public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Sebastian Pop <sebpop@gmail.com>
To: gcc-patches@gcc.gnu.org
Cc: Tobias Grosser <grosser@fim.uni-passau.de>,
	gcc-graphite <gcc-graphite@googlegroups.com>
Subject: [PATCH 19/65] Always insert out of SSA copies on edges except for loop->latch.
Date: Mon, 02 Aug 2010 20:22:00 -0000	[thread overview]
Message-ID: <1280780438-17543-20-git-send-email-sebpop@gmail.com> (raw)
In-Reply-To: <1280780438-17543-1-git-send-email-sebpop@gmail.com>

From: spop <spop@138bc75d-0d04-0410-961f-82ee72b054a4>

2010-06-23  Sebastian Pop  <sebastian.pop@amd.com>

	* graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Always insert out
	of SSA copies on edges except for loop->latch.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/branches/graphite@161292 138bc75d-0d04-0410-961f-82ee72b054a4
---
 gcc/ChangeLog               |    5 ++++
 gcc/ChangeLog.graphite      |    5 ++++
 gcc/graphite-sese-to-poly.c |   54 ++++---------------------------------------
 3 files changed, 15 insertions(+), 49 deletions(-)

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index bfdc49a..ec97649 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,5 +1,10 @@
 2010-08-02  Sebastian Pop  <sebastian.pop@amd.com>
 
+	* graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Always insert out
+	of SSA copies on edges except for loop->latch.
+
+2010-08-02  Sebastian Pop  <sebastian.pop@amd.com>
+
 	* graphite-sese-to-poly.c (rewrite_cross_bb_scalar_deps_out_of_ssa):
 	Split out of rewrite_reductions_out_of_ssa.
 	* graphite-sese-to-poly.h (rewrite_cross_bb_scalar_deps_out_of_ssa):
diff --git a/gcc/ChangeLog.graphite b/gcc/ChangeLog.graphite
index 0d64cc3..47cf924 100644
--- a/gcc/ChangeLog.graphite
+++ b/gcc/ChangeLog.graphite
@@ -1,5 +1,10 @@
 2010-06-23  Sebastian Pop  <sebastian.pop@amd.com>
 
+	* graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Always insert out
+	of SSA copies on edges except for loop->latch.
+
+2010-06-23  Sebastian Pop  <sebastian.pop@amd.com>
+
 	* graphite-sese-to-poly.c (rewrite_cross_bb_scalar_deps_out_of_ssa):
 	Split out of rewrite_reductions_out_of_ssa.
 	* graphite-sese-to-poly.h (rewrite_cross_bb_scalar_deps_out_of_ssa):
diff --git a/gcc/graphite-sese-to-poly.c b/gcc/graphite-sese-to-poly.c
index 8bcd684..75879ed 100644
--- a/gcc/graphite-sese-to-poly.c
+++ b/gcc/graphite-sese-to-poly.c
@@ -2246,58 +2246,14 @@ rewrite_phi_out_of_ssa (gimple_stmt_iterator *psi)
   for (i = 0; i < gimple_phi_num_args (phi); i++)
     {
       tree arg = gimple_phi_arg_def (phi, i);
+      edge e = gimple_phi_arg_edge (phi, i);
 
-      /* Try to avoid the insertion on edges as much as possible: this
-	 would avoid the insertion of code on loop latch edges, making
-	 the pattern matching of the vectorizer happy, or it would
-	 avoid the insertion of useless basic blocks.  Note that it is
-	 incorrect to insert out of SSA copies close by their
-	 definition when they are more than two loop levels apart:
-	 for example, starting from a double nested loop
-
-	 | a = ...
-	 | loop_1
-	 |  loop_2
-	 |    b = phi (a, c)
-	 |    c = ...
-	 |  end_2
-	 | end_1
-
-	 the following transform is incorrect
-
-	 | a = ...
-	 | Red[0] = a
-	 | loop_1
-	 |  loop_2
-	 |    b = Red[0]
-	 |    c = ...
-	 |    Red[0] = c
-	 |  end_2
-	 | end_1
-
-	 whereas inserting the copy on the incoming edge is correct
-
-	 | a = ...
-	 | loop_1
-	 |  Red[0] = a
-	 |  loop_2
-	 |    b = Red[0]
-	 |    c = ...
-	 |    Red[0] = c
-	 |  end_2
-	 | end_1
-      */
-      if (TREE_CODE (arg) == SSA_NAME
-	  && is_gimple_reg (arg)
-	  && gimple_bb (SSA_NAME_DEF_STMT (arg))
-	  && (flow_bb_inside_loop_p (bb->loop_father,
-				     gimple_bb (SSA_NAME_DEF_STMT (arg)))
-	      || flow_bb_inside_loop_p (loop_outer (bb->loop_father),
-					gimple_bb (SSA_NAME_DEF_STMT (arg)))))
+      /* Avoid the insertion of code in the loop latch to please the
+	 pattern matching of the vectorizer.  */
+      if (e->src == bb->loop_father->latch)
 	insert_out_of_ssa_copy (zero_dim_array, arg, SSA_NAME_DEF_STMT (arg));
       else
-	insert_out_of_ssa_copy_on_edge (gimple_phi_arg_edge (phi, i),
-					zero_dim_array, arg);
+	insert_out_of_ssa_copy_on_edge (e, zero_dim_array, arg);
     }
 
   var = force_gimple_operand (zero_dim_array, &stmts, true, NULL_TREE);
-- 
1.7.0.4

  parent reply	other threads:[~2010-08-02 20:22 UTC|newest]

