Allow to use with non UTF-8 locale by -Ku RUBYOPT.

Bump PKGREVISION.

(obache)

Bump PKGREVISION from default python change to 2.7.

(obache)

put an empty line around comments.

(obache)

tabfy

(obache)

Updated graphics/darktable to 1.0

(jakllsch)

Update to darktable-1.0.

changes since 0.9.3:
* new cameras supported
. Leica M9
. NX100/NX5/NX10/NX11
. Panasonic DMC-GX1
. Pentax K-r
. Canon Powershot S100
. Olympus XZ-1
. Olympus E-P3
. Sony DSLR A330
. Sony NEX-5N
. Canon EOS 1000D
. Canon EOS 600D
. Sony Alpha 390
. Fuji Finepix HS20EXR
* new and updated translations (we now have chinese!)
* new modules:
. shadows & highlights
. enhanced tone curve. now operates in a and b channels as well
* refactored modules:
. import
. snapshots (enable sliding separation line between before/after images)
. metadata
* new image cache
. faster concurrent access and insertion
. reduces needed memory
. more thumbnails stored on disk
. read embedded jpegs for creating thumbnails (faster folder import)
* increased general speed on sqlite3 (journaled, pagesize optimizations)
* reworked, modular ui
* keyboard shortcuts support - key accelerators (GSoC)
* unity launcher support (ubuntu)
* quicktool bar: exposure, presets and styles
* new color picker
* web gallery export now with next/prev buttons per image
* removed gconf: not used anymore, we have our own backend
* bugfixes

(jakllsch)

Set LICENSE. Quell a pkglint warning.

(wiz)

Updated security/py-cryptopp to 0.6.0.1206569328141510525648634803928199668821045408958

(gdt)

fix distinfo; patch-aa is gone

(gdt)

Set license (GPLv2, or boutique license not in pkgsrc/licenses).

(gdt)

Update to 0.6.0.1206569328141510525648634803928199668821045408958.
(Yes, that ridiculous version number really is what upstream calls it.)

No NEWS entry, but announcement includes:

2012-03-13  Zooko Wilcox-O'Hearn  <zooko@zooko.com>

      * src/pycryptopp/_version.py: release pycryptopp-0.6.0
      * add Ed25519 signatures (#75)
      * add XSalsa20 cipher (#40)
      * switch from darcs to git for revision control
      * pycryptopp version numbers now include a decimal encoding of *
      * reorganize the source tree and the version number generation
      * aesmodule.cpp: validate size of IV and throw exception if it
        is not 16 (#70)
      * fixed compile errors with gcc-4.7.0 (#78)
      * fixed compile errors concerning "CryptoPP::g_nullNameValuePairs" (#77)
      * suppress warnings from valgrind with new OpenSSL 1.0.1 on Fedora (#82)
      * raise Python exception instead of uncaught C++ exception
        (resulting in abort) when deserializing malformed RSA keys (#83)

(gdt)

fix build for proj-4.8.0

(plunky)

Updated misc/libreoffice3-bin to 3.5.1

(ryoon)

Update to 3.5.1

Changelog:
* Many bugfixes
* Many improvements
* Less Java dependency

In detail see:
http://wiki.documentfoundation.org/Releases/3.5.1/RC1
http://wiki.documentfoundation.org/Releases/3.5.1/RC2

(ryoon)

Fix this properly. Tested on NetBSD and SunOS.

(hans)

Fix error in setup.py which affects builing on Mac OS X.

(adam)

Changes 2.4.9:
* ldapobject.ReconnectLDAPObject.reconnect() now does kind of
  an internal locking to pause other threads while reconnecting
  is pending.
* Changes to bind- and startTLS-related operation methods of
  class ReconnectLDAPObject for more robustness
* New constant ldap.OPT_NAMES_DICT contains mapping from
  integer to variable name for all option-related constants.

(adam)

Added net/remmina version 0.9.3
Added net/remmina-plugins version 0.9.2

(asau)

+ remmina
+ remmina-plugins

(asau)

Import Remmina plugins 0.9.2 as net/remmina-plugins.

Plugins for Remmina, a remote desktop client written in GTK+.
Currently RDP, VNC, NX, XDMCP and SSH are supported.

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(asau)

Import Remmina 0.9.3 as net/remmina

Remmina is a remote desktop client written in GTK+, aiming to be
useful for system administrators and travellers, who need to
work with lots of remote computers in front of either large
monitors or tiny netbooks. Remmina supports multiple network
protocols in an integrated and consistant user interface.
Currently RDP, VNC, NX, XDMCP and SSH are supported.

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(asau)

Updated security/gnutls to 2.12.17

(adam)

Changes 2.12.17:
* libgnutls: Corrections in record packet parsing.
* libgnutls: Fixes in SRP authentication.
* libgnutls: Added function to force explicit reinitialization of PKCS 11
  modules. This is required on the child process after a fork.
* libgnutls: PKCS 11 objects that do not have ID no longer crash listing.
* API and ABI modifications: gnutls_pkcs11_reinit: Added

(adam)

print/cups-drivers-Magicolor5440DL: regenerate distinfo

For an unknown reason, the distinfo generated after patch-aa was last
modified was uncorrect, so it has been regenerated.

(marino)

Updated net/libquvi-scripts to 0.4.2nb1

(wiz)

Add patch fixing downloading some youtube URLs.
>From upstream.
Bump PKGREVISION.

(wiz)

fix BUILDLINK_INCDIRS & BUILDLINK_LIBDIRS - thanks wiz

(abs)

iodbc 3.52.7nb2 needed for conflict free life

(abs)

Also put libs in iodbc subdir, patch to ensure pkgconfigdir is not affected
bump PKGREVISION. Test built soprano

(abs)

Fix the symlink creation (if GTAR_PROGRAM_PREFIX != "g")

(fhajny)

Stop current memory usage being printed to stdout at process start if memory
ballooning if present.
Bump PKGREVISION

(sborrill)

Update CONFLICTS to iodbc<3.52.7nb1

(abs)

Set --includedir to avoid conflict with unixodbc, adjust BUILDLINK_INCDIRS

(abs)

Bump PKGREVISION from default python to 2.7.

(obache)

Bump PKGREVISION from default python to 2.7.

(obache)

Add thunderbird10 and thunderbird10-l10n

(ryoon)

Add firefox10 and firefox10-l10n

(ryoon)

Add xulrunner10

(ryoon)

Add recent changes.

(ryoon)

Import firefox10-l10n-10.0.3 as www/firefox10-l10n

* Sync with www/firefox10

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(ryoon)

Import thunderbird10-l10n-10.0.3 as mail/thunderbird10-l10n

* Sync with mail/thunderbird10

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(ryoon)

Improve MESSAGE

(ryoon)

Import thunderbird10-10.0.3 as mail/thunderbird10

* This is ESR (Extended Support Release) version.
* Fix security bugs

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(ryoon)

Update to 11.0

* Sync with www/thunderbird

(ryoon)

Import firefox10-10.0.3 as www/firefox10

* Switch to ESR branch
* Fix security bugs

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(ryoon)

Import xulrunner-10.0.3 as devel/xulrunner

* Successor of devel/xulrunner, xulrunner-10.0.2.
* This is ESR (Extended Support Release) version.
* Fix security bugs

Status:

Vendor Tag: TNF
Release Tags: pkgsrc-base

(ryoon)

Update to 11.0

* enigmail source is not provided for this version now.
* Switch to xulrunner 11 branch
* Fix security bugs
* Improve stability
* Many bug fixes

(ryoon)

Update to 11.0

* Improve MESSAGE
* Sync with firefox-11.0

(ryoon)

Bump PKGREVISION.

* buildlink to xulrunner10
* Try to fix build on DragonFly 3.0.1

(ryoon)

lang/ruby: Fix ruby193-base for DragonFly

The maintainers of ruby have changed the shared library naming scheme for
FreeBSD and DragonFly:
For ruby18,  it's libruby18.so.18  (last part = RUBY_VER)
For ruby19,  it's libruby19.so.19  (last part = RUBY_VER)
for ruby193, it's libruby193.so.191 (last part derived from API, not version)

The rubyversion.mk was never updated to reflect that, and as a result ruby
1.9.3 has never built on DragonFly.  This commit will allow
lang/ruby193-base package to build.

(marino)

Update to 4.9.3

* Extract from xulrunner-11.0
* Add some header files in buildlink.

