[PATCH 0/5] aarch64: Add ACLE intrinsics codegen support for lrcpc3 instructions

[PATCH 0/5] aarch64: Add ACLE intrinsics codegen support for lrcpc3 instructions

[Victor Do Nascimento]

Given the introduction of the third set of Release Consistency
processor consistent (RCpc) memory model-compliant instructions in
the form of FEAT_LRCPC3 as an optional extension from Armv8.2-a
onward, this patch series adds the RCPC3 ACLE Neon intrinsics,
thus enabling the use of the architectural feature in C.

These intrinsics enable the use of the new LDAP1 and STL1
instructions and are given single and twin-lane variants for unsigned,
signed and poly 64-bit values, in the form of the following
builtin-functions:

  * vldap1_lane_{u|s|p}64
  * vldap1q_lane_{u|s|p}64
  * ldp1_lane_{u|s|p}64
  * ldp1q_lane_{u|s|p}64

Bootstrapped and regression tested on aarch64-none-linux-gnu.

Victor Do Nascimento (5):
  aarch64: rcpc3: Add +rcpc3 extension
  aarch64: rcpc3: Add relevant iterators to handle Neon intrinsics
  aarch64: rcpc3: Add Neon ACLE intrinsics
  aarch64: rcpc3: add Neon ACLE wrapper functions to `arm_neon.h'
  aarch64: rcpc3: Add intrinsics tests

 .../aarch64/aarch64-option-extensions.def     |   1 +
 gcc/config/aarch64/aarch64-simd-builtins.def  |   7 +
 gcc/config/aarch64/aarch64-simd.md            |  65 +++++++++
 gcc/config/aarch64/aarch64.h                  |   4 +
 gcc/config/aarch64/aarch64.md                 |   2 +
 gcc/config/aarch64/arm_neon.h                 | 129 ++++++++++++++++++
 gcc/config/aarch64/iterators.md               |  25 ++--
 gcc/doc/invoke.texi                           |   4 +
 gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c |  47 +++++++
 9 files changed, 276 insertions(+), 8 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c

-- 
2.41.0

[PATCH 1/5] aarch64: rcpc3: Add +rcpc3 extension

[Victor Do Nascimento]

Given the optional LRCPC3 target support for Armv8.2-a cores onwards,
the +rcpc3 arch feature modifier is added to GCC's command-line options.

gcc/ChangeLog:

	* config/aarch64/aarch64-option-extensions.def (rcpc3): New.
	* config/aarch64/aarch64.h (AARCH64_ISA_RCPC3): Likewise.
	(TARGET_RCPC3): Likewise.
	* doc/invoke.texi (rcpc3): Document feature in AArch64 Options.
---
 gcc/config/aarch64/aarch64-option-extensions.def | 1 +
 gcc/config/aarch64/aarch64.h                     | 4 ++++
 gcc/doc/invoke.texi                              | 4 ++++
 3 files changed, 9 insertions(+)

diff --git a/gcc/config/aarch64/aarch64-option-extensions.def b/gcc/config/aarch64/aarch64-option-extensions.def
index 825f3bf7758..2ab94799d34 100644
--- a/gcc/config/aarch64/aarch64-option-extensions.def
+++ b/gcc/config/aarch64/aarch64-option-extensions.def
@@ -151,4 +151,5 @@ AARCH64_OPT_EXTENSION("mops", MOPS, (), (), (), "")
 
 AARCH64_OPT_EXTENSION("cssc", CSSC, (), (), (), "cssc")
 
+AARCH64_OPT_EXTENSION("rcpc3", RCPC3, (), (), (), "rcpc3")
 #undef AARCH64_OPT_EXTENSION
diff --git a/gcc/config/aarch64/aarch64.h b/gcc/config/aarch64/aarch64.h
index 2f0777a37ac..68bbaccef1a 100644
--- a/gcc/config/aarch64/aarch64.h
+++ b/gcc/config/aarch64/aarch64.h
@@ -213,6 +213,7 @@ enum class aarch64_feature : unsigned char {
 #define AARCH64_ISA_F64MM	   (aarch64_isa_flags & AARCH64_FL_F64MM)
 #define AARCH64_ISA_BF16	   (aarch64_isa_flags & AARCH64_FL_BF16)
 #define AARCH64_ISA_SB		   (aarch64_isa_flags & AARCH64_FL_SB)
+#define AARCH64_ISA_RCPC3	   (aarch64_isa_flags & AARCH64_FL_RCPC3)
 #define AARCH64_ISA_V8R		   (aarch64_isa_flags & AARCH64_FL_V8R)
 #define AARCH64_ISA_PAUTH	   (aarch64_isa_flags & AARCH64_FL_PAUTH)
 #define AARCH64_ISA_V9A		   (aarch64_isa_flags & AARCH64_FL_V9A)
@@ -344,6 +345,9 @@ enum class aarch64_feature : unsigned char {
    and sign-extending versions.*/
 #define TARGET_RCPC2 (AARCH64_ISA_RCPC8_4)
 
+/* RCPC3 LDAP1/STL1 loads/stores from Armv8.2-a.  */
+#define TARGET_RCPC3 (AARCH64_ISA_RCPC3)
+
 /* Apply the workaround for Cortex-A53 erratum 835769.  */
 #define TARGET_FIX_ERR_A53_835769	\
   ((aarch64_fix_a53_err835769 == 2)	\
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index 6e776a0faa1..ba28eb195ce 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -21028,6 +21028,10 @@ Enable the Flag Manipulation instructions Extension.
 Enable the Pointer Authentication Extension.
 @item cssc
 Enable the Common Short Sequence Compression instructions.
+@item rcpc3
+Enable the RCpc3 extension.  This enables the use of the LDAP1 and
+STL1 instructions for loads/stores of 64-bit values to and from SIMD
+register lanes, passing these on to the assembler.
 
 @end table
 
-- 
2.41.0

[PATCH 2/5] aarch64: rcpc3: Add relevant iterators to handle Neon intrinsics

[Victor Do Nascimento]

The LDAP1 and STL1 Neon ACLE intrinsics, operating on 64-bit data
values, operate on single-lane (Vt.1D) or twin-lane (Vt.2D) SIMD
register configurations, either in the DI or DF modes.  This leads to
the need for a mode iterator accounting for the V1DI, V1DF, V2DI and
V2DF modes.

