- Generated by
1.8.1.1
|
GCC Middle and Back End API Reference
|
Data Structures | |
| struct | bb_predicate_s |
| struct | ifc_dr |
Typedefs | |
| typedef struct bb_predicate_s * | bb_predicate_p |
Variables | |
| static basic_block * | ifc_bbs |
| typedef struct bb_predicate_s * bb_predicate_p |
Structure used to predicate basic blocks. This is attached to the ->aux field of the BBs in the loop to be if-converted.
|
inlinestatic |
Adds the sequence of statements STMTS to the sequence of statements of the predicate for basic block BB.
Referenced by add_to_predicate_list().
|
static |
Add the condition COND to the previous condition PREV_COND, and add this to the predicate list of the destination of edge E. LOOP is the loop to be if-converted.
References dump_file, and dump_flags.
|
inlinestatic |
Add condition NC to the predicate list of basic block BB.
Allow a TRUTH_NOT_EXPR around the main predicate.
References add_bb_predicate_gimplified_stmts(), force_gimple_operand_1(), and is_gimple_condexpr().
|
inlinestatic |
Returns true when the basic block BB has a predicate.
References basic_block_def::aux.
|
inlinestatic |
Returns the gimplified predicate for basic block BB.
References basic_block_def::aux, and is_gimple_condexpr().
Referenced by if_convertible_loop_p(), ifc_temp_var(), and insert_gimplified_predicates().
|
inlinestatic |
Returns the sequence of statements of the gimplification of the predicate for basic block BB.
References basic_block_def::aux.
|
static |
Return true if one of the successor edges of BB exits LOOP.
References dump_file, and print_gimple_stmt().
|
static |
Combine all the basic blocks from LOOP into one or two super basic blocks. Replace PHI nodes with conditional modify expressions.
Merge basic blocks: first remove all the edges in the loop,
except for those from the exit block. Connect this node to loop header.
Redirect non-exit edges to loop->latch.
If the loop does not have an exit, reconnect header and latch.
Make stmts member of loop->header.
Update stmt list.
If possible, merge loop header to the block with the exit edge.
This reduces the number of basic blocks to two, to please the
vectorizer that handles only loops with two nodes.
|
static |
Returns true if N is either a constant or a SSA_NAME.
Referenced by fold_build_cond_expr().
|
static |
Basic block BB has two predecessors. Using predecessor's bb predicate, set an appropriate condition COND for the PHI node replacement. Return the true block whose phi arguments are selected when cond is true. LOOP is the loop containing the if-converted region, GSI is the place to insert the code for the if-conversion.
Prefer an edge with a not negated predicate.
??? That's a very weak cost model. Check if the edge we take the condition from is not critical.
We know that at least one non-critical edge exists. Select non loop header bb.
Gimplify the condition to a valid cond-expr conditonal operand.
|
static |
Returns either a COND_EXPR or the folded expression if the folded expression is a MIN_EXPR, a MAX_EXPR, an ABS_EXPR, a constant or a SSA_NAME.
References constant_or_ssa_name().
Referenced by insert_gimplified_predicates().
|
static |
Returns the fold of predicate C1 OR C2 at location LOC.
|
inlinestatic |
Free the predicate of basic block BB.
Release the SSA_NAMEs created for the gimplification of the
predicate.
|
static |
Returns true when the if-conversion pass is enabled.
|
static |
Get body of a LOOP in suitable order for if-conversion. It is caller's responsibility to deallocate basic block list. If-conversion suitable order is, breadth first sort (BFS) order with an additional constraint: select a block only if all its predecessors are already selected.
This block is now visited.
Not done yet.
|
static |
Return true when BB is if-convertible. This routine does not check basic block's statements and phis. A basic block is not if-convertible if: - it is non-empty and it is after the exit block (in BFS order), - it is after the exit block but before the latch, - its edges are not normal. EXIT_BB is the basic block containing the exit of the LOOP. BB is inside LOOP.
Be less adventurous and handle only normal edges.
