Yes, only --useDouble=1.

No, there is no way to do both double and float, at least not without a lot more work.

I suggest you support only double. What is your speed difference?

Regards,
Mike


-----Original Message-----
>From: Anthony Kozar 
>Sent: Oct 28, 2008 11:19 PM
>To: Csound Developer list 
>Subject: Re: [Cs-dev] Compiling linear algebra ops
>
>Michael Gogins wrote on 10/28/08 9:00 AM:
>
>> The linear algebra opcodes are only for csound64, but they should work on all
>> platforms.
>
>I assume that this means I should only expect them to work with
>--useDouble=1 (i.e. you do not mean the --Word64 option) ?  I was compiling
>the float version.  However, SCons always tries to build the linear algebra
>opcodes if both Boost and gmm.h are found (both of which I installed)
>regardless of the state of the useDouble option.
>
>Is there any way to make these opcodes compatible with both floats and
>doubles?
>
>Here are the first few errors that I received in case I am misunderstanding
>something (using GCC 4.01):
>
>Opcodes/linear_algebra.cpp: In member function 'int
>la_i_qr_eigen_mr_t::init(CSOUND*)':
>
>Opcodes/linear_algebra.cpp:4117: error: ISO C++ says that these are
>ambiguous, even though the worst conversion for the first is better than the
>worst conversion for the second:
>
>/usr/local/include/gmm/gmm_dense_qr.h:552: note: candidate 1: void
>gmm::implicit_qr_algorithm(const MAT1&, const VECT&, const MAT2&, typename
>gmm::number_traitsgmm::linalg_traits::value_type>::magnitude_type, bool) [with MAT1 =
>gmm::dense_matrix, VECT = std::vector >,
>MAT2 = double]
>
>/usr/local/include/gmm/gmm_dense_qr.h:586: note: candidate 2: void
>gmm::implicit_qr_algorithm(const MAT1&, VECT&, typename
>gmm::number_traitsgmm::linalg_traits::value_type>::magnitude_type) [with MAT1 =
>gmm::dense_matrix, VECT = std::vector >]
>
>Opcodes/linear_algebra.cpp: In member function 'int
>la_k_qr_eigen_mr_t::kontrol(CSOUND*)':
>
>etc.
>
>
>
>-------------------------------------------------------------------------
>This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
>Build the coolest Linux based applications with Moblin SDK & win great prizes
>Grand prize is a trip for two to an Open Source event anywhere in the world
>http://moblin-contest.org/redirect.php?banner_id=100&url=/
>_______________________________________________
>Csound-devel mailing list
>Csound-devel@lists.sourceforge.net
>https://lists.sourceforge.net/lists/listinfo/csound-devel




-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Note that packages compiled against (using the csound API) the usual float
csound framework/library will not work with a double version.

> Yes, only --useDouble=1.
>
> No, there is no way to do both double and float, at least not without a
> lot more work.
>
> I suggest you support only double. What is your speed difference?
>
> Regards,
> Mike
>
>
> -----Original Message-----
>>From: Anthony Kozar 
>>Sent: Oct 28, 2008 11:19 PM
>>To: Csound Developer list 
>>Subject: Re: [Cs-dev] Compiling linear algebra ops
>>
>>Michael Gogins wrote on 10/28/08 9:00 AM:
>>
>>> The linear algebra opcodes are only for csound64, but they should work
>>> on all
>>> platforms.
>>
>>I assume that this means I should only expect them to work with
>>--useDouble=1 (i.e. you do not mean the --Word64 option) ?  I was
>> compiling
>>the float version.  However, SCons always tries to build the linear
>> algebra
>>opcodes if both Boost and gmm.h are found (both of which I installed)
>>regardless of the state of the useDouble option.
>>
>>Is there any way to make these opcodes compatible with both floats and
>>doubles?
>>
>>Here are the first few errors that I received in case I am
>> misunderstanding
>>something (using GCC 4.01):
>>
>>Opcodes/linear_algebra.cpp: In member function 'int
>>la_i_qr_eigen_mr_t::init(CSOUND*)':
>>
>>Opcodes/linear_algebra.cpp:4117: error: ISO C++ says that these are
>>ambiguous, even though the worst conversion for the first is better than
>> the
>>worst conversion for the second:
>>
>>/usr/local/include/gmm/gmm_dense_qr.h:552: note: candidate 1: void
>>gmm::implicit_qr_algorithm(const MAT1&, const VECT&, const MAT2&,
>> typename
>>gmm::number_traits>gmm::linalg_traits::value_type>::magnitude_type, bool) [with MAT1 =
>>gmm::dense_matrix, VECT = std::vector
>> >,
>>MAT2 = double]
>>
>>/usr/local/include/gmm/gmm_dense_qr.h:586: note: candidate 2: void
>>gmm::implicit_qr_algorithm(const MAT1&, VECT&, typename
>>gmm::number_traits>gmm::linalg_traits::value_type>::magnitude_type) [with MAT1 =
>>gmm::dense_matrix, VECT = std::vector
>> >]
>>
>>Opcodes/linear_algebra.cpp: In member function 'int
>>la_k_qr_eigen_mr_t::kontrol(CSOUND*)':
>>
>>etc.
>>
>>
>>
>>-------------------------------------------------------------------------
>>This SF.Net email is sponsored by the Moblin Your Move Developer's
>> challenge
>>Build the coolest Linux based applications with Moblin SDK & win great
>> prizes
>>Grand prize is a trip for two to an Open Source event anywhere in the
>> world
>>http://moblin-contest.org/redirect.php?banner_id=100&url=/
>>_______________________________________________
>>Csound-devel mailing list
>>Csound-devel@lists.sourceforge.net
>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>
>
>
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's
> challenge
> Build the coolest Linux based applications with Moblin SDK & win great
> prizes
> Grand prize is a trip for two to an Open Source event anywhere in the
> world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>



-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Sorry for this late reply.  I don't know what the speed difference is but
with -ftree-vectorize, floats _might_ be significantly faster on G4 and G5
machines (just speculating).

I would like to help Victor get a regular doubles-based package going for OS
X especially since I have read that new versions of blue need one.  However,
we also need to continue supporting floats-based packages on OS X as long as
certain key applications require it (MacCsound and maybe others).

Anthony

Jonatan Liljedahl wrote on 10/29/08 9:46 AM:

> Note that packages compiled against (using the csound API) the usual float
> csound framework/library will not work with a double version.
> 
>> Yes, only --useDouble=1.
>> 
>> No, there is no way to do both double and float, at least not without a
>> lot more work.
>> 
>> I suggest you support only double. What is your speed difference?
>> 
>> Regards,
>> Mike


-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net