Go to file
Greg Price 480c8d1991 cpython: Optimize dynamic symbol tables, for a 6% speedup.
I took a close look at how Debian builds the Python interpreter,
because I noticed it ran substantially faster than the one in nixpkgs
and I was curious why.

One thing that I found made a material difference in performance was
this pair of linker flags (passed to the compiler):

    -Wl,-O1 -Wl,-Bsymbolic-functions

In other words, effectively the linker gets passed the flags:

    -O1 -Bsymbolic-functions

Doing the same thing in nixpkgs turns out to make the interpreter
run about 6% faster, which is quite a big win for such an easy
change.  So, let's apply it.

---

I had not known there was a `-O1` flag for the *linker*!
But indeed there is.

These flags are unrelated to "link-time optimization" (LTO), despite
the latter's name.  LTO means doing classic compiler optimizations
on the actual code, at the linking step when it becomes possible to
do them with cross-object-file information.  These two flags, by
contrast, cause the linker to make certain optimizations within the
scope of its job as the linker.

Documentation is here, though sparse:
  https://sourceware.org/binutils/docs-2.31/ld/Options.html

The meaning of -O1 was explained in more detail in this LWN article:
  https://lwn.net/Articles/192624/
Apparently it makes the resulting symbol table use a bigger hash
table, so the load factor is smaller and lookups are faster.  Cool.

As for -Bsymbolic-functions, the documentation indicates that it's a
way of saving lookups through the symbol table entirely.  There can
apparently be situations where it changes the behavior of a program,
specifically if the program relies on linker tricks to provide
customization features:
  https://bugs.launchpad.net/ubuntu/+source/xfe/+bug/644645
  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=637184#35
But I'm pretty sure CPython doesn't permit that kind of trick: you
don't load a shared object that tries to redefine some symbol found
in the interpreter core.

The stronger reason I'm confident using -Bsymbolic-functions is
safe, though, is empirical.  Both Debian and Ubuntu have been
shipping a Python built this way since forever -- it was introduced
for the Python 2.4 and 2.5 in Ubuntu "hardy", and Debian "lenny",
released in 2008 and 2009.  In those 12 years they haven't seen a
need to drop this flag; and I've been unable to locate any reports
of trouble related to it, either on the Web in general or on the
Debian bug tracker.  (There are reports of a handful of other
programs breaking with it, but not Python/CPython.)  So that seems
like about as thorough testing as one could hope for.

---

As for the performance impact: I ran CPython upstream's preferred
benchmark suite, "pyperformance", in the same way as described in
the previous commit.  On top of that commit's change, the results
across the 60 benchmarks in the suite are:

The median is 6% faster.

The middle half (aka interquartile range) is from 4% to 8% faster.

Out of 60 benchmarks, 3 come out slower, by 1-4%.  At the other end,
5 are at least 10% faster, and one is 17% faster.

So, that's quite a material speedup!  I don't know how big the
effect of these flags is for other software; but certainly CPython
tends to do plenty of dynamic linking, as that's how it loads
extension modules, which are ubiquitous in the stdlib as well as
popular third-party libraries.  So perhaps that helps explain why
optimizing the dynamic linker has such an impact.
2020-05-13 21:24:30 -07:00
.github .github/CODEOWNERS: remove myself from the Haskell code owners 2020-02-21 10:01:41 +01:00
doc Revert "make-tarball.nix: Restore the 'commit' field in packages.json" 2020-03-29 20:57:50 +02:00
lib mongodb: 3.4.10 -> 4.0.4 2020-03-26 14:02:48 +01:00
maintainers Merge pull request #83590 from asbachb/update-netbeans-11-3 2020-03-28 23:01:59 -04:00
nixos nixos/azure: clarify how users work in basic example 2020-03-29 13:56:55 -07:00
pkgs cpython: Optimize dynamic symbol tables, for a 6% speedup. 2020-05-13 21:24:30 -07:00
.editorconfig Revert ".version: remove final newline" 2018-04-28 14:23:13 +02:00
.gitattributes gitattributes: disable merge=union in all-packages 2018-03-27 11:03:03 -05:00
.gitignore Replace androidenv by new implementation 2018-12-18 21:16:06 +01:00
.version 20.09 is Nightingale 2020-02-10 14:14:18 -05:00
COPYING COPYING: include 2020 2020-01-11 15:17:22 -08:00
default.nix Fix local path to release notes in error message 2018-10-08 05:43:15 -05:00
flake.nix flake.nix: Add note 2020-02-10 16:36:53 +01:00
README.md readme: add link to community chat options 2020-02-11 23:25:10 -08:00

NixOS logo

Code Triagers badge Open Collective supporters

Nixpkgs is a collection of over 40,000 software packages that can be installed with the Nix package manager. It also implements NixOS, a purely-functional Linux distribution.

Manuals

  • NixOS Manual - how to install, configure, and maintain a purely-functional Linux distribution
  • Nixpkgs Manual - contributing to Nixpkgs and using programming-language-specific Nix expressions
  • Nix Package Manager Manual - how to write Nix expressions (programs), and how to use Nix command line tools

Community

Other Project Repositories

The sources of all official Nix-related projects are in the NixOS organization on GitHub. Here are some of the main ones:

Continuous Integration and Distribution

Nixpkgs and NixOS are built and tested by our continuous integration system, Hydra.

Artifacts successfully built with Hydra are published to cache at https://cache.nixos.org/. When successful build and test criteria are met, the Nixpkgs expressions are distributed via Nix channels.

Contributing

Nixpkgs is among the most active projects on GitHub. While thousands of open issues and pull requests might seem a lot at first, it helps consider it in the context of the scope of the project. Nixpkgs describes how to build over 40,000 pieces of software and implements a Linux distribution. The GitHub Insights page gives a sense of the project activity.

Community contributions are always welcome through GitHub Issues and Pull Requests. When pull requests are made, our tooling automation bot, OfBorg will perform various checks to help ensure expression quality.

The Nixpkgs maintainers are people who have assigned themselves to maintain specific individual packages. We encourage people who care about a package to assign themselves as a maintainer. When a pull request is made against a package, OfBorg will notify the appropriate maintainer(s). The Nixpkgs committers are people who have been given permission to merge.

Most contributions are based on and merged into these branches:

  • master is the main branch where all small contributions go
  • staging is branched from master, changes that have a big impact on Hydra builds go to this branch
  • staging-next is branched from staging and only fixes to stabilize and security fixes with a big impact on Hydra builds should be contributed to this branch. This branch is merged into master when deemed of sufficiently high quality

For more information about contributing to the project, please visit the contributing page.

Donations

The infrastructure for NixOS and related projects is maintained by a nonprofit organization, the NixOS Foundation. To ensure the continuity and expansion of the NixOS infrastructure, we are looking for donations to our organization.

You can donate to the NixOS foundation by using Open Collective:

License

Nixpkgs is licensed under the MIT License.

Note: MIT license does not apply to the packages built by Nixpkgs, merely to the files in this repository (the Nix expressions, build scripts, NixOS modules, etc.). It also might not apply to patches included in Nixpkgs, which may be derivative works of the packages to which they apply. The aforementioned artifacts are all covered by the licenses of the respective packages.