Thread overview: 74+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-08-02 20:21 [PATCH 00/65] Merge from the Graphite branch to trunk Sebastian Pop
2010-08-02 20:21 ` [PATCH 05/65] Skeleton for the new SCoP detection Sebastian Pop
2010-08-02 20:21 ` [PATCH 09/65] Call compute_overall_effect_of_inner_loop from instantiate_scev_name Sebastian Pop
2010-08-02 20:21 ` [PATCH 07/65] Fix comments and indentation Sebastian Pop
2010-08-02 20:21 ` [PATCH 03/65] Use size_one_node Sebastian Pop
2010-08-02 20:21 ` [PATCH 06/65] Fix refined region tree nesting Sebastian Pop
2010-08-02 20:21 ` [PATCH 01/65] Fix invariant phi node removal Sebastian Pop
2010-08-02 20:21 ` [PATCH 08/65] Do not gather loop exit conditions on the basic blocks outside the loop Sebastian Pop
2010-08-02 20:21 ` [PATCH 02/65] Fix type of integer_one_node for fold_build of POINTER_PLUS_EXPR Sebastian Pop
2010-08-02 20:22 ` [PATCH 17/65] Add more debug counter guards Sebastian Pop
2010-08-02 20:22 ` [PATCH 10/65] Don't call pbb_to_depth_to_oldiv from compute_type_for_level Sebastian Pop
2010-08-02 20:22 ` [PATCH 21/65] Early return in rewrite_commutative_reductions_out_of_ssa when flag_associative_math is not set Sebastian Pop
2010-08-02 20:22 ` [PATCH 15/65] Remove rename_nb_iterations and rename_sese_parameters Sebastian Pop
2010-08-02 20:22 ` [PATCH 12/65] Detect commutative reductions in all the scops before rewriting out of SSA all the other scalar dependences Sebastian Pop
2010-08-02 20:22 ` [PATCH 04/65] Add analysis pass to build the refined program structure tree Sebastian Pop
2010-08-02 20:22 ` [PATCH 16/65] Schedule a copy_prop pass before graphite Sebastian Pop
2010-08-02 20:22 ` Sebastian Pop [this message]
2010-08-02 20:22 ` [PATCH 13/65] Remove insert_copyout and insert_copyin Sebastian Pop
2010-08-02 20:22 ` [PATCH 18/65] Split rewrite_cross_bb_scalar_deps_out_of_ssa out from rewrite_reductions_out_of_ssa Sebastian Pop
2010-08-02 20:22 ` [PATCH 14/65] Also rewrite out of SSA scalar dependences going outside the SCoP region Sebastian Pop
2010-11-10 12:23   ` H.J. Lu
2010-08-02 20:23 ` [PATCH 27/65] Fix 416.gamess miscompile: special case the rewrite of degenerate phi nodes Sebastian Pop
2010-08-02 20:23 ` [PATCH 23/65] Remove copy_renames Sebastian Pop
2010-08-02 20:23 ` [PATCH 28/65] Print bbs in refined region tree Sebastian Pop
2010-08-02 20:23 ` [PATCH 31/65] Partially removing cloog.h and graphite-clast-to-gimple.h where possible. Removing unused function check_poly_representation Sebastian Pop
2010-08-02 20:23 ` [PATCH 11/65] Launch dotty in background Sebastian Pop
2010-08-02 20:23 ` [PATCH 37/65] Also handle GIMPLE_CALLs in rewrite_cross_bb_scalar_deps Sebastian Pop
2010-08-02 20:23 ` [PATCH 32/65] Add testcase for miscompile in gamess Sebastian Pop
2010-08-02 20:23 ` [PATCH 38/65] Remove uses of loop->single_iv Sebastian Pop
2010-08-02 20:23 ` [PATCH 25/65] Remove dead code Sebastian Pop
2010-08-02 20:23 ` [PATCH 20/65] Fix testcase: call abort Sebastian Pop
2010-08-02 20:23 ` [PATCH 33/65] Fix flipped condition using mpz_sgn Sebastian Pop
2010-08-02 20:23 ` [PATCH 26/65] Fix 435.gromacs miscompile: call rewrite_close_phi_out_of_ssa from rewrite_cross_bb_phi_deps Sebastian Pop
2010-08-02 20:23 ` [PATCH 24/65] Replace map with rename_map Sebastian Pop
2010-08-02 20:23 ` [PATCH 30/65] Move cloog-related functions from graphite-ppl.{c,h} to new graphite-cloog-util.{c,h} Sebastian Pop
2010-08-02 20:23 ` [PATCH 40/65] Scevs could be expressions without chrecs and still be scev_analyzable_p Sebastian Pop
2010-08-02 20:24 ` [PATCH 51/65] Do not instantiate default definitions in instantiate_scev_name Sebastian Pop
2010-08-02 20:24 ` [PATCH 34/65] Remove insert_loop_close_phis Sebastian Pop
