Re: floating point operations in gcc 3.3.3, glibc 2.3.2

[Daniel Kegel <dank at kegel dot com>]

Kunjan Naik wrote:
>   We were able to successfully build a toolchain using Dan's
> crosstool-0.28-rc19 for x86 to ppc-750 and x86 to ppc405 architecture. GCC
> 3.3.3, GLIBC 2.3.2 and 2.6 kernel headers were used to build the toolchain.
> But we observed that all testcases (from GCC regression testsuite) for gcc
> 3.3.3 with any floating point operations are failing. This happens only for
> ppc-750 with hardware floating point unit support. 
>   
> The gcc regression test summary:
>  
> ----------------------------------------------------------------------------
>  LAST_UPDATED: gcc-3.3.3
> Target is powerpc-750-linux-gnu
> Host   is i686-host_pc-linux-gnu
> === gcc tests ===
> Running target powerpc-750-linux-gnu
>
> FAIL: gcc.c-torture/execute/920501-8.c execution,  -O0 
> ...
>                         === gcc Summary ===
> # of expected passes                 21423
> # of unexpected failures  61
> # of expected failures                 67
> # of unresolved testcases           28
> # of unsupported tests                185
>
> ----------------------------------------------------------------------------
>  
> Is this a bug in code generation for floating point operations?

Last time I ran a similar test (on the same hardware, probably)
was gcc-3.3.1, and it did pretty well:
http://www.kegel.com/crosstool/current/summaries/powerpc-750-linux-gnu-gcc-3.3.1-glibc-2.3.2.gcc.sum
That was running a 2.4 kernel, though.

Are you actually running a 2.6 kernel?

Also, how did you run the tests?
- Dan

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com