[linuxsh-dev] Question regarding -m4-nofpu usage

[M. R. B. <ma...@uw...>]

I remember seeing a brief thread on why -m4-nofpu was needed to compile
the kernel, but I don't remember the reasoning.  Regardless of the -m4...
option, the default gcc macros (gcc/config/sh/linux.h) don't properly
define the __SH4__ macros, etc. that would make sense with this option.

I understand the code generation selection macros within gcc, but this
seems like an obvious conflict of interest.  With -m4-nofpu defined in
arch/sh/Makefile, the kernel is forced to conditionalize for an SH3
target, which overrides target selection in `make config'.  This breaks
for me compiling for an SH4 target, but simply changing this to -m4 does
the trick.

Why is -m4-nofpu here?  And how does this benefit SH4 targets?  If
-D__SH4__ were added to the gcc option, would this option be better suited
as a kernel configuration option (e. g. CONFIG_SH_NOFPU)?

M. R.