq+ 6.46+- 0.00/s 1377% 880% 218% 41% 1% -- -21%
Sequence+ 8.14+- 0.00/s 1762% 1135% 301% 78% 27% 26% --
--- Lsystem Tetradragon for 12 iterations ---
Rate Stream+ LazyList+ CFStream+ Enum+ Core+ Seq+ Sequence+
Stream+ 0.307+-0.000/s -- -30% -78% -90% -93% -94% -95%
LazyList+ 0.435+-0.002/s 42% -- -68% -86% -90% -91% -92%
CFStream+ 1.37+- 0.00/s 348% 215% -- -56% -68% -72% -76%
Enum+ 3.10+- 0.00/s 910% 612% 126% -- -27% -36% -46%
Core+ 4.26+- 0.00/s 1291% 880% 211% 38% -- -12% -26%
Seq+ 4.86+- 0.00/s 1485% 1016% 254% 57% 14% -- -16%
Sequence+ 5.78+- 0.00/s 1784% 1227% 321% 86% 35% 19% --
On this benchmark, core's sequence is clearly less efficient than c-cube's, but already much better than cfstream (BTW, I'm not sure why cfstream performs any better than stream, this looks strange to me at first sight). I noticed there will be a talk at OUD'14 about core's sequence library. How about deciding whether to switch to it when we know a little more? Hopefully, the change should be easy (thanks to cfstream!)
[1]
https://github.com/Drup/LILiS