Commit the cleanup of the LLVM 1.4 release notes. This should clearly go

into the release :)

llvm-svn: 18599
This commit is contained in:
Chris Lattner 2004-12-07 08:04:13 +00:00
parent b95f8ab3ef
commit b5bb597405
1 changed files with 78 additions and 99 deletions

View File

@ -44,14 +44,14 @@
<p>This document contains the release notes for the LLVM compiler
infrastructure, release 1.4. Here we describe the status of LLVM, including any
known problems and bug fixes from the previous release. The most up-to-date
known problems and improvements from the previous release. The most up-to-date
version of this document can be found on the <a
href="http://llvm.cs.uiuc.edu/releases/1.4/">LLVM 1.4 web site</a>. If you are
not reading this on the LLVM web pages, you should probably go there because
this document may be updated after the release.</p>
<p>For more information about LLVM, including information about potentially more
current releases, please check out the <a href="http://llvm.cs.uiuc.edu">main
<p>For more information about LLVM, including information about the latest
release, please check out the <a href="http://llvm.cs.uiuc.edu">main LLVM
web site</a>. If you have questions or comments, the <a
href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM developer's mailing
list</a> is a good place to send them.</p>
@ -73,9 +73,18 @@ href="http://llvm.cs.uiuc.edu/releases/">releases page</a>.</p>
<p>This is the fifth public release of the LLVM compiler infrastructure.</p>
<p> At this time, LLVM is known to correctly compile and run all C &amp; C++
SPEC CPU95 &amp; 2000 benchmarks, the Olden benchmarks, and the Ptrdist
benchmarks, and <b>many</b> other C and C++ programs.</p>
<p> At this time, LLVM is known to correctly compile and a broad range of
C and C++ programs, including the SPEC CPU95 &amp; 2000 suite. This release
includes several major enhancements to the LLVM system, including a new
PowerPC JIT, enhancements to the C/C++ front-end to provide source line number
information in LLVM, a new <a href="CommandGuide/html/llvmc.html">compiler
driver</a>, and several other enhancements listed below. It also includes
bug fixes for those problems found since the 1.3 release.</p>
<p>Note that this release seperates the LLVM Program Testsuite out of the
main LLVM distribution into a seperate CVS repository and tarball. This
reduces the size of the main LLVM distribution. Also note that LLVM now
builds tools into llvm/Debug/bin by default instead of llvm/tools/Debug.</p>
</div>
@ -86,71 +95,69 @@ benchmarks, and <b>many</b> other C and C++ programs.</p>
<div class="doc_text">
<ol>
<li>LLVM now includes a JIT for the PowerPC target.</li>
<li>LLVM now optimizes global variables significantly more than it did
before.
<li>LLVM now includes a Just-In-Time compiler for the PowerPC target.</li>
<li>llvmgcc and llvmg++ now emit source line number information when '-g' is
passed in. This information can be used with llvm-db or other tools and
passes.</li>
<li>The test/Programs hierarchy <a href="http://llvm.cs.uiuc.edu/PR257">has
been moved out of the main LLVM tree</a> into a separate CVS repository and
tarball. This shrinks the distribution size of LLVM itself significantly.
</li>
<li>LLVM now optimizes global variables more aggressively than it did
before.</li>
<li>LLVM now includes the new '<tt>undef</tt>' value and
<a href="LangRef.html#i_unreachable"><tt>unreachable</tt></a> instruction,
which give the optimizer more information about the behavior of the
program.
</li>
<li>llvmgcc and llvmg++ now emit source line number information when '-g' is
passed in. This information can be used with llvm-db or other tools and
passes.
</li>
<li>The test/Programs hierarchy <a href="http://llvm.cs.uiuc.edu/PR257">has
been moved out of the main LLVM tree</a> into a separate CVS repository and
tarball. This shrinks the distribution size of LLVM itself significantly.
</li>
<li>Bytecode compression with bzip2 has been implemented. All bytecode files
generated by LLVM will now be compressed by default. Compression can be
disabled with the <tt>-disable-compression</tt> option to the tools that can
generate bytecode files.