At least one incoming edge has to be non-critical as otherwise edge
predicates are not equal to basic-block predicates of the edge
source.
|
static |
Return true when STMT is if-convertible. GIMPLE_ASSIGN statement is not if-convertible if, - it is not movable, - it could trap, - LHS is not var decl.
Some of these constrains might be too conservative.
References dump_file, and dump_flags.
|
static |
Return true when LOOP is if-convertible. LOOP is if-convertible if: - it is innermost, - it has two or more basic blocks, - it has only one exit, - loop header is not the exit edge, - if its basic blocks and phi nodes are if convertible.
Handle only innermost loop.
If only one block, no need for if-conversion.
More than one loop exit is too much to handle.
If one of the loop header's edge is an exit edge then do not
apply if-conversion.
References bb_predicate(), force_gimple_operand_gsi_1(), GSI_SAME_STMT, is_gimple_condexpr(), basic_block_def::preds, edge_def::src, basic_block_def::succs, and unshare_expr().
|
static |
Return true when LOOP is if-convertible. This is a helper function for if_convertible_loop_p. REFS and DDRS are initialized and freed in if_convertible_loop_p.
Don't if-convert the loop when the data dependences cannot be
computed: the loop won't be vectorized in that case. Allow statements that can be handled during if-conversion.
Check the if-convertibility of statements in predicated BBs.
|
static |
Return true when PHI is if-convertible. PHI is part of loop LOOP and it belongs to basic block BB. PHI is not if-convertible if: - it has more than 2 arguments. When the flag_tree_loop_if_convert_stores is not set, PHI is not if-convertible if: - a virtual PHI is immediately used in another PHI node, - there is a virtual PHI in a BB other than the loop->header.
When the flag_tree_loop_if_convert_stores is not set, check
that there are no memory writes in the branches of the loop to be
if-converted.
|
static |
Return true when STMT is if-convertible. A statement is if-convertible if: - it is an if-convertible GIMPLE_ASSIGN, - it is a GIMPLE_LABEL or a GIMPLE_COND.
We can only vectorize some builtins at the moment,
so restrict if-conversion to those. Don't know what to do with 'em so don't do anything.
|
static |
Returns a new SSA_NAME of type TYPE that is assigned the value of the expression EXPR. Inserts the statement created for this computation before GSI and leaves the iterator GSI at the same statement.
References bb_predicate(), and is_true_predicate().
Referenced by insert_gimplified_predicates().
|
static |
Wrapper around gimple_could_trap_p refined for the needs of the if-conversion. Try to prove that the memory accesses of STMT could not trap in the innermost loop containing STMT.
References dump_file, and dump_flags.
|
static |
Return true when the memory references of STMT won't trap in the if-converted code. There are two things that we have to check for: - writes to memory occur to writable memory: if-conversion of memory writes transforms the conditional memory writes into unconditional writes, i.e. "if (cond) A[i] = foo" is transformed into "A[i] = cond ? foo : A[i]", and as the write to memory may not be executed at all in the original code, it may be a readonly memory. To check that A is not const-qualified, we check that there exists at least an unconditional write to A in the current function. - reads or writes to memory are valid memory accesses for every iteration. To check that the memory accesses are correctly formed and that we are allowed to read and write in these locations, we check that the memory accesses to be if-converted occur at every iteration unconditionally.
References dump_file, and print_gimple_stmt().
|
inlinestatic |
Initializes to TRUE the predicate of basic block BB.
|
static |
Insert in each basic block of LOOP the statements produced by the gimplification of the predicates.
Do not insert statements for a basic block that is not
predicated. Also make sure that the predicate of the
basic block is set to true. Insert the predicate of the BB just after the label,
as the if-conversion of memory writes will use this
predicate. Insert the predicate of the BB at the end of the BB
as this would reduce the register pressure: the only
use of this predicate will be in successor BBs. Once the sequence is code generated, set it to NULL.
