RE: FTensor on cygwin - partial success

RE: FTensor on cygwin - partial success

[Billinghurst, David (CRTS)]

FTensor works with g++ 3.1 on cygwin.  Previously I compiled the 
speed tests with 2.95.  Oops.

Thanks Walter.

-----Original Message-----
From: Walter Landry [mailto:wlandry@ucsd.edu]
Sent: Tuesday, 30 April 2002 5:24 
To: Billinghurst, David (CRTS)
Cc: gcc@gcc.gnu.org
Subject: Re: FTensor on cygwin - partial success


"Billinghurst, David (CRTS)" <David.Billinghurst@riotinto.com> wrote:
> Just built FTensor on cygwin with todays CVS gcc-3.1, as described in 
> http://gcc.gnu.org/gcc-3.1/testing-ftensor.html
> 
> The compiler conformance tests all pass but I get failures on the 
> performance tests.  Is anyone else seeing this?
> 
> For example:
> 
> cd tests/speed
> make little
> 
> g++ -ftemplate-depth-100 -Drestrict= -w -O3 -finline-functions -finline-limit-1000000 -ffast-math -fno-rtti -fno-exceptions    little.C   -o little
> little.C: In function `int main()':
> little.C:48: no match for call to `(FTensor::Tensor1<double,3>) (const FTensor::Index<'i',3> &)'
> ../../Tensor1/Tensor1_value.h:31: candidates are: double & FTensor::Tensor1<double,3>::operator ()(int)
> ../../Tensor1/Tensor1_value.h:43:                 double FTensor::Tensor1<double,3>::operator ()(int) const

That's wierd.  These errors look a lot like the errors I got when
trying to compile FTensor with gcc 2.95 or 3.0.  The stuff it does in
that file, little.C, should be equivalent to what is happening in
tests/conformance/test_T1.C.  Can you try compiling that file with the
same options?  If it takes too long or uses too much memory, you can
try commenting out everything past line 28 except for line 56.  It
should compile the same sort of things, although the tests will fail
when you run.

Thanks,
Walter

Re: FTensor on cygwin - partial success

[Walter Landry]

"Billinghurst, David (CRTS)" <David.Billinghurst@riotinto.com> wrote:
> Just built FTensor on cygwin with todays CVS gcc-3.1, as described in 
> http://gcc.gnu.org/gcc-3.1/testing-ftensor.html
> 
> The compiler conformance tests all pass but I get failures on the 
> performance tests.  Is anyone else seeing this?
> 
> For example:
> 
> cd tests/speed
> make little
> 
> g++ -ftemplate-depth-100 -Drestrict= -w -O3 -finline-functions -finline-limit-1000000 -ffast-math -fno-rtti -fno-exceptions    little.C   -o little
> little.C: In function `int main()':
> little.C:48: no match for call to `(FTensor::Tensor1<double,3>) (const FTensor::Index<'i',3> &)'
> ../../Tensor1/Tensor1_value.h:31: candidates are: double & FTensor::Tensor1<double,3>::operator ()(int)
> ../../Tensor1/Tensor1_value.h:43:                 double FTensor::Tensor1<double,3>::operator ()(int) const

That's wierd.  These errors look a lot like the errors I got when
trying to compile FTensor with gcc 2.95 or 3.0.  The stuff it does in
that file, little.C, should be equivalent to what is happening in
tests/conformance/test_T1.C.  Can you try compiling that file with the
same options?  If it takes too long or uses too much memory, you can
try commenting out everything past line 28 except for line 56.  It
should compile the same sort of things, although the tests will fail
when you run.

Thanks,
Walter

Re: FTensor on cygwin - partial success

[Janis Johnson]

On Tue, Apr 30, 2002 at 12:09:29AM +1000, Billinghurst, David (CRTS) wrote:
> Just built FTensor on cygwin with todays CVS gcc-3.1, as described in 
> http://gcc.gnu.org/gcc-3.1/testing-ftensor.html
> 
> The compiler conformance tests all pass but I get failures on the 
> performance tests.  Is anyone else seeing this?
> 
> For example:
> 
> cd tests/speed
> make little
> 
> g++ -ftemplate-depth-100 -Drestrict= -w -O3 -finline-functions -finline-limit-1000000 -ffast-math -fno-rtti -fno-exceptions    little.C   -o little
> little.C: In function `int main()':
> little.C:48: no match for call to `(FTensor::Tensor1<double,3>) (const FTensor::Index<'i',3> &)'
> ../../Tensor1/Tensor1_value.h:31: candidates are: double & FTensor::Tensor1<double,3>::operator ()(int)
> ../../Tensor1/Tensor1_value.h:43:                 double FTensor::Tensor1<double,3>::operator ()(int) const
> 
> and so on.
> 
> Happy to supply the test cases, but probably easier to grab to original and follow the instructions.

I just tried that on ia64-unknown-linux-gnu with a compiler built from
CVS as of ten hours ago and it worked fine.  Very odd.

Janis

FTensor on cygwin - partial success

[Billinghurst, David (CRTS)]

Just built FTensor on cygwin with todays CVS gcc-3.1, as described in 
http://gcc.gnu.org/gcc-3.1/testing-ftensor.html

The compiler conformance tests all pass but I get failures on the 
performance tests.  Is anyone else seeing this?

For example:

cd tests/speed
make little

g++ -ftemplate-depth-100 -Drestrict= -w -O3 -finline-functions -finline-limit-1000000 -ffast-math -fno-rtti -fno-exceptions    little.C   -o little
little.C: In function `int main()':
little.C:48: no match for call to `(FTensor::Tensor1<double,3>) (const FTensor::Index<'i',3> &)'
../../Tensor1/Tensor1_value.h:31: candidates are: double & FTensor::Tensor1<double,3>::operator ()(int)
../../Tensor1/Tensor1_value.h:43:                 double FTensor::Tensor1<double,3>::operator ()(int) const

and so on.

Happy to supply the test cases, but probably easier to grab to original and follow the instructions.