Re: State of the overlay tree branch?

[Sebastian Sturm <s.sturm@arkona-technologies.de>]

for the record, this is what I get with the (non-noverlay, non-patched) 
emacs-27 master branch after calling `(set-buffer-multibyte nil)`. Guess 
this merely confirms what you already found out, but for me this is a 
viable workaround until the issue is properly resolved. Will keep 
looking for an open source test file over the weekend

emacs-27 master, ~6200 overlays, enable-multibyte-characters = nil:
iteration 1: 0.000300
iteration 2: 0.000268
iteration 3: 0.000267
iteration 4: 0.000268
iteration 5: 0.000267
iteration 6: 0.000266
iteration 7: 0.000263
iteration 8: 0.000263
iteration 9: 0.000277
iteration 10: 0.000269
iteration 11: 0.000264
iteration 12: 0.000266
iteration 13: 0.000263
iteration 14: 0.000265
iteration 15: 0.000273
iteration 16: 0.000267
iteration 17: 0.000266
iteration 18: 0.000266
iteration 19: 0.000265
iteration 20: 0.000263


On 03/22/2018 08:54 PM, Sebastian Sturm wrote:
>  >
>  > It should do that without you asking.  I mean, it won't show you
>  > `goto-char` and `point-min` kind of things since these are "inlined"
>  > (actually turned into their own byte-code), but `count-lines` should
>  > definitely be there.
>  >
>  >>               - let 18368  86%
>  >>                  line-number-at-pos 18348  86%
>  >
>  > It's very odd that there's no `count-lines` down here (and `count-lines`
>  > is a perfectly normal Elisp function that's not inlined or otherwise
>  > treated specially), since it should be where most of the time is spent!
> 
> when first evaluating simple.el as Eli suggested, the profiler indeed 
> shows that pretty much all time is spent somewhere within `count-lines`. 
> Unfortunately the patch didn't seem to help though I'll double-check at 
> home (also, it seemed to improve baseline performance, but I'm not sure 
> if I should regard that as a random fluctuation, will need to do more 
> measurements).
> Also, this time even the first measurement with a small number of 
> overlays was extremely slow, not sure what to make of that.
> 
> In general however, the trend seems clear to me --- for this one file, 
> overlays hurt `line-number-at-pos` performance very much, except when 
> using the noverlay branch.
> Are there other things I should look for in my file that might 
> negatively affect performance (strange codepoints or anything else that 
> might perhaps upset `buf_charpos_to_bytepos`)?
> 
> baseline (emacs-27 master #667cdf42..., approx. 5-10 overlays created by 
> flycheck):
> iteration 1: 0.012696
> iteration 2: 0.002595
> iteration 3: 0.002606
> iteration 4: 0.002601
> iteration 5: 0.002649
> iteration 6: 0.002605
> iteration 7: 0.002594
> iteration 8: 0.002601
> iteration 9: 0.002603
> iteration 10: 0.002601
> iteration 11: 0.002606
> iteration 12: 0.002626
> iteration 13: 0.002603
> iteration 14: 0.002642
> iteration 15: 0.002599
> iteration 16: 0.002598
> iteration 17: 0.002600
> iteration 18: 0.002601
> iteration 19: 0.002599
> iteration 20: 0.002608
> 
> emacs-27 master, approx. 6200 overlays created by (test-highlight)
> iteration 1: 0.022795
> iteration 2: 0.015560
> iteration 3: 0.015697
> iteration 4: 0.015913
> iteration 5: 0.015894
> iteration 6: 0.016063
> iteration 7: 0.015928
> iteration 8: 0.015890
> iteration 9: 0.015278
> iteration 10: 0.015515
> iteration 11: 0.015327
> iteration 12: 0.015326
> iteration 13: 0.015574
> iteration 14: 0.015319
> iteration 15: 0.015370
> iteration 16: 0.015354
> iteration 17: 0.015333
> iteration 18: 0.015312
> iteration 19: 0.015481
> iteration 20: 0.015411
> 
> emacs-27 master + patch "distance+10", ~10 overlays created by flycheck:
> iteration 1: 0.002938
> iteration 2: 0.001132
> iteration 3: 0.000550
> iteration 4: 0.000468
> iteration 5: 0.000470
> iteration 6: 0.000451
> iteration 7: 0.000449
> iteration 8: 0.000451
> iteration 9: 0.000450
> iteration 10: 0.000449
> iteration 11: 0.000448
> iteration 12: 0.000452
> iteration 13: 0.000451
> iteration 14: 0.000443
> iteration 15: 0.000448
> iteration 16: 0.000443
> iteration 17: 0.000444
> iteration 18: 0.000445
> iteration 19: 0.000445
> iteration 20: 0.000445
> 
> emacs-27 master + patch "distance+10", ~6200 overlays created by 
> (test-highlight):
> iteration 1: 0.019673
> iteration 2: 0.014469
> iteration 3: 0.014491
> iteration 4: 0.014430
> iteration 5: 0.014493
> iteration 6: 0.014704
> iteration 7: 0.014741
> iteration 8: 0.014536
> iteration 9: 0.014433
> iteration 10: 0.014469
> iteration 11: 0.014429
> iteration 12: 0.014509
> iteration 13: 0.014484
> iteration 14: 0.014487
> iteration 15: 0.014524
> iteration 16: 0.014449
> iteration 17: 0.014501
> iteration 18: 0.014469
> iteration 19: 0.014429
> iteration 20: 0.014507
> 
> noverlay branch (#886933...), ~10 overlays:
> iteration 1: 0.002370
> iteration 2: 0.001191
> iteration 3: 0.001189
> iteration 4: 0.001162
> iteration 5: 0.001095
> iteration 6: 0.001096
> iteration 7: 0.000522
> iteration 8: 0.000531
> iteration 9: 0.000369
> iteration 10: 0.000365
> iteration 11: 0.000365
> iteration 12: 0.000366
> iteration 13: 0.000365
> iteration 14: 0.000380
> iteration 15: 0.000367
> iteration 16: 0.000365
> iteration 17: 0.000365
> iteration 18: 0.000365
> iteration 19: 0.000363
> iteration 20: 0.000368
> 
> noverlay branch, ~6200 overlays:
> iteration 1: 0.001878
> iteration 2: 0.001831
> iteration 3: 0.001838
> iteration 4: 0.001826
> iteration 5: 0.001027
> iteration 6: 0.000722
> iteration 7: 0.000531
> iteration 8: 0.000479
> iteration 9: 0.000480
> iteration 10: 0.000479
> iteration 11: 0.000479
> iteration 12: 0.000479
> iteration 13: 0.000478
> iteration 14: 0.000479
> iteration 15: 0.000478
> iteration 16: 0.000478
> iteration 17: 0.000482
> iteration 18: 0.000479
> iteration 19: 0.000480
> iteration 20: 0.000476
>