This patch therefore introduces the new V12DIF mode iterator with
which to generate functions operating on signed 64-bit integer and
float values and V12DIUP for generating the unsigned and
polynomial-type counterparts.  Along with this, we modify the
associated mode attributes accordingly in order to allow for the
implementation of the relevant backend patterns for the intrinsics.

gcc/ChangeLog:

	* config/aarch64/iterators.md (V12DIF): New.
	(V12DUP): Likewise.
	(VEL): Add support for all V12DIF-associated modes.
	(Vetype): Add support for V1DI and V1DF.
	(Vel): Likewise.
---
 gcc/config/aarch64/iterators.md | 25 +++++++++++++++++--------
 1 file changed, 17 insertions(+), 8 deletions(-)

diff --git a/gcc/config/aarch64/iterators.md b/gcc/config/aarch64/iterators.md
index f9e2210095e..471438e27be 100644
--- a/gcc/config/aarch64/iterators.md
+++ b/gcc/config/aarch64/iterators.md
@@ -314,6 +314,12 @@
 ;; All byte modes.
 (define_mode_iterator VB [V8QI V16QI])
 
+;; 1 and 2 lane DI and DF modes.
+(define_mode_iterator V12DIF [V1DI V1DF V2DI V2DF])
+
+;; 1 and 2 lane DI mode for unsigned and poly types.
+(define_mode_iterator V12DIUP [V1DI V2DI])
+
 ;; 2 and 4 lane SI modes.
 (define_mode_iterator VS [V2SI V4SI])
 
@@ -1195,10 +1201,10 @@
 (define_mode_attr Vetype [(V8QI "b") (V16QI "b")
 			  (V4HI "h") (V8HI  "h")
 			  (V2SI "s") (V4SI  "s")
-			  (V2DI "d")
+			  (V2DI "d") (V1DI  "d")
 			  (V4HF "h") (V8HF  "h")
 			  (V2SF "s") (V4SF  "s")
-			  (V2DF "d")
+			  (V2DF "d") (V1DF  "d")
 			  (V2x8QI "b") (V2x4HI "h")
 			  (V2x2SI "s") (V2x1DI "d")
 			  (V2x4HF "h") (V2x2SF "s")
@@ -1358,10 +1364,12 @@
 (define_mode_attr VEL [(V8QI  "QI") (V16QI "QI")
 		       (V4HI "HI") (V8HI  "HI")
 		       (V2SI "SI") (V4SI  "SI")
-		       (DI   "DI") (V2DI  "DI")
+		       (DI   "DI") (V1DI  "DI")
+		       (V2DI  "DI")
 		       (V4HF "HF") (V8HF  "HF")
 		       (V2SF "SF") (V4SF  "SF")
-		       (DF   "DF") (V2DF  "DF")
+		       (DF   "DF") (V1DF  "DF")
+		       (V2DF  "DF")
 		       (SI   "SI") (HI    "HI")
 		       (QI   "QI")
 		       (V4BF "BF") (V8BF "BF")
@@ -1378,12 +1386,13 @@
 (define_mode_attr Vel [(V8QI "qi") (V16QI "qi")
 		       (V4HI "hi") (V8HI "hi")
 		       (V2SI "si") (V4SI "si")
-		       (DI   "di") (V2DI "di")
+		       (DI   "di") (V1DI "si")
+		       (V2DI "di")
 		       (V4HF "hf") (V8HF "hf")
 		       (V2SF "sf") (V4SF "sf")
-		       (V2DF "df") (DF   "df")
-		       (SI   "si") (HI   "hi")
-		       (QI   "qi")
+		       (V1DF "df") (V2DF "df")
+		       (DF   "df") (SI   "si")
+		       (HI   "hi") (QI   "qi")
 		       (V4BF "bf") (V8BF "bf")
 		       (VNx16QI "qi") (VNx8QI "qi") (VNx4QI "qi") (VNx2QI "qi")
 		       (VNx8HI "hi") (VNx4HI "hi") (VNx2HI "hi")
-- 
2.41.0

[PATCH 3/5] aarch64: rcpc3: Add Neon ACLE intrinsics

[Victor Do Nascimento]

Register the target specific builtins in `aarch64-simd-builtins.def'
and implement their associated backend patterns in `aarch64-simd.md'.

gcc/ChangeLog:

	* config/aarch64/aarch64-simd-builtins.def
	(vec_ldap1_lane): New.
	(vec_stl1_lane): Likewise.
	* config/aarch64/aarch64-simd.md
	(aarch64_vec_stl1_lanes<mode>_lane<Vel>): New.
	(aarch64_vec_stl1_lane<mode>): Likewise.
	(aarch64_vec_ldap1_lanes<mode>_lane<Vel>): Likewise.
	(aarch64_vec_ldap1_lane<mode>): Likewise.
---
 gcc/config/aarch64/aarch64-simd-builtins.def |  7 +++
 gcc/config/aarch64/aarch64-simd.md           | 65 ++++++++++++++++++++
 gcc/config/aarch64/aarch64.md                |  2 +
 3 files changed, 74 insertions(+)

diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index e2b94ad8247..0ae6c4ad41a 100644
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -43,6 +43,13 @@
    help describe the attributes (for example, pure) for the intrinsic
    function.  */
 
+  BUILTIN_V12DIF (LOADSTRUCT_LANE, vec_ldap1_lane, 0, ALL)
+  BUILTIN_V12DIUP (LOADSTRUCT_LANE_U, vec_ldap1_lane, 0, ALL)
+  BUILTIN_V12DIUP (LOADSTRUCT_LANE_P, vec_ldap1_lane, 0, ALL)
+  BUILTIN_V12DIF (STORESTRUCT_LANE, vec_stl1_lane, 0, ALL)
+  BUILTIN_V12DIUP (STORESTRUCT_LANE_U, vec_stl1_lane, 0, ALL)
+  BUILTIN_V12DIUP (STORESTRUCT_LANE_P, vec_stl1_lane, 0, ALL)
+
   BUILTIN_VDC (BINOP, combine, 0, AUTO_FP)
   BUILTIN_VD_I (BINOPU, combine, 0, NONE)
   BUILTIN_VDC_P (BINOPP, combine, 0, NONE)
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index 81ff5bad03d..79697336f61 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -7697,6 +7697,71 @@
   DONE;
 })
 
