* Hiremath, Vaibhav [off-list ref] [120710 02:15]:

On Tue, Jul 10, 2012 at 14:11:08, Tony Lindgren wrote:

quoted

* Hiremath, Vaibhav [off-list ref] [120627 13:37]:

quoted

--- a/arch/arm/plat-omap/include/plat/cpu.h
+++ b/arch/arm/plat-omap/include/plat/cpu.h

@@ -42,7 +42,14 @@
 #define OMAP2_DEVICE_TYPE_GP           3
 #define OMAP2_DEVICE_TYPE_BAD          4

+#ifdef CONFIG_ARCH_OMAP2PLUS
 int omap_type(void);
+#else
+inline static int omap_type(void)
+{
+	/* Always return GP, since it is not being used anyway for omap1 */
+	return OMAP2_DEVICE_TYPE_GP;
+}
+#endif

 /*
  * omap_rev bits:

Just to follow-up on this, looks like we need to postpone this
for v3.7. While at it, we need to make sure all soc_is_omapxxxx()
are false by default.

Just FYI, I had submitted V2 of this patch.

quoted

What you're suggesting above would also depend on some MULTI_ARCH
option set if multiple ARM architectures are selected. Above patch
would only optimize things when ARCH_OMAP2PLUS alone is selected.

Isn't this only applicable for ARCH_OMAP2PLUS and OMAP1?? And above patch 
take cares of both, as OMAP1 and 2 will not be build together, since 
compiler flags itself will be different.

Right, but we also need to deal with
 

Can you elaborate on how things will break if multiple ARM architectures are 
selected?

I am not sure whether I am following you completely.

We have quite a few initcalls that exit early with various soc_is, cpu_is,
cpu_class_is tests that would fail for multi zImage kernels. Some of these
have been patched away with additional calls in MACHINE_START, but some
still remain.

We probably always need to return false for all soc_is_omapxxxx() unless
the SoC detection is initialized in id.c.

Regards,

Tony