Add wider_subreg_mode helper functions

Message ID 87o9r0t7wt.fsf@linaro.org
State New
Headers show
Series
  • Add wider_subreg_mode helper functions
Related show

Commit Message

Richard Sandiford Aug. 28, 2017, 8:26 a.m.
This patch adds helper functions that say which of the two modes
involved in a subreg is the larger, preferring the outer mode in
the event of a tie.  It also converts IRA and reload to track modes
instead of byte sizes, since this is slightly more convenient when
variable-sized modes are added later.

Tested on aarch64-linux-gnu and x86_64-linux-gnu, and by checking
there was no change in the testsuite assembly output for at least
one target per CPU.  OK to install?

Richard


2017-08-28  Richard Sandiford  <richard.sandiford@linaro.org>
	    Alan Hayward  <alan.hayward@arm.com>
	    David Sherwood  <david.sherwood@arm.com>

gcc/
	* rtl.h (wider_subreg_mode): New function.
	* ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
	rather than an unsigned int *.
	* ira-color.c (regno_max_ref_width): Replace with...
	(regno_max_ref_mode): ...this new variable.
	(coalesced_pseudo_reg_slot_compare): Update accordingly.
	Use wider_subreg_mode.
	(ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
	rather than an unsigned int *.
	* lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
	(process_alt_operands): Likewise.
	(invariant_p): Likewise.
	* lra-spills.c (assign_mem_slot): Likewise.
	(add_pseudo_to_slot): Likewise.
	* lra.c (collect_non_operand_hard_regs): Likewise.
	(add_regs_to_insn_regno_info): Likewise.
	* reload1.c (regno_max_ref_width): Replace with...
	(regno_max_ref_mode): ...this new variable.
	(reload): Update accordingly.  Update call to
	ira_sort_regnos_for_alter_reg.
	(alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
	(init_eliminable_invariants): Update to use regno_max_ref_mode.
	(scan_paradoxical_subregs): Likewise.

Comments

Jeff Law Oct. 13, 2017, 3:59 p.m. | #1
On 08/28/2017 02:26 AM, Richard Sandiford wrote:
> This patch adds helper functions that say which of the two modes

> involved in a subreg is the larger, preferring the outer mode in

> the event of a tie.  It also converts IRA and reload to track modes

> instead of byte sizes, since this is slightly more convenient when

> variable-sized modes are added later.

> 

> Tested on aarch64-linux-gnu and x86_64-linux-gnu, and by checking

> there was no change in the testsuite assembly output for at least

> one target per CPU.  OK to install?

> 

> Richard

> 

> 

> 2017-08-28  Richard Sandiford  <richard.sandiford@linaro.org>

> 	    Alan Hayward  <alan.hayward@arm.com>

> 	    David Sherwood  <david.sherwood@arm.com>

> 

> gcc/

> 	* rtl.h (wider_subreg_mode): New function.

> 	* ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *

> 	rather than an unsigned int *.

> 	* ira-color.c (regno_max_ref_width): Replace with...

> 	(regno_max_ref_mode): ...this new variable.

> 	(coalesced_pseudo_reg_slot_compare): Update accordingly.

> 	Use wider_subreg_mode.

> 	(ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *

> 	rather than an unsigned int *.

> 	* lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.

> 	(process_alt_operands): Likewise.

> 	(invariant_p): Likewise.

> 	* lra-spills.c (assign_mem_slot): Likewise.

> 	(add_pseudo_to_slot): Likewise.

> 	* lra.c (collect_non_operand_hard_regs): Likewise.

> 	(add_regs_to_insn_regno_info): Likewise.

> 	* reload1.c (regno_max_ref_width): Replace with...

> 	(regno_max_ref_mode): ...this new variable.

> 	(reload): Update accordingly.  Update call to

> 	ira_sort_regnos_for_alter_reg.

> 	(alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.

> 	(init_eliminable_invariants): Update to use regno_max_ref_mode.

> 	(scan_paradoxical_subregs): Likewise.

