And I guess 1 means a lot more work but would be the better option apart
from that?!
The compiler is able to cope. Have made some updates to the D runtime
and Phobos library to weed out any issues relating to such platforms
(NaCL is another system which uses 32bit pointers on x86_64 arch),
Should go into upstream druntime shouldn't it?
http://d.puremagic.com/issues/show_bug.cgi?id=8148
gdmd [a lot of sources] -lib -odgenerated -oflibscid.a
... which if run, produces the "generated" subdir but places the output
file libscid.a in the current directory. This happens consistently,
even if the "generated" dir already exists.
https://bitbucket.org/goshawk/gdc/issue/283/gdm
gdmd -O -release -inline
produces a much faster executable (about twice as fast) as the typical
gcc-style options I would use,
gdc -O2 [or -O3]
Can anyone advise on appropriate gdc options to pick up the same speed
level?
Basically -frelease is missing.
Use -vdmd with gdmd to see t
I'll have to check tonight. Can you reduce these down to minimal test
cases?
Think it's done.
6 new issues opened.
DustMite gogogo.
Yep that's what I'll do next ;-)
Hmm indeed, lots of crashes when building the unittests.
Reducing std.container ICE now.
One failure seems related to #307, but there are some unrelated ones.
Here's the output (using GCC 4.6.2):
http://pastebin.com/PtNtTHG9
I'll have to check tonight. Can you reduce these down to minimal test
cases?
DustMite gogogo.
Which distro?
Is TLS support and/or a workaround for those crashes something
that's on the radar?
It's a known issue and being worked on I think.
You don't overwrite anything.
The install instructions suggest /opt/gdc as the install dir.
File a bug report. And remember to reduce it (perhaps with DustMite).
Things on this list will just get lost.
This page is in some sore need of updating. :)
http://prowiki.org/wiki4d/wiki.cgi?GdcHacking
Aren't you the only one who really knows the internals?
If you want cross-module inlining, compile all sources in one command.
gdc file1.d file2.d -o myapp
So 'gdc -c file1.d' won't inline small functions from file2? Why?
On Tuesday, 10 January 2012 at 01:52:57 UTC, Andrej Mitrovic
wrote:
Turns out i was right to be worried about ebp - see the vtst()
function below...
On DMD head I get an ICE, yay!
Internal error: backend/cod3.c 446
Like this?:
http://d.puremagic.com/issues/show_bug.cgi?id=7254
Just tested http://code.google.com/p/dgui/
gdc doesn't understand backslashes in response files.
And there are some strange errors that don't occur with dmd 2.057
even though it's frontend stuff. Maybe someone else could have a
try.
On Wednesday, 4 January 2012 at 21:40:28 UTC, Iain Buclaw wrote:
If I recall correctly, there are GCC binding to LLVM, are there
not?
The post 4.2 one is DragonEgg.
It's a gcc plugin that replaces GCC's optimizers and code
generators with those from llvm.
If I include additional modules on the compilation command line like
this:
gdc test.d /opt/gdc/include/d2/4.6.2/std/internal/uni.d
/opt/gdc/include/d2/4.6.2/std/internal/uni_tab.d
where test.d is my program, all symbols resolve OK.
Does this indicate a problem in the building of libgphobos
If I had more experience I'd try to build one myself, only I just got
GDC to compile a month ago and I still don't know how to install it
(like the LDC is installed)
I don't understand. Building it is the harder part (though imho it's not
hard at all on Linux).
Just run make install or make
Am 08.12.2011, 20:04 Uhr, schrieb Iain Buclaw :
Will also want to get things moving fast on 64bit support passing
testsuite in same manner as 32bit currently does. Now I have a new
capable machine, is all test test test to get that sorted out.
Need help with testing?
God I never even noticed
iain@dm1-oneiric:~/src/gdc/src$ hg status
? d/dmd2/cppmangle.c
? d/phobos2/std/internal/processinit.d
? d/phobos2/std/internal/windows/advapi32.d
I shall never learn my lesson of checking and adding new files before
upload. :)
Also had to re-setup gdc.
btw, is there a way to do that other th
dsimcha wrote:
make -j 16
-.-
make[2]: *** No rule to make target `d/cppmangle.dmd.o', needed by
`cc1d'. Stop.
make[2]: Leaving directory
`/var/tmp/gdcStuff/gdc/dev/gcc-4.6.1/objdir/gcc'
Same here. gcc 4.6.2
Why is D1 still the default?
Because this is the first release where I felt D2 was capable of being
the default and I forgot about it until writing the post. It also
requires some reworking of the changes that enable dual compilers.
but why is there a zip version anyway?
I posted with a zi
Thx!
* Release includes both D versions.
* -v1(default) compiles for D1.
* -v2 compiles for D2.
* The switch must be used for linking as well.
Why is D1 still the default?
* 64bit and up to date GDC forthcoming.
Imho the 64 bit version is even more important since gdc is the o
This did not help with my particular issue.
For some reason I'm getting those cross-compile errors.
Are there any CFLAGS you use?
I've tried building w/gcc-4.6.1 and gcc-4.4.6.
No. I only needed to set these 2 variables to make it work on Ubuntu 11.10
export $CPATH=/usr/include/x86_64-linux-gnu
Of course it's
export CPATH=/usr/include/x86_64-linux-gnu
export LIBRARY_PATH=/usr/lib/x86_64-linux-gnu
export $CPATH=/usr/include/x86_64-linux-gnu
I had similar problems on debian testing x86_64. Apparently some files
have been moved and this causes those problems. For example,
/usr/include/asm/errno.h is
now at /usr/include/x86_64-linux-gnu/asm/errno.h. Plain gcc failed to
compile too.
I tried messing around with symlinks and environme
Looks like you can't even compile zlib.
You might try downloading zlib and compiling it to check whether that
works at all.
I tried to follow the instructions as closely as possible and get
[...]
This doesn't show the real error, it should be somewhere above in the log.
https://bitbucket.org/goshawk/gdc/wiki/Home#!installation
For example, if I am creating a binary .deb or .rpm, I can compile
everything I want so that it expects to run in '/usr' (prefix), but dump
the files in '/home/vnayar/myproject/rpm' (DESTDIR).
After I run my command, I'll have stuff like:
/home/vnayar/myproject/rpm/
+ usr/bin/mybin
+ us
* default branch. DMD 1.067/2.052
* Contains both D1(default/-v1) and D2(-v2).
Thx for your work!
Btw the default branch already contains 2.054.
What about Iain's ppa: https://launchpad.net/~ibuclaw/+archive/ppa
// Assume __v4sf is defined by the compiler
pragma(set_attribute, _mm_add_ps, always_inline, artificial);
__v4sf _mm_add_ps (__v4sf __A, __v4sf __B)
{
return __builtin_ia32_addps(__A, __B);
}
2 notes:
Isn't it pragma(GNU_set_attribute?
And you should be able to do
pragma(GNU_attribute, alwa
I've got a working cross compiler now. The changes for phobos were
rather small, I'll do some testing whether my changes actually work and
if everything is fine I'll send those patches to the phobos mailing list.
The second gcc patch wasn't needed, and the default gc compiled fine.
So the ru
I wonder how much work it would be to support OpenMP in gdc seeing how most of
the magic actually happens in that libgomp library. Could that lib be used at
all (problems with GC etc?)
38 matches
Mail list logo
