2013-10-29 Andrew Pinski <apinski@cavium.com>
* tree-ssa-ifcombine.c: Include rtl.h and tm_p.h.
(ifcombine_ifandif): Handle cases where
maybe_fold_and_comparisons fails, combining the branches
anyways.
(tree_ssa_ifcombine): Inverse the order of
the basic block walk, increases the number of combinings.
* gimple.h (gsi_start_nondebug_after_labels_bb): New function.
2013-10-29 Andrew Pinski <apinski@cavium.com>
Zhenqiang Chen <zhenqiang.chen@linaro.org>
* gcc.dg/tree-ssa/ssa-ifcombine-ccmp-1.c: New test case.
* gcc.dg/tree-ssa/ssa-ifcombine-ccmp-2.c: New test case.
* gcc.dg/tree-ssa/ssa-ifcombine-ccmp-3.c: New test case.
* gcc.dg/tree-ssa/ssa-ifcombine-ccmp-4.c: New test case.
* gcc.dg/tree-ssa/ssa-ifcombine-ccmp-5.c: New test case.
* gcc.dg/tree-ssa/ssa-ifcombine-ccmp-6.c: New test case.
* gcc.dg/tree-ssa/phi-opt-9.c: Use a function call to prevent
conditional move to be used.
* gcc.dg/tree-ssa/ssa-dom-thread-3.c: Remove.
Co-Authored-By: Zhenqiang Chen <zhenqiang.chen@linaro.org>
From-SVN: r204194
+2013-10-29 Andrew Pinski <apinski@cavium.com>
+
+ * tree-ssa-ifcombine.c: Include rtl.h and tm_p.h.
+ (ifcombine_ifandif): Handle cases where
+ maybe_fold_and_comparisons fails, combining the branches
+ anyways.
+ (tree_ssa_ifcombine): Inverse the order of
+ the basic block walk, increases the number of combinings.
+ * gimple.h (gsi_start_nondebug_after_labels_bb): New function.
+
2013-10-29 Mike Stump <mikestump@comcast.net>
* machmode.def (PARTIAL_INT_MODE): Add precision and name.
return i;
}
+/* Return a new iterator pointing to the first non-debug non-label statement in
+ basic block BB. */
+
+static inline gimple_stmt_iterator
+gsi_start_nondebug_after_labels_bb (basic_block bb)
+{
+ gimple_stmt_iterator i = gsi_after_labels (bb);
+
+ if (!gsi_end_p (i) && is_gimple_debug (gsi_stmt (i)))
+ gsi_next_nondebug (&i);
+
+ return i;
+}
+
/* Return a new iterator pointing to the last non-debug statement in
basic block BB. */
+2013-10-29 Andrew Pinski <apinski@cavium.com>
+ Zhenqiang Chen <zhenqiang.chen@linaro.org>
+
+ * gcc.dg/tree-ssa/ssa-ifcombine-ccmp-1.c: New test case.
+ * gcc.dg/tree-ssa/ssa-ifcombine-ccmp-2.c: New test case.
+ * gcc.dg/tree-ssa/ssa-ifcombine-ccmp-3.c: New test case.
+ * gcc.dg/tree-ssa/ssa-ifcombine-ccmp-4.c: New test case.
+ * gcc.dg/tree-ssa/ssa-ifcombine-ccmp-5.c: New test case.
+ * gcc.dg/tree-ssa/ssa-ifcombine-ccmp-6.c: New test case.
+ * gcc.dg/tree-ssa/phi-opt-9.c: Use a function call to prevent
+ conditional move to be used.
+ * gcc.dg/tree-ssa/ssa-dom-thread-3.c: Remove.
