[Gnu-arch-users] Re: give us a hand with arch

[Miles Bader]

"Mark A. Flacy" <address@hidden> writes:
> My silly suggestion would be for someone (especially one of the someones
> who think that the single-file arch-id is a doubleplusgood idea) to make a
> branch and implement the single file version of .arch-ids.
> 
> Then run some tests and *see* if the Darkly Hinted Efficiency Improvements
> materialize.  Nothing convinces quite like hard data.

Um, which `Darkly Hinted Efficiency Improvements' are those?

I've done somewhat precise calculations of the potential space savings,
and it's quite substantial on some trees, especially those with lots of
smallish files.  Time-savings probably would be less noticable, because
of the recent inode-state-caching speedups.

Here's the meat of my earlier post on the subject:

----------------------------------------------------------------

Some highlights (for the current .arch-ids implementation):

  uClinux-dist       -- 143MB of id tag files, 40% of source size
  gcc                --  71MB of id tag files, 38% of source size
  linux-2.6.0-test2  --  62MB of id tag files, 30% of source size
  uClibc             -- only 7.6MB of id tag files, but 57% of source size!

I suppose ++pristine-trees dwarfs this, of course... :-/

linux-2.6.0-test2-moo
   arch id tag size:    60B
   block size:          4KB
   num files:         15952
   num dirs:            905
   av. files per dir:    17
   total source size: 203MB
   file average size:  13KB
   per-dir .arch-ids files:
      total overhead:    3.8MB
      overhead per file:  248B
      relative overhead:  1.9%
   per-file .arch-ids/* tag files:
      total overhead:     62MB
      overhead per file:   4KB
      relative overhead: 30.6%
      bloat factor:       16.5

emacs
   arch id tag size:    60B
   block size:          4KB
   num files:          2416
   num dirs:             57
   av. files per dir:    42
   total source size:  89MB
   file average size:  38KB
   per-dir .arch-ids files:
      total overhead:    344KB
      overhead per file:  145B
      relative overhead:  0.4%
   per-file .arch-ids/* tag files:
      total overhead:    9.4MB
      overhead per file:   4KB
      relative overhead: 10.5%
      bloat factor:       28.1

gcc
   arch id tag size:    60B
   block size:          4KB
   num files:         18426
   num dirs:            873
   av. files per dir:    21
   total source size: 188MB
   file average size:  10KB
   per-dir .arch-ids files:
      total overhead:    4.0MB
      overhead per file:  228B
      relative overhead:  2.1%
   per-file .arch-ids/* tag files:
      total overhead:     71MB
      overhead per file:   4KB
      relative overhead: 38.2%
      bloat factor:       18.0

binutils-2.13.2.1
   arch id tag size:    60B
   block size:          4KB
   num files:          4955
   num dirs:            126
   av. files per dir:    39
   total source size:  68MB
   file average size:  14KB
   per-dir .arch-ids files:
      total overhead:    648KB
      overhead per file:  133B
      relative overhead:  0.9%
   per-file .arch-ids/* tag files:
      total overhead:     19MB
      overhead per file:   4KB
      relative overhead: 28.1%
      bloat factor:       30.6

uClibc
   arch id tag size:    60B
   block size:          4KB
   num files:          1958
   num dirs:            154
   av. files per dir:    12
   total source size:  13MB
   file average size: 7.0KB
   per-dir .arch-ids files:
      total overhead:    644KB
      overhead per file:  336B
      relative overhead:  4.7%
   per-file .arch-ids/* tag files:
      total overhead:    7.6MB
      overhead per file:   4KB
      relative overhead: 57.4%
      bloat factor:       12.2

uClinux-dist-cvs
   arch id tag size:    60B
   block size:          4KB
   num files:         36669
   num dirs:           2685
   av. files per dir:    13
   total source size: 360MB
   file average size:  10KB
   per-dir .arch-ids files:
      total overhead:     10MB
      overhead per file:  311B
      relative overhead:  3.0%
   per-file .arch-ids/* tag files:
      total overhead:    143MB
      overhead per file:   4KB
      relative overhead: 39.7%
      bloat factor:       13.2

-Miles
-- 
I'm beginning to think that life is just one long Yoko Ono album; no rhyme
or reason, just a lot of incoherent shrieks and then it's over.  --Ian Wolff