OK.
jeff

Patch

Index: gcc/rtl.h
===================================================================
--- gcc/rtl.h	2017-08-28 09:23:51.181220876 +0100
+++ gcc/rtl.h	2017-08-28 09:23:56.230220860 +0100
@@ -2840,6 +2840,24 @@  subreg_lowpart_offset (machine_mode oute
 				     GET_MODE_SIZE (innermode));
 }
 
+/* Given that a subreg has outer mode OUTERMODE and inner mode INNERMODE,
+   return the mode that is big enough to hold both the outer and inner
+   values.  Prefer the outer mode in the event of a tie.  */
+
+inline machine_mode
+wider_subreg_mode (machine_mode outermode, machine_mode innermode)
+{
+  return partial_subreg_p (outermode, innermode) ? innermode : outermode;
+}
+
+/* Likewise for subreg X.  */
+
+inline machine_mode
+wider_subreg_mode (const_rtx x)
+{
+  return wider_subreg_mode (GET_MODE (x), GET_MODE (SUBREG_REG (x)));
+}
+
 extern unsigned int subreg_size_highpart_offset (unsigned int, unsigned int);
 
 /* Return the SUBREG_BYTE for an OUTERMODE highpart of an INNERMODE value.  */
Index: gcc/ira.h
===================================================================
--- gcc/ira.h	2017-08-28 09:22:51.676686496 +0100
+++ gcc/ira.h	2017-08-28 09:23:56.227520860 +0100
@@ -195,7 +195,7 @@  extern void ira_set_pseudo_classes (bool
 extern void ira_expand_reg_equiv (void);
 extern void ira_update_equiv_info_by_shuffle_insn (int, int, rtx_insn *);
 
-extern void ira_sort_regnos_for_alter_reg (int *, int, unsigned int *);
+extern void ira_sort_regnos_for_alter_reg (int *, int, machine_mode *);
 extern void ira_mark_allocation_change (int);
 extern void ira_mark_memory_move_deletion (int, int);
 extern bool ira_reassign_pseudos (int *, int, HARD_REG_SET, HARD_REG_SET *,
Index: gcc/ira-color.c
===================================================================
--- gcc/ira-color.c	2017-08-28 09:22:51.676686496 +0100
+++ gcc/ira-color.c	2017-08-28 09:23:56.226620860 +0100
@@ -3908,7 +3908,7 @@  coalesced_pseudo_reg_freq_compare (const
 
 /* Widest width in which each pseudo reg is referred to (via subreg).
    It is used for sorting pseudo registers.  */
-static unsigned int *regno_max_ref_width;
+static machine_mode *regno_max_ref_mode;
 
 /* Sort pseudos according their slot numbers (putting ones with
   smaller numbers first, or last when the frame pointer is not
@@ -3921,7 +3921,7 @@  coalesced_pseudo_reg_slot_compare (const
   ira_allocno_t a1 = ira_regno_allocno_map[regno1];
   ira_allocno_t a2 = ira_regno_allocno_map[regno2];
   int diff, slot_num1, slot_num2;
-  int total_size1, total_size2;
+  machine_mode mode1, mode2;
 
