Skip to content
GitLab
Menu
Projects
Groups
Snippets
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Sign in / Register
Toggle navigation
Menu
Open sidebar
Glasgow Haskell Compiler
GHC
Commits
b94d8b0b
Commit
b94d8b0b
authored
Sep 09, 2010
by
Simon Marlow
Browse files
updates to the release notes
parent
9fdce342
Changes
1
Hide whitespace changes
Inline
Side-by-side
docs/users_guide/6.14.1-notes.xml
View file @
b94d8b0b
...
...
@@ -13,10 +13,70 @@
<itemizedlist>
<listitem>
<para>
Due to changes in the runtime system, version 2.0 or later
(preferably version 2.2 or later) of the
<literal>
parallel
</literal>
library should be used, or
parallelism will be lost.
GHC now defaults to the Haskell 2010 language standard.
</para>
<para>
Libraries are not quite so straightforward. By default, GHC
provides access to the
<literal>
base
</literal>
package,
which includes the Haskell 2010 libraries, albeit with a few
minor differences. For those who want to write strictly
standards-conforming code we also provide
the
<literal>
haskell2010
</literal>
package which provides
the precise APIs specified by Haskell 2010, but because the
module names in this package overlap with those in
the
<literal>
base
</literal>
package it is not possible to
use both
<literal>
haskell2010
</literal>
and
<literal>
base
</literal>
at the same time (this also
applies to the
<literal>
array
</literal>
package). Hence to use
the Haskell 2010 libraries you should hide
the
<literal>
base
</literal>
and
<literal>
array
</literal>
packages, for example with GHCi:
<screen>
$ ghci -package haskell2010 -hide-package base -hide-package array
</screen>
If you are using Cabal it isn't necessary to
hide
<literal>
base
</literal>
and
<literal>
array
</literal>
explicitly, just don't include them in your
<literal>
build-depends
</literal>
.
</para>
</listitem>
<listitem>
<para>
On POSIX platforms, there is a new I/O manager based on
epoll/kqueue/poll, which allows multithreaded I/O code to
scale to a much larger number (100k+) of threads.
</para>
</listitem>
<listitem>
<para>
The inliner has been overhauled, which should in general
give better performance while reducing unnecessary code-size
explosion.
</para>
</listitem>
<listitem>
<para>
Large parts of the runtime system have been overhauled, in
particular the machinery related to blocking and wakeup of
threads and exception throwing (
<literal>
throwTo
</literal>
).
Several instances of pathological performance have been
fixed, especially where large numbers of threads are
involved.
</para>
</listitem>
<listitem>
<para>
Due to changes in the runtime system, if you are
using
<literal>
Control.Parallel.Strategies
</literal>
from
the
<literal>
parallel
</literal>
package, please upgrade to
at least version 2 (preferably version 3). The
implementation of Strategies
in
<literal>
parallel-1.x
</literal>
will lose parallelism
with GHC 6.14.1.
</para>
</listitem>
...
...
@@ -213,14 +273,6 @@ import SpecConstr
</para>
</listitem>
<listitem>
<para>
On POSIX platforms, the RTS now has a new I/O manager based
on epoll/kqueue/poll, which allows it to scale to a much
larger number (100k+) of open file descriptors.
</para>
</listitem>
<listitem>
<para>
The RTS now exports a function
<literal>
setKeepCAFs
</literal>
...
...
@@ -386,7 +438,7 @@ import SpecConstr
<listitem>
<para>
The registerised via-C backend, and the
<literal>
-fvia-C
</literal>
, have been deprecated. The poor
<literal>
-fvia-C
</literal>
flag
, have been deprecated. The poor
floating-point performance in the x86 native code generator
has now been fixed, so we don't believe there is still any
reason to use the via-C backend.
...
...
@@ -459,7 +511,7 @@ import SpecConstr
<itemizedlist>
<listitem>
<para>
GHCi now understands
the
layout
of
multiline
-
commands, so
GHCi now understands layout
in
multi
-
line
commands, so
this now works:
</para>
<programlisting>
...
...
Write
Preview
Supports
Markdown
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment