Replace FMA_EXPR with one internal fn per optab

There are four optabs for various forms of fused multiply-add:
fma, fms, fnma and fnms.  Of these, only fma had a direct gimple
representation.  For the other three we relied on special pattern-
matching during expand, although tree-ssa-math-opts.c did have
some code to try to second-guess what expand would do.

This patch removes the old FMA_EXPR representation of fma and
introduces four new internal functions, one for each optab.
IFN_FMA is tied to BUILT_IN_FMA* while the other three are
independent directly-mapped internal functions.  It's then
possible to do the pattern-matching in match.pd and
tree-ssa-math-opts.c (via folding) can select the exact
FMA-based operation.

The BRIG & HSA parts are a best guess, but seem relatively simple.

2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>

gcc/
	* doc/sourcebuild.texi (scalar_all_fma): Document.
	* tree.def (FMA_EXPR): Delete.
	* internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
	* internal-fn.c (ternary_direct): New macro.
	(expand_ternary_optab_fn): Likewise.
	(direct_ternary_optab_supported_p): Likewise.
	* Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
	* builtins.c (fold_builtin_fma): Delete.
	(fold_builtin_3): Don't call it.
	* cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
	* expr.c (expand_expr_real_2): Likewise.
	* fold-const.c (operand_equal_p): Likewise.
	(fold_ternary_loc): Likewise.
	* gimple-pretty-print.c (dump_ternary_rhs): Likewise.
	* gimple.c (DEFTREECODE): Likewise.
	* gimplify.c (gimplify_expr): Likewise.
	* optabs-tree.c (optab_for_tree_code): Likewise.
	* tree-cfg.c (verify_gimple_assign_ternary): Likewise.
	* tree-eh.c (operation_could_trap_p): Likewise.
	(stmt_could_throw_1_p): Likewise.
	* tree-inline.c (estimate_operator_cost): Likewise.
	* tree-pretty-print.c (dump_generic_node): Likewise.
	(op_code_prio): Likewise.
	* tree-ssa-loop-im.c (stmt_cost): Likewise.
	* tree-ssa-operands.c (get_expr_operands): Likewise.
	* tree.c (commutative_ternary_tree_code, add_expr): Likewise.
	* fold-const-call.h (fold_fma): Delete.
	* fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
	CFN_FNMA and CFN_FNMS.
	(fold_fma): Delete.
	* genmatch.c (combined_fn): New enum.
	(commutative_ternary_tree_code): Remove FMA_EXPR handling.
	(commutative_op): New function.
	(commutate): Use it.  Handle more than 2 operands.
	(dt_operand::gen_gimple_expr): Use commutative_op.
	(parser::parse_expr): Allow :c to be used with non-binary
	operators if the commutative operand is known.
	* gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
	CFN_FMS, CFN_FNMA and CFN_FNMS.
	(backprop::process_assign_use): Remove FMA_EXPR handling.
	* hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
	(gen_hsa_fma): New function.
	(gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
	IFN_FNMA and IFN_FNMS.
	* match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
	* gimple-fold.h (follow_all_ssa_edges): Declare.
	* gimple-fold.c (follow_all_ssa_edges): New function.
	* tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
	gimple_build interface and use follow_all_ssa_edges to fold the result.
	(convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
	instead of checking for optabs directly.
	* config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
	rather than FMA_EXPRs.
	* config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
	call to IFN_FMA instead of an FMA_EXPR.

gcc/brig/
	* brigfrontend/brig-function.cc
	(brig_function::get_builtin_for_hsa_opcode): Use BUILT_IN_FMA
	for BRIG_OPCODE_FMA.
	(brig_function::get_tree_code_for_hsa_opcode): Treat BUILT_IN_FMA
	as a call.

gcc/c/
	* gimple-parser.c (c_parser_gimple_postfix_expression): Remove
	__FMA_EXPR handlng.

gcc/cp/
	* constexpr.c (cxx_eval_constant_expression): Remove FMA_EXPR handling.
	(potential_constant_expression_1): Likewise.

gcc/testsuite/
	* lib/target-supports.exp (check_effective_target_scalar_all_fma):
	New proc.
	* gcc.dg/fma-1.c: New test.
	* gcc.dg/fma-2.c: Likewise.
	* gcc.dg/fma-3.c: Likewise.
	* gcc.dg/fma-4.c: Likewise.
	* gcc.dg/fma-5.c: Likewise.
	* gcc.dg/fma-6.c: Likewise.
	* gcc.dg/fma-7.c: Likewise.
	* gcc.dg/gimplefe-26.c: Use .FMA instead of __FMA and require
	scalar_all_fma.
	* gfortran.dg/reassoc_7.f: Pass -ffp-contract=off.
	* gfortran.dg/reassoc_8.f: Likewise.
	* gfortran.dg/reassoc_9.f: Likewise.
	* gfortran.dg/reassoc_10.f: Likewise.

