Peter Barada wrote:
> I figured out how to make it work using LEGITIMIZE_RELOAD_ADDRESS(at
> least for gcc-3.4.3) via:
I seem to recall rth opposing any use of LEGITIMIZE_RELOAD_ADDRESS
to ensure *correctness* of generated code; L_R_A's only purpose
is supposed to provide additional optimizations
>Peter Barada wrote:
>> I'd like to make the reload look like:
>> (set (reg:SI y) (plus:SI (reg_SI 16) (const_int 32832)))
>> (set (reg:DF x) (mem:DF (reg:SI y)))
>
>Reload already knows how to make this transformation, so it should not
>be necessary to resort to LEGITIMIZE_RELOAD_ADDRESS. This
Peter Barada wrote:
I'd like to make the reload look like:
(set (reg:SI y) (plus:SI (reg_SI 16) (const_int 32832)))
(set (reg:DF x) (mem:DF (reg:SI y)))
Reload already knows how to make this transformation, so it should not
be necessary to resort to LEGITIMIZE_RELOAD_ADDRESS. This is only
needed
I'm in the midst of fixing the m68k prologue/epilogue code for
ColdFire and its FPU, and stumbled across a problem. The following
code when compiled with -O2 -mcfv4e -fomit-frame-pointer (with the v4e cod in):
double func(int i1, int i2, int i3, int i4, double a, double b)
{
int stuff[8192];
