Re: reduction of GP registers in MIPS backend

["kernel coder" <lhrkernelcoder at gmail dot com>]

> This violates an assumption in mips.c:override_options that GPRs come in pairs.
> Thus you end up with (reg/v:DI 24 $24 [orig:55 res2 ] [55]), which does not
> satisfy the 'd' constraint.

Would you please give a bit more explaination of "GPRs come in pairs"
and  where this assumption is being checked in function
override_options.

thanks,
shaz

On 7/31/06, Joern RENNECKE <joern.rennecke@st.com> wrote:
> In http://gcc.gnu.org/ml/gcc/2006-07/msg00632.html, you wrote:
>
> dp-bit.c: In function '__muldf3':
> dp-bit.c:953: error: insn does not satisfy its constraints:
> (insn 677 231 616 19 dp-bit.c:871 (set (reg/v:DI 24 $24 [orig:55 res2 ] [55])
>        (reg:DI 2 $2)) 3 {*movdi_32bit} (nil)
>    (nil))
> dp-bit.c:953: internal compiler error: in
> reload_cse_simplify_operands, at postreload.c:394
>
> please read the documentation in gcc/gcc/doc/ ; this is explained in tm.texi,
> under HARD_REGNO_MODE_OK.
>
> > { 0x01ffffff, 0x00000000, },  /* integer registers */ \
>
> This violates an assumption in mips.c:override_options that GPRs come in pairs.
> Thus you end up with (reg/v:DI 24 $24 [orig:55 res2 ] [55]), which does not
> satisfy the 'd' constraint.