(ryoon)

Update to 3.13.3

* Extract from xulrunner-11.0.

(ryoon)

Update to 11.0

* Follow devel/xulrunner update.

(ryoon)

Update to 11.0

* Switch to xulrunner 11 branch
* Fix security bugs
* Improve stability
* Many bug fixes

(ryoon)

fix a entry format (hi plunky)

(obache)

Fix definition of NaN on VAX.

(jklos)

Updated lang/gcc44 to 4.4.7

(sbd)

Update to GCC 4.4.7

GCC 4.4.7 is a bug-fix release containing fixes for regressions and serious
bugs in GCC 4.4.6.  This release marks the end of the maintainance of
the GCC 4.4 series.

This is the list of problem reports (PRs) from GCC's bug tracking system
that are known to be fixed in the 4.4.7 release. This list might not be
complete (that is, it is possible that some PRs that have been fixed are
not listed here).
http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2

(sbd)

More dragonfly cleanup.

(sbd)

fix PLIST

(plunky)

Add this package is marked NOT-FOR-DRAGONFLY remove dragonfly patches
and pre-configure.

Also remove MESSAGE_SUBST.

(sbd)

Pickup maintainership of gcc4 series packages.

This means I will also take care of adding newer gcc version.

(sbd)

pdfjam update.

(markd)

Update to 2.08

2.08: fixed a bug in one of the tests
2.07: two other common graphics formats (JPG and PNG) are now explicitly
allowed as input files (i.e., not only PDF files are allowed as inputs).
2.06: changed the pdfbook script to include --booklet true as the default
behaviour (thanks to Julien Bossert for this good suggestion).
2.05: changes to the pdfbook script [the --right-edge-binding option is
now redundant, and there's a new --short-edge option for binding along
the short edge of pages instead of the long edge (thanks to Marco Pessotto
for this)]. The --preamble option to pdfjam is enhanced, to allow multiple
instances which get concatenated. Also various minor corrections to man
pages.
2.04: various minor improvements suggested by Debian maintainers (thanks
to Eduard Bloch for these). The main things are: addition of the --version
option; liberalisation of pdfjam to allow files in JPEG format to be
specified as input, as well as PDF (I don't know why or if this might
work! but some people have said it does); tidying of the man files; and
more use of exec, to avoid forking.
2.03: fixed a bug which caused problems when your /bin/sh is the zsh shell;
fixed a bug which prevented the correct representation of many UTF-8
characters in pdfinfo data.

(markd)

Not MAKE_JOBS_SAFE.

(joerg)

Updated geography/proj to 4.8.0

(gdt)

Update to 4.8.0.  As a byproduct, update proj-datumgrid to 1.5, which
has an additional grid shift file.  Add a test target; regression
tests now pass at least on NetBSD/i386 5.1_STABLE.

4.8.0 Release Notes
-------------------

o Added the Natural Earth projection.

o Added HEALPIX, rHEALPIX and Icosahedral Snyder Equal Area projections.

o nad2bin now produces "CTable2" format grid shift files by default which
  are platform independent.

o nad2nad removed, use cs2cs for datum shift operations.

o projects.h no longer installed as a public include file.  Please try to
  only use proj_api.h.

o Add pj_get_spheroid_defn() accessor.

o Added an alternate version of pj_init() that takes a projCtx (execution
  context) structure to address multithreading issues with error management
  and to provide a support for application hookable error reporting and
  logging.

o Upgrade to EPSG 7.9.  Some changes in ideal datum selection.

o JNI bindings reworked, org.proj4.Projections deprecated in favor of
  org.proj4.PJ.

o Added preliminary vertical datum support.

o Fix various multithreading issues, particular in datum grid handling code.

o Added support for the +axis= option for alternate axis orientations as
  part of a coordinate system (used for TM South Orientated support).

o +proj=omerc implementatioin replaced with code from libproj4.  +rot_conv
  flag no longer works, and some coordinate systems (ie. Malaysian) will
  need to use +gamma instead.  "epsg" init file updated accordingly.

(gdt)

Updated lang/gcc46 to 4.6.3nb1

(wiz)

Pass proper configure arguments so that it is happy with GNU tools available
on NetBSD; removes two dependencies.
Unlimit before running tests, reduces test failures.
Add t-crtstuff to tmake_file on NetBSD as well.

gcc46 should work much better now on NetBSD.

All from Kai-Uwe Eckhardt in private mail.

Bump PKGREVISION.

(wiz)

Fix build on newer NetBSD.

(joerg)

Fix build with newer glib.

(joerg)

Don't install cat pages. Bump revision.

(joerg)

Don't install cat pages. Bump revision.

(joerg)

Remove cat page entries in PLIST

(joerg)

Fix linkage for native X11. Still broken for modular Xorg.

(joerg)

Not MAKE_JOBS_SAFE

(joerg)

Fix build on NetBSD-current, per patch from Kai-Uwe Eckhardt in PR 46045.
Set LICENSE while here.

(wiz)

configure script expects darwin-ppc-cc and not darwin-powerpc-cc.
'should be ok' joerg@

(cegger)

Add a buildlink3.mk

(abs)

Updated devel/snappy to 1.0.5

(abs)

Updated devel/snappy to 1.0.5

------------------------------------------------------------------------
r60 | snappy.mirrorbot@gmail.com | 2012-02-23 18:00:36 +0100 (Thu, 23 Feb 2012) | 57 lines

For 32-bit platforms, do not try to accelerate multiple neighboring
32-bit loads with a 64-bit load during compression (it's not a win).

The main target for this optimization is ARM, but 32-bit x86 gets
a small gain, too, although there is noise in the microbenchmarks.
It's a no-op for 64-bit x86. It does not affect decompression.

Microbenchmark results on a Cortex-A9 1GHz, using g++ 4.6.2 (from
Ubuntu/Linaro), -O2 -DNDEBUG -Wa,-march=armv7a -mtune=cortex-a9
-mthumb-interwork, minimum 1000 iterations:

  Benchmark            Time(ns)    CPU(ns) Iterations
  ---------------------------------------------------
  BM_ZFlat/0            1158277    1160000      1000 84.2MB/s  html (23.57 %)    [ +4.3%]
  BM_ZFlat/1          14861782  14860000      1000 45.1MB/s  urls (50.89 %)    [ +1.1%]
  BM_ZFlat/2            393595    390000      1000 310.5MB/s  jpg (99.88 %)    [ +0.0%]
  BM_ZFlat/3            650583    650000      1000 138.4MB/s  pdf (82.13 %)    [ +3.1%]
  BM_ZFlat/4            4661480    4660000      1000 83.8MB/s  html4 (23.55 %)  [ +4.3%]
  BM_ZFlat/5            491973    490000      1000 47.9MB/s  cp (48.12 %)      [ +2.0%]
  BM_ZFlat/6            193575    192678      1038 55.2MB/s  c (42.40 %)      [ +9.0%]
  BM_ZFlat/7              62343      62754      3187 56.5MB/s  lsp (48.37 %)    [ +2.6%]
  BM_ZFlat/8          17708468  17710000      1000 55.5MB/s  xls (41.34 %)    [ -0.3%]
  BM_ZFlat/9            3755345    3760000      1000 38.6MB/s  txt1 (59.81 %)    [ +8.2%]
  BM_ZFlat/10          3324217    3320000      1000 36.0MB/s  txt2 (64.07 %)    [ +4.2%]
  BM_ZFlat/11          10139932  10140000      1000 40.1MB/s  txt3 (57.11 %)    [ +6.4%]
  BM_ZFlat/12          13532109  13530000      1000 34.0MB/s  txt4 (68.35 %)    [ +5.0%]
  BM_ZFlat/13          4690847    4690000      1000 104.4MB/s  bin (18.21 %)    [ +4.1%]
  BM_ZFlat/14            830682    830000      1000 43.9MB/s  sum (51.88 %)    [ +1.2%]
  BM_ZFlat/15            84784      85011      2235 47.4MB/s  man (59.36 %)    [ +1.1%]
  BM_ZFlat/16          1293254    1290000      1000 87.7MB/s  pb (23.15 %)      [ +2.3%]
  BM_ZFlat/17          2775155    2780000      1000 63.2MB/s  gaviota (38.27 %) [+12.2%]

Core i7 in 32-bit mode (only one run and 100 iterations, though, so noisy):

