[PATCH 2/5] ipa-inline: disallow inlining into SIMT regions

[Alexander Monakov <amonakov at ispras dot ru>]

This patch prevents inlining into SIMT code by introducing a new loop
property 'in_simtreg' and using ANNOTATE_EXPR (_, 'simtreg') to carry this
property between omp-low and the cfg pass (this is needed only for SIMD
reduction helper loops; for main bodies of SIMD loops omp-expand sets
loop->in_simtreg directly).

Technically the gimplify.c hunk is not needed since the frontends wouldn't
produce annot_expr_simtreg_kind.

Eventually inlining should be possible to do by remapping callee's variables
to use storage provided by IFN_GOMP_SIMT_ENTER ().

	* cfgloop.h (struct loop): New field 'in_simtreg'.  Use it...
        * ipa-inline.c (can_inline_edge_p): ...here to disallow inlining.
        * omp-expand.c (expand_omp_simd): Set loop->in_simtreg.
        * omp-low.c (lower_rec_input_clauses): Annotate condition of SIMT
        reduction loops with annot_expr_simtreg_kind.
        * tree-core.h (enum annot_expr_kind): New entry
        'annot_expr_simtreg_kind'.
        * tree-cfg.c (replace_loop_annotate_in_block): Handle
        annot_expr_simtreg_kind.
        (replace_loop_annotate): Ditto.
        * tree-pretty-print.c (dump_generic_node): Ditto.
        * gimplify.c (gimple_boolify): Ditto.

---
 gcc/cfgloop.h           |  3 +++
 gcc/gimplify.c          |  1 +
 gcc/ipa-inline.c        | 14 ++++++++++++++
 gcc/omp-expand.c        |  1 +
 gcc/omp-low.c           |  9 ++++++++-
 gcc/tree-cfg.c          |  4 ++++
 gcc/tree-core.h         |  1 +
 gcc/tree-pretty-print.c |  3 +++
 8 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/gcc/cfgloop.h b/gcc/cfgloop.h
index 0448a61..25b441c 100644
--- a/gcc/cfgloop.h
+++ b/gcc/cfgloop.h
@@ -220,6 +220,9 @@ struct GTY ((chain_next ("%h.next"))) loop {
   /* True if the loop is part of an oacc kernels region.  */
   bool in_oacc_kernels_region;
 
+  /* True if the loop corresponds to a SIMT variant of OpenMP SIMD region.  */
+  bool in_simtreg;
+
   /* For SIMD loops, this is a unique identifier of the loop, referenced
      by IFN_GOMP_SIMD_VF, IFN_GOMP_SIMD_LANE and IFN_GOMP_SIMD_LAST_LANE
      builtins.  */
diff --git a/gcc/gimplify.c b/gcc/gimplify.c
index a300133..a09037b 100644
--- a/gcc/gimplify.c
+++ b/gcc/gimplify.c
@@ -3682,6 +3682,7 @@ gimple_boolify (tree expr)
 	case annot_expr_ivdep_kind:
 	case annot_expr_no_vector_kind:
 	case annot_expr_vector_kind:
+	case annot_expr_simtreg_kind:
 	  TREE_OPERAND (expr, 0) = gimple_boolify (TREE_OPERAND (expr, 0));
 	  if (TREE_CODE (type) != BOOLEAN_TYPE)
 	    TREE_TYPE (expr) = boolean_type_node;
diff --git a/gcc/ipa-inline.c b/gcc/ipa-inline.c
index 5f2371c..dd4c0948 100644
--- a/gcc/ipa-inline.c
+++ b/gcc/ipa-inline.c
@@ -101,6 +101,7 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-pass.h"
 #include "gimple-ssa.h"
 #include "cgraph.h"
+#include "cfgloop.h"
 #include "lto-streamer.h"
 #include "trans-mem.h"
 #include "calls.h"
@@ -374,6 +375,19 @@ can_inline_edge_p (struct cgraph_edge *e, bool report,
       e->inline_failed = CIF_ATTRIBUTE_MISMATCH;
       inlinable = false;
     }
+  /* Don't inline into SIMT variants of OpenMP SIMD regions: inlining needs to
+     remap addressable variables to use storage provided by IFN_SIMT_ENTER.  */
+  else if (flag_openmp && gimple_has_body_p (caller->decl))
+    {
+      struct loop *l;
+      for (l = gimple_bb (e->call_stmt)->loop_father; l; l = loop_outer (l))
+	if (l->in_simtreg)
+	  {
+	    e->inline_failed = CIF_UNSPECIFIED;
+	    inlinable = false;
+	    break;
+	  }
+    }
   /* Check if caller growth allows the inlining.  */
   else if (!DECL_DISREGARD_INLINE_LIMITS (callee->decl)
 	   && !disregard_limits
diff --git a/gcc/omp-expand.c b/gcc/omp-expand.c
index 1f1055c..6a29df6 100644
--- a/gcc/omp-expand.c
+++ b/gcc/omp-expand.c
@@ -4794,6 +4794,7 @@ expand_omp_simd (struct omp_region *region, struct omp_for_data *fd)
       loop->latch = cont_bb;
       add_loop (loop, l1_bb->loop_father);
       loop->safelen = safelen_int;
+      loop->in_simtreg = is_simt;
       if (simduid)
 	{
 	  loop->simduid = OMP_CLAUSE__SIMDUID__DECL (simduid);
diff --git a/gcc/omp-low.c b/gcc/omp-low.c
index 13d9b6b..a5f8bf65 100644
--- a/gcc/omp-low.c
+++ b/gcc/omp-low.c
@@ -4510,7 +4510,14 @@ lower_rec_input_clauses (tree clauses, gimple_seq *ilist, gimple_seq *dlist,
 	  gimple_seq_add_stmt (dlist, g);
 
