GCC 4.0 was released yesterday. Read all about it.
GCC 4.0 Released
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
GCC 4.0 was released yesterday. Read all about it.
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
They were promised for 3.4 but don’t work in 3.4. Do they work now?
a wonderful day for the free software movement!
The package is broken.
If you try compiling it in the same gcc-4.0.0 dir doing a make bootstrap then you end up in errors.
You need to untar gcc-4.0.0 and then create a new dir named gcc bysides it and then enter the gcc dir and from within there.
../gcc-4.0.0/configure –prefix=/usr
make bootstrap
There are a few other errors comming up too but it continues compiling anyways stuff missing etc.
Before we get a bunch of complaints about the fact that most binaries generated by GCC 4.0 are only marginally faster (and some a bit slower) than those compiled with 3.4, let me point out a few things that I’ve gathered from casually browsing the GCC development lists. I’m neither a GCC contributor nor a compiler expert.
Prior to GCC 4.0, the implementation of optimizations was mostly language-specific; there was little or no integration of optimization techniques across all languages. The main goal of the 4.0 release is to roll out a new, unified optimization framework (Tree-SSA), and to begin converting the old, fragmented optimization strategies to the unified framework.
Major improvements to the quality of the generated code aren’t expected to arrive until later versions, when GCC contributors will have had a chance to really begin to leverage the new optimization infrastructure instead of just migrating to it.
So, although GCC 4.0 brings fairly dramatic benefits to compilation speed, the speed of generated binaries isn’t expected to be markedly better than 3.4; that latter speedup isn’t expected until later installments in the 4.x series.
> The package is broken.
> If you try compiling it in the same gcc-4.0.0 dir doing a
> make bootstrap then you end up in errors.
Gcc has always been that way. Same deal with glibc.
> Gcc has always been that way. Same deal with glibc.
Well I usually made quite good impressions with gcc..
untar’ing
cd’ing
./configure –prefix=/usr && make bootstrap
Usually worked like a charm but this time its quite broken. If you use SRCDIR == COMPILEDIR then it fails but if you create a different dir besides it SRCDIR != COMPILEDIR then it works, though I detected a bunch of error 1 failures in some of the subcomponents that get compiled, while the compile proceeds the stuff that failed stays failed (from the quick looks it looked like some documents or so).
Now the people spent so much time in creating gcc 4.0.0, even announced RC1 and RC2 but still they rushed it out in an urge and hurry without proper testing.
Thats sadly a bit disappointing.
To correct a note above. In gcc 3.x the optimizations were all done at the RTL level, which is something that looks lispish, but is very close to the assembly for a given machine However, with GCC 4.x and up there are optimizations done at the tree level, which looks more like the programming language and optimizations done at the RTL level, which is more like the machine the code will run on.
Read the installation docs,
YOU ARE SUPPOSED to use
OBJDIR != SOURCEDIR.
If you don’t do that, you are on your own.
So how well does this GCC conform to the C++ standard?
> Read the installation docs,
> YOU ARE SUPPOSED to use
> OBJDIR != SOURCEDIR.
> If you don’t do that, you are on your own.
There should then be a proper check for this like in glibc, advising one to use OBJDIR != SRCDIR, after compiling gcc OBJDIR == SRCDIR for years I expect it to work out of box as it should work out of the box anyways. OBJDIR != SRCDIR are usually only required if you want to build multiple instances of the same tool from within different directories. But otoh why are other errors there ?
OBJDIR is not supposed to be the same as SRCDIR for ages. I’m sure you never bothered to read the installation docs, and probably didn’t know to use “make bootstrap”. You probably didn’t even compile the stage? compilers with optimizations, either.
How long until some distro uses this? (like Ubuntu)
> I’m sure you never bothered to read the installation
> docs, and probably didn’t know to use “make bootstrap”.
The last time I read the Installation manuals it was advised to configure, make bootstrap and make install the stuff. I don’t see that anything else is required here. There is no need to compile a special stage or even add extra optimizations. What worked perfectly for the past 9 years should still work. I think it’s you who requires proper reading of the manuals and not me.
how fast does it compile compared to previous releases and other compilers?
how fast does the final code run?
