diff options
| author | Mohammad Akhlaghi <mohammad@akhlaghi.org> | 2019-07-07 17:12:48 +0100 |
|---|---|---|
| committer | Mohammad Akhlaghi <mohammad@akhlaghi.org> | 2019-07-07 17:12:48 +0100 |
| commit | d2887a9239992fb8e62b0506000b0a8e7d48aaca (patch) | |
| tree | bc1157598442b02be6061880ecfb2a9227b68ee9 /reproduce/software/make/high-level.mk | |
| parent | 468bd2a8c5f3b8d94eddeef641ce90d863b7bebc (diff) | |
FFTW single-precision library built
Until now, we were using the `--enable-single' configure-time option of
FFTW to build its single-precision library. The reason was that according
to the output of `./configure --help', this was equivalent to
`--enable-float'. However, the single precision library wasn't being built.
As a result, on systems that already had it, SExtractor would use the
system's library and on other systems, it would simply not pass the
configure step.
With this commit, we are now using `--enable-float' which fixes this
problem and installs the `libfftw3f*' libraries (showing that it is not
equal to `--enable-single'!). Also, some optimization flags were added to
hopefully make it faster.
This issue was found thanks to Zahra Sharbaf and Hamed Altafi.
This fixes bug #56588.
Diffstat (limited to 'reproduce/software/make/high-level.mk')
| -rw-r--r-- | reproduce/software/make/high-level.mk | 15 |
1 files changed, 7 insertions, 8 deletions
diff --git a/reproduce/software/make/high-level.mk b/reproduce/software/make/high-level.mk index 1c306ac..4b8bcfd 100644 --- a/reproduce/software/make/high-level.mk +++ b/reproduce/software/make/high-level.mk @@ -364,16 +364,15 @@ $(ibidir)/cairo: $(tdir)/cairo-$(cairo-version).tar.xz \ && echo "Cairo $(cairo-version)" > $@ $(ibidir)/fftw: $(tdir)/fftw-$(fftw-version).tar.gz - # In order to build single and double precission libraries of - # `fftw', installation of `fftw' is done twice. First time is to - # build single precission float libraries and second time is for - # building the default double precission float libraries + # FFTW's single and double precission libraries must be built + # independently: for the the single-precision library, we need to + # add the `--enable-float' option. We will build this first, then + # the default double-precision library. + confop="--enable-shared --enable-threads --enable-avx --enable-sse2" $(call gbuild, $<, fftw-$(fftw-version), static, \ - --enable-shared enable-threads \ - --enable-single --enable-type-prefix) \ + $$confop --enable-float) \ && $(call gbuild, $<, fftw-$(fftw-version), static, \ - --enable-shared --enable-threads \ - --enable-type-prefix) \ + $$confop) \ && cp $(dtexdir)/fftw.tex $(ictdir)/ \ && echo "FFTW $(fftw-version) \citep{fftw}" > $@ |