+
2013-10-29 Tobias Burnus <burnus@net-b.de>
PR fortran/44350
/* { dg-options "-O -fdump-tree-optimized" } */
int g(int,int);
+int h(int);
int f(int t, int c)
{
int d = 0;
int e = 0;
if (t)
{
- d = c+1;
+ d = h(c);
e = t;
}
else d = 0, e = 0;
+++ /dev/null
-/* { dg-do compile } */
-/* { dg-options "-O2 -fdump-tree-dom1-details -fno-short-enums" } */
-
-extern void abort (void) __attribute__ ((__noreturn__));
-union tree_node;
-typedef union tree_node *tree;
-enum tree_code
-{
- VAR_DECL,
- SSA_NAME,
- MAX_TREE_CODES
-};
-extern unsigned char tree_contains_struct[MAX_TREE_CODES][64];
-struct tree_base
-{
- enum tree_code code:16;
-};
-enum tree_node_structure_enum
-{
- TS_DECL_COMMON
-};
-struct tree_ssa_name
-{
- tree var;
-};
-union tree_node
-{
- struct tree_base base;
- struct tree_ssa_name ssa_name;
-};
-long
-expand_one_var (tree var, unsigned char toplevel, unsigned char really_expand)
-{
- tree origvar = var;
- var = var->ssa_name.var;
- if (((enum tree_code) (origvar)->base.code) == SSA_NAME
- && !((var->base.code != VAR_DECL)))
- abort ();
- if ((var->base.code) != VAR_DECL && ((origvar)->base.code) != SSA_NAME)
- ;
- else if (tree_contains_struct[(var->base.code)][(TS_DECL_COMMON)] != 1)
- abort ();
-}
-/* We should thread the jump, through an intermediate block. */
-/* { dg-final { scan-tree-dump-times "Threaded" 2 "dom1"} } */
-/* { dg-final { scan-tree-dump-times "Registering jump thread: \\(.*\\) incoming edge; \\(.*\\) joiner; \\(.*\\) nocopy;" 1 "dom1"} } */
-/* { dg-final { cleanup-tree-dump "dom1" } } */
-
--- /dev/null
+/* { dg-do compile { target { ! "m68k*-*-* mmix*-*-* mep*-*-* bfin*-*-* v850*-*-* picochip*-*-* moxie*-*-* cris*-*-* m32c*-*-* fr30*-*-* mcore*-*-* powerpc*-*-* xtensa*-*-*"} } } */
+
+/* { dg-options "-O2 -g -fdump-tree-optimized" } */
+/* { dg-additional-options "-mbranch-cost=2" { target avr-*-* } } */
+
+int t (int a, int b)
+{
+ if (a > 0)
+ if (b > 0)
+ return 0;
+ return 1;
+}
+/* { dg-final { scan-tree-dump "\&" "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */
--- /dev/null
+/* { dg-do compile { target { ! "m68k*-*-* mmix*-*-* mep*-*-* bfin*-*-* v850*-*-* picochip*-*-* moxie*-*-* cris*-*-* m32c*-*-* fr30*-*-* mcore*-*-* powerpc*-*-* xtensa*-*-*"} } } */
+
+/* { dg-options "-O2 -g -fdump-tree-optimized" } */
+/* { dg-additional-options "-mbranch-cost=2" { target avr-*-* } } */
+
+int t (int a, int b)
+{
+ if (a > 0)
+ goto L1;
+ if (b > 0)
+ goto L1;
+ return 0;
+L1:
+ return 1;
+}
+/* { dg-final { scan-tree-dump "\|" "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */
--- /dev/null
+/* { dg-do compile { target { ! "m68k*-*-* mmix*-*-* mep*-*-* bfin*-*-* v850*-*-* picochip*-*-* moxie*-*-* cris*-*-* m32c*-*-* fr30*-*-* mcore*-*-* powerpc*-*-* xtensa*-*-*"} } } */
+
+/* { dg-options "-O2 -g -fdump-tree-optimized" } */
+/* { dg-additional-options "-mbranch-cost=2" { target avr-*-* } } */
+
+int t (int a, int b)
+{
+ if (a > 0)
+ goto L1;
+ else
+ goto L2;
+L1:
+ if (b > 0)
+ goto L2;
+ return 5;
+L2:
+ return 6;
+}
+/* { dg-final { scan-tree-dump "\|" "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */
--- /dev/null
+/* { dg-do compile { target { ! "m68k*-*-* mmix*-*-* mep*-*-* bfin*-*-* v850*-*-* picochip*-*-* moxie*-*-* cris*-*-* m32c*-*-* fr30*-*-* mcore*-*-* powerpc*-*-* xtensa*-*-*"} } } */
+
+/* { dg-options "-O2 -g -fdump-tree-optimized" } */
+/* { dg-additional-options "-mbranch-cost=2" { target avr-*-* } } */
+
+int t (int a, int b)
+{
+ if (a > 0)
+ goto L1;
+ if (b > 0)
+ goto L2;
+L1:
+ return 0;
+L2:
+ return 1;
+}