+;; Patterns for rcpc3 vector lane loads and stores.
+
+(define_insn "aarch64_vec_stl1_lanes<mode>_lane<Vel>"
+  [(set (match_operand:BLK 0 "aarch64_simd_struct_operand" "=Q")
+	(unspec:BLK [(match_operand:V12DIF 1 "register_operand" "w")
+		     (match_operand:SI 2 "immediate_operand" "i")]
+		     UNSPEC_STL1_LANE))]
+  "TARGET_RCPC3"
+  {
+    operands[2] = aarch64_endian_lane_rtx (<MODE>mode,
+					   INTVAL (operands[2]));
+    return "stl1\\t{%S1.<Vetype>}[%2], %0";
+  }
+  [(set_attr "type" "neon_store2_one_lane")]
+)
+
+(define_expand "aarch64_vec_stl1_lane<mode>"
+ [(match_operand:DI 0 "register_operand")
+  (match_operand:V12DIF 1 "register_operand")
+  (match_operand:SI 2 "immediate_operand")]
+  "TARGET_RCPC3"
+{
+  rtx mem = gen_rtx_MEM (BLKmode, operands[0]);
+  set_mem_size (mem, GET_MODE_SIZE (GET_MODE_INNER (<MODE>mode)));
+
+  aarch64_simd_lane_bounds (operands[2], 0,
+			    GET_MODE_NUNITS (<MODE>mode).to_constant (), NULL);
+  emit_insn (gen_aarch64_vec_stl1_lanes<mode>_lane<Vel> (mem,
+					operands[1], operands[2]));
+  DONE;
+})
+
+(define_insn "aarch64_vec_ldap1_lanes<mode>_lane<Vel>"
+  [(set (match_operand:V12DIF 0 "register_operand" "=w")
+	(unspec:V12DIF [
+		(match_operand:BLK 1 "aarch64_simd_struct_operand" "Q")
+		(match_operand:V12DIF 2 "register_operand" "0")
+		(match_operand:SI 3 "immediate_operand" "i")]
+		UNSPEC_LDAP1_LANE))]
+  "TARGET_RCPC3"
+  {
+    operands[3] = aarch64_endian_lane_rtx (<MODE>mode,
+					   INTVAL (operands[3]));
+    return "ldap1\\t{%S0.<Vetype>}[%3], %1";
+  }
+  [(set_attr "type" "neon_load2_one_lane")]
+)
+
+(define_expand "aarch64_vec_ldap1_lane<mode>"
+  [(match_operand:V12DIF 0 "register_operand")
+	(match_operand:DI 1 "register_operand")
+	(match_operand:V12DIF 2 "register_operand")
+	(match_operand:SI 3 "immediate_operand")]
+  "TARGET_RCPC3"
+{
+  rtx mem = gen_rtx_MEM (BLKmode, operands[1]);
+  set_mem_size (mem, GET_MODE_SIZE (GET_MODE_INNER (<MODE>mode)));
+
+  aarch64_simd_lane_bounds (operands[3], 0,
+			    GET_MODE_NUNITS (<MODE>mode).to_constant (), NULL);
+  emit_insn (gen_aarch64_vec_ldap1_lanes<mode>_lane<Vel> (operands[0],
+				mem, operands[2], operands[3]));
+  DONE;
+})
+
 (define_insn_and_split "aarch64_rev_reglist<mode>"
 [(set (match_operand:VSTRUCT_QD 0 "register_operand" "=&w")
 	(unspec:VSTRUCT_QD
diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md
index 5bb8c772be8..fb6de3b1fbf 100644
--- a/gcc/config/aarch64/aarch64.md
+++ b/gcc/config/aarch64/aarch64.md
@@ -295,6 +295,8 @@
     UNSPEC_LD1RO
     UNSPEC_SALT_ADDR
     UNSPECV_PATCHABLE_AREA
+    UNSPEC_LDAP1_LANE
+    UNSPEC_STL1_LANE
 ])
 
 (define_c_enum "unspecv" [
-- 
2.41.0

[PATCH 4/5] aarch64: rcpc3: add Neon ACLE wrapper functions to `arm_neon.h'

[Victor Do Nascimento]

Create the necessary mappings from the ACLE-defined Neon intrinsics
names[1] to the internal builtin function names.

[1] https://arm-software.github.io/acle/neon_intrinsics/advsimd.html

gcc/ChangeLog:

	* gcc/config/aarch64/arm_neon.h (vldap1_lane_u64): New.
	(vldap1q_lane_u64): Likewise.
	(vldap1_lane_s64): Likewise.
	(vldap1q_lane_s64): Likewise.
	(vldap1_lane_f64): Likewise.
	(vldap1q_lane_f64): Likewise.
	(vldap1_lane_p64): Likewise.
	(vldap1q_lane_p64): Likewise.
	(vstl1_lane_u64): Likewise.
	(vstl1q_lane_u64): Likewise.
	(vstl1_lane_s64): Likewise.
	(vstl1q_lane_s64): Likewise.
	(vstl1_lane_f64): Likewise.
	(vstl1q_lane_f64): Likewise.
	(vstl1_lane_p64): Likewise.
	(vstl1q_lane_p64): Likewise.
---
 gcc/config/aarch64/arm_neon.h | 129 ++++++++++++++++++++++++++++++++++
 1 file changed, 129 insertions(+)

diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 349f3167699..ef0d75e07ce 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -13446,6 +13446,135 @@ vld1q_lane_u64 (const uint64_t *__src, uint64x2_t __vec, const int __lane)
   return __aarch64_vset_lane_any (*__src, __vec, __lane);
 }
 
+#pragma GCC push_options
+#pragma GCC target ("+nothing+rcpc3+simd")
+
+/* vldap1_lane.  */
+
+__extension__ extern __inline uint64x1_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1_lane_u64 (const uint64_t *__src, uint64x1_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev1di_usus (
+	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
+}
+
+__extension__ extern __inline uint64x2_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1q_lane_u64 (const uint64_t *__src, uint64x2_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev2di_usus (
+	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
+}
+
+__extension__ extern __inline int64x1_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1_lane_s64 (const int64_t *__src, int64x1_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev1di (__src, __vec, __lane);
+}
+
+__extension__ extern __inline int64x2_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1q_lane_s64 (const int64_t *__src, int64x2_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev2di (__src, __vec, __lane);
+}
+
+__extension__ extern __inline float64x1_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1_lane_f64 (const float64_t *__src, float64x1_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev1df (__src, __vec, __lane);
+}
+
+__extension__ extern __inline float64x2_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1q_lane_f64 (const float64_t *__src, float64x2_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev2df (__src, __vec, __lane);
+}
+
+__extension__ extern __inline poly64x1_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1_lane_p64 (const poly64_t *__src, poly64x1_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev1di_psps (
+	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
+}
+
+__extension__ extern __inline poly64x2_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vldap1q_lane_p64 (const poly64_t *__src, poly64x2_t __vec, const int __lane)
+{
+  return __builtin_aarch64_vec_ldap1_lanev2di_psps (
+	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
+}
+
+/* vstl1_lane.  */
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1_lane_u64 (const uint64_t *__src, uint64x1_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev1di_sus ((__builtin_aarch64_simd_di *) __src,
+					   __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1q_lane_u64 (uint64_t *__src, uint64x2_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev2di_sus ((__builtin_aarch64_simd_di *) __src,
+					   __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1_lane_s64 (int64_t *__src, int64x1_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev1di (__src, __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1q_lane_s64 (int64_t *__src, int64x2_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev2di (__src, __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1_lane_f64 (float64_t *__src, float64x1_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev1df (__src, __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1q_lane_f64 (float64_t *__src, float64x2_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev2df (__src, __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1_lane_p64 (poly64_t *__src, poly64x1_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev1di_sps ((__builtin_aarch64_simd_di *) __src,
+					   __vec, __lane);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vstl1q_lane_p64 (poly64_t *__src, poly64x2_t __vec, const int __lane)
+{
+  __builtin_aarch64_vec_stl1_lanev2di_sps ((__builtin_aarch64_simd_di *) __src,
+					   __vec, __lane);
+}
+
+#pragma GCC pop_options
+
 /* vldn */
 