Citing from gcc/doc/install.texi
First, we @strong{highly} recommend that GCC be built into a
separate directory than the sources which does @strong{not} reside
within the source tree. This is how we generally build GCC; building
where @var{srcdir} == @var{objdir} should still work, but doesn’t
get extensive testing; building where @var{objdir} is a subdirectory
of @var{srcdir} is unsupported.
No further comments.
> No further comments.
I was up to quote the same. And now read your own stuff what you have quoted:
> This is how we generally build GCC; building where
> srcdir == objdir should still work
From your own quote!!!!! While it might not be exessively tested and not supported its still mentioned that it should work.
As you read in the further quote that I made it’s still the old game.
tar xfjv gcc-4.0.0.tar.bz2
cd gcc-4.0.0
./configure –prefix=/usr
make bootstrap && make install
Even this is written in the further installation instructions as it was for the past years. I don’t need to compile some steps magic and other stuff and since I want to compile a native compiler for my own host system instructing make bootstrap is all there is.
Now go back in the corner and stop playing the expert. You obviously aren’t.
;———–
And now from further quote:
Installing GCC: Configuration
First, we highly recommend that GCC be built into a separate directory than the sources which does not reside within the source tree. This is how we generally build GCC; building where srcdir == objdir should still work, but doesn’t get extensive testing; building where objdir is a subdirectory of srcdir is unsupported.
Installing GCC: Building
Now that GCC is configured, you are ready to build the compiler and runtime libraries.
Some commands executed when making the compiler may fail (return a nonzero status) and be ignored by make. These failures, which are often due to files that were not found, are expected, and can safely be ignored.
Building a native compiler
For a native build issue the command make bootstrap. This will build the entire GCC system, which includes the following steps:
Well, Fedora is using it in their next release which will be out June 6th.
lighten up people
IIRC gcc is gnu software and accordind to the GPL:
NO WARRANTY
11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM “AS IS” WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
I guess that includes this unfortunate incident with the compilation process
.
anyway, have a nice day …
… and if it doesn’t work the way you want it, fix it.
> guess that includes this unfortunate incident with the
> compilation process and if it doesn’t work the way you
> want it, fix it.
Kalispera, while you are right it’s still a bad excuse.
would be great if the link is for a release note, not a binary download!
cheers,
Can’t get it to compile on NetBSD 2.0.2 following the instructions given..not too surprised though i have found there are almost always steps and dependecies left out the documentation. Ever try to get abiword to compile on a NetBSD system…what a hoot.
How do you do when you fail to compile abiword on NetBSD?
Since you seem to be a “non-developer” I assume you are using pkgsrc from one of the quaterly branches, which are stable and known good. Especially on i386, since it is the most used platform.
If you are not using pkgsrc, please explain why. It is the preferred way in NetBSD to get third party software installed, you know.
hmm.. kinda makes you wonder when Ubuntu is going to move support for 486 and Pentium 1 machines into a seperate distribution so the rest of us can get the benefit of an i686 compiled distribution. With the autovectorization in gcc 4.0 we should see some serious speed improvements.
QuantumG, I think you need to wait for 4.1 to see “serious speedup” from autovectorization… Read:
http://gcc.gnu.org/projects/tree-ssa/vectorization.html#status4.0
That’s what actually went into 4.0. Other stuffs should wait.
GCC 4.0 is a release goal for Breezy Badger (the next Ubuntu release).
I recall hearing a rumor that GCC 4 was supposed to finally get the Apple patches to allow Objective-C++ (Objective-C and C++ in the same function.) I didn’t see that anywhere in the change list.
That could be a pretty big deal for GNUStep: not only does it make Objective-C much more powerful and much easier to use, but it improves portability between OS X and everything else (since Cocoa has had it from the beginning.)
I don’t think the autovectorization stuff in 4.0 is really going to affect much. Look at all the enhancements to it going into 4.1:
http://gcc.gnu.org/wiki/Autovectorization%20Enhancements
ObjectiveC++ is not in 4.0:
http://gcc.gnu.org/wiki/What%20will%20be%20in%204.0
Therefore, GCC 4.0 will not be released without Objective-C++, unless something seems to have gone horribly awry. (Update) Things went horribly awry, so Objective-C++ has not made it into GCC 4.0.
any word on what “horribly awry” entails?
No idea.
Worked for me in 3.4.
How do you do when you fail to compile abiword on NetBSD?