</li>
<li>A generic <a href="http://llvm.cs.uiuc.edu/PR353">compiler driver</a> and
an associated <a href="CommandGuide/html/llvm-ld.html">generic linker</a> have
been implemented. The compiler driver is generic because it can be configured
<li>A generic <a href="CommandGuide/html/llvmc.html">compiler driver</a>
(llvmc) and
an associated <a href="CommandGuide/html/llvm-ld.html">generic linker</a>
(llvm-ld) have been added. The compiler driver is generic because it can be
configured
to pre-process, translate, optimize, assemble, and link code from any source
language. This aids compiler writers because all that is needed is a
source-to-bytecode or source-to-assembly translator and a configuration file.
The linker is generic because it allows dynamically loadable optimization
modules to be executed for link-time optimization. Language specific
link-time optimization modules can be created and executed automatically.
language with an LLVM front-end. This makes it easier for compiler writers
to hide the multiple steps required to compile a program (compiling,
optimizing, linking runtime libraries, etc) in one simple command.
</li>
<li>The <a href="http://llvm.cs.uiuc.edu/PR263">dependent libraries</a>
feature has been implemented. This allows front end compilers to indicate in
the bytecode which libraries the bytecode needs to be linked with. Both the
C/C++ front end and Stacker support generating the required libraries. The
Linker now supports using this information to ensure required libaries are
linker now supports using this information to ensure required libaries are
linked into the module. This minimizes the need to use the <tt>-l</tt> option
when using <a href="CommandGuide/html/llvmc.html"><tt>llvmc</tt></a>
</li>
<li>The LLVM makefiles have been improved to build LLVM faster (2x) and
includes new targets (like dist-check, uninstall). One important change is
associated with <a href="http://llvm.cs.uiuc.edu/PR456">PR456</a>. The
libraries and tools will now be built into <tt>$builddir/Debug/{bin,lib}</tt>
instead of <tt>$builddir/tools/Debug</tt> and <tt>$builddir/lib/Debug</tt>.
Similarly for <tt>Release</tt> and <tt>Profile</tt> builds.
<li>The LLVM makefiles have been improved to build LLVM much faster and
includes new targets (like dist-check, uninstall). One important user-visible
change is that libraries and tools will now be built into
<tt>$builddir/Debug/{bin,lib}</tt>
instead of <tt>$builddir/tools/Debug</tt> and <tt>$builddir/lib/Debug</tt>
(Similarly for <tt>Release</tt> and <tt>Profile</tt> builds).
</li>
<li>The LLVM source code is much more compatible with Microsoft Visual C++,
including the JIT and runtime-code generation, though the entire system
may not work with it.
</li>
<li>The target-to-JIT interfaces <a href="http://llvm.cs.uiuc.edu/PR283">are
<li>The JIT-Target interfaces <a href="http://llvm.cs.uiuc.edu/PR283">are
now much simpler</a> and more powerful.
</li>
<li>LLVM now provides llvm-ar and llvm-ranlib tools for working with archives
of LLVM bytecode files.</li>
<li>zlib and libpng are <a href="http://llvm.cs.uiuc.edu/PR417">no longer
included in the main LLVM tarball</a>.</li>
<li>The LLVM code generator now generates asm writers for the target from
an abstract target description, instead of requiring them to be hand
written.</li>
<li>The LLVM code generator now automatically generates assembly code writers
from an abstract target descriptions, eliminating the need to write
assembly printers manually.</li>
<li>LLVM regression and feature tests can now be run with DejaGNU.</li>
<li>llvmgcc and llvmg++ now emit source-level line number information, making
it possible to map from LLVM code back to source. This is currently used
by llvm-db.</li>
<li>Floating point intensive programs on X86 systems run much faster
with the LLC code generator and JIT than in 1.3.</li>
</ol>
</div>
@ -187,14 +194,12 @@ issues were fixed:</a>
<div class="doc_text">
<ol>
<li><a href="http://llvm.cs.uiuc.edu/PR409">
[core/asmparser] ConstantFP::isValueValidForType Broken</a>
<li>The linker no longer <a href="http://llvm.cs.uiuc.edu/PR426">emits many
useless warnings</a> when linking C++ programs.</a>
</li>
<li><a href="http://llvm.cs.uiuc.edu/PR426">[llvmg++] Tons of warnings
are spewed when linking to libstdc++</a>
</li>
<li><a href="http://llvm.cs.uiuc.edu/PR352">include/{Support,Config} ->
include/llvm/{Support,Config}</a>
<li>The LLVM <a href="http://llvm.cs.uiuc.edu/PR352">#include namespace</a>
has been made consistent. Files in <tt>llvm/include/{Support,Config}</tt>
are now located in <tt>llvm/include/llvm/{Support,Config}</tt>.</a>
</li>
<li>The names of the libraries generated by compiling LLVM source have been
changed to ensure they do not conflict with other packages upon installation.