 __extension__ extern __inline int64x1x2_t
-- 
2.41.0

[PATCH 5/5] aarch64: rcpc3: Add intrinsics tests

[Victor Do Nascimento]

Add unit test to ensure that added intrinsics compile to the correct
`LDAP1 {Vt.D}[lane],[Xn]' and `STL1 {Vt.d}[lane],[Xn]' instructions.

gcc/testsuite/ChangeLog:

	* gcc.target/aarch64/acle/rcpc3.c: New.
---
 gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c | 47 +++++++++++++++++++
 1 file changed, 47 insertions(+)
 create mode 100644 gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c

diff --git a/gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c b/gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c
new file mode 100644
index 00000000000..689d047ab91
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c
@@ -0,0 +1,47 @@
+/* Test the rcpc3 ACLE intrinsics.  */
+/* { dg-do compile } */
+/* { dg-options "-O2 -march=armv8.2-a+rcpc3" } */
+#include <stdint.h>
+#include <arm_neon.h>
+
+#define TEST_LDAP(TYPE, T)						\
+  TYPE##x##1_t T##1_test (TYPE##_t const * ptr, TYPE##x##1_t src) {	\
+    return vldap1_lane_##T##64 (ptr, src, 0);				\
+  }
+
+#define TEST_LDAPQ(TYPE, T)						\
+  TYPE##x##2_t T##2_test (TYPE##_t const * ptr, TYPE##x##2_t src) {	\
+    return vldap1q_lane_##T##64 (ptr, src, 1);				\
+  }
+
+#define TEST_STL(TYPE, T)				    \
+  void T##1s_test (TYPE##_t * ptr, TYPE##x##1_t src) {	    \
+    vstl1_lane_##T##64 (ptr, src, 0);			    \
+  }
+
+#define TEST_STLQ(TYPE, T)		    \
+  void T##2s_test (TYPE##_t * ptr, TYPE##x##2_t src) {	    \
+    vstl1q_lane_##T##64 (ptr, src, 1);			    \
+  }
+
+TEST_LDAP (uint64, u);
+TEST_LDAP (int64, s);
+TEST_LDAP (float64, f);
+TEST_LDAP (poly64, p);
+/* { dg-final { scan-assembler-times {ldap1\t\{v\d.d\}\[0\], \[x\d\]} 4 } } */
+TEST_LDAPQ (uint64, u);
+TEST_LDAPQ (int64, s);
+TEST_LDAPQ (float64, f);
+TEST_LDAPQ (poly64, p);
+/* { dg-final { scan-assembler-times {ldap1\t\{v\d.d\}\[1\], \[x\d\]} 4 } } */
+
+TEST_STL (uint64, u);
+TEST_STL (int64, s);
+TEST_STL (float64, f);
+TEST_STL (poly64, p);
+/* { dg-final { scan-assembler-times {stl1\t\{v\d.d\}\[0\], \[x\d\]} 4 } } */
+TEST_STLQ (uint64, u);
+TEST_STLQ (int64, s);
+TEST_STLQ (float64, f);
+TEST_STLQ (poly64, p);
+/* { dg-final { scan-assembler-times {stl1\t\{v\d.d\}\[1\], \[x\d\]} 4 } } */
-- 
2.41.0

Re: [PATCH 2/5] aarch64: rcpc3: Add relevant iterators to handle Neon intrinsics

[Richard Sandiford]

Victor Do Nascimento <victor.donascimento@arm.com> writes:
> The LDAP1 and STL1 Neon ACLE intrinsics, operating on 64-bit data
> values, operate on single-lane (Vt.1D) or twin-lane (Vt.2D) SIMD
> register configurations, either in the DI or DF modes.  This leads to
> the need for a mode iterator accounting for the V1DI, V1DF, V2DI and
> V2DF modes.
>
> This patch therefore introduces the new V12DIF mode iterator with
> which to generate functions operating on signed 64-bit integer and
> float values and V12DIUP for generating the unsigned and
> polynomial-type counterparts.  Along with this, we modify the
> associated mode attributes accordingly in order to allow for the
> implementation of the relevant backend patterns for the intrinsics.
>
> gcc/ChangeLog:
>
> 	* config/aarch64/iterators.md (V12DIF): New.
> 	(V12DUP): Likewise.
> 	(VEL): Add support for all V12DIF-associated modes.
> 	(Vetype): Add support for V1DI and V1DF.
> 	(Vel): Likewise.
> ---
>  gcc/config/aarch64/iterators.md | 25 +++++++++++++++++--------
>  1 file changed, 17 insertions(+), 8 deletions(-)
>
> diff --git a/gcc/config/aarch64/iterators.md b/gcc/config/aarch64/iterators.md
> index f9e2210095e..471438e27be 100644
> --- a/gcc/config/aarch64/iterators.md
> +++ b/gcc/config/aarch64/iterators.md
> @@ -314,6 +314,12 @@
>  ;; All byte modes.
>  (define_mode_iterator VB [V8QI V16QI])
>  
> +;; 1 and 2 lane DI and DF modes.
> +(define_mode_iterator V12DIF [V1DI V1DF V2DI V2DF])
> +
> +;; 1 and 2 lane DI mode for unsigned and poly types.
> +(define_mode_iterator V12DIUP [V1DI V2DI])

Probably easiest just to call it V12DI, without the UP.  The same
iterator could be useful in other situations.