   if (a1 == NULL || ALLOCNO_HARD_REGNO (a1) >= 0)
     {
@@ -3936,11 +3936,11 @@  coalesced_pseudo_reg_slot_compare (const
   if ((diff = slot_num1 - slot_num2) != 0)
     return (frame_pointer_needed
 	    || (!FRAME_GROWS_DOWNWARD) == STACK_GROWS_DOWNWARD ? diff : -diff);
-  total_size1 = MAX (PSEUDO_REGNO_BYTES (regno1),
-		     regno_max_ref_width[regno1]);
-  total_size2 = MAX (PSEUDO_REGNO_BYTES (regno2),
-		     regno_max_ref_width[regno2]);
-  if ((diff = total_size2 - total_size1) != 0)
+  mode1 = wider_subreg_mode (PSEUDO_REGNO_MODE (regno1),
+			     regno_max_ref_mode[regno1]);
+  mode2 = wider_subreg_mode (PSEUDO_REGNO_MODE (regno2),
+			     regno_max_ref_mode[regno2]);
+  if ((diff = GET_MODE_SIZE (mode2) - GET_MODE_SIZE (mode1)) != 0)
     return diff;
   return regno1 - regno2;
 }
@@ -4144,7 +4144,7 @@  coalesce_spill_slots (ira_allocno_t *spi
    reload.  */
 void
 ira_sort_regnos_for_alter_reg (int *pseudo_regnos, int n,
-			       unsigned int *reg_max_ref_width)
+			       machine_mode *reg_max_ref_mode)
 {
   int max_regno = max_reg_num ();
   int i, regno, num, slot_num;
@@ -4225,10 +4225,14 @@  ira_sort_regnos_for_alter_reg (int *pseu
 	  ira_assert (ALLOCNO_HARD_REGNO (a) < 0);
 	  ALLOCNO_HARD_REGNO (a) = -slot_num;
 	  if (internal_flag_ira_verbose > 3 && ira_dump_file != NULL)
-	    fprintf (ira_dump_file, " a%dr%d(%d,%d)",
-		     ALLOCNO_NUM (a), ALLOCNO_REGNO (a), ALLOCNO_FREQ (a),
-		     MAX (PSEUDO_REGNO_BYTES (ALLOCNO_REGNO (a)),
-			  reg_max_ref_width[ALLOCNO_REGNO (a)]));
+	    {
+	      machine_mode mode = wider_subreg_mode
+		(PSEUDO_REGNO_MODE (ALLOCNO_REGNO (a)),
+		 reg_max_ref_mode[ALLOCNO_REGNO (a)]);
+	      fprintf (ira_dump_file, " a%dr%d(%d,%d)",
+		       ALLOCNO_NUM (a), ALLOCNO_REGNO (a), ALLOCNO_FREQ (a),
+		       GET_MODE_SIZE (mode));
+	    }
 
 	  if (a == allocno)
 	    break;
@@ -4239,7 +4243,7 @@  ira_sort_regnos_for_alter_reg (int *pseu
   ira_spilled_reg_stack_slots_num = slot_num - 1;
   ira_free (spilled_coalesced_allocnos);
   /* Sort regnos according the slot numbers.  */
-  regno_max_ref_width = reg_max_ref_width;
+  regno_max_ref_mode = reg_max_ref_mode;
   qsort (pseudo_regnos, n, sizeof (int), coalesced_pseudo_reg_slot_compare);
   FOR_EACH_ALLOCNO (a, ai)
     ALLOCNO_ADD_DATA (a) = NULL;
Index: gcc/lra-constraints.c
===================================================================
--- gcc/lra-constraints.c	2017-08-28 09:23:40.551863444 +0100
+++ gcc/lra-constraints.c	2017-08-28 09:23:56.227520860 +0100
@@ -1772,10 +1772,9 @@  uses_hard_regs_p (rtx x, HARD_REG_SET se
   mode = GET_MODE (x);
   if (code == SUBREG)
     {
+      mode = wider_subreg_mode (x);
       x = SUBREG_REG (x);
       code = GET_CODE (x);
-      if (GET_MODE_SIZE (GET_MODE (x)) > GET_MODE_SIZE (mode))
-	mode = GET_MODE (x);
     }
 
   if (REG_P (x))
@@ -1953,10 +1952,8 @@  process_alt_operands (int only_alternati
       biggest_mode[nop] = GET_MODE (op);
       if (GET_CODE (op) == SUBREG)
 	{
+	  biggest_mode[nop] = wider_subreg_mode (op);
 	  operand_reg[nop] = reg = SUBREG_REG (op);
-	  if (GET_MODE_SIZE (biggest_mode[nop])
-	      < GET_MODE_SIZE (GET_MODE (reg)))
-	    biggest_mode[nop] = GET_MODE (reg);
 	}
       if (! REG_P (reg))
 	operand_reg[nop] = NULL_RTX;
@@ -5665,8 +5662,7 @@  invariant_p (const_rtx x)
     {
       x = SUBREG_REG (x);
       code = GET_CODE (x);
-      if (GET_MODE_SIZE (GET_MODE (x)) > GET_MODE_SIZE (mode))
-	mode = GET_MODE (x);
+      mode = wider_subreg_mode (mode, GET_MODE (x));
     }
 
