[ARM] Vectorise bswap* in aarch32.

	* config/arm/neon.md (neon_bswap<mode>): New pattern.
	* config/arm/arm.c (neon_itype): Add NEON_BSWAP.
	(arm_init_neon_builtins): Handle NEON_BSWAP.
	Define required type nodes.
	(arm_expand_neon_builtin): Handle NEON_BSWAP.
	(arm_builtin_vectorized_function): Handle BUILTIN_BSWAP builtins.
	* config/arm/arm_neon_builtins.def (bswap): Define builtins.
	* config/arm/iterators.md (VDQHSD): New mode iterator.

	* lib/target-supports.exp (check_effective_target_vect_bswap):
	Specify arm*-*-* support.

From-SVN: r210967

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index d336d136240fd53fede8a92999b4579677f0154f..8845c8858c2d0f14bb90487991f16c6dff170e06 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,14 @@
+2014-05-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
+
+       * config/arm/neon.md (neon_bswap<mode>): New pattern.
+       * config/arm/arm.c (neon_itype): Add NEON_BSWAP.
+       (arm_init_neon_builtins): Handle NEON_BSWAP.
+       Define required type nodes.
+       (arm_expand_neon_builtin): Handle NEON_BSWAP.
+       (arm_builtin_vectorized_function): Handle BUILTIN_BSWAP builtins.
+       * config/arm/arm_neon_builtins.def (bswap): Define builtins.
+       * config/arm/iterators.md (VDQHSD): New mode iterator.
+
 2014-05-27  Richard Biener  <rguenther@suse.de>
 
        * tree-vrp.c (vrp_evaluate_conditional_warnv_with_ops_using_ranges):

diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
index ccad5484ff519c5c304d54875ec960a61dee869a..48c8091bd6da5165f881745ec017b42ac5398ba2 100644 (file)
--- a/gcc/config/arm/arm.c
+++ b/gcc/config/arm/arm.c
@@ -23065,6 +23065,7 @@ typedef enum {
   NEON_BINOP,
   NEON_TERNOP,
   NEON_UNOP,
+  NEON_BSWAP,
   NEON_GETLANE,
   NEON_SETLANE,
   NEON_CREATE,
@@ -23531,14 +23532,19 @@ arm_init_neon_builtins (void)
 
   tree V8QI_type_node;
   tree V4HI_type_node;
+  tree V4UHI_type_node;
   tree V4HF_type_node;
   tree V2SI_type_node;
+  tree V2USI_type_node;
   tree V2SF_type_node;
   tree V16QI_type_node;
   tree V8HI_type_node;
+  tree V8UHI_type_node;
   tree V4SI_type_node;
+  tree V4USI_type_node;
   tree V4SF_type_node;
   tree V2DI_type_node;
+  tree V2UDI_type_node;
 
   tree intUQI_type_node;
   tree intUHI_type_node;
@@ -23613,16 +23619,26 @@ arm_init_neon_builtins (void)
   const_intDI_pointer_node = build_pointer_type (const_intDI_node);
   const_float_pointer_node = build_pointer_type (const_float_node);
 
+  /* Unsigned integer types for various mode sizes.  */
+  intUQI_type_node = make_unsigned_type (GET_MODE_PRECISION (QImode));
+  intUHI_type_node = make_unsigned_type (GET_MODE_PRECISION (HImode));
+  intUSI_type_node = make_unsigned_type (GET_MODE_PRECISION (SImode));
+  intUDI_type_node = make_unsigned_type (GET_MODE_PRECISION (DImode));
+  neon_intUTI_type_node = make_unsigned_type (GET_MODE_PRECISION (TImode));
   /* Now create vector types based on our NEON element types.  */
   /* 64-bit vectors.  */
   V8QI_type_node =
     build_vector_type_for_mode (neon_intQI_type_node, V8QImode);
   V4HI_type_node =
     build_vector_type_for_mode (neon_intHI_type_node, V4HImode);
+  V4UHI_type_node =
+    build_vector_type_for_mode (intUHI_type_node, V4HImode);
   V4HF_type_node =
     build_vector_type_for_mode (neon_floatHF_type_node, V4HFmode);
   V2SI_type_node =
     build_vector_type_for_mode (neon_intSI_type_node, V2SImode);
+  V2USI_type_node =
+    build_vector_type_for_mode (intUSI_type_node, V2SImode);
   V2SF_type_node =
     build_vector_type_for_mode (neon_float_type_node, V2SFmode);
   /* 128-bit vectors.  */
@@ -23630,19 +23646,18 @@ arm_init_neon_builtins (void)
     build_vector_type_for_mode (neon_intQI_type_node, V16QImode);
   V8HI_type_node =
     build_vector_type_for_mode (neon_intHI_type_node, V8HImode);
+  V8UHI_type_node =
+    build_vector_type_for_mode (intUHI_type_node, V8HImode);
   V4SI_type_node =
     build_vector_type_for_mode (neon_intSI_type_node, V4SImode);
+  V4USI_type_node =
+    build_vector_type_for_mode (intUSI_type_node, V4SImode);
   V4SF_type_node =
     build_vector_type_for_mode (neon_float_type_node, V4SFmode);
   V2DI_type_node =
     build_vector_type_for_mode (neon_intDI_type_node, V2DImode);
-
-  /* Unsigned integer types for various mode sizes.  */
-  intUQI_type_node = make_unsigned_type (GET_MODE_PRECISION (QImode));
-  intUHI_type_node = make_unsigned_type (GET_MODE_PRECISION (HImode));
-  intUSI_type_node = make_unsigned_type (GET_MODE_PRECISION (SImode));
-  intUDI_type_node = make_unsigned_type (GET_MODE_PRECISION (DImode));
-  neon_intUTI_type_node = make_unsigned_type (GET_MODE_PRECISION (TImode));
+  V2UDI_type_node =
+    build_vector_type_for_mode (intUDI_type_node, V2DImode);
 
