Loading ...
Sorry, an error occurred while loading the content.

Re: gcc ipk package.

Expand Messages
  • r2004roellig
    Hi Lennert et all, based on your kernel patch suggestion for the Netwinder floating point issues I have performed these steps: - recompiled unslung 3.16 with
    Message 1 of 20 , Jan 1, 2005
    • 0 Attachment
      Hi Lennert et all,

      based on your kernel patch suggestion for the Netwinder floating
      point issues I have performed these steps:

      - recompiled unslung 3.16 with patched kernel (OE build) -> PASS
      - gave hello-fp.c another try (esp. sign and abs tests) -> PASS
      - compiled perl with native compiler -> PASS
      - expecuted perl test suite -> PASS 100.0%

      Although perl might not be a 100% test for floating point calculations
      I think fp looks pretty robust now. Of course more comprehensive fp tests
      are required. However at this point I would say this is a great break
      through! Thanks for your help.

      I'll request, that one of the NSLU2 OE developers will put the
      patch into OE!

      More on Perl (linked against shlibs with dynaloader support) to
      follow in another message.

      Thanks,
      Rene.


      --- In nslu2-linux@yahoogroups.com, "r2004roellig" <r2004roellig@y...>
      wrote:
      >
      > Hi Lennert,
      >
      > this sounds like a good explanation for the floating point sign issues
      > with NWFPE hardfloat. Due to those issues I have compiled my cross
      > compilers to do softfloat only...
      >
      > Your patch is for kernel 2.6.x, however in this portion of the kernel
      > there are no differences to NSLU's 2.4.22 kernel. Thus I think the
      > patch should work for 2.4.22 too.
      >
      > Here is what I'll do next:
      > - recompile unslung 3.16 with patched kernel
      > - give hello-fp.c another try
      > - if successful compile perl with native toolchain
      > - if perl compiles & runs flawless @ NSLU with patched kernel
      > I suggest that the patch should be driven into OE. Maybe then
      > all openslug and unslung 4.x developers/users could benefit from
      > working fp.
      >
      > To run the tests will take me until Sunday. I need to go party now :)
      >
      > Rene.
      >
      >
      >
      > --- In nslu2-linux@yahoogroups.com, Lennert Buytenhek <buytenh@w...>
      > wrote:
      > > On Thu, Dec 30, 2004 at 08:54:05PM -0000, r2004roellig wrote:
      > >
      > > > The expected output is:
      > > >
      > http://www.familie-roellig.de/public/nslu2/toolchain/hello-fp-output.txt
      > >
      > > Aha! It doesn't negate properly. The 'abs' operation is probably
      > > also broken.
      > >
      > > I would ask you to apply this kernel patch:
      > >
      > >
      >
      http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2004-October/024909.html
      > >
      > > But that's going to be rather hard in case of the NSLU2. Anyone got a
      > > 2.6 kernel running on the thing yet?
      > >
      > >
      > > cheers,
      > > Lennert
    • Rod Whitby
      ... Done. Can you test the 3.17 built from OE? When someone reports that 3.17 built from source fixes the problem, then we will release an updated binary. --
      Message 2 of 20 , Jan 1, 2005
      • 0 Attachment
        On Sun, 02 Jan 2005 00:13:48 -0000, r2004roellig <r2004roellig@...> wrote:
        > I'll request, that one of the NSLU2 OE developers will put the
        > patch into OE!

        Done.

        Can you test the 3.17 built from OE? When someone reports that 3.17
        built from source fixes the problem, then we will release an updated
        binary.

        -- Rod
      Your message has been successfully submitted and would be delivered to recipients shortly.