> +
>  ;; 2 and 4 lane SI modes.
>  (define_mode_iterator VS [V2SI V4SI])
>  
> @@ -1195,10 +1201,10 @@
>  (define_mode_attr Vetype [(V8QI "b") (V16QI "b")
>  			  (V4HI "h") (V8HI  "h")
>  			  (V2SI "s") (V4SI  "s")
> -			  (V2DI "d")
> +			  (V2DI "d") (V1DI  "d")
>  			  (V4HF "h") (V8HF  "h")
>  			  (V2SF "s") (V4SF  "s")
> -			  (V2DF "d")
> +			  (V2DF "d") (V1DF  "d")
>  			  (V2x8QI "b") (V2x4HI "h")
>  			  (V2x2SI "s") (V2x1DI "d")
>  			  (V2x4HF "h") (V2x2SF "s")
> @@ -1358,10 +1364,12 @@
>  (define_mode_attr VEL [(V8QI  "QI") (V16QI "QI")
>  		       (V4HI "HI") (V8HI  "HI")
>  		       (V2SI "SI") (V4SI  "SI")
> -		       (DI   "DI") (V2DI  "DI")
> +		       (DI   "DI") (V1DI  "DI")
> +		       (V2DI  "DI")

Very, very minor, but: would be good to have one fewer space before "DI",
so that the quotes line up.

>  		       (V4HF "HF") (V8HF  "HF")
>  		       (V2SF "SF") (V4SF  "SF")
> -		       (DF   "DF") (V2DF  "DF")
> +		       (DF   "DF") (V1DF  "DF")
> +		       (V2DF  "DF")

Same here.

OK for trunk with those changes, thanks.

Richard

>  		       (SI   "SI") (HI    "HI")
>  		       (QI   "QI")
>  		       (V4BF "BF") (V8BF "BF")
> @@ -1378,12 +1386,13 @@
>  (define_mode_attr Vel [(V8QI "qi") (V16QI "qi")
>  		       (V4HI "hi") (V8HI "hi")
>  		       (V2SI "si") (V4SI "si")
> -		       (DI   "di") (V2DI "di")
> +		       (DI   "di") (V1DI "si")
> +		       (V2DI "di")
>  		       (V4HF "hf") (V8HF "hf")
>  		       (V2SF "sf") (V4SF "sf")
> -		       (V2DF "df") (DF   "df")
> -		       (SI   "si") (HI   "hi")
> -		       (QI   "qi")
> +		       (V1DF "df") (V2DF "df")
> +		       (DF   "df") (SI   "si")
> +		       (HI   "hi") (QI   "qi")
>  		       (V4BF "bf") (V8BF "bf")
>  		       (VNx16QI "qi") (VNx8QI "qi") (VNx4QI "qi") (VNx2QI "qi")
>  		       (VNx8HI "hi") (VNx4HI "hi") (VNx2HI "hi")

Re: [PATCH 3/5] aarch64: rcpc3: Add Neon ACLE intrinsics

[Richard Sandiford]

Victor Do Nascimento <victor.donascimento@arm.com> writes:
> Register the target specific builtins in `aarch64-simd-builtins.def'
> and implement their associated backend patterns in `aarch64-simd.md'.
>
> gcc/ChangeLog:
>
> 	* config/aarch64/aarch64-simd-builtins.def
> 	(vec_ldap1_lane): New.
> 	(vec_stl1_lane): Likewise.
> 	* config/aarch64/aarch64-simd.md
> 	(aarch64_vec_stl1_lanes<mode>_lane<Vel>): New.
> 	(aarch64_vec_stl1_lane<mode>): Likewise.
> 	(aarch64_vec_ldap1_lanes<mode>_lane<Vel>): Likewise.
> 	(aarch64_vec_ldap1_lane<mode>): Likewise.

OK, thanks.

Richard

> ---
>  gcc/config/aarch64/aarch64-simd-builtins.def |  7 +++
>  gcc/config/aarch64/aarch64-simd.md           | 65 ++++++++++++++++++++
>  gcc/config/aarch64/aarch64.md                |  2 +
>  3 files changed, 74 insertions(+)
>
> diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
> index e2b94ad8247..0ae6c4ad41a 100644
> --- a/gcc/config/aarch64/aarch64-simd-builtins.def
> +++ b/gcc/config/aarch64/aarch64-simd-builtins.def
> @@ -43,6 +43,13 @@
>     help describe the attributes (for example, pure) for the intrinsic
>     function.  */
>  
> +  BUILTIN_V12DIF (LOADSTRUCT_LANE, vec_ldap1_lane, 0, ALL)
> +  BUILTIN_V12DIUP (LOADSTRUCT_LANE_U, vec_ldap1_lane, 0, ALL)
> +  BUILTIN_V12DIUP (LOADSTRUCT_LANE_P, vec_ldap1_lane, 0, ALL)
> +  BUILTIN_V12DIF (STORESTRUCT_LANE, vec_stl1_lane, 0, ALL)
> +  BUILTIN_V12DIUP (STORESTRUCT_LANE_U, vec_stl1_lane, 0, ALL)
> +  BUILTIN_V12DIUP (STORESTRUCT_LANE_P, vec_stl1_lane, 0, ALL)
> +
>    BUILTIN_VDC (BINOP, combine, 0, AUTO_FP)
>    BUILTIN_VD_I (BINOPU, combine, 0, NONE)
>    BUILTIN_VDC_P (BINOPP, combine, 0, NONE)
> diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
> index 81ff5bad03d..79697336f61 100644
> --- a/gcc/config/aarch64/aarch64-simd.md
> +++ b/gcc/config/aarch64/aarch64-simd.md
> @@ -7697,6 +7697,71 @@
>    DONE;
>  })
>  
> +;; Patterns for rcpc3 vector lane loads and stores.
> +
> +(define_insn "aarch64_vec_stl1_lanes<mode>_lane<Vel>"
> +  [(set (match_operand:BLK 0 "aarch64_simd_struct_operand" "=Q")
> +	(unspec:BLK [(match_operand:V12DIF 1 "register_operand" "w")
> +		     (match_operand:SI 2 "immediate_operand" "i")]
> +		     UNSPEC_STL1_LANE))]
> +  "TARGET_RCPC3"
> +  {
> +    operands[2] = aarch64_endian_lane_rtx (<MODE>mode,
> +					   INTVAL (operands[2]));
> +    return "stl1\\t{%S1.<Vetype>}[%2], %0";
> +  }
> +  [(set_attr "type" "neon_store2_one_lane")]
> +)
> +
> +(define_expand "aarch64_vec_stl1_lane<mode>"
> + [(match_operand:DI 0 "register_operand")
> +  (match_operand:V12DIF 1 "register_operand")
> +  (match_operand:SI 2 "immediate_operand")]
> +  "TARGET_RCPC3"
> +{
> +  rtx mem = gen_rtx_MEM (BLKmode, operands[0]);
> +  set_mem_size (mem, GET_MODE_SIZE (GET_MODE_INNER (<MODE>mode)));
> +
> +  aarch64_simd_lane_bounds (operands[2], 0,
> +			    GET_MODE_NUNITS (<MODE>mode).to_constant (), NULL);
> +  emit_insn (gen_aarch64_vec_stl1_lanes<mode>_lane<Vel> (mem,
> +					operands[1], operands[2]));
> +  DONE;
> +})
> +
> +(define_insn "aarch64_vec_ldap1_lanes<mode>_lane<Vel>"
> +  [(set (match_operand:V12DIF 0 "register_operand" "=w")
> +	(unspec:V12DIF [
> +		(match_operand:BLK 1 "aarch64_simd_struct_operand" "Q")
> +		(match_operand:V12DIF 2 "register_operand" "0")
> +		(match_operand:SI 3 "immediate_operand" "i")]
> +		UNSPEC_LDAP1_LANE))]
> +  "TARGET_RCPC3"
> +  {
> +    operands[3] = aarch64_endian_lane_rtx (<MODE>mode,
> +					   INTVAL (operands[3]));
> +    return "ldap1\\t{%S0.<Vetype>}[%3], %1";
> +  }
> +  [(set_attr "type" "neon_load2_one_lane")]
> +)
> +
> +(define_expand "aarch64_vec_ldap1_lane<mode>"
> +  [(match_operand:V12DIF 0 "register_operand")
> +	(match_operand:DI 1 "register_operand")
> +	(match_operand:V12DIF 2 "register_operand")
> +	(match_operand:SI 3 "immediate_operand")]
> +  "TARGET_RCPC3"
> +{
> +  rtx mem = gen_rtx_MEM (BLKmode, operands[1]);
> +  set_mem_size (mem, GET_MODE_SIZE (GET_MODE_INNER (<MODE>mode)));
> +
> +  aarch64_simd_lane_bounds (operands[3], 0,
> +			    GET_MODE_NUNITS (<MODE>mode).to_constant (), NULL);
> +  emit_insn (gen_aarch64_vec_ldap1_lanes<mode>_lane<Vel> (operands[0],
> +				mem, operands[2], operands[3]));
> +  DONE;
> +})
> +
>  (define_insn_and_split "aarch64_rev_reglist<mode>"
>  [(set (match_operand:VSTRUCT_QD 0 "register_operand" "=&w")
>  	(unspec:VSTRUCT_QD
> diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md
> index 5bb8c772be8..fb6de3b1fbf 100644
> --- a/gcc/config/aarch64/aarch64.md
> +++ b/gcc/config/aarch64/aarch64.md
> @@ -295,6 +295,8 @@
>      UNSPEC_LD1RO
>      UNSPEC_SALT_ADDR
>      UNSPECV_PATCHABLE_AREA
> +    UNSPEC_LDAP1_LANE
> +    UNSPEC_STL1_LANE
>  ])
>  
>  (define_c_enum "unspecv" [

