[AArch64][6/14] ARMv8.2-A FP16 reduction vector intrinsics

[Jiong Wang <jiong dot wang at foss dot arm dot com>]

This patch add ARMv8.2-A FP16 reduction vector intrinsics.

gcc/
2016-07-07  Jiong Wang <jiong.wang@arm.com>

        * config/aarch64/arm_neon.h (vmaxv_f16): New.
        (vmaxvq_f16): Likewise.
        (vminv_f16): Likewise.
        (vminvq_f16): Likewise.
        (vmaxnmv_f16): Likewise.
        (vmaxnmvq_f16): Likewise.
        (vminnmv_f16): Likewise.
        (vminnmvq_f16): Likewise.
        * config/aarch64/iterators.md (vp): Support HF modes.

>From 514e5d195867d2f53fac50804748976626748f81 Mon Sep 17 00:00:00 2001
From: Jiong Wang <jiong.wang@arm.com>
Date: Wed, 8 Jun 2016 10:23:17 +0100
Subject: [PATCH 06/14] [6/14] ARMv8.2 FP16 reduction vector intrinsics

---
 gcc/config/aarch64/aarch64-simd-builtins.def |  8 ++---
 gcc/config/aarch64/aarch64-simd.md           | 12 +++----
 gcc/config/aarch64/arm_neon.h                | 50 ++++++++++++++++++++++++++++
 gcc/config/aarch64/iterators.md              |  7 ++--
 4 files changed, 65 insertions(+), 12 deletions(-)

diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index 6ff5063..64c5f86 100644
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -234,12 +234,12 @@
   BUILTIN_VALL (UNOP, reduc_plus_scal_, 10)
 
   /* Implemented by reduc_<maxmin_uns>_scal_<mode> (producing scalar).  */
-  BUILTIN_VDQIF (UNOP, reduc_smax_scal_, 10)
-  BUILTIN_VDQIF (UNOP, reduc_smin_scal_, 10)
+  BUILTIN_VDQIF_F16 (UNOP, reduc_smax_scal_, 10)
+  BUILTIN_VDQIF_F16 (UNOP, reduc_smin_scal_, 10)
   BUILTIN_VDQ_BHSI (UNOPU, reduc_umax_scal_, 10)
   BUILTIN_VDQ_BHSI (UNOPU, reduc_umin_scal_, 10)
-  BUILTIN_VDQF (UNOP, reduc_smax_nan_scal_, 10)
-  BUILTIN_VDQF (UNOP, reduc_smin_nan_scal_, 10)
+  BUILTIN_VHSDF (UNOP, reduc_smax_nan_scal_, 10)
+  BUILTIN_VHSDF (UNOP, reduc_smin_nan_scal_, 10)
 
   /* Implemented by <maxmin><mode>3.
      smax variants map to fmaxnm,
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index c0600df..d5b25fa 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -2073,8 +2073,8 @@
 ;; gimple_fold'd to the REDUC_(MAX|MIN)_EXPR tree code.  (This is FP smax/smin).
 (define_expand "reduc_<maxmin_uns>_scal_<mode>"
   [(match_operand:<VEL> 0 "register_operand")
-   (unspec:VDQF [(match_operand:VDQF 1 "register_operand")]
-		FMAXMINV)]
+   (unspec:VHSDF [(match_operand:VHSDF 1 "register_operand")]
+		  FMAXMINV)]
   "TARGET_SIMD"
   {
     rtx elt = GEN_INT (ENDIAN_LANE_N (<MODE>mode, 0));
@@ -2121,12 +2121,12 @@
 )
 
 (define_insn "aarch64_reduc_<maxmin_uns>_internal<mode>"
- [(set (match_operand:VDQF 0 "register_operand" "=w")
-       (unspec:VDQF [(match_operand:VDQF 1 "register_operand" "w")]
-		    FMAXMINV))]
+ [(set (match_operand:VHSDF 0 "register_operand" "=w")
+       (unspec:VHSDF [(match_operand:VHSDF 1 "register_operand" "w")]
+		      FMAXMINV))]
  "TARGET_SIMD"
  "<maxmin_uns_op><vp>\\t%<Vetype>0, %1.<Vtype>"
-  [(set_attr "type" "neon_fp_reduc_minmax_<Vetype><q>")]
+  [(set_attr "type" "neon_fp_reduc_minmax_<stype><q>")]
 )
 
 ;; aarch64_simd_bsl may compile to any of bsl/bif/bit depending on register
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index b09a3a7..f3e5d0e 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -26638,6 +26638,56 @@ vmulxq_n_f16 (float16x8_t __a, float16_t __b)
   return vmulxq_f16 (__a, vdupq_n_f16 (__b));
 }
 
+/* ARMv8.2-A FP16 reduction vector intrinsics.  */
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vmaxv_f16 (float16x4_t __a)
+{
+  return __builtin_aarch64_reduc_smax_nan_scal_v4hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vmaxvq_f16 (float16x8_t __a)
+{
+  return __builtin_aarch64_reduc_smax_nan_scal_v8hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vminv_f16 (float16x4_t __a)
+{
+  return __builtin_aarch64_reduc_smin_nan_scal_v4hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vminvq_f16 (float16x8_t __a)
+{
+  return __builtin_aarch64_reduc_smin_nan_scal_v8hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vmaxnmv_f16 (float16x4_t __a)
+{
+  return __builtin_aarch64_reduc_smax_scal_v4hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vmaxnmvq_f16 (float16x8_t __a)
+{
+  return __builtin_aarch64_reduc_smax_scal_v8hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vminnmv_f16 (float16x4_t __a)
+{
+  return __builtin_aarch64_reduc_smin_scal_v4hf (__a);
+}
+
+__extension__ static __inline float16_t __attribute__ ((__always_inline__))
+vminnmvq_f16 (float16x8_t __a)
+{
+  return __builtin_aarch64_reduc_smin_scal_v8hf (__a);
+}
+
 #pragma GCC pop_options
 
 #undef __aarch64_vget_lane_any
diff --git a/gcc/config/aarch64/iterators.md b/gcc/config/aarch64/iterators.md
index 8d4dc6c..011b937 100644
--- a/gcc/config/aarch64/iterators.md
+++ b/gcc/config/aarch64/iterators.md
@@ -159,6 +159,8 @@
 
 ;; Vector modes except double int.
 (define_mode_iterator VDQIF [V8QI V16QI V4HI V8HI V2SI V4SI V2SF V4SF V2DF])
+(define_mode_iterator VDQIF_F16 [V8QI V16QI V4HI V8HI V2SI V4SI
+                                 V4HF V8HF V2SF V4SF V2DF])
 
 ;; Vector modes for S type.
 (define_mode_iterator VDQ_SI [V2SI V4SI])
@@ -760,8 +762,9 @@
 (define_mode_attr vp [(V8QI "v") (V16QI "v")
 		      (V4HI "v") (V8HI  "v")
 		      (V2SI "p") (V4SI  "v")
-		      (V2DI  "p") (V2DF  "p")
-		      (V2SF "p") (V4SF  "v")])
+		      (V2DI "p") (V2DF  "p")
+		      (V2SF "p") (V4SF  "v")
+		      (V4HF "v") (V8HF  "v")])
 
 (define_mode_attr vsi2qi [(V2SI "v8qi") (V4SI "v16qi")])
 (define_mode_attr VSI2QI [(V2SI "V8QI") (V4SI "V16QI")])
-- 
2.5.0