From-SVN: r260348

1 files changed, 77 insertions, 24 deletions

diff --git a/gcc/genmatch.c b/gcc/genmatch.c
index be6efe3bf12..c794e4d30cf 100644
--- a/gcc/genmatch.c
+++ b/gcc/genmatch.c
@@ -241,6 +241,20 @@ enum internal_fn {
   IFN_LAST
 };
 
+enum combined_fn {
+#define DEF_BUILTIN(ENUM, N, C, T, LT, B, F, NA, AT, IM, COND) \
+  CFN_##ENUM = int (ENUM),
+#include "builtins.def"
+
+#define DEF_INTERNAL_FN(CODE, FLAGS, FNSPEC) \
+  CFN_##CODE = int (END_BUILTINS) + int (IFN_##CODE),
+#include "internal-fn.def"
+
+  CFN_LAST
+};
+
+#include "case-cfn-macros.h"
+
 /* Return true if CODE represents a commutative tree code.  Otherwise
    return false.  */
 bool
@@ -288,7 +302,6 @@ commutative_ternary_tree_code (enum tree_code code)
     case WIDEN_MULT_PLUS_EXPR:
     case WIDEN_MULT_MINUS_EXPR:
     case DOT_PROD_EXPR:
-    case FMA_EXPR:
       return true;
 
     default:
@@ -450,6 +463,44 @@ is_a_helper <user_id *>::test (id_base *id)
   return id->kind == id_base::USER;
 }
 
+/* If ID has a pair of consecutive, commutative operands, return the
+   index of the first, otherwise return -1.  */
+
+static int
+commutative_op (id_base *id)
+{
+  if (operator_id *code = dyn_cast <operator_id *> (id))
+    {
+      if (commutative_tree_code (code->code)
+	  || commutative_ternary_tree_code (code->code))
+	return 0;
+      return -1;
+    }
+  if (fn_id *fn = dyn_cast <fn_id *> (id))
+    switch (fn->fn)
+      {
+      CASE_CFN_FMA:
+      case CFN_FMS:
+      case CFN_FNMA:
+      case CFN_FNMS:
+	return 0;
+
+      default:
+	return -1;
+      }
+  if (user_id *uid = dyn_cast<user_id *> (id))
+    {
+      int res = commutative_op (uid->substitutes[0]);
+      if (res < 0)
+	return 0;
+      for (unsigned i = 1; i < uid->substitutes.length (); ++i)
+	if (res != commutative_op (uid->substitutes[i]))
+	  return -1;
+      return res;
+    }
+  return -1;
+}
+
 /* Add a predicate identifier to the hash.  */
 
 static predicate_id *
@@ -946,6 +997,9 @@ commutate (operand *op, vec<vec<user_id *> > &for_vec)
   if (!e->is_commutative)
     return ret;
 
+  /* The operation is always binary if it isn't inherently commutative.  */
+  int natural_opno = commutative_op (e->operation);
+  unsigned int opno = natural_opno >= 0 ? natural_opno : 0;
   for (unsigned i = 0; i < result.length (); ++i)
     {
       expr *ne = new expr (e);
@@ -994,9 +1048,11 @@ commutate (operand *op, vec<vec<user_id *> > &for_vec)
 	    }
 	}
       ne->is_commutative = false;
-      // result[i].length () is 2 since e->operation is binary
-      for (unsigned j = result[i].length (); j; --j)
-	ne->append_op (result[i][j-1]);
+      for (unsigned j = 0; j < result[i].length (); ++j)
+	{
+	  int old_j = (j == opno ? opno + 1 : j == opno + 1 ? opno : j);
+	  ne->append_op (result[i][old_j]);
+	}
       ret.safe_push (ne);
     }
 
@@ -2759,24 +2815,18 @@ dt_operand::gen_gimple_expr (FILE *f, int indent)
   /* While the toplevel operands are canonicalized by the caller
      after valueizing operands of sub-expressions we have to
      re-canonicalize operand order.  */
-  if (operator_id *code = dyn_cast <operator_id *> (id))
+  int opno = commutative_op (id);
+  if (opno >= 0)
     {
-      /* ???  We can't canonicalize tcc_comparison operands here
-         because that requires changing the comparison code which
-	 we already matched...  */
-      if (commutative_tree_code (code->code)
-	  || commutative_ternary_tree_code (code->code))
-	{
-	  char child_opname0[20], child_opname1[20];
-	  gen_opname (child_opname0, 0);
-	  gen_opname (child_opname1, 1);
-	  fprintf_indent (f, indent,
-			  "if (tree_swap_operands_p (%s, %s))\n",
-			  child_opname0, child_opname1);
-	  fprintf_indent (f, indent,
-			  "  std::swap (%s, %s);\n",
-			  child_opname0, child_opname1);
-	}
+      char child_opname0[20], child_opname1[20];
+      gen_opname (child_opname0, opno);
+      gen_opname (child_opname1, opno + 1);
+      fprintf_indent (f, indent,
+		      "if (tree_swap_operands_p (%s, %s))\n",
+		      child_opname0, child_opname1);
+      fprintf_indent (f, indent,
+		      "  std::swap (%s, %s);\n",
+		      child_opname0, child_opname1);
     }
 
   return n_braces;
@@ -4217,11 +4267,14 @@ parser::parse_expr ()
 		      e->operation->id, e->operation->nargs, e->ops.length ());
 	  if (is_commutative)
 	    {
-	      if (e->ops.length () == 2)
+	      if (e->ops.length () == 2
+		  || commutative_op (e->operation) >= 0)
 		e->is_commutative = true;
 	      else
-		fatal_at (token, "only binary operators or function with "
-			  "two arguments can be marked commutative");
+		fatal_at (token, "only binary operators or functions with "
+			  "two arguments can be marked commutative, "
+			  "unless the operation is known to be inherently "
+			  "commutative");
 	    }
 	  e->expr_type = expr_type;
 	  return op;