@ -202,23 +207,13 @@ issues were fixed:</a>
the library <tt>libasmparser.a</tt> in 1.3 has become
<tt>libLLVMAsmParser.a</tt> in release 1.4.
</li>
<li><a href="http://llvm.cs.uiuc.edu/PR459">[llvmg++] C++ frontend is expanding
lots of unused inline functions</a></li>
</ol>
<li>The C++ frontend no longer expands and emits <a
href="http://llvm.cs.uiuc.edu/PR459">all inline functions, even if they
are unused</a>. It now properly tracks which functions are needed and
only compiles those.</li>
</div>
<!--=========================================================================-->
<div class="doc_subsubsection">
In this release, the following build problems were fixed:
</div>
<div class="doc_text">
<ol>
<li><a href="http://llvm.cs.uiuc.edu/PR256">[autoconf] further standardizing
autoconf usage</a>. Various improvements in the configure.ac script were
made as well as the makefile system.
</li>
<li>Many improvements in the <a href="http://llvm.cs.uiuc.edu/PR256">autoconf
and makefile systems</a> have been implemented.</li>
</ol>
</div>
@ -230,9 +225,12 @@ improvements:</a>
<div class="doc_text">
<ol>
<li><a href="http://llvm.cs.uiuc.edu/PR362">Ugly code generated for
std::min/std::max</a>
</li>
<li>The optimizer produces <a href="http://llvm.cs.uiuc.edu/PR362">more
efficient code for std::min/std::max</a> and other similar functions.</li>
<li>The X86 backend generates substantially faster code for floating point
intensive programs.</li>
<li>The PowerPC backend generates more efficient code in many common
scenarios.</li>
</ol>
</div>
@ -247,10 +245,6 @@ were fixed:</a>
<p>Bugs fixed in the LLVM Core:</p>
<ol>
<li><a href="http://llvm.cs.uiuc.edu/PR139">[Linker] gccld does not link
objects/archives in order specified on the command line</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR227">[X86] llc output for functions
w/certain names tickles GNU 'as' bugs</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR420">[licm] LICM invalidates alias
analysis info and uses broken information</a> (optimizer crash)</li>
<li><a href="http://llvm.cs.uiuc.edu/PR422">[asmwriter] Asmwriter is really
@ -259,6 +253,8 @@ were fixed:</a>
completely broken in LLVM 1.3</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR430">[bcwriter] Empty compaction
tables defined</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR227">[X86] llc output for functions
w/certain names tickles GNU 'as' bugs</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR472">[cbackend] Static globals are
prototyped as 'extern'</a></li>
</ol>
@ -274,7 +270,7 @@ were fixed:</a>
bitfield which does not increase struct size</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR424">[llvmgcc] llvmgcc emits invalid
constant exprs</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR421">[llvmg++] Crash in initializing
<li><a href="http://llvm.cs.uiuc.edu/PR421">[llvmg++] Crash initializing
array with constructors in hard EH situations</a></li>
<li><a href="http://llvm.cs.uiuc.edu/PR397">[llvm-gcc] Inline function
redefinitions error due to 'asm' function rename</a></li>
@ -309,8 +305,9 @@ were fixed:</a>
<li>Intel and AMD machines running Red Hat Linux and FreeBSD (and probably
other unix-like systems).</li>
<li>Sun UltraSPARC workstations running Solaris 8.</li>
<li>Intel and AMD machines running on Win32 with the Cygwin libraries.</li>
<li>PowerPC-based Mac OS X boxes, running 10.2 and above.</li>
<li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
support is available for native builds with Visual C++).</li>
<li>PowerPC-based Mac OS X systems, running 10.2 and above.</li>
</ul>
<p>The core LLVM infrastructure uses
@ -319,6 +316,7 @@ to the machine and operating system on which it is built. However, minor
porting may be required to get LLVM to work on new platforms. We welcome your
portability patches and reports of successful builds or error messages.</p>
<!--