Since you seem to be a “non-developer” I assume you are using pkgsrc from one of the quaterly branches, which are stable and known good. Especially on i386, since it is the most used platform.
If you are not using pkgsrc, please explain why. It is the preferred way in NetBSD to get third party software installed, you know.
Maybe because GCC 4.0 is not yet in pkgsrc? Anyway, the pkgsrc developers themselves also have to build things outside pkgsrc the first time(s).
If gcc/abiwords buildsystem is not self-contained (missing dependencies), then that’s their developers fault, not the pkgsrc/users fault.
too bad they don’t do it yet :/
Worked for me in 3.4.
You are lucky then.
Here is an extract from the offical GCC 3.4 page at http://gcc.gnu.org/gcc-3.4/changes.html :
There are some known defects in the current precompiled header implementation that will result in compiler crashes … Therefore, precompiled headers should be considered a “technology preview” in this release
And indeed I experienced those crashes.
So I really hope that precompiler headers work in 4.0 !
Hi!
Do anybody know the advantages of the other free C/C++ compiler over the GCC?
I mean for example
– Tiny C http://fabrice.bellard.free.fr/tcc/
– TenDRA http://www.tendra.org/
– The BSD-Compiler for FreeBSD, OpenBSD und NetBSD
– under Windows OpenWatcom http://www.openwatcom.org/
– under 64bit Computer on Linux Open64 http://open64.sourceforge.net/
I think, some advantages will the other compiler have. Otherwise nobody would working longer on the other compiler and all would working together on GCC.
But which are the advantages of the other free Compiler? Do anybody know more about it?
Have anybody performancetests and so?
Greatings
theuserbl
It would be nice to know what the Apple GCC 4.0 includes – did they merge and work on the autovectorisation for the PowerPC seperately, and will have those changes merged in at a later date (at GCC steering committees discretion)? I’m curious as to whether the GCC 4.0 improvements that Apple has instore will be noticeable from regular use.
theuserbl, I don’t know about others, but Tiny C’s advantage is clear. It’s FAST: about or more than 10 times faster than gcc without optimization, and still generates reasonably good code.
I have a program that generates rather large and messy C code (several megabytes), and TCC really helped in reducing edit-compile-run cycle.
“The BSD-Compiler for FreeBSD, OpenBSD und NetBSD ”
BSDs use gcc.
I’ve tried it on Fedora Core 4 test 2, and their GCC 4.0 build was able to compile C++ code quite a big faster than GCC 3.3. Hopefully precompiled headers work too, so developing C++ apps on Linux will be fun again.
Last time I checked, tcc allowed two things gcc does not
* use #!/usr/bin/tcc (script interpeter) line on the top of the source code -> writin small “scripts” in C
* supported runtime boundary checking.
#! is used only in interpreted languages, perl python bash.
gcc is a compiler, so no way to use it as an interpreter, why mixing compiler with interpreter? and really i see no advantages to use c as scripting language, seeing how low level is.
example, i prefer to learn perl and use its regex tools than script a c regex by hand.
according to me script c is misleading, in fact csh is no more wide used.
“supported runtime boundary checking.”
GCC 4 now support that with libmudflap, as well as memory leak checking, uninitialized memory accesses, etc.
See http://gcc.gnu.org/wiki/Mudflap%20Pointer%20Debugging
Is there a version for win32?
As usual, there will be a win32 version running with cygwin (http://www.cygwin.com) and with ming (http://www.mingw.org).
You can install cygwin or mingw/msys and build it yourself if you don’t want to wait until they release the binaries. MinGW is better if you want to do native windows development, as it provides the required headers to access the win32 api and doesn’t require a non-commercially redistributable dll at runtime.
You can select the 686/k7 etc optimized kernel from the package manager. By shipping with 383 by default they allow a great range of possible users
Got it to compile just fine. I had to compile it in a directory other than the root of the source. I compiled iton FC2 and NetBSD without issue and already recompile a few programs i use for server duties just for fun.
@ Mark Ping:
> So how well does this GCC conform to the C++ standard?
They *still* didn’t bother to implement ‘export’, so you still have to write templates in the header files instead of just declaring them there.
They claim that “no-one” requests that feature, which might be because *all* compiler vendors (except EDG) blissfully ignore that this has been part of the standard ever since 1998 – lest users actually demand it being implemented.