headers
Jonathan Wilson | 1 Sep 2003 03:58
Picon

Does GCC support TLS or SEH?

Does the GCC core contain support for Thread Local Storage or Structured 
Exception Handling?

I would like to see if it is fesable to implement windows specific TLS and 
SEH and if there is some sort of hooks in the GCC core already, it would 
make it easier.

Both Watcom and Microsofts linkers have a feature used for constructors and 
also for TLS whereby a segment named .tls, a segment named .tls$ and a 
segment named .tls$$$ will be combined into 1 segment called .tls with .tls 
first then .tls$ and .tls$$$
Does the GCC linker have a similar feature?
If not, how does GCC handle constructors?
Permalink | Reply | 1 comment |
headers
Wu Yongwei | 1 Sep 2003 04:27

Re: G++ 3.3.1 Specialization bug?

Carlo Wood wrote:
> > #include <stddef.h>
>
> #include <cstddef>

No.  Please notice that I used "size_t" instead of "std::size_t".  Some
compilers do not support "std::size_t" well and I do not want the code
to work only in GCC.  Otherwise things will be simpler.

>
> > #if __GNUC__ == 3 && __GNUC_MINOR__ > 0
> > #include <ext/hash_map>
> > namespace sgi {
> >     using __gnu_cxx::hash;
> >     using __gnu_cxx::hash_map;
> > }
>
> #define SGI_HASH_NAMESPACE __gnu_cxx

Yes, this is really the solution I use now (though I use the name SGI_EXT).

> [snip]

Macros can solve the problem for now.  But I want to see whether a more
"decent" solution exists.  __gnu_cxx is really uglifying things (yes, I
know there are reasons for this :-) ).

Best regards,

Wu Yongwei
(Continue reading)

Permalink | Reply | 12 comments |
headers
Richard Kenner | 1 Sep 2003 05:23
Picon

Re: Strict aliasing and converting binary patterns between integer and float

    I need to convert back and forth between an unsigned int and the float
    with the same binary pattern (admitting I knows they're the same size
    - 32bits - in the first place).  

    WHile this very simple example may still work with -fstrict-aliasing,
    as soon as you have a more complex function the value of v2 is not
    always what is expected.

    So my question is, what is a reliable way to do this conversion with
    gcc and strict aliasing enabled?

Use a union.
Permalink | Reply | 0 comments |
headers
Andrew Pinski | 1 Sep 2003 06:17
Picon

Re: Does GCC support TLS or SEH?


On Sunday, Aug 31, 2003, at 18:58 US/Pacific, Jonathan Wilson wrote:

> Does the GCC core contain support for Thread Local Storage or 
> Structured Exception Handling?

GCC supports TLS on some targets like i686-pc-linux-gnu with certain 
versions of
glibc and the binutils but not on Win32 yet, if you implement it, it 
will some faster.

>
> I would like to see if it is fesable to implement windows specific TLS 
> and SEH and if there is some sort of hooks in the GCC core already, it 
> would make it easier.

As I said TLS is there but not for Win32, I do not know anything about 
SEH
at all (I think it is signals thrown as exceptions right?)

Thanks,
Andrew Pinski

>
> Both Watcom and Microsofts linkers have a feature used for 
> constructors and also for TLS whereby a segment named .tls, a segment 
> named .tls$ and a segment named .tls$$$ will be combined into 1 
> segment called .tls with .tls first then .tls$ and .tls$$$
> Does the GCC linker have a similar feature?
> If not, how does GCC handle constructors?
(Continue reading)

Permalink | Reply | 0 comments |
headers
Andrew Pinski | 1 Sep 2003 07:35
Picon

Re: tree-ssa merge from mainline?


On Sunday, Aug 31, 2003, at 12:33 US/Pacific, Nathanael Nerode wrote:

> There's been a lot of straightforward systematic improvements on
> mainline recently.
>
> How recently has tree-ssa been merged?  Should it perhaps be merged 
> soon
> to minimize future difficulty?  :-)

It was merged from 20030817 last so less than 3 weeks.

I agree (but I have no say) that it should be merged again but the 
mainline is having problems
for most targets right now, in fact I cannot even bootstrap on 
i686-pc-linux-gnu (which most
people test their patches on) it is that bad.

I think the bootstrap problems are related to GC but I have no way to 
prove this though, I was
traveling all today, Sunday, and this is the first time I could answer 
any email and I have been
trying to debug the failure but I am getting no where.  (Note RTH, I 
will send you the preprocessed
source once I get settled in).

