diff mbox series

, Add future cost placeholder to PowerPC

Message ID 20190801194756.GA11962@ibm-toto.the-meissners.org
State New
Headers show
Series , Add future cost placeholder to PowerPC | expand

Commit Message

Michael Meissner Aug. 1, 2019, 7:47 p.m. UTC 
Add future_cost placeholder.

Currently, the -mcpu=future uses the power9 costs.  This patch separates out
the cost to a separate structure so that in the future we can tweak the costs
as needed.

I have bootstraped versions of this patch multiple times on a power8 little
endian system.  Can I check this into the FSF trunk?

2019-08-01  Michael Meissner  <meissner@linux.ibm.com>

	* config/rs6000/rs6000.c (future_cost): Add separate cost
	structure for a 'future' machine.
	(rs6000_option_override_internal): If -mcpu=future, use the future
	costs instead of power9.

Comments

Segher Boessenkool Aug. 1, 2019, 9:26 p.m. UTC | #1 
On Thu, Aug 01, 2019 at 03:47:56PM -0400, Michael Meissner wrote:
> Add future_cost placeholder.
> 
> Currently, the -mcpu=future uses the power9 costs.  This patch separates out
> the cost to a separate structure so that in the future we can tweak the costs
> as needed.

Is there any reason for this?  At least when it actually points to the
power9 costs it is clear it is just a placeholder.


Segher
Michael Meissner Aug. 2, 2019, 8:43 p.m. UTC | #2 
On Thu, Aug 01, 2019 at 04:26:11PM -0500, Segher Boessenkool wrote:
> On Thu, Aug 01, 2019 at 03:47:56PM -0400, Michael Meissner wrote:
> > Add future_cost placeholder.
> > 
> > Currently, the -mcpu=future uses the power9 costs.  This patch separates out
> > the cost to a separate structure so that in the future we can tweak the costs
> > as needed.
> 
> Is there any reason for this?  At least when it actually points to the
> power9 costs it is clear it is just a placeholder.

The main reason was you had asked for it many, many months ago on the internal
branch.  But until we actually need to change the values, it doesn't need to be
changed immediately.
diff mbox series

Patch

Index: gcc/config/rs6000/rs6000.c
===================================================================
--- gcc/config/rs6000/rs6000.c	(revision 273954)
+++ gcc/config/rs6000/rs6000.c	(working copy)
@@ -1079,6 +1079,27 @@  struct processor_costs power9_cost = {
   COSTS_N_INSNS (3),	/* SF->DF convert */
 };
 
+/* Instruction costs on FUTURE processors.  At present, this is a place holder,
+   until more accurate information can be provided.  */
+static const
+struct processor_costs future_cost = {
+  COSTS_N_INSNS (3),	/* mulsi */
+  COSTS_N_INSNS (3),	/* mulsi_const */
+  COSTS_N_INSNS (3),	/* mulsi_const9 */
+  COSTS_N_INSNS (3),	/* muldi */
+  COSTS_N_INSNS (8),	/* divsi */
+  COSTS_N_INSNS (12),	/* divdi */
+  COSTS_N_INSNS (3),	/* fp */
+  COSTS_N_INSNS (3),	/* dmul */
+  COSTS_N_INSNS (13),	/* sdiv */
+  COSTS_N_INSNS (18),	/* ddiv */
+  128,			/* cache line size */
+  32,			/* l1 cache */
+  512,			/* l2 cache */
+  8,			/* prefetch streams */
+  COSTS_N_INSNS (3),	/* SF->DF convert */
+};
+
 /* Instruction costs on POWER A2 processors.  */
 static const
 struct processor_costs ppca2_cost = {
@@ -4487,10 +4508,13 @@  rs6000_option_override_internal (bool gl
 	break;
 
       case PROCESSOR_POWER9:
-      case PROCESSOR_FUTURE:
 	rs6000_cost = &power9_cost;
 	break;
 
+      case PROCESSOR_FUTURE:
+	rs6000_cost = &future_cost;
+	break;
+
       case PROCESSOR_PPCA2:
 	rs6000_cost = &ppca2_cost;
 	break;