References bb_predicate(), fold_build_cond_expr(), force_gimple_operand_gsi_1(), gimple_assign_lhs(), gimple_assign_rhs1(), gimple_assign_set_rhs1(), gimple_assign_single_p(), gimple_vdef(), gsi_end_p(), gsi_next(), GSI_SAME_STMT, gsi_start_bb(), gsi_stmt(), ifc_temp_var(), is_gimple_condexpr(), is_true_predicate(), loop::num_nodes, swap(), unshare_expr(), and update_stmt().
|
inlinestatic |
Returns true when BB has a predicate that is not trivial: true or NULL_TREE.
|
inlinestatic |
Return true when COND is a true predicate.
References gimple_assign_rhs1(), gimple_assign_rhs_code(), is_gimple_assign(), and tcc_comparison.
Referenced by ifc_temp_var(), and insert_gimplified_predicates().
|
static |
Tree if-conversion pass management.
| gimple_opt_pass* make_pass_if_conversion | ( | ) |
|
static |
Returns true when the memory references of STMT are read or written unconditionally. In other words, this function returns true when for every data reference A in STMT there exist other accesses to a data reference with the same base with predicates that add up (OR-up) to the true predicate: this ensures that the data reference A is touched (read or written) on every iteration of the if-converted loop.
|
static |
Parses the predicate COND and returns its comparison code and operands OP0 and OP1.
|
static |
Return true when all predecessor blocks of BB are visited. The VISITED bitmap keeps track of the visited blocks.
|
static |
Replaces in LOOP all the scalar phi nodes other than those in the LOOP->header block with conditional modify expressions.
BB has two predecessors. Using predecessor's aux field, set
appropriate condition for the PHI node replacement.
|
static |
Returns true when the analysis of the predicates for all the basic blocks in LOOP succeeded. predicate_bbs first allocates the predicates of the basic blocks. These fields are then initialized with the tree expressions representing the predicates under which a basic block is executed in the LOOP. As the loop->header is executed at each iteration, it has the "true" predicate. Other statements executed under a condition are predicated with that condition, for example | if (x) | S1; | else | S2; S1 will be predicated with "x", and S2 will be predicated with "!x".
The loop latch is always executed and has no extra conditions
to be processed: skip it. Add new condition into destination's predicate list.
If C is true, then TRUE_EDGE is taken.
If C is false, then FALSE_EDGE is taken.
Not handled yet in if-conversion.
If current bb has only one successor, then consider it as an
unconditional goto. The successor bb inherits the predicate of its
predecessor. If there is no predicate in the predecessor
bb, then consider the successor bb as always executed. The loop header is always executed.
|
static |
@verbatim
Predicate each write to memory in LOOP.
This function transforms control flow constructs containing memory writes of the form:
| for (i = 0; i < N; i++) | if (cond) | A[i] = expr;
into the following form that does not contain control flow:
| for (i = 0; i < N; i++) | A[i] = cond ? expr : A[i];
The original CFG looks like this:
| bb_0 | i = 0 | end_bb_0 | | bb_1 | if (i < N) goto bb_5 else goto bb_2 | end_bb_1 | | bb_2 | cond = some_computation; | if (cond) goto bb_3 else goto bb_4 | end_bb_2 | | bb_3 | A[i] = expr; | goto bb_4 | end_bb_3 | | bb_4 | goto bb_1 | end_bb_4
insert_gimplified_predicates inserts the computation of the COND expression at the beginning of the destination basic block:
| bb_0 | i = 0 | end_bb_0 | | bb_1 | if (i < N) goto bb_5 else goto bb_2 | end_bb_1 | | bb_2 | cond = some_computation; | if (cond) goto bb_3 else goto bb_4 | end_bb_2 | | bb_3 | cond = some_computation; | A[i] = expr; | goto bb_4 | end_bb_3 | | bb_4 | goto bb_1 | end_bb_4
predicate_mem_writes is then predicating the memory write as follows:
| bb_0 | i = 0 | end_bb_0 | | bb_1 | if (i < N) goto bb_5 else goto bb_2 | end_bb_1 | | bb_2 | if (cond) goto bb_3 else goto bb_4 | end_bb_2 | | bb_3 | cond = some_computation; | A[i] = cond ? expr : A[i]; | goto bb_4 | end_bb_3 | | bb_4 | goto bb_1 | end_bb_4
and finally combine_blocks removes the basic block boundaries making the loop vectorizable:
| bb_0 | i = 0 | if (i < N) goto bb_5 else goto bb_1 | end_bb_0 | | bb_1 | cond = some_computation; | A[i] = cond ? expr : A[i]; | if (i < N) goto bb_5 else goto bb_4 | end_bb_1 | | bb_4 | goto bb_1 | end_bb_4
|
static |
Replace a scalar PHI node with a COND_EXPR using COND as condition.
