jghub

to be on the save side I ran the test with Version ABIJM 93v- 2014-12-24, too. result: 93v- about 10% slower than 93u+. _that_ I can understand: 93v- was in...

300% is huge. 10% is 10%. 93v would have been a major release relative to 93u, not a minor one, in the view of D. Korn as far as I...

> @jghub I suggest you read #42. You seem to think I'm actively trying to destroy ksh as a viable shell for writing scripts in the 21st century. I have...

repeated benchmarks under 5.0.0-29-generic #31-Ubuntu SMP Thu Sep 12 13:05:32 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux ` name /bin/ksh ksh2020 (head of master) braces.ksh 0m3.77s 0m7.00s gsub.ksh 0m0.13s 0m0.11s numfor.ksh...

just tested this. looks better but definitely not good enough: ` ---------------------------------------------------------- name /bin/ksh ksh2020def ksh2020minsize braces.ksh 0m3.87s 0m9.17s 0m6.11s numfor.ksh 0m0.70s 0m2.18s 0m1.36s numloop.ksh 0m0.62s 0m1.24s 0m0.61s ` the...

as explained, these are really very simple benchmarks, each testing essentially a single property: ``` #!/bin/ksh cat >numfor.ksh

update: as of commit g5abcbd06 (current head) performance of ksh2020 remains down by a factor of 2-3 compared to ksh93u+ and ksh93v- in several critical features like incrementing numeric for...

@dannyweldon, @siteshwar : FYI information, I did some further benchmarking, also adding one for recursive function calls. result: ``` --------------------------------------------------------------------- name /bin/ksh ksh93v- ksh2020nms ksh2020ms ksh2020rel --------------------------------------------------------------------- braces.ksh 0m0.26s 0m0.29s...

first of all: I welcome any initiative to finally start looking into the issue at a technical level (instead of dodging the issue and acting silly when faced with criticism)....

see above... 1. setting the historical/factual record straight. your claim above: > I never "declared it "solved" by using --buildtype=minsize." I said that addressed most of the performance decrease. fact:...