Re: [PATCH 1/5] aarch64: rcpc3: Add +rcpc3 extension

[Richard Sandiford]

Victor Do Nascimento <victor.donascimento@arm.com> writes:
> Given the optional LRCPC3 target support for Armv8.2-a cores onwards,
> the +rcpc3 arch feature modifier is added to GCC's command-line options.
>
> gcc/ChangeLog:
>
> 	* config/aarch64/aarch64-option-extensions.def (rcpc3): New.
> 	* config/aarch64/aarch64.h (AARCH64_ISA_RCPC3): Likewise.
> 	(TARGET_RCPC3): Likewise.
> 	* doc/invoke.texi (rcpc3): Document feature in AArch64 Options.
> ---
>  gcc/config/aarch64/aarch64-option-extensions.def | 1 +
>  gcc/config/aarch64/aarch64.h                     | 4 ++++
>  gcc/doc/invoke.texi                              | 4 ++++
>  3 files changed, 9 insertions(+)
>
> diff --git a/gcc/config/aarch64/aarch64-option-extensions.def b/gcc/config/aarch64/aarch64-option-extensions.def
> index 825f3bf7758..2ab94799d34 100644
> --- a/gcc/config/aarch64/aarch64-option-extensions.def
> +++ b/gcc/config/aarch64/aarch64-option-extensions.def
> @@ -151,4 +151,5 @@ AARCH64_OPT_EXTENSION("mops", MOPS, (), (), (), "")
>  
>  AARCH64_OPT_EXTENSION("cssc", CSSC, (), (), (), "cssc")
>  
> +AARCH64_OPT_EXTENSION("rcpc3", RCPC3, (), (), (), "rcpc3")
>  #undef AARCH64_OPT_EXTENSION
> diff --git a/gcc/config/aarch64/aarch64.h b/gcc/config/aarch64/aarch64.h
> index 2f0777a37ac..68bbaccef1a 100644
> --- a/gcc/config/aarch64/aarch64.h
> +++ b/gcc/config/aarch64/aarch64.h
> @@ -213,6 +213,7 @@ enum class aarch64_feature : unsigned char {
>  #define AARCH64_ISA_F64MM	   (aarch64_isa_flags & AARCH64_FL_F64MM)
>  #define AARCH64_ISA_BF16	   (aarch64_isa_flags & AARCH64_FL_BF16)
>  #define AARCH64_ISA_SB		   (aarch64_isa_flags & AARCH64_FL_SB)
> +#define AARCH64_ISA_RCPC3	   (aarch64_isa_flags & AARCH64_FL_RCPC3)
>  #define AARCH64_ISA_V8R		   (aarch64_isa_flags & AARCH64_FL_V8R)
>  #define AARCH64_ISA_PAUTH	   (aarch64_isa_flags & AARCH64_FL_PAUTH)
>  #define AARCH64_ISA_V9A		   (aarch64_isa_flags & AARCH64_FL_V9A)
> @@ -344,6 +345,9 @@ enum class aarch64_feature : unsigned char {
>     and sign-extending versions.*/
>  #define TARGET_RCPC2 (AARCH64_ISA_RCPC8_4)
>  
> +/* RCPC3 LDAP1/STL1 loads/stores from Armv8.2-a.  */
> +#define TARGET_RCPC3 (AARCH64_ISA_RCPC3)

The extension is more general than that, so maybe just:

/* RCPC3 (Release Consistency) extensions, optional from Armv8.2-A.  */

> +
>  /* Apply the workaround for Cortex-A53 erratum 835769.  */
>  #define TARGET_FIX_ERR_A53_835769	\
>    ((aarch64_fix_a53_err835769 == 2)	\
> diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
> index 6e776a0faa1..ba28eb195ce 100644
> --- a/gcc/doc/invoke.texi
> +++ b/gcc/doc/invoke.texi
> @@ -21028,6 +21028,10 @@ Enable the Flag Manipulation instructions Extension.
>  Enable the Pointer Authentication Extension.
>  @item cssc
>  Enable the Common Short Sequence Compression instructions.
> +@item rcpc3
> +Enable the RCpc3 extension.  This enables the use of the LDAP1 and
> +STL1 instructions for loads/stores of 64-bit values to and from SIMD
> +register lanes, passing these on to the assembler.