<p>Note that the LLVM build system does not currently support directories with
spaces on them when running on Win32/cygwin. We strongly recommend running
LLVM and the C frontend out of a top-level directory without spaces (e.g.,
@ -327,6 +325,7 @@ cygwin packages. By default, many important tools are not installed that
are needed by the LLVM build process or test suite (e.g., /bin/time). Finally,
please make sure that there are no directories with spaces in them in your
PATH environment variable.</p>
-->
</div>
@ -360,11 +359,11 @@ useful to some people. In particular, if you would like to work on one of these
components, please contact us on the llvmdev list.</p>
<ul>
<li>The following passes are incomplete or buggy: <tt>-pgmdep, -memdep,
-ipmodref, -cee, -branch-combine, -instloops, -paths</tt></li>
<li>The <tt>-pre</tt> pass is incomplete (there are cases it doesn't handle that
it should) and not thoroughly tested.</li>
<li>The <tt>llvm-db</tt> tool is in a very early stage of development.</li>
<li>The following passes are incomplete or buggy, and may be removed in future
releases: <tt>-pgmdep, -memdep, -ipmodref, -cee, -branch-combine,
-instloops, -paths, -pre</tt></li>
<li>The <tt>llvm-db</tt> tool is in a very early stage of development, but can
be used to step through programs and inspect the stack.</li>
<li>The "iterative scan" register allocator (enabled with -regalloc=iterativescan)
is not stable.</li>
</ul>
@ -385,22 +384,10 @@ components, please contact us on the llvmdev list.</p>
such, execution of a threaded program could cause these data structures to be
corrupted.
</li>
<li>Linking in static archive files (.a files) is slow by default because
there is no symbol table in the archive. To remedy this, run
<a href="CommandGuide/html/llvm-ranlib.html"><tt>llvm-ranlib</tt></a> on the
archive to add an LLVM symbol table.
</li>
<li>The gccld program <a href="http://llvm.cs.uiuc.edu/PR139">does not link
objects/archives in the order specified on the command line.</a>
</li>
<li><a href="http://llvm.cs.uiuc.edu/PR240">The lower-invoke pass does not
mark values live across a setjmp as volatile</a>. This missing feature
only affects targets whose setjmp/longjmp libraries do not save and restore
the entire register file.</li>
<li><a href="http://llvm.cs.uiuc.edu/PR427">[bytecode] Assertion on V1
Bytecode Files</a>. This bug won't be fixed because V1 bytecode had its own
problems, no one is using V1 bytecode any more, and the fix is non-trivial.
</li>
</ul>
</div>
@ -426,9 +413,6 @@ components, please contact us on the llvmdev list.</p>
<li>Initialization of global union variables can only be done <a
href="http://llvm.cs.uiuc.edu/PR162">with the largest union member</a>.</li>
<li><a href="http://llvm.cs.uiuc.edu/PR244">[llvm-gcc] Error when an implicitly
external function is re-declared as static</a></li>
</ul>
</div>
@ -581,10 +565,9 @@ lists, please let us know (also including whether or not they work).</p>
<div class="doc_text">
<p>For this release, the C++ front-end is considered to be fully functional but
has not been tested as thoroughly as the C front-end. It has been tested and
works for a number of non-trivial programs, but there may be lurking bugs.
Please report any bugs or problems.</p>
<p>For this release, the C++ front-end is considered to be fully. It has been
tested and works for a number of non-trivial programs, including LLVM
itself.</p>
</div>
@ -696,10 +679,6 @@ Analysis rules. As such, special options may be necessary to compile the code
(for example, GCC requires the <tt>-fno-strict-aliasing</tt> option). This
problem probably cannot be fixed.</li>
<li><a href="http://llvm.cs.uiuc.edu/PR33">Initializers for global variables</a>
cannot include special floating point numbers like Not-A-Number or
Infinity.</li>
<li><a href="http://llvm.cs.uiuc.edu/PR56">Zero arg vararg functions are not
supported</a>. This should not affect LLVM produced by the C or C++
frontends.</li>