#66 atlas new archdef for ppc64le

[mnormand]

atlas new archdef for ppc64le

with suggested attached atlas-new_archdef_for_ppc64le.patch

[mnormand]

I did not received any comments on this patch proposal.
Does it mean your are expecting something else to accept the change ?

[R. Clint Whaley]

I have not adopted this patch yet, and will have to figure it out when I apply the rest of the patches. We only want an ARCH name difference when it changes the kernels. Since I assume this will definitely change the compiler flags, and may also change any assembly (not sure if the ABI changes enough to effect PowerPC asssembly?), I'm not sure yet what is needed here.

[R. Clint Whaley]

This patch applied in basefiles, will be in 3.11.30.

I take it the macros powerpc64 & ORDER_LITTLE_ENDIAN are autodefined by your patched gcc? Have those patches been incorp to mainline gcc?

[mnormand]

-------- Original Message --------
Subject: [math-atlas:patches] #66 atlas new archdef for ppc64le
From: R. Clint Whaley rwhaley@users.sf.net
To: [math-atlas:patches] 66@patches.math-atlas.p.re.sf.net
Date: 19/08/2014 23:59

This patch applied in basefiles, will be in 3.11.30.

I take it the macros powerpc64 & ORDER_LITTLE_ENDIAN are autodefined by your patched gcc? Have those patches been incorp to mainline gcc?

Correct the two are defined as part of the new gcc 4.9 that support powerpc64le-linux platform https://gcc.gnu.org/gcc-4.9/changes.html

--
Michel Normand

[Dan Horak]

The correct check for little endian powerpc64 is

#if defined(__powerpc64__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)

[R. Clint Whaley]

OK, I think this should fixed as of 3.11.35 which fully support vsx & little endian linux/pwr8. Please reopen if not the case.