   if (MEM_P (x))
Index: gcc/lra-spills.c
===================================================================
--- gcc/lra-spills.c	2017-08-28 09:23:42.709863597 +0100
+++ gcc/lra-spills.c	2017-08-28 09:23:56.227520860 +0100
@@ -134,8 +134,7 @@  assign_mem_slot (int i)
   machine_mode mode = GET_MODE (regno_reg_rtx[i]);
   HOST_WIDE_INT inherent_size = PSEUDO_REGNO_BYTES (i);
   machine_mode wider_mode
-    = (GET_MODE_SIZE (mode) >= GET_MODE_SIZE (lra_reg_info[i].biggest_mode)
-       ? mode : lra_reg_info[i].biggest_mode);
+    = wider_subreg_mode (mode, lra_reg_info[i].biggest_mode);
   HOST_WIDE_INT total_size = GET_MODE_SIZE (wider_mode);
   HOST_WIDE_INT adjust = 0;
 
@@ -311,10 +310,8 @@  add_pseudo_to_slot (int regno, int slot_
      and a total size which provides room for paradoxical subregs.
      We need to make sure the size and alignment of the slot are
      sufficient for both.  */
-  machine_mode mode = (GET_MODE_SIZE (PSEUDO_REGNO_MODE (regno))
-		       >= GET_MODE_SIZE (lra_reg_info[regno].biggest_mode)
-		       ? PSEUDO_REGNO_MODE (regno)
-		       : lra_reg_info[regno].biggest_mode);
+  machine_mode mode = wider_subreg_mode (PSEUDO_REGNO_MODE (regno),
+					 lra_reg_info[regno].biggest_mode);
   unsigned int align = spill_slot_alignment (mode);
   slots[slot_num].align = MAX (slots[slot_num].align, align);
   slots[slot_num].size = MAX (slots[slot_num].size, GET_MODE_SIZE (mode));
Index: gcc/lra.c
===================================================================
--- gcc/lra.c	2017-08-28 09:23:38.966863332 +0100
+++ gcc/lra.c	2017-08-28 09:23:56.229320860 +0100
@@ -831,12 +831,11 @@  collect_non_operand_hard_regs (rtx *x, l
   subreg_p = false;
   if (code == SUBREG)
     {
+      mode = wider_subreg_mode (op);
       if (read_modify_subreg_p (op))
 	subreg_p = true;
       op = SUBREG_REG (op);
       code = GET_CODE (op);
-      if (GET_MODE_SIZE (mode) < GET_MODE_SIZE (GET_MODE (op)))
-	mode = GET_MODE (op);
     }
   if (REG_P (op))
     {
@@ -1426,12 +1425,11 @@  add_regs_to_insn_regno_info (lra_insn_re
   subreg_p = false;
   if (GET_CODE (x) == SUBREG)
     {
+      mode = wider_subreg_mode (x);
       if (read_modify_subreg_p (x))
 	subreg_p = true;
       x = SUBREG_REG (x);
       code = GET_CODE (x);
-      if (GET_MODE_SIZE (mode) < GET_MODE_SIZE (GET_MODE (x)))
-	mode = GET_MODE (x);
     }
   if (REG_P (x))
     {
Index: gcc/reload1.c
===================================================================
--- gcc/reload1.c	2017-08-28 09:23:42.710863597 +0100
+++ gcc/reload1.c	2017-08-28 09:23:56.229320860 +0100
@@ -97,8 +97,8 @@  #define spill_indirect_levels			\
    in the current insn.  */
 static HARD_REG_SET reg_is_output_reload;
 
