Thread (6 messages) flat view 6 messages, 3 authors, 2013-11-21

Re: [RFC][PATCH] powerpc/CoreNet64: compile with CONFIG_E{5, 6}500_CPU well

From: “tiejun.chen” <hidden>
Date: 2013-11-21 12:15:37
Subsystem: linux for powerpc (32-bit and 64-bit), the rest · Maintainers: Madhavan Srinivasan, Linus Torvalds

On 11/21/2013 12:41 AM, Scott Wood wrote:
On Wed, 2013-11-20 at 16:35 +0800, Tiejun Chen wrote:
quoted
CONFIG_ALTIVEC is always enabled for CoreNet64.
In the defconfig perhaps, but this isn't a generally true statement.
Yes, but I think we should avoid this probable scenario :)
quoted
And if we select CONFIG_E{5,6}500_CPU this may introduce -mcpu=e500mc64
into $CFLAGS. But Altivec and Spe options not allowed with
e500mc64, so :
Sigh.
quoted
   CC      arch/powerpc/lib/xor_vmx.o
arch/powerpc/lib/xor_vmx.c:1:0: error: AltiVec not supported in this target
make[1]: *** [arch/powerpc/lib/xor_vmx.o] Error 1
make: *** [arch/powerpc/lib] Error 2

Signed-off-by: Tiejun Chen <redacted>
---
  arch/powerpc/lib/Makefile |    3 +++
  1 file changed, 3 insertions(+)
diff --git a/arch/powerpc/lib/Makefile b/arch/powerpc/lib/Makefile
index 95a20e1..641a77d 100644
--- a/arch/powerpc/lib/Makefile
+++ b/arch/powerpc/lib/Makefile
@@ -40,5 +40,8 @@ obj-y			+= code-patching.o
  obj-y			+= feature-fixups.o
  obj-$(CONFIG_FTR_FIXUP_SELFTEST) += feature-fixups-test.o

+# Altivec and Spe options not allowed with e500mc64 in GCC.
+ifeq ($(call cc-option-yn,-mcpu=e500mc64),n)
  obj-$(CONFIG_ALTIVEC)	+= xor_vmx.o
  CFLAGS_xor_vmx.o += -maltivec -mabi=altivec
+endif
This does not seem like the right fix.  What if GCC supports both
-mcpu=e500mc64 and -mcpu=e6500, and we're using the latter?  Or for that
I can understand what you mean, but in current kernel, -mcpu=e500mc64 should be 
excluded from -mcpu=e6500,

arch/powerpc/Makefile:

E5500_CPU := $(call cc-option,-mcpu=e500mc64,-mcpu=powerpc64)
CFLAGS-$(CONFIG_E5500_CPU) += $(E5500_CPU)
CFLAGS-$(CONFIG_E6500_CPU) += $(call cc-option,-mcpu=e6500,$(E5500_CPU))

But unfortunately, another place also use the same option,

lib/raid6/Makefile:

raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o
...
ifeq ($(CONFIG_ALTIVEC),y)
altivec_flags := -maltivec -mabi=altivec
endif

Looks we have to do something in this common Makefile file as well, but it may 
be a bit ugly if still judge some cpu-specific flags...

So what about this version?
diff --git a/arch/powerpc/Makefile b/arch/powerpc/Makefile
index 607acf5..872a85c 100644
--- a/arch/powerpc/Makefile
+++ b/arch/powerpc/Makefile
@@ -127,7 +127,12 @@ CFLAGS-$(CONFIG_POWER5_CPU) += $(call cc-option,-mcpu=power5)
  CFLAGS-$(CONFIG_POWER6_CPU) += $(call cc-option,-mcpu=power6)
  CFLAGS-$(CONFIG_POWER7_CPU) += $(call cc-option,-mcpu=power7)

+# Altivec and Spe options not allowed with e500mc64 in GCC.
+ifeq ($(CONFIG_ALTIVEC),)
  E5500_CPU := $(call cc-option,-mcpu=e500mc64,-mcpu=powerpc64)
+else
+E5500_CPU := -mcpu=powerpc64
+endif
  CFLAGS-$(CONFIG_E5500_CPU) += $(E5500_CPU)
  CFLAGS-$(CONFIG_E6500_CPU) += $(call cc-option,-mcpu=e6500,$(E5500_CPU))
Tiejun
matter, if we're using -mcpu=whatever-ibm-chip-has-this?

Plus, wouldn't you need to do something to prevent code in that file
from being called?

-Scott


Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help