www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Hot for dmd 64bit

reply dwilson <darenw darenscotwilson.com> writes:
Beside praying and pestering, what can we D non-experts do to help get a stable
64-bit dmd available?

Killer D features are strings, slick built in dynamics arrays, no headers files
to keep in sync, and the
other nice features often praised by others.   I'm not sure yet that D is my
favorite language, but it's in
the list of top three.

Killing D (at least for me) is the limit choices for compiling on 64-bit Linux
with D2 and preferably Phobos
instead of Tango.  My setup, for reasons I haven't investigated deeply, can't
run 32-bit anything, and I do
intend to work on huge arrays of data, a few GB in RAM.  As for Phobos, it's
obviously more Mars-related
than "Tango" :)
Jan 26 2011
next sibling parent Daniel Gibson <metalcaedes gmail.com> writes:
Am 27.01.2011 07:32, schrieb dwilson:

 Beside praying and pestering, what can we D non-experts do to help get a
stable 64-bit dmd available?
 
 Killer D features are strings, slick built in dynamics arrays, no headers
files to keep in sync, and the
 other nice features often praised by others.   I'm not sure yet that D is my
favorite language, but it's in
 the list of top three.
 
 Killing D (at least for me) is the limit choices for compiling on 64-bit Linux
with D2 and preferably Phobos
 instead of Tango.  My setup, for reasons I haven't investigated deeply, can't
run 32-bit anything, and I do
 intend to work on huge arrays of data, a few GB in RAM.  As for Phobos, it's
obviously more Mars-related
 than "Tango" :)
 
 


If you're using Linux you could test GDC (
https://bitbucket.org/goshawk/gdc/wiki/Home ) or LDC2 (
https://bitbucket.org/prokhin_alexey/ldc2 ).

Cheers,
- Daniel
Jan 26 2011
prev sibling next sibling parent Brad Roberts <braddr puremagic.com> writes:
On 1/26/2011 10:32 PM, dwilson wrote:

 Beside praying and pestering, what can we D non-experts do to help get a
stable 64-bit dmd available?
 
 Killer D features are strings, slick built in dynamics arrays, no headers
files to keep in sync, and the
 other nice features often praised by others.   I'm not sure yet that D is my
favorite language, but it's in
 the list of top three.
 
 Killing D (at least for me) is the limit choices for compiling on 64-bit Linux
with D2 and preferably Phobos
 instead of Tango.  My setup, for reasons I haven't investigated deeply, can't
run 32-bit anything, and I do
 intend to work on huge arrays of data, a few GB in RAM.  As for Phobos, it's
obviously more Mars-related
 than "Tango" :)
 


We're getting awfully near to a d2 dmd that passes all of it's current tests
(d1 achieved that state about 2 weeks ago).
 As a rough guess, I'd say another few weeks and we'll reach that milestone. 
There are 8 failing tests in the public
version of the dmd d2 test suite (not sure about the non-public suite, probably
some undiagnosed issues in there too) +
18 failures in the phobos test suite.  Bugs have been fixed at a rate of about
1 per day.  So.. you can do the math.

Once that's done, it'll be time to cast a wider net.. ie, a seriously alpha
quality release.  Chances are super high
that lots of problem will be encountered and lots of bug reports will roll in. 
They'll be fixed.  More releases will
occur.  More will be fixed.. and at some point, it'll be usable enough.  The
hard part will be getting good bug reports,
very small reproducible test cases.

I won't hazard a guess at how long it'll take before dmd 64 bit support is as
stable as the 32 bit support, other than
'longer than everyone would probably like'.

Hope that helps,
Brad
Jan 26 2011
prev sibling next sibling parent reply Trass3r <un known.com> writes:
 Killing D (at least for me) is the limit choices for compiling on 64-bit Linux
with D2 and preferably Phobos instead of Tango.


I do think GDC is the most stable D2 x64 solution at the moment.
Especially since you're on Linux it shouldn't be too hard to compile it (while
it still is a PITA on Windows)
Jan 27 2011
parent darenw <darenw darenscotwilson.com> writes:
Will try it.  I had tried gdc in December, but had some sort of trouble with it,
and moved on to trying ldc rather than fuss with it much.  Perhaps it's worth a
little more fuss to get running.
Jan 27 2011
prev sibling next sibling parent reply retard <re tard.com.invalid> writes:
Thu, 27 Jan 2011 06:32:58 +0000, dwilson wrote:


 Beside praying and pestering, what can we D non-experts do to help get a
 stable 64-bit dmd available?
 
 Killer D features are strings, slick built in dynamics arrays, no
 headers files to keep in sync, and the other nice features often praised
 by others.   I'm not sure yet that D is my favorite language, but it's
 in the list of top three.
 