-/* Widest width in which each pseudo reg is referred to (via subreg).  */
-static unsigned int *reg_max_ref_width;
+/* Widest mode in which each pseudo reg is referred to (via subreg).  */
+static machine_mode *reg_max_ref_mode;
 
 /* Vector to remember old contents of reg_renumber before spilling.  */
 static short *reg_old_renumber;
@@ -832,7 +832,7 @@  reload (rtx_insn *first, int global)
   if (ira_conflicts_p)
     /* Ask IRA to order pseudo-registers for better stack slot
        sharing.  */
-    ira_sort_regnos_for_alter_reg (temp_pseudo_reg_arr, n, reg_max_ref_width);
+    ira_sort_regnos_for_alter_reg (temp_pseudo_reg_arr, n, reg_max_ref_mode);
 
   for (i = 0; i < n; i++)
     alter_reg (temp_pseudo_reg_arr[i], -1, false);
@@ -1254,7 +1254,7 @@  reload (rtx_insn *first, int global)
   /* Indicate that we no longer have known memory locations or constants.  */
   free_reg_equiv ();
 
-  free (reg_max_ref_width);
+  free (reg_max_ref_mode);
   free (reg_old_renumber);
   free (pseudo_previous_regs);
   free (pseudo_forbidden_regs);
@@ -2146,8 +2146,9 @@  alter_reg (int i, int from_reg, bool don
       machine_mode mode = GET_MODE (regno_reg_rtx[i]);
       unsigned int inherent_size = PSEUDO_REGNO_BYTES (i);
       unsigned int inherent_align = GET_MODE_ALIGNMENT (mode);
-      unsigned int total_size = MAX (inherent_size, reg_max_ref_width[i]);
-      unsigned int min_align = reg_max_ref_width[i] * BITS_PER_UNIT;
+      machine_mode wider_mode = wider_subreg_mode (mode, reg_max_ref_mode[i]);
+      unsigned int total_size = GET_MODE_SIZE (wider_mode);
+      unsigned int min_align = GET_MODE_BITSIZE (reg_max_ref_mode[i]);
       int adjust = 0;
 
       something_was_spilled = true;
@@ -4085,9 +4086,9 @@  init_eliminable_invariants (rtx_insn *fi
 
   grow_reg_equivs ();
   if (do_subregs)
-    reg_max_ref_width = XCNEWVEC (unsigned int, max_regno);
+    reg_max_ref_mode = XCNEWVEC (machine_mode, max_regno);
   else
-    reg_max_ref_width = NULL;
+    reg_max_ref_mode = NULL;
 
   num_eliminable_invariants = 0;
 
@@ -4406,7 +4407,7 @@  finish_spills (int global)
   return something_changed;
 }
 
-/* Find all paradoxical subregs within X and update reg_max_ref_width.  */
+/* Find all paradoxical subregs within X and update reg_max_ref_mode.  */
 
 static void
 scan_paradoxical_subregs (rtx x)
@@ -4429,13 +4430,14 @@  scan_paradoxical_subregs (rtx x)
       return;
 
     case SUBREG:
-      if (REG_P (SUBREG_REG (x))
-	  && (GET_MODE_SIZE (GET_MODE (x))
-	      > reg_max_ref_width[REGNO (SUBREG_REG (x))]))
+      if (REG_P (SUBREG_REG (x)))
 	{
-	  reg_max_ref_width[REGNO (SUBREG_REG (x))]
-	    = GET_MODE_SIZE (GET_MODE (x));
-	  mark_home_live_1 (REGNO (SUBREG_REG (x)), GET_MODE (x));
+	  unsigned int regno = REGNO (SUBREG_REG (x));
+	  if (partial_subreg_p (reg_max_ref_mode[regno], GET_MODE (x)))
+	    {
+	      reg_max_ref_mode[regno] = GET_MODE (x);
+	      mark_home_live_1 (regno, GET_MODE (x));
+	    }
 	}
       return;