From d14cf89b94299d6d66c150fbbb9899a5dd91e7d4 Mon Sep 17 00:00:00 2001
From: Kyrylo Tkachov <kyrylo.tkachov@arm.com>
Date: Tue, 2 Feb 2021 14:27:34 +0000
Subject: [PATCH] aarch64: Reimplement vrsqrte* intrinsics with builtins

Another very simple move from inline asm to builtins.
Only two intrinsics this time.

gcc/ChangeLog:

	* config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
	* config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
	* config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
	(vrsqrteq_u32): Likewise.
---
 gcc/config/aarch64/aarch64-simd-builtins.def |  3 +++
 gcc/config/aarch64/aarch64-simd.md           |  8 ++++++++
 gcc/config/aarch64/arm_neon.h                | 14 ++------------
 3 files changed, 13 insertions(+), 12 deletions(-)

diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index 2f50ecc6d2f..48e481c0f96 100644
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -705,6 +705,9 @@
   /* Implemented by aarch64_rsqrts<mode>.  */
   BUILTIN_VHSDF_HSDF (BINOP, rsqrts, 0, FP)
 
+  /* Implemented by aarch64_ursqrte<mode>.  */
+  BUILTIN_VDQ_SI (UNOPU, ursqrte, 0, NONE)
+
   /* Implemented by fabd<mode>3.  */
   BUILTIN_VHSDF_HSDF (BINOP, fabd, 3, FP)
 
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index eea5c87576d..767d6739794 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -755,6 +755,14 @@
   DONE;
 })
 
+(define_insn "aarch64_ursqrte<mode>"
+[(set (match_operand:VDQ_SI 0 "register_operand" "=w")
+      (unspec:VDQ_SI [(match_operand:VDQ_SI 1 "register_operand" "w")]
+		   UNSPEC_RSQRTE))]
+"TARGET_SIMD"
+"ursqrte\\t%<v>0<Vmtype>, %<v>1<Vmtype>"
+[(set_attr "type" "neon_fp_rsqrte_<stype><q>")])
+
 (define_insn "*aarch64_mul3_elt_to_64v2df"
   [(set (match_operand:DF 0 "register_operand" "=w")
      (mult:DF
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index e32f7ab9141..4b905d9d731 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -9387,24 +9387,14 @@ __extension__ extern __inline uint32x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vrsqrte_u32 (uint32x2_t __a)
 {
-  uint32x2_t __result;
-  __asm__ ("ursqrte %0.2s,%1.2s"
-           : "=w"(__result)
-           : "w"(__a)
-           : /* No clobbers */);
-  return __result;
+  return __builtin_aarch64_ursqrtev2si_uu (__a);
 }
 
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vrsqrteq_u32 (uint32x4_t __a)
 {
-  uint32x4_t __result;
-  __asm__ ("ursqrte %0.4s,%1.4s"
-           : "=w"(__result)
-           : "w"(__a)
-           : /* No clobbers */);
-  return __result;
+  return __builtin_aarch64_ursqrtev4si_uu (__a);
 }
 
 __extension__ extern __inline int8x16_t
-- 
2.30.2