Version 28 (modified by rrnewton, 18 months ago) (diff)


Release plans for GHC 7.8


The planned timeline for 7.8 is to have a feature freeze on the 14 Sept 2013, followed by a period of frantic bug fixing (interrupted by ICFP), aiming for a release during November 2013.


We would like to fix all of the high and highest priority tickets in the 7.8.1 milestone, but there are currently a lot of them so this seems optimistic. Please feel free to take a ticket and help us!

Completed new features

The features already completed are documented in the release notes: docs/users_guide/7.8.1-notes.xml

Pending new features

The following new features are planned for 7.8. They are at varying degrees of completeness, and may not all make it in.

  • Nicolas Frisby is adding a few Core-to-Core optimisations. They usually slightly improve performance, but can sometimes make a big difference, both good and bad. They are off by default.
    • -flate-dmd will run demand analysis near the end of the pipeline. cf #7782
    • -fdicts-strict will make all dictionary arguments strict if they are certainly not part of a knot.
    • -ffun-to-thunk reverts a new behavior that prevents GHC from creating sharing (cf af12cf66d1a416a135cb98b86717aba2cd247e1a)

  • Pedro and Richard were working on tidying up the poly-kinded Typable, and propositional equality (gcast and friends) story.
  • Geoff Mainland is working generalising the support for SSE-like instructions.
  • Unboxed Booleans. Patches merged.. Author: Jan Stolarek. See #6135.
  • Austin Seipp would like to do some official ARMv7 binary releases with a working stage2 compiler and GHCi, but there are probably some bugs waiting here.
  • Dynamic GHCi (#3658). This is working in HEAD, and enabled if DYNAMIC_GHC_PROGRAMS=YES. Currently it's enabled by default if dynamic libraries are supported, except for FreeBSD and Windows. On FreeBSD the reason it's disabled is due to a bug in FreeBSD's rtld. This has been fixed, but we're waiting for the fix to make it into releases. This might be in time for 7.8, but certainly will be for 7.10. See #7819. On Windows, there are a couple of build time annoyances: -dynamic-too doesn't work on Windows, and linking takes a very long time when dynamic linking is used. There's no technical reason why it couldn't be enabled, though. The plan is/was to use dynamic GHCi on as many platforms as possible in 7.8, and to remove support for non-dynamic-ghci in HEAD soon after. See discussion in #8039, however.
  • Ryan Newton has added a larger set of atomic memory primops than were previously available. The code can be found on the atomics branch). The set of PrimOps may continue to expand slightly, but will definitely be ready for 7.8. In future work, Carter Schonwald will provide optimized (inline) versions of these PrimOps, at least for the LLVM backend.

Features that will definitely not make it