 
   (*lang_hooks.types.register_builtin_type) (intUQI_type_node,
@@ -23677,8 +23692,6 @@ arm_init_neon_builtins (void)
 
   if (TARGET_CRYPTO && TARGET_HARD_FLOAT)
   {
-    tree V4USI_type_node =
-      build_vector_type_for_mode (intUSI_type_node, V4SImode);
 
     tree V16UQI_type_node =
       build_vector_type_for_mode (intUQI_type_node, V16QImode);
@@ -24023,6 +24036,31 @@ arm_init_neon_builtins (void)
            ftype = build_function_type_list (return_type, eltype, NULL);
            break;
          }
+       case NEON_BSWAP:
+       {
+           tree eltype = NULL_TREE;
+           switch (insn_data[d->code].operand[1].mode)
+           {
+             case V4HImode:
+               eltype = V4UHI_type_node;
+               break;
+             case V8HImode:
+               eltype = V8UHI_type_node;
+               break;
+             case V2SImode:
+               eltype = V2USI_type_node;
+               break;
+             case V4SImode:
+               eltype = V4USI_type_node;
+               break;
+             case V2DImode:
+               eltype = V2UDI_type_node;
+               break;
+             default: gcc_unreachable ();
+           }
+           ftype = build_function_type_list (eltype, eltype, NULL);
+           break;
+       }
        default:
          gcc_unreachable ();
        }
@@ -25197,6 +25235,7 @@ arm_expand_neon_builtin (int fcode, tree exp, rtx target)
     case NEON_SPLIT:
     case NEON_FLOAT_WIDEN:
     case NEON_FLOAT_NARROW:
+    case NEON_BSWAP:
     case NEON_REINTERP:
       return arm_expand_neon_args (target, icode, 1, type_mode, exp, fcode,
         NEON_ARG_COPY_TO_REG, NEON_ARG_STOP);
@@ -29677,8 +29716,7 @@ arm_builtin_vectorized_function (tree fndecl, tree type_out, tree type_in)
   int in_n, out_n;
 
   if (TREE_CODE (type_out) != VECTOR_TYPE
-      || TREE_CODE (type_in) != VECTOR_TYPE
-      || !(TARGET_NEON && TARGET_FPU_ARMV8 && flag_unsafe_math_optimizations))
+      || TREE_CODE (type_in) != VECTOR_TYPE)
     return NULL_TREE;
 
   out_mode = TYPE_MODE (TREE_TYPE (type_out));
@@ -29690,7 +29728,13 @@ arm_builtin_vectorized_function (tree fndecl, tree type_out, tree type_in)
    decl of the vectorized builtin for the appropriate vector mode.
    NULL_TREE is returned if no such builtin is available.  */
 #undef ARM_CHECK_BUILTIN_MODE
-#define ARM_CHECK_BUILTIN_MODE(C) \
+#define ARM_CHECK_BUILTIN_MODE(C)    \
+  (TARGET_NEON && TARGET_FPU_ARMV8   \
+   && flag_unsafe_math_optimizations \
+   && ARM_CHECK_BUILTIN_MODE_1 (C))
+
+#undef ARM_CHECK_BUILTIN_MODE_1
+#define ARM_CHECK_BUILTIN_MODE_1(C) \
   (out_mode == SFmode && out_n == C \
    && in_mode == SFmode && in_n == C)
 
@@ -29715,6 +29759,30 @@ arm_builtin_vectorized_function (tree fndecl, tree type_out, tree type_in)
             return ARM_FIND_VRINT_VARIANT (vrintz);
           case BUILT_IN_ROUNDF:
             return ARM_FIND_VRINT_VARIANT (vrinta);
+#undef ARM_CHECK_BUILTIN_MODE
+#define ARM_CHECK_BUILTIN_MODE(C, N) \
+  (out_mode == N##Imode && out_n == C \
+   && in_mode == N##Imode && in_n == C)
+          case BUILT_IN_BSWAP16:
+            if (ARM_CHECK_BUILTIN_MODE (4, H))
+              return arm_builtin_decl (ARM_BUILTIN_NEON_bswapv4hi, false);
+            else if (ARM_CHECK_BUILTIN_MODE (8, H))
+              return arm_builtin_decl (ARM_BUILTIN_NEON_bswapv8hi, false);
+            else
+              return NULL_TREE;
+          case BUILT_IN_BSWAP32:
+            if (ARM_CHECK_BUILTIN_MODE (2, S))
+              return arm_builtin_decl (ARM_BUILTIN_NEON_bswapv2si, false);
+            else if (ARM_CHECK_BUILTIN_MODE (4, S))
+              return arm_builtin_decl (ARM_BUILTIN_NEON_bswapv4si, false);
+            else
+              return NULL_TREE;
+          case BUILT_IN_BSWAP64:
+            if (ARM_CHECK_BUILTIN_MODE (2, D))
+              return arm_builtin_decl (ARM_BUILTIN_NEON_bswapv2di, false);
+            else
+              return NULL_TREE;
+
           default:
             return NULL_TREE;
         }

