ra-colorize.c:1068: `inv_reg_alloc_order' undeclared

ra-colorize.c:1068: `inv_reg_alloc_order' undeclared

[John David Anglin]

The following error occurs building a vax-ultrix cross:

gcc -c -DIN_GCC -DCROSS_COMPILE   -g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wtraditional -pedantic -Wno-long-long  -DHAVE_CONFIG_H    -I. -I. -I../../gcc/gcc -I../../gcc/gcc/. -I../../gcc/gcc/config -I../../gcc/gcc/../include ../../gcc/gcc/ra-colorize.c -o ra-colorize.o
../../gcc/gcc/ra-colorize.c: In function `get_free_reg':
../../gcc/gcc/ra-colorize.c:1068: `inv_reg_alloc_order' undeclared (first use in this function)
../../gcc/gcc/ra-colorize.c:1068: (Each undeclared identifier is reported only once

The error occurs because inv_reg_alloc_order is only defined when
REG_ALLOC_ORDER is defined.  None of the vax ports define this macro.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

[fix] for: ra-colorize.c:1068: `inv_reg_alloc_order' undeclared

[Michael Matz]

Hi,

On Wed, 24 Jul 2002, John David Anglin wrote:

> The error occurs because inv_reg_alloc_order is only defined when
> REG_ALLOC_ORDER is defined.  None of the vax ports define this macro.

Fixed thusly (will checkin after bootstrap/regtest):


Ciao,
Michael.
-- 
        * ra-colorize.c (inv_reg_alloc_order): New macro.
        (free_reg): Use it.

Index: ra-colorize.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/ra-colorize.c,v
retrieving revision 1.2
diff -u -p -r1.2 ra-colorize.c
--- ra-colorize.c	15 Jul 2002 14:07:05 -0000	1.2
+++ ra-colorize.c	24 Jul 2002 20:09:14 -0000
@@ -1031,6 +1031,13 @@ color_usable_p (c, dont_begin_colors, fr
   return 0;
 }

+/* I don't want to clutter up the actual code with ifdef's.  */
+#ifdef REG_ALLOC_ORDER
+#define inv_reg_alloc_order(c) inv_reg_alloc_order[c]
+#else
+#define inv_reg_alloc_order(c) c
+#endif
+
 /* Searches in FREE_COLORS for a block of hardregs of the right length
    for MODE, which doesn't begin at a hardreg mentioned in DONT_BEGIN_COLORS.
    If it needs more than one hardreg it prefers blocks beginning
@@ -1065,16 +1072,16 @@ get_free_reg (dont_begin_colors, free_co
 	  {
 	    if (size < 2 || (c & 1) == 0)
 	      {
-		if (inv_reg_alloc_order[c] < pref_reg_order)
+		if (inv_reg_alloc_order (c) < pref_reg_order)
 		  {
 		    pref_reg = c;
-		    pref_reg_order = inv_reg_alloc_order[c];
+		    pref_reg_order = inv_reg_alloc_order (c);
 		  }
 	      }
-	    else if (inv_reg_alloc_order[c] < last_resort_reg_order)
+	    else if (inv_reg_alloc_order (c) < last_resort_reg_order)
 	      {
 		last_resort_reg = c;
-		last_resort_reg_order = inv_reg_alloc_order[c];
+		last_resort_reg_order = inv_reg_alloc_order (c);
 	      }
 	  }
 	else

Re: [fix] for: ra-colorize.c:1068: `inv_reg_alloc_order' undeclared

[John David Anglin]

> +/* I don't want to clutter up the actual code with ifdef's.  */
> +#ifdef REG_ALLOC_ORDER
> +#define inv_reg_alloc_order(c) inv_reg_alloc_order[c]
> +#else
> +#define inv_reg_alloc_order(c) c

Should the above be

#define inv_reg_alloc_order(c) FIRST_PSEUDO_REGISTER - 1 - (c)

to obtain the inverse of lowest to highest ordering?

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: [fix] for: ra-colorize.c:1068: `inv_reg_alloc_order' undeclared

[Michael Matz]

Hi,

On Wed, 24 Jul 2002, John David Anglin wrote:

> > +/* I don't want to clutter up the actual code with ifdef's.  */
> > +#ifdef REG_ALLOC_ORDER
> > +#define inv_reg_alloc_order(c) inv_reg_alloc_order[c]
> > +#else
> > +#define inv_reg_alloc_order(c) c
>
> Should the above be
>
> #define inv_reg_alloc_order(c) FIRST_PSEUDO_REGISTER - 1 - (c)
>
> to obtain the inverse of lowest to highest ordering?

No, it's not inverse ordering, but inverse mapping, so that
inv_order[order[i]] == i.  As this platform doesn't define REG_ALLOC_ORDER
it makes sense to begin allocating from the beginning, i.e. order[i]==i.
Ergo we have inv_order[order[i]] == inv_order[i] == i ;-)


Ciao,
Michael.