  Benchmark            Time(ns)    CPU(ns) Iterations
  ---------------------------------------------------
  BM_ZFlat/0            227582    223464      3043 437.0MB/s  html (23.57 %)    [ +7.4%]
  BM_ZFlat/1            2982430    2918455        233 229.4MB/s  urls (50.89 %)    [ +2.9%]
  BM_ZFlat/2              46967      46658      15217 2.5GB/s  jpg (99.88 %)      [ +0.0%]
  BM_ZFlat/3            115298    114864      5833 783.2MB/s  pdf (82.13 %)    [ +1.5%]
  BM_ZFlat/4            913440    899743        778 434.2MB/s  html4 (23.55 %)  [ +0.3%]
  BM_ZFlat/5            110302    108571      7000 216.1MB/s  cp (48.12 %)      [ +0.0%]
  BM_ZFlat/6              44409      43372      15909 245.2MB/s  c (42.40 %)      [ +0.8%]
  BM_ZFlat/7              15713      15643      46667 226.9MB/s  lsp (48.37 %)    [ +2.7%]
  BM_ZFlat/8            2625539    2602230        269 377.4MB/s  xls (41.34 %)    [ +1.4%]
  BM_ZFlat/9            808884    811429        875 178.8MB/s  txt1 (59.81 %)    [ -3.9%]
  BM_ZFlat/10            709532    700000      1000 170.5MB/s  txt2 (64.07 %)    [ +0.0%]
  BM_ZFlat/11          2177682    2162162        333 188.2MB/s  txt3 (57.11 %)    [ -1.4%]
  BM_ZFlat/12          2849640    2840000        250 161.8MB/s  txt4 (68.35 %)    [ -1.4%]
  BM_ZFlat/13            849760    835476        778 585.8MB/s  bin (18.21 %)    [ +1.2%]
  BM_ZFlat/14            165940    164571      4375 221.6MB/s  sum (51.88 %)    [ +1.4%]
  BM_ZFlat/15            20939      20571      35000 196.0MB/s  man (59.36 %)    [ +2.1%]
  BM_ZFlat/16            239209    236544      2917 478.1MB/s  pb (23.15 %)      [ +4.2%]
  BM_ZFlat/17            616206    610000      1000 288.2MB/s  gaviota (38.27 %) [ -1.6%]

R=sanjay

------------------------------------------------------------------------
r59 | snappy.mirrorbot@gmail.com | 2012-02-21 18:02:17 +0100 (Tue, 21 Feb 2012) | 107 lines

Enable the use of unaligned loads and stores for ARM-based architectures
where they are available (ARMv7 and higher). This gives a significant
speed boost on ARM, both for compression and decompression.
It should not affect x86 at all.

There are more changes possible to speed up ARM, but it might not be
that easy to do without hurting x86 or making the code uglier.
Also, we de not try to use NEON yet.

Microbenchmark results on a Cortex-A9 1GHz, using g++ 4.6.2 (from Ubuntu/Linaro),
-O2 -DNDEBUG -Wa,-march=armv7a -mtune=cortex-a9 -mthumb-interwork:

Benchmark            Time(ns)    CPU(ns) Iterations
---------------------------------------------------
BM_UFlat/0            524806    529100        378 184.6MB/s  html            [+33.6%]
BM_UFlat/1            5139790    5200000        100 128.8MB/s  urls            [+28.8%]
BM_UFlat/2              86540      84166      1901 1.4GB/s  jpg              [ +0.6%]
BM_UFlat/3            215351    210176        904 428.0MB/s  pdf            [+29.8%]
BM_UFlat/4            2144490    2100000        100 186.0MB/s  html4          [+33.3%]
BM_UFlat/5            194482    190000      1000 123.5MB/s  cp              [+36.2%]
BM_UFlat/6              91843      90175      2107 117.9MB/s  c              [+38.6%]
BM_UFlat/7              28535      28426      6684 124.8MB/s  lsp            [+34.7%]
BM_UFlat/8            9206600    9200000        100 106.7MB/s  xls            [+42.4%]
BM_UFlat/9            1865273    1886792        106 76.9MB/s  txt1            [+32.5%]
BM_UFlat/10          1576809    1587301        126 75.2MB/s  txt2            [+32.3%]
BM_UFlat/11          4968450    4900000        100 83.1MB/s  txt3            [+32.7%]
BM_UFlat/12          6673970    6700000        100 68.6MB/s  txt4            [+32.8%]
BM_UFlat/13          2391470    2400000        100 203.9MB/s  bin            [+29.2%]
BM_UFlat/14            334601    344827        522 105.8MB/s  sum            [+30.6%]
BM_UFlat/15            37404      38080      5252 105.9MB/s  man            [+33.8%]
BM_UFlat/16            535470    540540        370 209.2MB/s  pb              [+31.2%]
BM_UFlat/17          1875245    1886792        106 93.2MB/s  gaviota          [+37.8%]
BM_UValidate/0        178425    179533      1114 543.9MB/s  html            [ +2.7%]
BM_UValidate/1        2100450    2000000        100 334.8MB/s  urls            [ +5.0%]
BM_UValidate/2          1039      1044    172413 113.3GB/s  jpg            [ +3.4%]
BM_UValidate/3          59423      59470      3363 1.5GB/s  pdf              [ +7.8%]
BM_UValidate/4        760716    766283        261 509.8MB/s  html4          [ +6.5%]
BM_ZFlat/0            1204632    1204819        166 81.1MB/s  html (23.57 %)  [+32.8%]
BM_ZFlat/1          15656190  15600000        100 42.9MB/s  urls (50.89 %)  [+27.6%]
BM_ZFlat/2            403336    410677        487 294.8MB/s  jpg (99.88 %)  [+16.5%]
BM_ZFlat/3            664073    671140        298 134.0MB/s  pdf (82.13 %)  [+28.4%]
BM_ZFlat/4            4961940    4900000        100 79.7MB/s  html4 (23.55 %)  [+30.6%]
BM_ZFlat/5            500664    501253        399 46.8MB/s  cp (48.12 %)    [+33.4%]
BM_ZFlat/6            217276    215982        926 49.2MB/s  c (42.40 %)      [+25.0%]
BM_ZFlat/7              64122      65487      3054 54.2MB/s  lsp (48.37 %)    [+36.1%]
BM_ZFlat/8          18045730  18000000        100 54.6MB/s  xls (41.34 %)    [+34.4%]
BM_ZFlat/9            4051530    4000000        100 36.3MB/s  txt1 (59.81 %)  [+25.0%]
BM_ZFlat/10          3451800    3500000        100 34.1MB/s  txt2 (64.07 %)  [+25.7%]
BM_ZFlat/11          11052340  11100000        100 36.7MB/s  txt3 (57.11 %)  [+24.3%]
BM_ZFlat/12          14538690  14600000        100 31.5MB/s  txt4 (68.35 %)  [+24.7%]
BM_ZFlat/13          5041850    5000000        100 97.9MB/s  bin (18.21 %)    [+32.0%]
BM_ZFlat/14            908840    909090        220 40.1MB/s  sum (51.88 %)    [+22.2%]
BM_ZFlat/15            86921      86206      1972 46.8MB/s  man (59.36 %)    [+42.2%]
BM_ZFlat/16          1312315    1315789        152 86.0MB/s  pb (23.15 %)    [+34.5%]
BM_ZFlat/17          3173120    3200000        100 54.9MB/s  gaviota (38.27%) [+28.1%]

The move from 64-bit to 32-bit operations for the copies also affected 32-bit x86;
positive on the decompression side, and slightly negative on the compression side
(unless that is noise; I only ran once):