Thanks,
Andrew Pinski
(Continue reading)

Permalink | Reply | 0 comments |
headers
William S. | 1 Sep 2003 07:38
Picon
Picon
Favicon

[tree-ssa] Bootstrap on IRIX 6.5

I have been unsuccessful in being able to bootstrap
tree-ssa on an Octane R12k with IRIX 6.5.

Has anyone done this and if so is there a procedure?

I have used:

'./configure --prefix=/usr/gFORTRAN --with-gmp=/usr/GMP --enable-languages=c,f95'

and

'gmake CC=gcc bootstrap' (from within /usr/objdir)

--

-- 
Bill
Amsterdam, NL
Permalink | Reply | 3 comments |
headers
Andrew Pinski | 1 Sep 2003 07:46
Picon

Re: [tree-ssa] Bootstrap on IRIX 6.5


On Sunday, Aug 31, 2003, at 22:38 US/Pacific, William S. wrote:

> I have been unsuccessful in being able to bootstrap
> tree-ssa on an Octane R12k with IRIX 6.5.
>
> Has anyone done this and if so is there a procedure?
>
> I have used:
>
> './configure --prefix=/usr/gFORTRAN --with-gmp=/usr/GMP 
> --enable-languages=c,f95'
>
> and
>
> 'gmake CC=gcc bootstrap' (from within /usr/objdir)
>

You should be using a different obj dir first.
Second I would read the instructions for the regular releases also:
<http://gcc.gnu.org/install/specific.html#mips-sgi-irix6>
If that all fails report it as a bug.

Thanks,
Andrew Pinski
Permalink | Reply | 1 comment |
headers
Daniel Berlin | 1 Sep 2003 07:47

Re: [tree-ssa] Bootstrap on IRIX 6.5


On Mon, 1 Sep 2003, William S. wrote:

> I have been unsuccessful in being able to bootstrap
> tree-ssa on an Octane R12k with IRIX 6.5.
>
> Has anyone done this and if so is there a procedure?
>

Based on the fact that there are test results posted for it for irix 6.5,
i would say yes.

See http://gcc.gnu.org/ml/gcc-testresults/2003-08/msg01414.html for
example

Maybe david billinghurst would be kind enough to try to help with whatever
problem you are having.

> I have used:
>
> './configure --prefix=/usr/gFORTRAN --with-gmp=/usr/GMP --enable-languages=c,f95'
>
> and
>
> 'gmake CC=gcc bootstrap' (from within /usr/objdir)
>
> --
> Bill
> Amsterdam, NL
>
(Continue reading)

Permalink | Reply | 0 comments |
headers
William S. | 1 Sep 2003 07:53
Picon
Picon
Favicon

Re: [tree-ssa] Bootstrap on IRIX 6.5

Sorry, that was a typo. It should have been:

'gmake CC=gcc bootstrap' (from within /objdir)

On Sun, Aug 31, 2003 at 10:46:33PM -0700, Andrew Pinski wrote:
> 
> On Sunday, Aug 31, 2003, at 22:38 US/Pacific, William S. wrote:
> 
> >'gmake CC=gcc bootstrap' (from within /usr/objdir)
> >
> 
> You should be using a different obj dir first.
> 

--

-- 
Bill
Amsterdam, NL
Permalink | Reply | 0 comments |
headers
Billinghurst, David (CRTS | 1 Sep 2003 08:38
Favicon

RE: [tree-ssa] Bootstrap on IRIX 6.5

> From: Daniel Berlin 
> Sent: Monday, 1 September 2003 3:47 PM
> To: William S.
> Cc: gcc
> Subject: Re: [tree-ssa] Bootstrap on IRIX 6.5
> 
> On Mon, 1 Sep 2003, William S. wrote:
>
>> I have been unsuccessful in being able to bootstrap
>> tree-ssa on an Octane R12k with IRIX 6.5.
>>
>> Has anyone done this and if so is there a procedure?

I haven't had a successful tree-ssa branch on irix6.5 since  
	LAST_UPDATED: Tue Aug 26 13:15:23 GMT 2003
although my build script tries each night.  I haven't had time
to see what is failing.  I have a couple of work arounds in 
my tree to build libmudflap and test libgfortran.

Perhaps you could start by checking out the tree from a date 
I have tested - see gcc-testresults mailing list - and we can work
through any issues.
Permalink | Reply | 0 comments |

Gmane