[Pkg-julia-devel] dsfmt ftbfs on ppc64el

[Graham Inggs]

On 8 November 2015 at 19:01, Peter Colberg <peter at colberg.org> wrote:
> Yes, that is the easiest solution. It was worth a try, but the Altivec
> variant of the recursion in dSFMT-common.h assumes big endian byte order.

Shall we enable altivec for powerpc?  I sent a mail earlier but it
didn't make it to the list, see below.

> Before I forget, after uploading the revised utf8proc and dsfmt, could
> you also push the debian/* git tags for versions -1 and -2?

Will do.


On 8 November 2015 at 14:38, Graham Inggs <ginggs at debian.org> wrote:
> Just out of interest, I repeated the benchmark test on powerpc, results below.
> It appears there would be an advantage to changing lines 11-13 of
> debian/rules to the following for now:
>
> ifneq (,$(filter $(DEB_HOST_ARCH),ppc64 powerpc))
> CONFIGURE_FLAGS += --enable-altivec
> endif
>
>
> without --enable altivec:
>
> (sid_powerpc-dchroot)ginggs at partch:~/dsfmt/dsfmt-2.2.3+dfsg$ ./test-M19937 -s
> consumed time for generating 100000000 randoms.
> ST BLOCK [0, 1) AVE: 278ms.
> ST BLOCK (0, 1] AVE: 277ms.
> ST BLOCK (0, 1) AVE: 545ms.
> ST BLOCK [1, 2) AVE: 256ms.
> ST SEQ [0, 1) 1 AVE: 417ms.
> ST SEQ [0, 1) 2 AVE: 429ms.
> total = 500014655.815776
> ST SEQ (0, 1] 1 AVE: 447ms.
> ST SEQ (0, 1] 2 AVE: 377ms.
> total = 500035344.184224
> ST SEQ (0, 1) 1 AVE:1365ms.
> ST SEQ (0, 1) 2 AVE:1314ms.
> total = 500014655.815776
> ST SEQ [1, 2) 1 AVE: 417ms.
> ST SEQ [1, 2) 2 AVE: 391ms.
> total = 1500064655.815183
>
>
> with --enable-altivec:
>
> (sid_powerpc-dchroot)ginggs at partch:~/dsfmt/dsfmt-2.2.3+dfsg$ ./test-M19937 -s
> consumed time for generating 100000000 randoms.
> ST BLOCK [0, 1) AVE: 178ms.
> ST BLOCK (0, 1] AVE: 178ms.
> ST BLOCK (0, 1) AVE: 631ms.
> ST BLOCK [1, 2) AVE: 134ms.
> ST SEQ [0, 1) 1 AVE: 300ms.
> ST SEQ [0, 1) 2 AVE: 311ms.
> total = 500014655.815776
> ST SEQ (0, 1] 1 AVE: 329ms.
> ST SEQ (0, 1] 2 AVE: 259ms.
> total = 500035344.184224
> ST SEQ (0, 1) 1 AVE:1276ms.
> ST SEQ (0, 1) 2 AVE:1201ms.
> total = 500014655.815776
> ST SEQ [1, 2) 1 AVE: 299ms.
> ST SEQ [1, 2) 2 AVE: 274ms.
> total = 1500064655.815183