Benchmark              Time(ns)    CPU(ns) Iterations
-----------------------------------------------------
BM_UFlat/0                86279      86140      7778 1.1GB/s  html            [ +7.5%]
BM_UFlat/1              839265    822622        778 813.9MB/s  urls          [ +9.4%]
BM_UFlat/2                9180      9143      87500 12.9GB/s  jpg            [ +1.2%]
BM_UFlat/3                35080      35000      20000 2.5GB/s  pdf              [+10.1%]
BM_UFlat/4              350318    345000      2000 1.1GB/s  html4            [ +7.0%]
BM_UFlat/5                33808      33472      21212 701.0MB/s  cp            [ +9.0%]
BM_UFlat/6                15201      15214      46667 698.9MB/s  c              [+14.9%]
BM_UFlat/7                4652      4651    159091 762.9MB/s  lsp            [ +7.5%]
BM_UFlat/8              1285551    1282528        538 765.7MB/s  xls            [+10.7%]
BM_UFlat/9              282510    281690      2414 514.9MB/s  txt1          [+13.6%]
BM_UFlat/10              243494    239286      2800 498.9MB/s  txt2          [+14.4%]
BM_UFlat/11              743625    740000      1000 550.0MB/s  txt3          [+14.3%]
BM_UFlat/12              999441    989717        778 464.3MB/s  txt4          [+16.1%]
BM_UFlat/13              412402    410076      1707 1.2GB/s  bin              [ +7.3%]
BM_UFlat/14              54876      54000      10000 675.3MB/s  sum            [+13.0%]
BM_UFlat/15                6146      6100    100000 660.8MB/s  man            [+14.8%]
BM_UFlat/16              90496      90286      8750 1.2GB/s  pb              [ +4.0%]
BM_UFlat/17              292650    292000      2500 602.0MB/s  gaviota        [+18.1%]
BM_UValidate/0            49620      49699      14286 1.9GB/s  html            [ +0.0%]
BM_UValidate/1          501371    500000      1000 1.3GB/s  urls            [ +0.0%]
BM_UValidate/2              232        227    3043478 521.5GB/s  jpg            [ +1.3%]
BM_UValidate/3            17250      17143      43750 5.1GB/s  pdf              [ -1.3%]
BM_UValidate/4          198643    200000      3500 1.9GB/s  html4            [ -0.9%]
BM_ZFlat/0              227128    229415      3182 425.7MB/s  html (23.57 %) [ -1.4%]
BM_ZFlat/1              2970089    2960000        250 226.2MB/s  urls (50.89 %) [ -1.9%]
BM_ZFlat/2                45683      44999      15556 2.6GB/s  jpg (99.88 %)    [ +2.2%]
BM_ZFlat/3              114661    113136      6364 795.1MB/s  pdf (82.13 %)  [ -1.5%]
BM_ZFlat/4              919702    914286        875 427.2MB/s  html4 (23.55%) [ -1.3%]
BM_ZFlat/5              108189    108422      6364 216.4MB/s  cp (48.12 %)  [ -1.2%]
BM_ZFlat/6                44525      44000      15909 241.7MB/s  c (42.40 %)    [ -2.9%]
BM_ZFlat/7                15973      15857      46667 223.8MB/s  lsp (48.37 %)  [ +0.0%]
BM_ZFlat/8              2677888    2639405        269 372.1MB/s  xls (41.34 %)  [ -1.4%]
BM_ZFlat/9              800715    780000      1000 186.0MB/s  txt1 (59.81 %) [ -0.4%]
BM_ZFlat/10              700089    700000      1000 170.5MB/s  txt2 (64.07 %) [ -2.9%]
BM_ZFlat/11            2159356    2138365        318 190.3MB/s  txt3 (57.11 %) [ -0.3%]
BM_ZFlat/12            2796143    2779923        259 165.3MB/s  txt4 (68.35 %) [ -1.4%]
BM_ZFlat/13              856458    835476        778 585.8MB/s  bin (18.21 %)  [ -0.1%]
BM_ZFlat/14              166908    166857      4375 218.6MB/s  sum (51.88 %)  [ -1.4%]
BM_ZFlat/15              21181      20857      35000 193.3MB/s  man (59.36 %)  [ -0.8%]
BM_ZFlat/16              244009    239973      2917 471.3MB/s  pb (23.15 %)  [ -1.4%]
BM_ZFlat/17              596362    590000      1000 297.9MB/s  gaviota (38.27%) [ +0.0%]

R=sanjay

------------------------------------------------------------------------
r58 | snappy.mirrorbot@gmail.com | 2012-02-11 23:11:22 +0100 (Sat, 11 Feb 2012) | 9 lines

Lower the size allocated in the "corrupted input" unit test from 256 MB
to 2 MB. This fixes issues with running the unit test on platforms with
little RAM (e.g. some ARM boards).

Also, reactivate the 2 MB test for 64-bit platforms; there's no good
reason why it shouldn't be.

R=sanjay

------------------------------------------------------------------------
r57 | snappy.mirrorbot@gmail.com | 2012-01-08 18:55:48 +0100 (Sun, 08 Jan 2012) | 2 lines

Minor refactoring to accomodate changes in Google's internal code tree.

------------------------------------------------------------------------
r56 | snappy.mirrorbot@gmail.com | 2012-01-04 14:10:46 +0100 (Wed, 04 Jan 2012) | 19 lines

Fix public issue r57: Fix most warnings with -Wall, mostly signed/unsigned
warnings. There are still some in the unit test, but the main .cc file should
be clean. We haven't enabled -Wall for the default build, since the unit test
is still not clean.

This also fixes a real bug in the open-source implementation of
ReadFileToStringOrDie(); it would not detect errors correctly.

I had to go through some pains to avoid performance loss as the types
were changed; I think there might still be some with 32-bit if and only if LFS
is enabled (ie., size_t is 64-bit), but for regular 32-bit and 64-bit I can't
see any losses, and I've diffed the generated GCC assembler between the old and
new code without seeing any significant choices. If anything, it's ever so
slightly faster.

This may or may not enable compression of very large blocks (>2^32 bytes)
when size_t is 64-bit, but I haven't checked, and it is still not a supported
case.

------------------------------------------------------------------------
r55 | snappy.mirrorbot@gmail.com | 2012-01-04 11:46:39 +0100 (Wed, 04 Jan 2012) | 6 lines

Add a framing format description. We do not have any implementation of this at
the current point, but there seems to be enough of a general interest in the
topic (cf. public bug #34).

R=csilvers,sanjay

------------------------------------------------------------------------
r54 | snappy.mirrorbot@gmail.com | 2011-12-05 22:27:26 +0100 (Mon, 05 Dec 2011) | 81 lines

Speed up decompression by moving the refill check to the end of the loop.

This seems to work because in most of the branches, the compiler can evaluate
���ip_limit_ - ip��� in a more efficient way than reloading ip_limit_ from memory
(either by already having the entire expression in a register, or reconstructing
it from ���avail���, or something else). Memory loads, even from L1, are seemingly
costly in the big picture at the current decompression speeds.

Microbenchmarks (64-bit, opt mode):

Westmere (Intel Core i7):

  Benchmark    Time(ns)    CPU(ns) Iterations
  --------------------------------------------
  BM_UFlat/0      74492      74491    187894 1.3GB/s  html      [ +5.9%]
  BM_UFlat/1      712268    712263      19644 940.0MB/s  urls    [ +3.8%]
  BM_UFlat/2      10591      10590    1000000 11.2GB/s  jpg      [ -6.8%]
  BM_UFlat/3      29643      29643    469915 3.0GB/s  pdf      [ +7.9%]
  BM_UFlat/4      304669    304667      45930 1.3GB/s  html4    [ +4.8%]
  BM_UFlat/5      28508      28507    490077 823.1MB/s  cp      [ +4.0%]
  BM_UFlat/6      12415      12415    1000000 856.5MB/s  c      [ +8.6%]
  BM_UFlat/7        3415      3415    4084723 1039.0MB/s  lsp    [+18.0%]
  BM_UFlat/8      979569    979563      14261 1002.5MB/s  xls    [ +5.8%]
  BM_UFlat/9      230150    230148      60934 630.2MB/s  txt1    [ +5.2%]
  BM_UFlat/10    197167    197166      71135 605.5MB/s  txt2    [ +4.7%]
  BM_UFlat/11    607394    607390      23041 670.1MB/s  txt3    [ +5.6%]
  BM_UFlat/12    808502    808496      17316 568.4MB/s  txt4    [ +5.0%]
  BM_UFlat/13    372791    372788      37564 1.3GB/s  bin      [ +3.3%]
  BM_UFlat/14      44541      44541    313969 818.8MB/s  sum    [ +5.7%]
  BM_UFlat/15      4833      4833    2898697 834.1MB/s  man    [ +4.8%]
  BM_UFlat/16      79855      79855    175356 1.4GB/s  pb        [ +4.8%]
  BM_UFlat/17    245845    245843      56838 715.0MB/s  gaviota [ +5.8%]

Clovertown (Intel Core 2):