 	  gimple_seq_add_stmt (dlist, gimple_build_label (header));
-	  g = gimple_build_cond (LT_EXPR, simt_lane, simt_vf, body, end);
+	  t = create_tmp_var (boolean_type_node);
+	  g = gimple_build_assign (t, LT_EXPR, simt_lane, simt_vf);
+	  gimple_seq_add_stmt (dlist, g);
+	  tree ann = build_int_cst (integer_type_node, annot_expr_simtreg_kind);
+	  g = gimple_build_call_internal (IFN_ANNOTATE, 2, t, ann);
+	  gimple_call_set_lhs (g, t);
+	  gimple_seq_add_stmt (dlist, g);
+	  g = gimple_build_cond (NE_EXPR, t, boolean_false_node, body, end);
 	  gimple_seq_add_stmt (dlist, g);
 
 	  gimple_seq_add_stmt (dlist, gimple_build_label (end));
diff --git a/gcc/tree-cfg.c b/gcc/tree-cfg.c
index d4a7db8..c1392be 100644
--- a/gcc/tree-cfg.c
+++ b/gcc/tree-cfg.c
@@ -286,6 +286,9 @@ replace_loop_annotate_in_block (basic_block bb, struct loop *loop)
 	  loop->force_vectorize = true;
 	  cfun->has_force_vectorize_loops = true;
 	  break;
+	case annot_expr_simtreg_kind:
+	  loop->in_simtreg = true;
+	  break;
 	default:
 	  gcc_unreachable ();
 	}
@@ -335,6 +338,7 @@ replace_loop_annotate (void)
 	    case annot_expr_ivdep_kind:
 	    case annot_expr_no_vector_kind:
 	    case annot_expr_vector_kind:
+	    case annot_expr_simtreg_kind:
 	      break;
 	    default:
 	      gcc_unreachable ();
diff --git a/gcc/tree-core.h b/gcc/tree-core.h
index eec2d4f3..568cb1e 100644
--- a/gcc/tree-core.h
+++ b/gcc/tree-core.h
@@ -843,6 +843,7 @@ enum annot_expr_kind {
   annot_expr_ivdep_kind,
   annot_expr_no_vector_kind,
   annot_expr_vector_kind,
+  annot_expr_simtreg_kind,
   annot_expr_kind_last
 };
 
diff --git a/gcc/tree-pretty-print.c b/gcc/tree-pretty-print.c
index 5b3e23e..e0693dc 100644
--- a/gcc/tree-pretty-print.c
+++ b/gcc/tree-pretty-print.c
@@ -2569,6 +2569,9 @@ dump_generic_node (pretty_printer *pp, tree node, int spc, int flags,
 	case annot_expr_vector_kind:
 	  pp_string (pp, ", vector");
 	  break;
+	case annot_expr_simtreg_kind:
+	  pp_string (pp, ", simtreg");
+	  break;
 	default:
 	  gcc_unreachable ();
 	}
-- 
1.8.3.1