 Killing D (at least for me) is the limit choices for compiling on 64-bit
 Linux with D2 and preferably Phobos instead of Tango.  My setup, for
 reasons I haven't investigated deeply, can't run 32-bit anything, and I
 do intend to work on huge arrays of data, a few GB in RAM.  As for
 Phobos, it's obviously more Mars-related than "Tango" :)


Didn't Walter say about one year ago that it only takes 1-2 months to 
finish the 64-bit port.
Jan 27 2011
next sibling parent so <so so.do> writes:
 Didn't Walter say about one year ago that it only takes 1-2 months to
 finish the 64-bit port.


He probably meant 1~2. :)
This is quite common for programmers, no?
Jan 27 2011
prev sibling parent Walter Bright <newshound2 digitalmars.com> writes:
retard wrote:

 Didn't Walter say about one year ago that it only takes 1-2 months to 
 finish the 64-bit port.


Yeah, well, I work on a lot of other stuff at the same time. The most recent 
distraction was the move to github.
Jan 27 2011
prev sibling parent reply Eric Poggel <dnewsgroup2 yage3d.net> writes:
On 1/27/2011 1:32 AM, dwilson wrote:

 Beside praying and pestering, what can we D non-experts do to help get a
stable 64-bit dmd available?

 Killer D features are strings, slick built in dynamics arrays, no headers
files to keep in sync, and the
 other nice features often praised by others.   I'm not sure yet that D is my
favorite language, but it's in
 the list of top three.

 Killing D (at least for me) is the limit choices for compiling on 64-bit Linux
with D2 and preferably Phobos
 instead of Tango.  My setup, for reasons I haven't investigated deeply, can't
run 32-bit anything, and I do
 intend to work on huge arrays of data, a few GB in RAM.  As for Phobos, it's
obviously more Mars-related
 than "Tango" :)


When 64-bit dmd arrives for Windows, what will be used for a linker? 
Will optlink be ported also?
Jan 27 2011
next sibling parent reply Daniel Gibson <metalcaedes gmail.com> writes:
Am 27.01.2011 19:56, schrieb Eric Poggel:

 On 1/27/2011 1:32 AM, dwilson wrote:

 Beside praying and pestering, what can we D non-experts do to help get a
 stable 64-bit dmd available?

 Killer D features are strings, slick built in dynamics arrays, no headers
 files to keep in sync, and the
 other nice features often praised by others.   I'm not sure yet that D is my
 favorite language, but it's in
 the list of top three.

 Killing D (at least for me) is the limit choices for compiling on 64-bit Linux
 with D2 and preferably Phobos
 instead of Tango.  My setup, for reasons I haven't investigated deeply, can't
 run 32-bit anything, and I do
 intend to work on huge arrays of data, a few GB in RAM.  As for Phobos, it's
 obviously more Mars-related
 than "Tango" :)


 
 When 64-bit dmd arrives for Windows, what will be used for a linker? Will
 optlink be ported also?


If it will depend on optlink it will probably take ages until a win64 version is
ready, because optlink is (mostly) written in assembler..

I guess it would be smart to output object files compatible to the windows port
of the gnu linker (from MingW), so this can be used and you can link in object
files from GCC/MinGW (otherwise DMC would need a 64bit port as well - don't know
if that is planned).
Jan 27 2011
parent Trass3r <un known.com> writes:
 I guess it would be smart to output object files compatible to the  
 windows port
 of the gnu linker (from MingW), so this can be used and you can link in  
 object
 files from GCC/MinGW (otherwise DMC would need a 64bit port as well -  
 don't know
 if that is planned).


I think dmc uses the very same backend dmd does.
But a change to COFF would indeed be great.
Jan 27 2011
prev sibling next sibling parent Trass3r <un known.com> writes:
 When 64-bit dmd arrives for Windows, what will be used for a linker?  
 Will optlink be ported also?


We recently had a thread if a linker called "UniLink" could be a solution.
Jan 27 2011
prev sibling parent Bruno Medeiros <brunodomedeiros+spam com.gmail> writes:
On 27/01/2011 18:56, Eric Poggel wrote:

 On 1/27/2011 1:32 AM, dwilson wrote:

 Beside praying and pestering, what can we D non-experts do to help get
 a stable 64-bit dmd available?

 Killer D features are strings, slick built in dynamics arrays, no
 headers files to keep in sync, and the
 other nice features often praised by others. I'm not sure yet that D
 is my favorite language, but it's in
 the list of top three.

 Killing D (at least for me) is the limit choices for compiling on
 64-bit Linux with D2 and preferably Phobos
 instead of Tango. My setup, for reasons I haven't investigated deeply,
 can't run 32-bit anything, and I do
 intend to work on huge arrays of data, a few GB in RAM. As for Phobos,
 it's obviously more Mars-related
 than "Tango" :)


 When 64-bit dmd arrives for Windows, what will be used for a linker?
 Will optlink be ported also?


I'm curious about this as well. Will it require a new linker? :/

-- 
Bruno Medeiros - Software Engineer
Feb 10 2011