  Benchmark    Time(ns)    CPU(ns) Iterations
  --------------------------------------------
  BM_UFlat/0      107911    107890    100000 905.1MB/s  html    [ +2.2%]
  BM_UFlat/1    1011237    1011041      10000 662.3MB/s  urls    [ +2.5%]
  BM_UFlat/2      26775      26770    523089 4.4GB/s  jpg      [ +0.0%]
  BM_UFlat/3      48103      48095    290618 1.8GB/s  pdf      [ +3.4%]
  BM_UFlat/4      437724    437644      31937 892.6MB/s  html4  [ +2.1%]
  BM_UFlat/5      39607      39600    358284 592.5MB/s  cp      [ +2.4%]
  BM_UFlat/6      18227      18224    768191 583.5MB/s  c      [ +2.7%]
  BM_UFlat/7        5171      5170    2709437 686.4MB/s  lsp    [ +3.9%]
  BM_UFlat/8    1560291    1559989      8970 629.5MB/s  xls    [ +3.6%]
  BM_UFlat/9      335401    335343      41731 432.5MB/s  txt1    [ +3.0%]
  BM_UFlat/10    287014    286963      48758 416.0MB/s  txt2    [ +2.8%]
  BM_UFlat/11    888522    888356      15752 458.1MB/s  txt3    [ +2.9%]
  BM_UFlat/12    1186600    1186378      10000 387.3MB/s  txt4    [ +3.1%]
  BM_UFlat/13    572295    572188      24468 855.4MB/s  bin    [ +2.1%]
  BM_UFlat/14      64060      64049    218401 569.4MB/s  sum    [ +4.1%]
  BM_UFlat/15      7264      7263    1916168 555.0MB/s  man    [ +1.4%]
  BM_UFlat/16    108853    108836    100000 1039.1MB/s  pb    [ +1.7%]
  BM_UFlat/17    364289    364223      38419 482.6MB/s  gaviota [ +4.9%]

Barcelona (AMD Opteron):

  Benchmark    Time(ns)    CPU(ns) Iterations
  --------------------------------------------
  BM_UFlat/0      103900    103871    100000 940.2MB/s  html    [ +8.3%]
  BM_UFlat/1    1000435    1000107      10000 669.5MB/s  urls    [ +6.6%]
  BM_UFlat/2      24659      24652    567362 4.8GB/s  jpg      [ +0.1%]
  BM_UFlat/3      48206      48193    291121 1.8GB/s  pdf      [ +5.0%]
  BM_UFlat/4      421980    421850      33174 926.0MB/s  html4  [ +7.3%]
  BM_UFlat/5      40368      40357    346994 581.4MB/s  cp      [ +8.7%]
  BM_UFlat/6      19836      19830    708695 536.2MB/s  c      [ +8.0%]
  BM_UFlat/7        6100      6098    2292774 581.9MB/s  lsp    [ +9.0%]
  BM_UFlat/8    1693093    1692514      8261 580.2MB/s  xls    [ +8.0%]
  BM_UFlat/9      365991    365886      38225 396.4MB/s  txt1    [ +7.1%]
  BM_UFlat/10    311330    311238      44950 383.6MB/s  txt2    [ +7.6%]
  BM_UFlat/11    975037    974737      14376 417.5MB/s  txt3    [ +6.9%]
  BM_UFlat/12    1303558    1303175      10000 352.6MB/s  txt4    [ +7.3%]
  BM_UFlat/13    517448    517290      27144 946.2MB/s  bin    [ +5.5%]
  BM_UFlat/14      66537      66518    210352 548.3MB/s  sum    [ +7.5%]
  BM_UFlat/15      7976      7974    1760383 505.6MB/s  man    [ +5.6%]
  BM_UFlat/16    103121    103092    100000 1097.0MB/s  pb    [ +8.7%]
  BM_UFlat/17    391431    391314      35733 449.2MB/s  gaviota [ +6.5%]

R=sanjay

------------------------------------------------------------------------
r53 | snappy.mirrorbot@gmail.com | 2011-11-23 12:14:17 +0100 (Wed, 23 Nov 2011) | 88 lines

Speed up decompression by making the fast path for literals faster.

We do the fast-path step as soon as possible; in fact, as soon as we know the
literal length. Since we usually hit the fast path, we can then skip the checks
for long literals and available input space (beyond what the fast path check
already does).

Note that this changes the decompression Writer API; however, it does not
change the ABI, since writers are always templatized and as such never
cross compilation units. The new API is slightly more general, in that it
doesn't hard-code the value 16. Note that we also take care to check
for len <= 16 first, since the other two checks almost always succeed
(so we don't want to waste time checking for them until we have to).

The improvements are most marked on Nehalem, but are generally positive
on other platforms as well. All microbenchmarks are 64-bit, opt.

Clovertown (Core 2):

  Benchmark    Time(ns)    CPU(ns) Iterations
  --------------------------------------------
  BM_UFlat/0      110226    110224    100000 886.0MB/s  html    [ +1.5%]
  BM_UFlat/1    1036523    1036508      10000 646.0MB/s  urls    [ -0.8%]
  BM_UFlat/2      26775      26775    522570 4.4GB/s  jpg      [ +0.0%]
  BM_UFlat/3      49738      49737    280974 1.8GB/s  pdf      [ +0.3%]
  BM_UFlat/4      446790    446792      31334 874.3MB/s  html4  [ +0.8%]
  BM_UFlat/5      40561      40562    350424 578.5MB/s  cp      [ +1.3%]
  BM_UFlat/6      18722      18722    746903 568.0MB/s  c      [ +1.4%]
  BM_UFlat/7        5373      5373    2608632 660.5MB/s  lsp    [ +8.3%]
  BM_UFlat/8    1615716    1615718      8670 607.8MB/s  xls    [ +2.0%]
  BM_UFlat/9      345278    345281      40481 420.1MB/s  txt1    [ +1.4%]
  BM_UFlat/10    294855    294855      47452 404.9MB/s  txt2    [ +1.6%]
  BM_UFlat/11    914263    914263      15316 445.2MB/s  txt3    [ +1.1%]
  BM_UFlat/12    1222694    1222691      10000 375.8MB/s  txt4    [ +1.4%]
  BM_UFlat/13    584495    584489      23954 837.4MB/s  bin    [ -0.6%]
  BM_UFlat/14      66662      66662    210123 547.1MB/s  sum    [ +1.2%]
  BM_UFlat/15      7368      7368    1881856 547.1MB/s  man    [ +4.0%]
  BM_UFlat/16    110727    110726    100000 1021.4MB/s  pb    [ +2.3%]
  BM_UFlat/17    382138    382141      36616 460.0MB/s  gaviota [ -0.7%]

Westmere (Core i7):

  Benchmark    Time(ns)    CPU(ns) Iterations
  --------------------------------------------
  BM_UFlat/0      78861      78853    177703 1.2GB/s  html      [ +2.1%]
  BM_UFlat/1      739560    739491      18912 905.4MB/s  urls    [ +3.4%]
  BM_UFlat/2        9867      9866    1419014 12.0GB/s  jpg      [ +3.4%]
  BM_UFlat/3      31989      31986    438385 2.7GB/s  pdf      [ +0.2%]
  BM_UFlat/4      319406    319380      43771 1.2GB/s  html4    [ +1.9%]
  BM_UFlat/5      29639      29636    472862 791.7MB/s  cp      [ +5.2%]
  BM_UFlat/6      13478      13477    1000000 789.0MB/s  c      [ +2.3%]
  BM_UFlat/7        4030      4029    3475364 880.7MB/s  lsp    [ +8.7%]
  BM_UFlat/8    1036585    1036492      10000 947.5MB/s  xls    [ +6.9%]
  BM_UFlat/9      242127    242105      57838 599.1MB/s  txt1    [ +3.0%]
  BM_UFlat/10    206499    206480      67595 578.2MB/s  txt2    [ +3.4%]
  BM_UFlat/11    641635    641570      21811 634.4MB/s  txt3    [ +2.4%]
  BM_UFlat/12    848847    848769      16443 541.4MB/s  txt4    [ +3.1%]
  BM_UFlat/13    384968    384938      36366 1.2GB/s  bin      [ +0.3%]
  BM_UFlat/14      47106      47101    297770 774.3MB/s  sum    [ +4.4%]
  BM_UFlat/15      5063      5063    2772202 796.2MB/s  man    [ +7.7%]
  BM_UFlat/16      83663      83656    167697 1.3GB/s  pb        [ +1.8%]
  BM_UFlat/17    260224    260198      53823 675.6MB/s  gaviota [ -0.5%]