This routine does not handle PHI nodes with more than two
arguments.
For example,
S1: A = PHI <x1(1), x2(5)>
is converted into,
S2: A = cond ? x1 : x2;
The generated code is inserted at GSI that points to the top of
basic block's statement list. When COND is true, phi arg from
TRUE_BB is selected. Do not handle virtual phi nodes.
Use condition that is not TRUTH_NOT_EXPR in conditional modify expr.
Build new RHS using selected condition and arguments.
|
static |
Remove all GIMPLE_CONDs and GIMPLE_LABELs of all the basic blocks other than the exit and latch of the LOOP. Also resets the GIMPLE_DEBUG information.
??? Should there be conditional GIMPLE_DEBUG_BINDs?
References CDI_DOMINATORS, loop::header, loop::latch, loop_exit_edge_p(), make_edge(), redirect_edge_and_branch(), set_immediate_dominator(), and basic_block_def::succs.
|
inlinestatic |
Free the predicate of BB and reinitialize it with the true predicate.
|
inlinestatic |
Sets the gimplified predicate COND for basic block BB.
References basic_block_def::aux.
|
inlinestatic |
Sets the sequence of statements STMTS of the gimplification of the predicate for basic block BB.
|
static |
If-convert LOOP when it is legal. For the moment this pass has no profitability analysis. Returns true when something changed.
Now all statements are if-convertible. Combine all the basic
blocks into one huge basic block doing the if-conversion
on-the-fly.
|
static |
Returns true when the memory references of STMT are unconditionally written. In other words, this function returns true when for every data reference A written in STMT, there exist other writes to the same data reference with predicates that add up (OR-up) to the true predicate: this ensures that the data reference A is written on every iteration of the if-converted loop.
|
static |
@verbatim
If-conversion for vectorizer. Copyright (C) 2004-2013 Free Software Foundation, Inc. Contributed by Devang Patel dpatel@apple.com
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version.
GCC is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with GCC; see the file COPYING3. If not see http://www.gnu.org/licenses/.
This pass implements a tree level if-conversion of loops. Its
initial goal is to help the vectorizer to vectorize loops with
conditions.
A short description of if-conversion:
o Decide if a loop is if-convertible or not.
o Walk all loop basic blocks in breadth first order (BFS order).
o Remove conditional statements (at the end of basic block)
and propagate condition into destination basic blocks'
predicate list.
o Replace modify expression with conditional modify expression
using current basic block's condition.
o Merge all basic blocks
o Replace phi nodes with conditional modify expr
o Merge all basic blocks into header
Sample transformation:
INPUT
-----
# i_23 = PHI <0(0), i_18(10)>;
<L0>:;
j_15 = A[i_23];
if (j_15 > 41) goto <L1>; else goto <L17>;
<L17>:;
goto <bb 3> (<L3>);
<L1>:;
# iftmp.2_4 = PHI <0(8), 42(2)>;
<L3>:;
A[i_23] = iftmp.2_4;
i_18 = i_23 + 1;
if (i_18 <= 15) goto <L19>; else goto <L18>;
<L19>:;
goto <bb 1> (<L0>);
<L18>:;
OUTPUT
------
# i_23 = PHI <0(0), i_18(10)>;
<L0>:;
j_15 = A[i_23];
<L3>:;
iftmp.2_4 = j_15 > 41 ? 42 : 0;
A[i_23] = iftmp.2_4;
i_18 = i_23 + 1;
if (i_18 <= 15) goto <L19>; else goto <L18>;
<L19>:;
goto <bb 1> (<L0>);
<L18>:;
List of basic blocks in if-conversion-suitable order.
1.8.1.1