diff --git a/gcc/config/arm/arm_neon_builtins.def b/gcc/config/arm/arm_neon_builtins.def
index 3f64479c2195e9d23f2516822fd6b57e406b6312..f4531f36e7aa0bff4d33bb78eb68a480bf389d47 100644 (file)
--- a/gcc/config/arm/arm_neon_builtins.def
+++ b/gcc/config/arm/arm_neon_builtins.def
@@ -88,6 +88,7 @@ VAR8 (UNOP, vneg, v8qi, v4hi, v2si, v2sf, v16qi, v8hi, v4si, v4sf),
 VAR6 (UNOP, vqneg, v8qi, v4hi, v2si, v16qi, v8hi, v4si),
 VAR6 (UNOP, vcls, v8qi, v4hi, v2si, v16qi, v8hi, v4si),
 VAR6 (UNOP, vclz, v8qi, v4hi, v2si, v16qi, v8hi, v4si),
+VAR5 (BSWAP, bswap, v4hi, v8hi, v2si, v4si, v2di),
 VAR2 (UNOP, vcnt, v8qi, v16qi),
 VAR4 (UNOP, vrecpe, v2si, v2sf, v4si, v4sf),
 VAR4 (UNOP, vrsqrte, v2si, v2sf, v4si, v4sf),

diff --git a/gcc/config/arm/iterators.md b/gcc/config/arm/iterators.md
index aebab934049871b218b9c525e62afd1d703e6172..892e5f328b968c84442830dd7a2850e672fd7d69 100644 (file)
--- a/gcc/config/arm/iterators.md
+++ b/gcc/config/arm/iterators.md
@@ -116,6 +116,9 @@
 ;; Vector modes including 64-bit integer elements, but no floats.
 (define_mode_iterator VDQIX [V8QI V16QI V4HI V8HI V2SI V4SI DI V2DI])
 
+;; Vector modes for H, S and D types.
+(define_mode_iterator VDQHSD [V4HI V8HI V2SI V4SI V2DI])
+
 ;; Vector modes for float->int conversions.
 (define_mode_iterator VCVTF [V2SF V4SF])
 

diff --git a/gcc/config/arm/neon.md b/gcc/config/arm/neon.md
index 22180e5820d06d96d4986e7d4ff450e5de1d91e6..83970618a6987c565a37a3facd5e669e7ca537be 100644 (file)
--- a/gcc/config/arm/neon.md
+++ b/gcc/config/arm/neon.md
@@ -2547,6 +2547,14 @@
   [(set_attr "type" "neon_qabs<q>")]
 )
 
+(define_insn "neon_bswap<mode>"
+  [(set (match_operand:VDQHSD 0 "register_operand" "=w")
+        (bswap:VDQHSD (match_operand:VDQHSD 1 "register_operand" "w")))]
+  "TARGET_NEON"
+  "vrev<V_sz_elem>.8\\t%<V_reg>0, %<V_reg>1"
+  [(set_attr "type" "neon_rev<q>")]
+)
+
 (define_expand "neon_vneg<mode>"
   [(match_operand:VDQW 0 "s_register_operand" "")
    (match_operand:VDQW 1 "s_register_operand" "")

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index d4e9ba83a56e56194a54b41940a9acdfc1f1b4e8..3491f6ec087f9577bbc5a35ea651065d40b12c3b 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2014-05-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
+
+       * lib/target-supports.exp (check_effective_target_vect_bswap):
+       Specify arm*-*-* support.
+
 2014-05-27  Dominique d'Humieres <dominiq@lps.ens.fr>
 
        PR testsuite/61319

diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp
index 07f045aa477c8b6a38ecbfe6db631fa1258c1ca5..5ec16088355d82b4a71f23d76f42bb0b32155dae 100644 (file)
--- a/gcc/testsuite/lib/target-supports.exp
+++ b/gcc/testsuite/lib/target-supports.exp
@@ -3306,7 +3306,10 @@ proc check_effective_target_vect_bswap { } {
        verbose "check_effective_target_vect_bswap: using cached result" 2
     } else {
        set et_vect_bswap_saved 0
-       if { [istarget aarch64*-*-*] } {
+       if { [istarget aarch64*-*-*]
+             || ([istarget arm*-*-*]
+                && [check_effective_target_arm_neon])
+          } {
           set et_vect_bswap_saved 1
        }
     }