Similarly here, it's probably enough to say:

Enable the RCpc3 (Release Consistency) extension.

OK with those changes, thanks.

Richard

Re: [PATCH 4/5] aarch64: rcpc3: add Neon ACLE wrapper functions to `arm_neon.h'

[Richard Sandiford]

Victor Do Nascimento <victor.donascimento@arm.com> writes:
> Create the necessary mappings from the ACLE-defined Neon intrinsics
> names[1] to the internal builtin function names.
>
> [1] https://arm-software.github.io/acle/neon_intrinsics/advsimd.html
>
> gcc/ChangeLog:
>
> 	* gcc/config/aarch64/arm_neon.h (vldap1_lane_u64): New.
> 	(vldap1q_lane_u64): Likewise.
> 	(vldap1_lane_s64): Likewise.
> 	(vldap1q_lane_s64): Likewise.
> 	(vldap1_lane_f64): Likewise.
> 	(vldap1q_lane_f64): Likewise.
> 	(vldap1_lane_p64): Likewise.
> 	(vldap1q_lane_p64): Likewise.
> 	(vstl1_lane_u64): Likewise.
> 	(vstl1q_lane_u64): Likewise.
> 	(vstl1_lane_s64): Likewise.
> 	(vstl1q_lane_s64): Likewise.
> 	(vstl1_lane_f64): Likewise.
> 	(vstl1q_lane_f64): Likewise.
> 	(vstl1_lane_p64): Likewise.
> 	(vstl1q_lane_p64): Likewise.

OK, thanks.

Richard

> ---
>  gcc/config/aarch64/arm_neon.h | 129 ++++++++++++++++++++++++++++++++++
>  1 file changed, 129 insertions(+)
>
> diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
> index 349f3167699..ef0d75e07ce 100644
> --- a/gcc/config/aarch64/arm_neon.h
> +++ b/gcc/config/aarch64/arm_neon.h
> @@ -13446,6 +13446,135 @@ vld1q_lane_u64 (const uint64_t *__src, uint64x2_t __vec, const int __lane)
>    return __aarch64_vset_lane_any (*__src, __vec, __lane);
>  }
>  
> +#pragma GCC push_options
> +#pragma GCC target ("+nothing+rcpc3+simd")
> +
> +/* vldap1_lane.  */
> +
> +__extension__ extern __inline uint64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_u64 (const uint64_t *__src, uint64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1di_usus (
> +	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline uint64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_u64 (const uint64_t *__src, uint64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2di_usus (
> +	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline int64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_s64 (const int64_t *__src, int64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline int64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_s64 (const int64_t *__src, int64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline float64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_f64 (const float64_t *__src, float64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline float64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_f64 (const float64_t *__src, float64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline poly64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_p64 (const poly64_t *__src, poly64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1di_psps (
> +	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline poly64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_p64 (const poly64_t *__src, poly64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2di_psps (
> +	  (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +/* vstl1_lane.  */
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_u64 (const uint64_t *__src, uint64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1di_sus ((__builtin_aarch64_simd_di *) __src,
> +					   __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_u64 (uint64_t *__src, uint64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2di_sus ((__builtin_aarch64_simd_di *) __src,
> +					   __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_s64 (int64_t *__src, int64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_s64 (int64_t *__src, int64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_f64 (float64_t *__src, float64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_f64 (float64_t *__src, float64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_p64 (poly64_t *__src, poly64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1di_sps ((__builtin_aarch64_simd_di *) __src,
> +					   __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_p64 (poly64_t *__src, poly64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2di_sps ((__builtin_aarch64_simd_di *) __src,
> +					   __vec, __lane);
> +}
> +
> +#pragma GCC pop_options
> +
>  /* vldn */
>  
>  __extension__ extern __inline int64x1x2_t

Re: [PATCH 5/5] aarch64: rcpc3: Add intrinsics tests

[Richard Sandiford]

Victor Do Nascimento <victor.donascimento@arm.com> writes:
> Add unit test to ensure that added intrinsics compile to the correct
> `LDAP1 {Vt.D}[lane],[Xn]' and `STL1 {Vt.d}[lane],[Xn]' instructions.
>
> gcc/testsuite/ChangeLog:
>
> 	* gcc.target/aarch64/acle/rcpc3.c: New.

OK, thanks.

Richard

> ---
>  gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c | 47 +++++++++++++++++++
>  1 file changed, 47 insertions(+)
>  create mode 100644 gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c
>
> diff --git a/gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c b/gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c
> new file mode 100644
> index 00000000000..689d047ab91
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/acle/rcpc3.c
> @@ -0,0 +1,47 @@
> +/* Test the rcpc3 ACLE intrinsics.  */
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -march=armv8.2-a+rcpc3" } */
> +#include <stdint.h>
> +#include <arm_neon.h>
> +
> +#define TEST_LDAP(TYPE, T)						\
> +  TYPE##x##1_t T##1_test (TYPE##_t const * ptr, TYPE##x##1_t src) {	\
> +    return vldap1_lane_##T##64 (ptr, src, 0);				\
> +  }
> +
> +#define TEST_LDAPQ(TYPE, T)						\
> +  TYPE##x##2_t T##2_test (TYPE##_t const * ptr, TYPE##x##2_t src) {	\
> +    return vldap1q_lane_##T##64 (ptr, src, 1);				\
> +  }
> +
> +#define TEST_STL(TYPE, T)				    \
> +  void T##1s_test (TYPE##_t * ptr, TYPE##x##1_t src) {	    \
> +    vstl1_lane_##T##64 (ptr, src, 0);			    \
> +  }
> +
> +#define TEST_STLQ(TYPE, T)		    \
> +  void T##2s_test (TYPE##_t * ptr, TYPE##x##2_t src) {	    \
> +    vstl1q_lane_##T##64 (ptr, src, 1);			    \
> +  }
> +
> +TEST_LDAP (uint64, u);
> +TEST_LDAP (int64, s);
> +TEST_LDAP (float64, f);
> +TEST_LDAP (poly64, p);
> +/* { dg-final { scan-assembler-times {ldap1\t\{v\d.d\}\[0\], \[x\d\]} 4 } } */
> +TEST_LDAPQ (uint64, u);
> +TEST_LDAPQ (int64, s);
> +TEST_LDAPQ (float64, f);
> +TEST_LDAPQ (poly64, p);
> +/* { dg-final { scan-assembler-times {ldap1\t\{v\d.d\}\[1\], \[x\d\]} 4 } } */
> +
> +TEST_STL (uint64, u);
> +TEST_STL (int64, s);
> +TEST_STL (float64, f);
> +TEST_STL (poly64, p);
> +/* { dg-final { scan-assembler-times {stl1\t\{v\d.d\}\[0\], \[x\d\]} 4 } } */
> +TEST_STLQ (uint64, u);
> +TEST_STLQ (int64, s);
> +TEST_STLQ (float64, f);
> +TEST_STLQ (poly64, p);
> +/* { dg-final { scan-assembler-times {stl1\t\{v\d.d\}\[1\], \[x\d\]} 4 } } */

