| Message ID | 5543378B.2090600@arm.com |
|---|---|
| State | New |
| Headers | show |
On 1 May 2015 at 09:21, Kyrill Tkachov <kyrylo.tkachov@arm.com> wrote: > 2015-05-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com> > > * config/aarch64/aarch64.c (aarch64_rtx_costs): Handle FLOAT and > UNSIGNED_FLOAT. OK /Marcus
commit 3a14409ce63cac482e48ab731cff0267f7b2ce53
Author: Kyrylo Tkachov <kyrylo.tkachov@arm.com>
Date: Wed Mar 11 12:04:13 2015 +0000
[AArch64] Handle FLOAT and UNSIGNED_FLOAT in rtx costs
diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
index d1635f4..8661896 100644
--- a/gcc/config/aarch64/aarch64.c
+++ b/gcc/config/aarch64/aarch64.c
@@ -6497,6 +6497,12 @@ cost_plus:
*cost += rtx_cost (op2, FMA, 2, speed);
return true;
+ case FLOAT:
+ case UNSIGNED_FLOAT:
+ if (speed)
+ *cost += extra_cost->fp[mode == DFmode].fromint;
+ return false;
+
case FLOAT_EXTEND:
if (speed)
*cost += extra_cost->fp[mode == DFmode].widen;
Hi all, This patch adds handling of the FLOAT and UNSIGNED_FLOAT cases in rtx costs that map down to the scvtf and ucvtf instructions respectively. We use the fromint cost field for them. Tested on aarch64-none-elf. Ok for trunk? Thanks, Kyrill 2015-05-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com> * config/aarch64/aarch64.c (aarch64_rtx_costs): Handle FLOAT and UNSIGNED_FLOAT.