2010-08-02 20:24 ` [PATCH 43/65] Correctly handle SSA_NAME_IS_DEFAULT_DEF in rewrite_close_phi_out_of_ssa Sebastian Pop
2010-08-02 20:24 ` [PATCH 39/65] Bump the size of scevs Sebastian Pop
2010-08-02 20:24 ` [PATCH 36/65] chrec_apply should only apply to the specified variable Sebastian Pop
2010-08-02 20:24 ` [PATCH 44/65] Do not rewrite out of SSA scalar phi nodes that can be scev_analyzable_p Sebastian Pop
2010-08-02 20:24 ` [PATCH 52/65] Use SSA_NAME_DEF_STMT only on SSA_NAMEs Sebastian Pop
2010-08-02 20:24 ` [PATCH 49/65] Add loop_exits_from_bb_p Sebastian Pop
2010-08-02 20:24 ` [PATCH 22/65] Run rewrite_cross_bb_phi_deps before rewrite_cross_bb_scalar_deps Sebastian Pop
2010-08-02 20:24 ` [PATCH 47/65] Do not translate out-of-SSA close phi nodes that can be analyzed with scev Sebastian Pop
2010-08-02 20:24 ` [PATCH 46/65] Add testcase for PR20742 Sebastian Pop
2010-08-02 20:24 ` [PATCH 29/65] Resolve CLooG's value_* macros to their respective mpz_* counterparts Sebastian Pop
2010-08-02 20:24 ` [PATCH 48/65] Dot refined regions Sebastian Pop
2010-08-02 20:24 ` [PATCH 35/65] Remove expand_scalar_variables_ hack Sebastian Pop
2010-08-02 20:25 ` [PATCH 59/65] Adapt to new CLAST structure within upstream CLooG version and retain compatibility to CLooG Legacy Sebastian Pop
2010-08-02 20:25 ` [PATCH 55/65] Fix tonto: call propagate_expr_outside_region Sebastian Pop
2010-10-26 19:37   ` H.J. Lu
2010-08-02 20:25 ` [PATCH 62/65] Remove cloog_initialize and cloog_finalize when using official CLooG releases Sebastian Pop
2010-08-02 20:25 ` [PATCH 61/65] Adapt to new CloogScattering abstraction, introduced in official CLooG versions (CLOOG_ORG) Sebastian Pop
2010-08-02 20:25 ` [PATCH 60/65] Introduce compatibility to newer CLooG's CloogState and mask the use with macros where possible Sebastian Pop
2010-08-02 20:25 ` [PATCH 64/65] Fix gamess: the only constant phi nodes with one argument are is_gimple_min_invariant and SSA_NAME_IS_DEFAULT_DEF Sebastian Pop
2010-08-02 20:25 ` [PATCH 54/65] Call scev_analyzable_p only on is_gimple_reg Sebastian Pop
2010-08-02 20:25 ` [PATCH 58/65] Make CLooG options compatible to newer CLooG releases and pass options to build_cloog_prog (CLOOG_ORG) Sebastian Pop
2010-08-02 20:25 ` [PATCH 53/65] Propagate constant values or parametric expressions outside the scop region Sebastian Pop
2010-08-02 20:25 ` [PATCH 42/65] Add testcase for PR42729 and fix handling of gimple_debug info Sebastian Pop
2010-08-02 20:25 ` [PATCH 57/65] Reintroduce necessary CLooG accessors to graphite. Masked by CLOOG_ORG Sebastian Pop
2010-08-02 20:25 ` [PATCH 45/65] Unshare the scev before code generating it Sebastian Pop
2010-08-02 20:25 ` [PATCH 65/65] Fix miscompile of 416.gamess Sebastian Pop
2010-08-02 20:25 ` [PATCH 41/65] Special case non close-phi nodes with one argument in rewrite_close_phi_out_of_ssa Sebastian Pop
2010-08-02 20:25 ` [PATCH 50/65] Enhance region checks Sebastian Pop
2010-08-02 20:25 ` [PATCH 56/65] Call scev_reset_htab when Graphite out-of-SSA has changed something Sebastian Pop
2010-08-02 20:25 ` [PATCH 63/65] Fix typos Sebastian Pop
2010-08-03  2:52 ` [PATCH 00/65] Merge from the Graphite branch to trunk Jack Howarth
2010-08-03  2:55   ` Jack Howarth
2010-08-03  3:07     ` Sebastian Pop
2010-08-07 10:26       ` Jack Howarth
2010-08-05 15:58 ` Jack Howarth
2010-08-11 20:54 ` Sebastian Pop

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1280780438-17543-20-git-send-email-sebpop@gmail.com \
    --to=sebpop@gmail.com \
    --cc=gcc-graphite@googlegroups.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=grosser@fim.uni-passau.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).