Barcelona (Opteron):

  Benchmark    Time(ns)    CPU(ns) Iterations
  --------------------------------------------
  BM_UFlat/0      112490    112457    100000 868.4MB/s  html    [ -0.4%]
  BM_UFlat/1    1066719    1066339      10000 627.9MB/s  urls    [ +1.0%]
  BM_UFlat/2      24679      24672    563802 4.8GB/s  jpg      [ +0.7%]
  BM_UFlat/3      50603      50589    277285 1.7GB/s  pdf      [ +2.6%]
  BM_UFlat/4      452982    452849      30900 862.6MB/s  html4  [ -0.2%]
  BM_UFlat/5      43860      43848    319554 535.1MB/s  cp      [ +1.2%]
  BM_UFlat/6      21419      21413    653573 496.6MB/s  c      [ +1.0%]
  BM_UFlat/7        6646      6645    2105405 534.1MB/s  lsp    [ +0.3%]
  BM_UFlat/8    1828487    1827886      7658 537.3MB/s  xls    [ +2.6%]
  BM_UFlat/9      391824    391714      35708 370.3MB/s  txt1    [ +2.2%]
  BM_UFlat/10    334913    334816      41885 356.6MB/s  txt2    [ +1.7%]
  BM_UFlat/11    1042062    1041674      10000 390.7MB/s  txt3    [ +1.1%]
  BM_UFlat/12    1398902    1398456      10000 328.6MB/s  txt4    [ +1.7%]
  BM_UFlat/13    545706    545530      25669 897.2MB/s  bin    [ -0.4%]
  BM_UFlat/14      71512      71505    196035 510.0MB/s  sum    [ +1.4%]
  BM_UFlat/15      8422      8421    1665036 478.7MB/s  man    [ +2.6%]
  BM_UFlat/16    112053    112048    100000 1009.3MB/s  pb    [ -0.4%]
  BM_UFlat/17    416723    416713      33612 421.8MB/s  gaviota [ -2.0%]

R=sanjay

------------------------------------------------------------------------
r52 | snappy.mirrorbot@gmail.com | 2011-11-08 15:46:39 +0100 (Tue, 08 Nov 2011) | 5 lines

Fix public issue #53: Update the README to the API we actually open-sourced
with.

R=sanjay

------------------------------------------------------------------------
r51 | snappy.mirrorbot@gmail.com | 2011-10-05 14:27:12 +0200 (Wed, 05 Oct 2011) | 5 lines

In the format description, use a clearer example to emphasize that varints are
stored in little-endian. Patch from Christian von Roques.

R=csilvers

------------------------------------------------------------------------
r50 | snappy.mirrorbot@gmail.com | 2011-09-15 21:34:06 +0200 (Thu, 15 Sep 2011) | 4 lines

Release Snappy 1.0.4.

R=sanjay

------------------------------------------------------------------------
r49 | snappy.mirrorbot@gmail.com | 2011-09-15 11:50:05 +0200 (Thu, 15 Sep 2011) | 5 lines

Fix public issue #50: Include generic byteswap macros.
Also include Solaris 10 and FreeBSD versions.

R=csilvers

------------------------------------------------------------------------
r48 | snappy.mirrorbot@gmail.com | 2011-08-10 20:57:27 +0200 (Wed, 10 Aug 2011) | 5 lines

Partially fix public issue 50: Remove an extra comma from the end of some
enum declarations, as it seems the Sun compiler does not like it.

Based on patch by Travis Vitek.

------------------------------------------------------------------------
r47 | snappy.mirrorbot@gmail.com | 2011-08-10 20:44:16 +0200 (Wed, 10 Aug 2011) | 4 lines

Use the right #ifdef test for sys/mman.h.

Based on patch by Travis Vitek.

------------------------------------------------------------------------
r46 | snappy.mirrorbot@gmail.com | 2011-08-10 03:22:09 +0200 (Wed, 10 Aug 2011) | 6 lines

Fix public issue #47: Small comment cleanups in the unit test.

Originally based on a patch by Patrick Pelletier.

R=sanjay

------------------------------------------------------------------------
r45 | snappy.mirrorbot@gmail.com | 2011-08-10 03:14:43 +0200 (Wed, 10 Aug 2011) | 8 lines

Fix public issue #46: Format description said "3-byte offset"
instead of "4-byte offset" for the longest copies.

Also fix an inconsistency in the heading for section 2.2.3.
Both patches by Patrick Pelletier.

R=csilvers

------------------------------------------------------------------------
r44 | snappy.mirrorbot@gmail.com | 2011-06-28 13:40:25 +0200 (Tue, 28 Jun 2011) | 8 lines

Fix public issue #44: Make the definition and declaration of CompressFragment
identical, even regarding cv-qualifiers.

This is required to work around a bug in the Solaris Studio C++ compiler
(it does not properly disregard cv-qualifiers when doing name mangling).

R=sanjay

------------------------------------------------------------------------
r43 | snappy.mirrorbot@gmail.com | 2011-06-04 12:19:05 +0200 (Sat, 04 Jun 2011) | 7 lines

Correct an inaccuracy in the Snappy format description.
(I stumbled into this when changing the way we decompress literals.)

R=csilvers

Revision created by MOE tool push_codebase.

------------------------------------------------------------------------
r42 | snappy.mirrorbot@gmail.com | 2011-06-03 22:53:06 +0200 (Fri, 03 Jun 2011) | 50 lines

Speed up decompression by removing a fast-path attempt.

Whenever we try to enter a copy fast-path, there is a certain cost in checking
that all the preconditions are in place, but it's normally offset by the fact
that we can usually take the cheaper path. However, in a certain path we've
already established that "avail < literal_length", which usually means that
either the available space is small, or the literal is big. Both will disqualify
us from taking the fast path, and thus we take the hit from the precondition
checking without gaining much from having a fast path. Thus, simply don't try
the fast path in this situation -- we're already on a slow path anyway
(one where we need to refill more data from the reader).

I'm a bit surprised at how much this gained; it could be that this path is
more common than I thought, or that the simpler structure somehow makes the
compiler happier. I haven't looked at the assembler, but it's a win across
the board on both Core 2, Core i7 and Opteron, at least for the cases we
typically care about. The gains seem to be the largest on Core i7, though.
Results from my Core i7 workstation:

  Benchmark            Time(ns)    CPU(ns) Iterations
  ---------------------------------------------------
  BM_UFlat/0              73337      73091    190996 1.3GB/s  html      [ +1.7%]
  BM_UFlat/1            696379    693501      20173 965.5MB/s  urls    [ +2.7%]
  BM_UFlat/2              9765      9734    1472135 12.1GB/s  jpg      [ +0.7%]
  BM_UFlat/3              29720      29621    472973 3.0GB/s  pdf      [ +1.8%]
  BM_UFlat/4            294636    293834      47782 1.3GB/s  html4    [ +2.3%]
  BM_UFlat/5              28399      28320    494700 828.5MB/s  cp      [ +3.5%]
  BM_UFlat/6              12795      12760    1000000 833.3MB/s  c      [ +1.2%]
  BM_UFlat/7              3984      3973    3526448 893.2MB/s  lsp    [ +5.7%]
  BM_UFlat/8            991996    989322      14141 992.6MB/s  xls    [ +3.3%]
  BM_UFlat/9            228620    227835      61404 636.6MB/s  txt1    [ +4.0%]
  BM_UFlat/10            197114    196494      72165 607.5MB/s  txt2    [ +3.5%]
  BM_UFlat/11            605240    603437      23217 674.4MB/s  txt3    [ +3.7%]
  BM_UFlat/12            804157    802016      17456 573.0MB/s  txt4    [ +3.9%]
  BM_UFlat/13            347860    346998      40346 1.4GB/s  bin      [ +1.2%]
  BM_UFlat/14            44684      44559    315315 818.4MB/s  sum    [ +2.3%]
  BM_UFlat/15              5120      5106    2739726 789.4MB/s  man    [ +3.3%]
  BM_UFlat/16            76591      76355    183486 1.4GB/s  pb        [ +2.8%]
  BM_UFlat/17            238564    237828      58824 739.1MB/s  gaviota [ +1.6%]
  BM_UValidate/0          42194      42060    333333 2.3GB/s  html      [ -0.1%]
  BM_UValidate/1        433182    432005      32407 1.5GB/s  urls      [ -0.1%]
  BM_UValidate/2            197        196  71428571 603.3GB/s  jpg    [ +0.5%]
  BM_UValidate/3          14494      14462    972222 6.1GB/s  pdf      [ +0.5%]
  BM_UValidate/4        168444    167836      83832 2.3GB/s  html4    [ +0.1%]

R=jeff

Revision created by MOE tool push_codebase.