+/* { dg-final { scan-tree-dump "\&" "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */
--- /dev/null
+/* { dg-do compile { target { ! "m68k*-*-* mmix*-*-* mep*-*-* bfin*-*-* v850*-*-* picochip*-*-* moxie*-*-* cris*-*-* m32c*-*-* fr30*-*-* mcore*-*-* powerpc*-*-* xtensa*-*-*"} } } */
+
+/* { dg-options "-O2 -g -fdump-tree-optimized" } */
+/* { dg-additional-options "-mbranch-cost=2" { target avr-*-* } } */
+
+int t (int a, int b, int c)
+{
+ if (a > 0 && b > 0 && c > 0)
+ return 0;
+ return 1;
+}
+/* { dg-final { scan-tree-dump-times "\&" 2 "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */
--- /dev/null
+/* { dg-do compile { target { ! "m68k*-*-* mmix*-*-* mep*-*-* bfin*-*-* v850*-*-* picochip*-*-* moxie*-*-* cris*-*-* m32c*-*-* fr30*-*-* mcore*-*-* powerpc*-*-* xtensa*-*-*"} } } */
+
+/* { dg-options "-O2 -g -fdump-tree-optimized" } */
+/* { dg-additional-options "-mbranch-cost=2" { target avr-*-* } } */
+
+int t (int a, int b, int c)
+{
+ if (a > 0 || b > 0 || c > 0)
+ return 0;
+ return 1;
+}
+/* { dg-final { scan-tree-dump-times "\\|" 2 "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */
#include "system.h"
#include "coretypes.h"
#include "tm.h"
+/* rtl is needed only because arm back-end requires it for
+ BRANCH_COST. */
+#include "rtl.h"
+#include "tm_p.h"
#include "tree.h"
#include "basic-block.h"
#include "tree-pretty-print.h"
#include "ssa-iterators.h"
#include "tree-pass.h"
+#ifndef LOGICAL_OP_NON_SHORT_CIRCUIT
+#define LOGICAL_OP_NON_SHORT_CIRCUIT \
+ (BRANCH_COST (optimize_function_for_speed_p (cfun), \
+ false) >= 2)
+#endif
+
/* This pass combines COND_EXPRs to simplify control flow. It
currently recognizes bit tests and comparisons in chains that
represent logical and or logical or of two COND_EXPRs.
outer_cond_code,
gimple_cond_lhs (outer_cond),
gimple_cond_rhs (outer_cond))))
- return false;
+ {
+ tree t1, t2;
+ gimple_stmt_iterator gsi;
+ if (!LOGICAL_OP_NON_SHORT_CIRCUIT)
+ return false;
+ /* Only do this optimization if the inner bb contains only the conditional. */
+ if (!gsi_one_before_end_p (gsi_start_nondebug_after_labels_bb (inner_cond_bb)))
+ return false;
+ t1 = fold_build2_loc (gimple_location (inner_cond),
+ inner_cond_code,
+ boolean_type_node,
+ gimple_cond_lhs (inner_cond),
+ gimple_cond_rhs (inner_cond));
+ t2 = fold_build2_loc (gimple_location (outer_cond),
+ outer_cond_code,
+ boolean_type_node,
+ gimple_cond_lhs (outer_cond),
+ gimple_cond_rhs (outer_cond));
+ t = fold_build2_loc (gimple_location (inner_cond),
+ TRUTH_AND_EXPR, boolean_type_node, t1, t2);
+ if (result_inv)
+ {
+ t = fold_build1 (TRUTH_NOT_EXPR, TREE_TYPE (t), t);
+ result_inv = false;
+ }
+ gsi = gsi_for_stmt (inner_cond);
+ t = force_gimple_operand_gsi_1 (&gsi, t, is_gimple_condexpr, NULL, true,
+ GSI_SAME_STMT);
+ }
if (result_inv)
t = fold_build1 (TRUTH_NOT_EXPR, TREE_TYPE (t), t);
t = canonicalize_cond_expr_cond (t);
bbs = single_pred_before_succ_order ();
calculate_dominance_info (CDI_DOMINATORS);
- for (i = 0; i < n_basic_blocks - NUM_FIXED_BLOCKS; ++i)
+ /* Search every basic block for COND_EXPR we may be able to optimize.
+
+ We walk the blocks in order that guarantees that a block with
+ a single predecessor is processed after the predecessor.
+ This ensures that we collapse outter ifs before visiting the
+ inner ones, and also that we do not try to visit a removed
+ block. This is opposite of PHI-OPT, because we cascade the
+ combining rather than cascading PHIs. */
+ for (i = n_basic_blocks - NUM_FIXED_BLOCKS - 1; i >= 0; i--)
{
basic_block bb = bbs[i];
gimple stmt = last_stmt (bb);
projects / platform / upstream / gcc.git / commitdiff