LoongArch: Optimize initializing fp resgister to zero

In LoongArch, currently uses instruction movgr2fr.{d|w} to move zero
from fixed-point register to floating-pointer regsiter for initializing
fp register to zero. When LSX or LASX is enabled, we can use instruction
vxor.v which has lower latency than instruction movgr2fr.{d|w} to set fp
register to zero directly.

gcc/ChangeLog:

	* config/loongarch/loongarch.cc (loongarch_output_move):
	Optimize instructions for initializing fp regsiter to zero.

gcc/testsuite/ChangeLog:

	* gcc.target/loongarch/mov-zero-1.c: New test.
	* gcc.target/loongarch/mov-zero-2.c: New test.

diff --git a/gcc/config/loongarch/loongarch.cc b/gcc/config/loongarch/loongarch.cc
index 89237c377e7707f312d82e87b3c21b878a2a1d11..d506354c48a0f5ead877b9475a2445f4e72d2959 100644 (file)
--- a/gcc/config/loongarch/loongarch.cc
+++ b/gcc/config/loongarch/loongarch.cc
@@ -4772,6 +4772,8 @@ loongarch_output_move (rtx dest, rtx src)
                      gcc_unreachable ();
                    }
                }
+             if (ISA_HAS_LSX && src == CONST0_RTX (GET_MODE (src)))
+               return "vxor.v\t%w0,%w0,%w0";
 
              return dbl_p ? "movgr2fr.d\t%0,%z1" : "movgr2fr.w\t%0,%z1";
            }

diff --git a/gcc/testsuite/gcc.target/loongarch/mov-zero-1.c b/gcc/testsuite/gcc.target/loongarch/mov-zero-1.c
new file mode 100644 (file)
index 0000000..4744f2f
--- /dev/null
+++ b/gcc/testsuite/gcc.target/loongarch/mov-zero-1.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mlsx" } */
+/* { dg-final { scan-assembler-times "vxor\\.v" 2 } } */
+
+double
+get_double_zero ()
+{
+  return 0;
+}
+
+float
+get_float_zero ()
+{
+  return 0;
+}

diff --git a/gcc/testsuite/gcc.target/loongarch/mov-zero-2.c b/gcc/testsuite/gcc.target/loongarch/mov-zero-2.c
new file mode 100644 (file)
index 0000000..6cb4805
--- /dev/null
+++ b/gcc/testsuite/gcc.target/loongarch/mov-zero-2.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-lsx" } */
+/* { dg-final { scan-assembler-times "movgr2fr" 2 } } */
+
+double
+get_double_zero ()
+{
+  return 0;
+}
+
+float
+get_float_zero ()
+{
+  return 0;
+}