[Differential] [Commented On] D30: merge: Removed sorting in casefolding detection, for a slight performance win

[alex_gaynor (Alex Gaynor)]

alex_gaynor added a comment.


  Going to note that I'm less confident in the correctness of the second change 
than I am the first.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D30

EMAIL PREFERENCES
  https://phab.mercurial-scm.org/settings/panel/emailpreferences/

To: alex_gaynor, durin42, dsp
Cc: quark, krbullock, dsp, mercurial-devel
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

[Differential] [Commented On] D30: merge: Removed sorting in casefolding detection, for a slight performance win

[alex_gaynor (Alex Gaynor)]

alex_gaynor added a comment.


  Not quite sure this is science, but 
  
  First set of numbers is `tip`, second is with my patch:
  
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
9.49 real10.23 user 3.13 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
4.74 real 7.59 user 2.73 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.45 real 7.32 user 2.66 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
4.83 real 7.50 user 2.66 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.64 real 7.55 user 2.69 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
5.09 real 7.87 user 2.67 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
5.39 real 8.13 user 2.74 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
5.27 real 8.00 user 2.70 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.45 real 7.68 user 2.58 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
4.93 real 7.72 user 2.61 sys
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.90 real 6.91 user 2.81 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
3.99 real 6.81 user 2.52 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.02 real 7.00 user 2.61 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
4.82 real 7.42 user 2.69 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.60 real 7.80 user 2.78 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
4.11 real 7.77 user 2.73 sys
~/p/mozilla-central ❯❯❯ time ~/projects/hg/hg up 
consolidate-sandbox-disable; time ~/projects/hg/hg up central
4324 files updated, 0 files merged, 491 files removed, 0 files unresolved
(activating bookmark consolidate-sandbox-disable)
4.24 real 7.74 user 2.77 sys
4654 files updated, 0 files merged, 161 files removed, 0 files unresolved
(leaving bookmark consolidate-sandbox-disable)
4.35 real 7.80 user 2.78 sys

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D30

EMAIL PREFERENCES
  

[Differential] [Commented On] D30: merge: Removed sorting in casefolding detection, for a slight performance win

[durin42 (Augie Fackler)]

durin42 added a comment.


  Ah, I see. Alex, do you have any idea what the performance numbers on this 
are like, anecdotally on your mozilla repo?

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D30

EMAIL PREFERENCES
  https://phab.mercurial-scm.org/settings/panel/emailpreferences/

To: alex_gaynor, durin42, dsp
Cc: quark, krbullock, dsp, mercurial-devel
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

[Differential] [Commented On] D30: merge: Removed sorting in casefolding detection, for a slight performance win

[quark (Jun Wu)]

quark added a comment.


  > Is it important that the error be deterministic?
  
  Good question. Deterministic error is just to make tests stable AFAIK. If the 
perf win is noticable, I personally prefer perf win. We can use `(glob)` and 
`(?)` in tests to make it support non-deterministic outputs.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D30

EMAIL PREFERENCES
  https://phab.mercurial-scm.org/settings/panel/emailpreferences/

To: alex_gaynor, durin42, dsp
Cc: quark, krbullock, dsp, mercurial-devel
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

[Differential] [Commented On] D30: merge: Removed sorting in casefolding detection, for a slight performance win

[alex_gaynor (Alex Gaynor)]

alex_gaynor added a comment.


  Is it important that the error be deterministic?
  
  I think we can improve the situation if you have exactly two collisions, by 
sorting `(f, foldmap[fold])` and `(lastfull, f)`, but I don't think it can 
easily be preserved for multiple collisions.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D30

EMAIL PREFERENCES
  https://phab.mercurial-scm.org/settings/panel/emailpreferences/

To: alex_gaynor, durin42, dsp
Cc: krbullock, dsp, mercurial-devel
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel