This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[patch,avr] Fix PR80929, work around middle-end PR81444


Hi, this patch restores some of the divmod handling.

It addresses two issues:

1) rtx_costs might be called with code = LSHIFTRT for a mul_highpart. This is the case when outer_code = TRUNCATE. This patch uses a new function to compute the costs for that case (also used for code = TRUNCATE).

2) Due to PR81444, the middle-end is calling the cost functions with strange RTXes like

  (truncate:HI (lshiftrt:PSI (mult:PSI (zero_extend:PSI (reg:TI 42))
                                       (zero_extend:PSI (reg:TI 42)))
                             (const_int 16 [0x10])))

i.e. completely messed up modes.  The correct RTX would be

  (truncate:HI (lshiftrt:SI (mult:SI (zero_extend:SI (reg:HI 42))
                                     (zero_extend:SI (reg:HI 42)))
                            (const_int 16 [0x10])))

The is worked around by also considering PSImode in the cost computation.

FYI, I am currently also just reg-testing a fix for PR81444, cf.

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81444#c1

Ok to apply the patch below (trunk and v7)?

Johann

	PR 80929
	* config/avr/avr.c (avr_mul_highpart_cost): New static function.
	(avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
	[LSHIFTRT, outer_code = TRUNCATE]: Same.
Index: config/avr/avr.c
===================================================================
--- config/avr/avr.c	(revision 250156)
+++ config/avr/avr.c	(working copy)
@@ -10709,6 +10709,33 @@ avr_memory_move_cost (machine_mode mode,
 }
 
 
+/* Cost for mul highpart.  X is a LSHIFTRT, i.e. the outer TRUNCATE is
+   already stripped off.  */
+
+static int
+avr_mul_highpart_cost (rtx x, int)
+{
+  if (AVR_HAVE_MUL
+      && LSHIFTRT == GET_CODE (x)
+      && MULT == GET_CODE (XEXP (x, 0))
+      && CONST_INT_P (XEXP (x, 1)))
+    {
+      // This is the wider mode.
+      machine_mode mode = GET_MODE (x);
+  
+      // The middle-end might still have PR81444, i.e. it is calling the cost
+      // functions with strange modes.  Fix this now by also considering
+      // PSImode (should actually be SImode instead).
+      if (HImode == mode || PSImode == mode || SImode == mode)
+        {
+          return COSTS_N_INSNS (2);
+        }
+    }
+
+  return 10000;
+}
+
+
 /* Mutually recursive subroutine of avr_rtx_cost for calculating the
    cost of an RTX operand given its context.  X is the rtx of the
    operand, MODE is its mode, and OUTER is the rtx_code of this
@@ -10748,7 +10775,7 @@ avr_operand_rtx_cost (rtx x, machine_mod
    In either case, *TOTAL contains the cost result.  */
 
 static bool
-avr_rtx_costs_1 (rtx x, machine_mode mode, int outer_code ATTRIBUTE_UNUSED,
+avr_rtx_costs_1 (rtx x, machine_mode mode, int outer_code,
                  int opno ATTRIBUTE_UNUSED, int *total, bool speed)
 {
   enum rtx_code code = GET_CODE (x);
@@ -11402,6 +11429,12 @@ avr_rtx_costs_1 (rtx x, machine_mode mod
       return true;
 
     case LSHIFTRT:
+      if (outer_code == TRUNCATE)
+        {
+          *total = avr_mul_highpart_cost (x, speed);
+          return true;
+        }
+
       switch (mode)
 	{
 	case QImode:
@@ -11579,16 +11612,10 @@ avr_rtx_costs_1 (rtx x, machine_mode mod
       return true;
 
     case TRUNCATE:
-      if (AVR_HAVE_MUL
-          && LSHIFTRT == GET_CODE (XEXP (x, 0))
-          && MULT == GET_CODE (XEXP (XEXP (x, 0), 0))
-          && CONST_INT_P (XEXP (XEXP (x, 0), 1)))
+      if (LSHIFTRT == GET_CODE (XEXP (x, 0)))
         {
-          if (QImode == mode || HImode == mode)
-            {
-              *total = COSTS_N_INSNS (2);
-              return true;
-            }
+          *total = avr_mul_highpart_cost (XEXP (x, 0), speed);
+          return true;
         }
       break;
 

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]