Warp 400-410 and up to warp 430, threshold at 180 keywords just using Continue. Note: Changed the fill of result tables that start with index 1, to have a function return at 0 if no match in mode 254. #Include once "D:\Basic\LZLE_.bi" ' WITH CONST MAX_ASIDE=500 & change Byte to Integer...

It would probably be possible to gain another 4-5% by tweaking the index structure to optimize the search, but this could lead to stability issues. This approach seems very cumbersome. A little code cleaning allowed for a slight gain. "Warp 380-390 and up to warp 410" and challenging from ...

edited : wrong topic

How I see the issue : there are several reasons that may explain the lack of tips and tricks regarding code optimization. Look on the forum; there are some particularly fast codes (Dodicat, Srvaldez, Juergen Kulhwein, etc.), and yet all these codes are very different. Coding style is a coherent way ...

From experience with debuggers—and I'm probably wrong because it's old—we can't be absolutely certain of transparent operation; undocumented operation or degraded mode of the processor or system, automatic dynamic optimization, an impromptu load, or whatever else would suffice. Who needs a powerful ...

Warp 360-380 and up to Warp 400, challenging till 195-205 substring to test for in string on exemple #2 (instead of w300-310 up to 335 and 230-240 kwds) I think this time I'm finally reaching the end of what I could do with the tool to try to improve speed on an InstrMulti. Perhaps some code with mu...

The idea is to continue de-prototyping the InstrMulti function. In the same way we replaced Instr and Mid with pointers, in the same way the less "low level" functions of lzle, that is to say those which use loops can usefully be replaced by pointers, in particular the .RootPrivate and .Ha...

Reason : code cleaning and performance improvments : Warp 300 on ex#2 (100000 val) and challenging Instr till 255 values 8) It is done gradually, but it is difficult to move too quickly because of the engine. Important : new properties as well as RootPrivate are not designed to be 'multi-context', t...

Cause : 1 fix and one precision Precision : A Recycled list can become circular, this is NOT a malfunction. Hard to find a bug wich was not a 'true' one :mrgreen: :) When "hacking" the index directly to use each character as key, the parse use is not controled by the default instruction se...

Meaning wanting to avoid a string copy would be to use rather a strptr instead bref ?

This latest version of InstrMulti fixes two (minor) bugs: duplicate occurrence detections due to unmanaged 250/500 overlaps, and the untested Start parameter. The performance degradation issue on long strings has been addressed in several ways: a Borg collective submission mentions that ByPass would...

For 100,000 strings to search in test #2, I get Mr. La Forge's previous warp factor of 200, but without the bugs or patches—thanks, Weasley! The Instr emotion chip from Mr. Data is no longer required, and the results are more precise and complete by giving all multiple occurrences. To do this, the ....

Okay, so I went after the 30-40% speed loss. There might still be some false positive use cases; the uBNoFalsePositive parameter is set to 0 by default, but it seems okay, at least in my tests. We could perhaps gain another 20 or 30%, but it's starting to get complicated. So unless there's a huge bu...

Okay, I'm posting a fix because there's a list consistency issue with excessive use of List.Up: it seems there's an unmanaged context with this property. Speed ​​must have degraded further. My tests seem to indicate that it should be possible to gain another 40-50% by adapting/correcting lzle, but p...

Reason : serious bug fixes. Much more deeply tested. Looks like it works now. edited #Include once "D:\Basic\LZLE_.bi" ' WITH CONST MAX_ASIDE=100 Function InstrMulti(Str_Input As String, ListeQC As List, uBStopFirstFound As uByte=0) As String Dim As String Str_Result="", str_tmp_...