------------------------------------------------------------------------
r41 | snappy.mirrorbot@gmail.com | 2011-06-03 22:47:14 +0200 (Fri, 03 Jun 2011) | 43 lines

Speed up decompression by not needing a lookup table for literal items.

Looking up into and decoding the values from char_table has long shown up as a
hotspot in the decompressor. While it turns out that it's hard to make a more
efficient decoder for the copy ops, the literals are simple enough that we can
decode them without needing a table lookup. (This means that 1/4 of the table
is now unused, although that in itself doesn't buy us anything.)

The gains are small, but definitely present; some tests win as much as 10%,
but 1-4% is more typical. These results are from Core i7, in 64-bit mode;
Core 2 and Opteron show similar results. (I've run with more iterations
than unusual to make sure the smaller gains don't drown entirely in noise.)

  Benchmark            Time(ns)    CPU(ns) Iterations
  ---------------------------------------------------
  BM_UFlat/0              74665      74428    182055 1.3GB/s  html      [ +3.1%]
  BM_UFlat/1            714106    711997      19663 940.4MB/s  urls    [ +4.4%]
  BM_UFlat/2              9820      9789    1427115 12.1GB/s  jpg      [ -1.2%]
  BM_UFlat/3              30461      30380    465116 2.9GB/s  pdf      [ +0.8%]
  BM_UFlat/4            301445    300568      46512 1.3GB/s  html4    [ +2.2%]
  BM_UFlat/5              29338      29263    479452 801.8MB/s  cp      [ +1.6%]
  BM_UFlat/6              13004      12970    1000000 819.9MB/s  c      [ +2.1%]
  BM_UFlat/7              4180      4168    3349282 851.4MB/s  lsp    [ +1.3%]
  BM_UFlat/8            1026149    1024000      10000 959.0MB/s  xls    [+10.7%]
  BM_UFlat/9            237441    236830      59072 612.4MB/s  txt1    [ +0.3%]
  BM_UFlat/10            203966    203298      69307 587.2MB/s  txt2    [ +0.8%]
  BM_UFlat/11            627230    625000      22400 651.2MB/s  txt3    [ +0.7%]
  BM_UFlat/12            836188    833979      16787 551.0MB/s  txt4    [ +1.3%]
  BM_UFlat/13            351904    350750      39886 1.4GB/s  bin      [ +3.8%]
  BM_UFlat/14            45685      45562    308370 800.4MB/s  sum    [ +5.9%]
  BM_UFlat/15              5286      5270    2656546 764.9MB/s  man    [ +1.5%]
  BM_UFlat/16            78774      78544    178117 1.4GB/s  pb        [ +4.3%]
  BM_UFlat/17            242270    241345      58091 728.3MB/s  gaviota [ +1.2%]
  BM_UValidate/0          42149      42000    333333 2.3GB/s  html      [ -3.0%]
  BM_UValidate/1        432741    431303      32483 1.5GB/s  urls      [ +7.8%]
  BM_UValidate/2            198        197  71428571 600.7GB/s  jpg    [+16.8%]
  BM_UValidate/3          14560      14521    965517 6.1GB/s  pdf      [ -4.1%]
  BM_UValidate/4        169065    168671      83832 2.3GB/s  html4    [ -2.9%]

R=jeff

Revision created by MOE tool push_codebase.

------------------------------------------------------------------------
r40 | snappy.mirrorbot@gmail.com | 2011-06-03 00:57:41 +0200 (Fri, 03 Jun 2011) | 2 lines

Release Snappy 1.0.3.

------------------------------------------------------------------------
r39 | snappy.mirrorbot@gmail.com | 2011-06-02 20:06:54 +0200 (Thu, 02 Jun 2011) | 11 lines

Remove an unneeded goto in the decompressor; it turns out that the
state of ip_ after decompression (or attempted decompresion) is
completely irrelevant, so we don't need the trailer.

Performance is, as expected, mostly flat -- there's a curious ~3-5%
loss in the "lsp" test, but that test case is so short it is hard to say
anything definitive about why (most likely, it's some sort of
unrelated effect).

R=jeff

------------------------------------------------------------------------
r38 | snappy.mirrorbot@gmail.com | 2011-06-02 19:59:40 +0200 (Thu, 02 Jun 2011) | 52 lines

Speed up decompression by caching ip_.

It is seemingly hard for the compiler to understand that ip_, the current input
pointer into the compressed data stream, can not alias on anything else, and
thus using it directly will incur memory traffic as it cannot be kept in a
register. The code already knew about this and cached it into a local
variable, but since Step() only decoded one tag, it had to move ip_ back into
place between every tag. This seems to have cost us a significant amount of
performance, so changing Step() into a function that decodes as much as it can
before it saves ip_ back and returns. (Note that Step() was already inlined,
so it is not the manual inlining that buys the performance here.)

The wins are about 3-6% for Core 2, 6-13% on Core i7 and 5-12% on Opteron
(for plain array-to-array decompression, in 64-bit opt mode).

There is a tiny difference in the behavior here; if an invalid literal is
encountered (ie., the writer refuses the Append() operation), ip_ will now
point to the byte past the tag byte, instead of where the literal was
originally thought to end. However, we don't use ip_ for anything after
DecompressAllTags() has returned, so this should not change external behavior
in any way.

Microbenchmark results for Core i7, 64-bit (Opteron results are similar):

Benchmark            Time(ns)    CPU(ns) Iterations
---------------------------------------------------
BM_UFlat/0              79134      79110      8835 1.2GB/s  html      [ +6.2%]
BM_UFlat/1            786126    786096        891 851.8MB/s  urls    [+10.0%]
BM_UFlat/2              9948      9948      69125 11.9GB/s  jpg      [ -1.3%]
BM_UFlat/3              31999      31998      21898 2.7GB/s  pdf      [ +6.5%]
BM_UFlat/4            318909    318829      2204 1.2GB/s  html4    [ +6.5%]
BM_UFlat/5              31384      31390      22363 747.5MB/s  cp      [ +9.2%]
BM_UFlat/6              14037      14034      49858 757.7MB/s  c      [+10.6%]
BM_UFlat/7              4612      4612    151395 769.5MB/s  lsp    [ +9.5%]
BM_UFlat/8            1203174    1203007        582 816.3MB/s  xls    [+19.3%]
BM_UFlat/9            253869    253955      2757 571.1MB/s  txt1    [+11.4%]
BM_UFlat/10            219292    219290      3194 544.4MB/s  txt2    [+12.1%]
BM_UFlat/11            672135    672131      1000 605.5MB/s  txt3    [+11.2%]
BM_UFlat/12            902512    902492        776 509.2MB/s  txt4    [+12.5%]
BM_UFlat/13            372110    371998      1881 1.3GB/s  bin      [ +5.8%]
BM_UFlat/14            50407      50407      10000 723.5MB/s  sum    [+13.5%]
BM_UFlat/15              5699      5701    100000 707.2MB/s  man    [+12.4%]
BM_UFlat/16            83448      83424      8383 1.3GB/s  pb        [ +5.7%]
BM_UFlat/17            256958    256963      2723 684.1MB/s  gaviota [ +7.9%]
BM_UValidate/0          42795      42796      16351 2.2GB/s  html      [+25.8%]
BM_UValidate/1        490672    490622      1427 1.3GB/s  urls      [+22.7%]
BM_UValidate/2            237        237    2950297 499.0GB/s  jpg    [+24.9%]
BM_UValidate/3          14610      14611      47901 6.0GB/s  pdf      [+26.8%]
BM_UValidate/4        171973    171990      4071 2.2GB/s  html4    [+25.7%]

------------------------------------------------------------------------
r37 | snappy.mirrorbot@gmail.com | 2011-05-17 10:48:25 +0200 (Tue, 17 May 2011) | 10 lines

Fix the numbering of the headlines in the Snappy format description.

