[v4,1/1] fftw: add compile precision option

From: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

fftw has options to select compile precision between single, long-double and
quad. These options are exclusives. This patch adds choice to select precision
option.

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>
---
Changes v3 -> v4:
 * add depends instruction for long-double procession. This option needs to
   have a toolchain with long-double trigonometric routines.
Changes v2 -> v3:
 * add depends instruction for quad-precision. This option needs to have a
   toolchain with libquadmath enabled (only x86, x86_64 and Itanium).
Changes v1 -> v2:
 * FFTW_CONFIGURE_OPTS -> FFTW_CONF_OPTS
---
 package/fftw/Config.in | 38 ++++++++++++++++++++++++++++++++++++++
 package/fftw/fftw.mk   |  9 +++++++++
 2 files changed, 47 insertions(+)

Gwenhael, All,

On 2015-03-15 17:15 +0100, Gwenhael Goavec-Merou spake thusly:
> fftw has options to select compile precision between single, long-double and
> quad. These options are exclusives. This patch adds choice to select precision
> option.
> 
> Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

So, I have tested this patch in different configurations:
  - x86, all of none, single, double and quad
  - arm, all of none single, and double

All did build successfuly.

However, I am a bit worried that we do not have a disbling condition for
'none'...

[--SNIP--]
> diff --git a/package/fftw/fftw.mk b/package/fftw/fftw.mk
> index 3b302df..80cc0dc 100644
> --- a/package/fftw/fftw.mk
> +++ b/package/fftw/fftw.mk
> @@ -10,4 +10,13 @@ FFTW_INSTALL_STAGING = YES
>  FFTW_LICENSE = GPLv2+
>  FFTW_LICENSE_FILES = COPYING
>  

Maybe we could have something like:

    # Disable everything, and then selectively enable what we want
    FFTW_CONF_OPTS = --disable-single --disable-long-double --disable-quad-precision

    ifeq ($(BR2_PACKAGE_FFTW_PRECISION_SINGLE),y)
    FFTW_CONF_OPTS = --enable-single
    endif

    ifeq ($(BR2_PACKAGE_FFTW_PRECISION_LONG_DOUBLE),y)
    FFTW_CONF_OPTS = --enable-long-double
    endif

    ifeq ($(BR2_PACKAGE_FFTW_PRECISION_QUAD),y)
    FFTW_CONF_OPTS = --enable-quad-precision
    endif

When we have both --disabe/--enable, the latter wins. So we can simply
disable everything, and then enable only the one we need.

I got it that they are mutually exclusive, but if the user selects
'none', we do not have any '--enable-XXX' so what happens?

Alternatively, you could do:

    FFTW_CONF_OPTS += $(if $(BR2_PACKAGE_FFTW_PRECISION_SINGLE),--enable,--disable)-single
    FFTW_CONF_OPTS += $(if $(BR2_PACKAGE_FFTW_PRECISION_LONG_DOUBLE),--enable,--disable)-long-double
    FFTW_CONF_OPTS += $(if $(BR2_PACKAGE_FFTW_PRECISION_QUAD),--enable,--disable)-quad-precision

Not sure which I prefer...

Regards,
Yann E. MORIN.

> +ifeq ($(BR2_PACKAGE_FFTW_PRECISION_SINGLE),y)
> +FFTW_CONF_OPTS = --enable-single
> +endif
> +ifeq ($(BR2_PACKAGE_FFTW_PRECISION_LONG_DOUBLE),y)
> +FFTW_CONF_OPTS = --enable-long-double
> +endif
> +ifeq ($(BR2_PACKAGE_FFTW_PRECISION_QUAD),y)
> +FFTW_CONF_OPTS = --enable-quad-precision
> +endif
>  $(eval $(autotools-package))
> -- 
> 2.0.5
> 
> _______________________________________________
> buildroot mailing list
> buildroot@busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

On Sun, 15 Mar 2015 20:26:10 +0100
"Yann E. MORIN" <yann.morin.1998@free.fr> wrote:

> Gwenhael, All,
> 
> On 2015-03-15 17:15 +0100, Gwenhael Goavec-Merou spake thusly:
> > fftw has options to select compile precision between single, long-double and
> > quad. These options are exclusives. This patch adds choice to select precision
> > option.
> > 
> > Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>
> 
> So, I have tested this patch in different configurations:
>   - x86, all of none, single, double and quad
>   - arm, all of none single, and double
> 
> All did build successfuly.
> 
> However, I am a bit worried that we do not have a disbling condition for
> 'none'...
>
The none have no existence because fftw by default has no options enabled.
This choice is only present to reflect the default behaviour. 
>
> [--SNIP--]
> > diff --git a/package/fftw/fftw.mk b/package/fftw/fftw.mk
> > index 3b302df..80cc0dc 100644
> > --- a/package/fftw/fftw.mk
> > +++ b/package/fftw/fftw.mk
> > @@ -10,4 +10,13 @@ FFTW_INSTALL_STAGING = YES
> >  FFTW_LICENSE = GPLv2+
> >  FFTW_LICENSE_FILES = COPYING
> >  
> 
> Maybe we could have something like:
> 
>     # Disable everything, and then selectively enable what we want
>     FFTW_CONF_OPTS = --disable-single --disable-long-double --disable-quad-precision
> 
>     ifeq ($(BR2_PACKAGE_FFTW_PRECISION_SINGLE),y)
>     FFTW_CONF_OPTS = --enable-single
>     endif
> 
>     ifeq ($(BR2_PACKAGE_FFTW_PRECISION_LONG_DOUBLE),y)
>     FFTW_CONF_OPTS = --enable-long-double
>     endif
> 
>     ifeq ($(BR2_PACKAGE_FFTW_PRECISION_QUAD),y)
>     FFTW_CONF_OPTS = --enable-quad-precision
>     endif
> 
> When we have both --disabe/--enable, the latter wins. So we can simply
> disable everything, and then enable only the one we need.
> 
> I got it that they are mutually exclusive, but if the user selects
> 'none', we do not have any '--enable-XXX' so what happens?
>
Nothing this is the default situation ;-) In fact fftw configure don't care
about disable-{single,quad,...}. Only enable-xxx is taken into account. This is
why I do nothing if none is selected.
>
> Alternatively, you could do:
> 
>     FFTW_CONF_OPTS += $(if $(BR2_PACKAGE_FFTW_PRECISION_SINGLE),--enable,--disable)-single
>     FFTW_CONF_OPTS += $(if $(BR2_PACKAGE_FFTW_PRECISION_LONG_DOUBLE),--enable,--disable)-long-double
>     FFTW_CONF_OPTS += $(if $(BR2_PACKAGE_FFTW_PRECISION_QUAD),--enable,--disable)-quad-precision
> 
> Not sure which I prefer...
> 
> Regards,
> Yann E. MORIN.
> 
Regards,
Gwen