Re: [PATCH 4/5] aarch64: rcpc3: add Neon ACLE wrapper functions to `arm_neon.h'

[Prathamesh Kulkarni]

On Thu, 9 Nov 2023 at 19:44, Victor Do Nascimento
<victor.donascimento@arm.com> wrote:
>
> Create the necessary mappings from the ACLE-defined Neon intrinsics
> names[1] to the internal builtin function names.
>
> [1] https://arm-software.github.io/acle/neon_intrinsics/advsimd.html
Hi Victor,
It seems this patch broke kernel build after the recent patch to
upgrade -Wincompatible-pointer-types to an error:

00:00:56 /home/tcwg-buildslave/workspace/tcwg_kernel_1/abe/builds/destdir/x86_64-pc-linux-gnu/lib/gcc/aarch64-linux-gnu/14.0.0/include/arm_neon.h:
In function ‘vldap1_lane_s64’:
00:00:56 /home/tcwg-buildslave/workspace/tcwg_kernel_1/abe/builds/destdir/x86_64-pc-linux-gnu/lib/gcc/aarch64-linux-gnu/14.0.0/include/arm_neon.h:13474:48:
error: passing argument 1 of ‘__builtin_aarch64_vec_ldap1_lanev1di’
from incompatible pointer type [-Wincompatible-pointer-types]
00:00:56 13474 |   return __builtin_aarch64_vec_ldap1_lanev1di (__src,
__vec, __lane);
00:00:56       |                                                ^~~~~
00:00:56       |                                                |
00:00:56       |                                                const
int64_t * {aka const long long int *}
00:00:56 /home/tcwg-buildslave/workspace/tcwg_kernel_1/abe/builds/destdir/x86_64-pc-linux-gnu/lib/gcc/aarch64-linux-gnu/14.0.0/include/arm_neon.h:13474:48:
note: expected ‘const long int *’ but argument is of type ‘const
int64_t *’ {aka ‘const long long int *’}

Looking cursorily at the code, should __src be casted to
(__builtin_aarch64_simd_di *) before passing it to
__builtin_aarch64_vec_ldap1_lanev1di ?
For more details, please see:
https://ci.linaro.org/job/tcwg_kernel--gnu-master-aarch64-next-defconfig-build/91/artifact/artifacts/notify/mail-body.txt/*view*/

Thanks,
Prathamesh


>
> gcc/ChangeLog:
>
>         * gcc/config/aarch64/arm_neon.h (vldap1_lane_u64): New.
>         (vldap1q_lane_u64): Likewise.
>         (vldap1_lane_s64): Likewise.
>         (vldap1q_lane_s64): Likewise.
>         (vldap1_lane_f64): Likewise.
>         (vldap1q_lane_f64): Likewise.
>         (vldap1_lane_p64): Likewise.
>         (vldap1q_lane_p64): Likewise.
>         (vstl1_lane_u64): Likewise.
>         (vstl1q_lane_u64): Likewise.
>         (vstl1_lane_s64): Likewise.
>         (vstl1q_lane_s64): Likewise.
>         (vstl1_lane_f64): Likewise.
>         (vstl1q_lane_f64): Likewise.
>         (vstl1_lane_p64): Likewise.
>         (vstl1q_lane_p64): Likewise.
> ---
>  gcc/config/aarch64/arm_neon.h | 129 ++++++++++++++++++++++++++++++++++
>  1 file changed, 129 insertions(+)
>
> diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
> index 349f3167699..ef0d75e07ce 100644
> --- a/gcc/config/aarch64/arm_neon.h
> +++ b/gcc/config/aarch64/arm_neon.h
> @@ -13446,6 +13446,135 @@ vld1q_lane_u64 (const uint64_t *__src, uint64x2_t __vec, const int __lane)
>    return __aarch64_vset_lane_any (*__src, __vec, __lane);
>  }
>
> +#pragma GCC push_options
> +#pragma GCC target ("+nothing+rcpc3+simd")
> +
> +/* vldap1_lane.  */
> +
> +__extension__ extern __inline uint64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_u64 (const uint64_t *__src, uint64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1di_usus (
> +         (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline uint64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_u64 (const uint64_t *__src, uint64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2di_usus (
> +         (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline int64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_s64 (const int64_t *__src, int64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline int64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_s64 (const int64_t *__src, int64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline float64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_f64 (const float64_t *__src, float64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline float64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_f64 (const float64_t *__src, float64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline poly64x1_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1_lane_p64 (const poly64_t *__src, poly64x1_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev1di_psps (
> +         (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline poly64x2_t
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vldap1q_lane_p64 (const poly64_t *__src, poly64x2_t __vec, const int __lane)
> +{
> +  return __builtin_aarch64_vec_ldap1_lanev2di_psps (
> +         (__builtin_aarch64_simd_di *) __src, __vec, __lane);
> +}
> +
> +/* vstl1_lane.  */
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_u64 (const uint64_t *__src, uint64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1di_sus ((__builtin_aarch64_simd_di *) __src,
> +                                          __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_u64 (uint64_t *__src, uint64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2di_sus ((__builtin_aarch64_simd_di *) __src,
> +                                          __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_s64 (int64_t *__src, int64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_s64 (int64_t *__src, int64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2di (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_f64 (float64_t *__src, float64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_f64 (float64_t *__src, float64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2df (__src, __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1_lane_p64 (poly64_t *__src, poly64x1_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev1di_sps ((__builtin_aarch64_simd_di *) __src,
> +                                          __vec, __lane);
> +}
> +
> +__extension__ extern __inline void
> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> +vstl1q_lane_p64 (poly64_t *__src, poly64x2_t __vec, const int __lane)
> +{
> +  __builtin_aarch64_vec_stl1_lanev2di_sps ((__builtin_aarch64_simd_di *) __src,
> +                                          __vec, __lane);
> +}
> +
> +#pragma GCC pop_options
> +
>  /* vldn */
>
>  __extension__ extern __inline int64x1x2_t
> --
> 2.41.0
>