R=csilvers
DELTA=4  (0 added, 0 deleted, 4 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1906

------------------------------------------------------------------------
r36 | snappy.mirrorbot@gmail.com | 2011-05-16 10:59:18 +0200 (Mon, 16 May 2011) | 12 lines

Fix public issue #32: Add compressed format documentation for Snappy.
This text is new, but an earlier version from Zeev Tarantov was used
as reference.

R=csilvers
DELTA=112  (111 added, 0 deleted, 1 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1867

------------------------------------------------------------------------
r35 | snappy.mirrorbot@gmail.com | 2011-05-09 23:29:02 +0200 (Mon, 09 May 2011) | 12 lines

Fix public issue #39: Pick out the median runs based on CPU time,
not real time. Also, use nth_element instead of sort, since we
only need one element.

R=csilvers
DELTA=5  (3 added, 0 deleted, 2 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1799

------------------------------------------------------------------------
r34 | snappy.mirrorbot@gmail.com | 2011-05-09 23:28:45 +0200 (Mon, 09 May 2011) | 19 lines

Fix public issue #38: Make the microbenchmark framework handle
properly cases where gettimeofday() can stand return the same
result twice (as sometimes on GNU/Hurd) or go backwards
(as when the user adjusts the clock). We avoid a division-by-zero,
and put a lower bound on the number of iterations -- the same
amount as we use to calibrate.

We should probably use CLOCK_MONOTONIC for platforms that support
it, to be robust against clock adjustments; we already use Windows'
monotonic timers. However, that's for a later changelist.

R=csilvers
DELTA=7  (5 added, 0 deleted, 2 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1798

------------------------------------------------------------------------
r33 | snappy.mirrorbot@gmail.com | 2011-05-04 01:22:52 +0200 (Wed, 04 May 2011) | 11 lines

Fix public issue #37: Only link snappy_unittest against -lz and other autodetected
libraries, not libsnappy.so (which doesn't need any such dependency).

R=csilvers
DELTA=20  (14 added, 0 deleted, 6 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1710

------------------------------------------------------------------------
r32 | snappy.mirrorbot@gmail.com | 2011-05-04 01:22:33 +0200 (Wed, 04 May 2011) | 11 lines

Release Snappy 1.0.2, to get the license change and various other fixes into
a release.

R=csilvers
DELTA=239  (236 added, 0 deleted, 3 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1709

------------------------------------------------------------------------
r31 | snappy.mirrorbot@gmail.com | 2011-04-26 14:34:55 +0200 (Tue, 26 Apr 2011) | 15 lines

Fix public issue #30: Stop using gettimeofday() altogether on Win32,
as MSVC doesn't include it. Replace with QueryPerformanceCounter(),
which is monotonic and probably reasonably high-resolution.
(Some machines have traditionally had bugs in QPC, but they should
be relatively rare these days, and there's really no much better
alternative that I know of.)

R=csilvers
DELTA=74  (55 added, 19 deleted, 0 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1556

------------------------------------------------------------------------
r30 | snappy.mirrorbot@gmail.com | 2011-04-26 14:34:37 +0200 (Tue, 26 Apr 2011) | 11 lines

Fix public issue #31: Don't reset PATH in autogen.sh; instead, do the trickery
we need for our own build system internally.

R=csilvers
DELTA=16  (13 added, 1 deleted, 2 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1555

------------------------------------------------------------------------
r29 | snappy.mirrorbot@gmail.com | 2011-04-16 00:55:56 +0200 (Sat, 16 Apr 2011) | 12 lines

When including <windows.h>, define WIN32_LEAN_AND_MEAN first,
so we won't pull in macro definitions of things like min() and max(),
which can conflict with <algorithm>.

R=csilvers
DELTA=1  (1 added, 0 deleted, 0 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1485

------------------------------------------------------------------------
r28 | snappy.mirrorbot@gmail.com | 2011-04-11 11:07:01 +0200 (Mon, 11 Apr 2011) | 15 lines

Fix public issue #29: Write CPU timing code for Windows, based on GetProcessTimes()
instead of getursage().

I thought I'd already committed this patch, so that the 1.0.1 release already
would have a Windows-compatible snappy_unittest, but I'd seemingly deleted it
instead, so this is a reconstruction.

R=csilvers
DELTA=43  (39 added, 3 deleted, 1 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1295

------------------------------------------------------------------------
r27 | snappy.mirrorbot@gmail.com | 2011-04-08 11:51:53 +0200 (Fri, 08 Apr 2011) | 22 lines

Include C bindings of Snappy, contributed by Martin Gieseking.

I've made a few changes since Martin's version; mostly style nits, but also
a semantic change -- most functions that return bool in the C++ version now
return an enum, to better match typical C (and zlib) semantics.

I've kept the copyright notice, since Martin is obviously the author here;
he has signed the contributor license agreement, though, so this should not
hinder Google's use in the future.

We'll need to update the libtool version number to match the added interface,
but as of http://www.gnu.org/software/libtool/manual/html_node/Updating-version-info.html
I'm going to wait until public release.

R=csilvers
DELTA=238  (233 added, 0 deleted, 5 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1294

------------------------------------------------------------------------
r26 | snappy.mirrorbot@gmail.com | 2011-04-07 18:36:43 +0200 (Thu, 07 Apr 2011) | 13 lines

Replace geo.protodata with a newer version.

The data compresses/decompresses slightly faster than the old data, and has
similar density.

R=lookingbill
DELTA=1  (0 added, 0 deleted, 1 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1288

------------------------------------------------------------------------
r25 | snappy.mirrorbot@gmail.com | 2011-03-30 22:27:53 +0200 (Wed, 30 Mar 2011) | 12 lines

Fix public issue #27: Add HAVE_CONFIG_H tests around the config.h
inclusion in snappy-stubs-internal.h, which eases compiling outside the
automake/autoconf framework.

R=csilvers
DELTA=5  (4 added, 1 deleted, 0 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1152

------------------------------------------------------------------------
r24 | snappy.mirrorbot@gmail.com | 2011-03-30 22:27:39 +0200 (Wed, 30 Mar 2011) | 13 lines

Fix public issue #26: Take memory allocation and reallocation entirely out of the
Measure() loop. This gives all algorithms a small speed boost, except Snappy which
already didn't do reallocation (so the measurements were slightly biased in its
favor).

R=csilvers
DELTA=92  (69 added, 9 deleted, 14 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1151

------------------------------------------------------------------------
r23 | snappy.mirrorbot@gmail.com | 2011-03-30 22:25:09 +0200 (Wed, 30 Mar 2011) | 18 lines

Renamed "namespace zippy" to "namespace snappy" to reduce
the differences from the opensource code.  Will make it easier
in the future to mix-and-match third-party code that uses
snappy with google code.

Currently, csearch shows that the only external user of
"namespace zippy" is some bigtable code that accesses
a TEST variable, which is temporarily kept in the zippy
namespace.

R=sesse
DELTA=123  (18 added, 3 deleted, 102 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1150

------------------------------------------------------------------------
r22 | snappy.mirrorbot@gmail.com | 2011-03-29 00:17:04 +0200 (Tue, 29 Mar 2011) | 11 lines

Put back the final few lines of what was truncated during the
license header change.

R=csilvers
DELTA=5  (4 added, 0 deleted, 1 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1094

------------------------------------------------------------------------
r21 | snappy.mirrorbot@gmail.com | 2011-03-26 03:34:34 +0100 (Sat, 26 Mar 2011) | 20 lines

Change on 2011-03-25 19:18:00-07:00 by sesse

Replace the Apache 2.0 license header by the BSD-type license header;
somehow a lot of the files were missed in the last round.

R=dannyb,csilvers
DELTA=147  (74 added, 2 deleted, 71 changed)

Change on 2011-03-25 19:25:07-07:00 by sesse

Unbreak the build; the relicensing removed a bit too much (only comments
were intended, but I also accidentially removed some of the top lines of
the actual source).

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1072

------------------------------------------------------------------------
r20 | snappy.mirrorbot@gmail.com | 2011-03-25 17:14:41 +0100 (Fri, 25 Mar 2011) | 10 lines

Change Snappy from the Apache 2.0 to a BSD-type license.

R=dannyb
DELTA=328  (80 added, 184 deleted, 64 changed)

Revision created by MOE tool push_codebase.
MOE_MIGRATION=1061

(abs)