代码拉取完成,页面将自动刷新
同步操作将从 src-openEuler/gcc 强制同步,此操作会覆盖自 Fork 仓库以来所做的任何修改,且无法恢复!!!
确定后同步将在后台操作,完成时将刷新页面,请耐心等待。
From c690da762e873d0f5c66ea084e420ba4842354a6 Mon Sep 17 00:00:00 2001
From: Jakub Jelinek <jakub@redhat.com>
Date: Wed, 4 Nov 2020 11:55:29 +0100
Subject: [PATCH 02/35] [Backport] phiopt: Optimize x ? 1024 : 0 to (int) x <<
10 [PR97690]
Reference: https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=3e190757fa332d327bee27495f37beb01155cfab
The following patch generalizes the x ? 1 : 0 -> (int) x optimization
to handle also left shifts by constant.
During x86_64-linux and i686-linux bootstraps + regtests it triggered
in 1514 unique non-LTO -m64 cases (sort -u on log mentioning
filename, function name and shift count) and 1866 -m32 cases.
Unfortunately, the patch regresses (before the tests have been adjusted):
+FAIL: gcc.dg/tree-ssa/ssa-ccp-11.c scan-tree-dump-times optimized "if " 0
+FAIL: gcc.dg/vect/bb-slp-pattern-2.c -flto -ffat-lto-objects scan-tree-dump-times slp1 "optimized: basic block" 1
+FAIL: gcc.dg/vect/bb-slp-pattern-2.c scan-tree-dump-times slp1 "optimized: basic block" 1
and in both cases it actually results in worse code.
> > We'd need some optimization that would go through all PHI edges and
> > compute if some use of the phi results don't actually compute a constant
> > across all the PHI edges - 1 & 0 and 0 & 1 is always 0.
> PRE should do this, IMHO only optimizing it at -O2 is fine.
> > Similarly, in the slp vectorization test there is:
> > a[0] = b[0] ? 1 : 7;
> note this, carefully avoiding the already "optimized" b[0] ? 1 : 0 ...
> So the option is to put : 7 in the 2, 4 an 8 case as well. The testcase
> wasn't added for any real-world case but is artificial I guess for
> COND_EXPR handling of invariants.
> But yeah, for things like SLP it means we eventually have to
> implement reverse transforms for all of this to make the lanes
> matching. But that's true anyway for things like x + 1 vs. x + 0
> or x / 3 vs. x / 2 or other simplifications we do.
2020-11-04 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/97690
* tree-ssa-phiopt.c (conditional_replacement): Also optimize
cond ? pow2p_cst : 0 as ((type) cond) << cst.
* gcc.dg/tree-ssa/phi-opt-22.c: New test.
* gcc.dg/tree-ssa/ssa-ccp-11.c: Use -O2 instead of -O1.
* gcc.dg/vect/bb-slp-pattern-2.c (foo): Use ? 2 : 7, ? 4 : 7 and
? 8 : 7 instead of ? 2 : 0, ? 4 : 0, ? 8 : 0.
---
gcc/testsuite/gcc.dg/tree-ssa/phi-opt-22.c | 11 ++++++
gcc/testsuite/gcc.dg/tree-ssa/ssa-ccp-11.c | 2 +-
gcc/testsuite/gcc.dg/vect/bb-slp-pattern-2.c | 6 ++--
gcc/tree-ssa-phiopt.c | 38 ++++++++++++++------
4 files changed, 43 insertions(+), 14 deletions(-)
create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-22.c
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-22.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-22.c
new file mode 100644
index 000000000..fd3706666
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-22.c
@@ -0,0 +1,11 @@
+/* PR tree-optimization/97690 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-phiopt2" } */
+
+int foo (_Bool d) { return d ? 2 : 0; }
+int bar (_Bool d) { return d ? 1 : 0; }
+int baz (_Bool d) { return d ? -__INT_MAX__ - 1 : 0; }
+int qux (_Bool d) { return d ? 1024 : 0; }
+
+/* { dg-final { scan-tree-dump-not "if" "phiopt2" } } */
+/* { dg-final { scan-tree-dump-times " << " 3 "phiopt2" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ssa-ccp-11.c b/gcc/testsuite/gcc.dg/tree-ssa/ssa-ccp-11.c
index 36b8e7fc8..d70ea5a01 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/ssa-ccp-11.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/ssa-ccp-11.c
@@ -1,5 +1,5 @@
/* { dg-do compile } */
-/* { dg-options "-O1 -fdump-tree-optimized" } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
/* Test for CPROP across a DAG. */
diff --git a/gcc/testsuite/gcc.dg/vect/bb-slp-pattern-2.c b/gcc/testsuite/gcc.dg/vect/bb-slp-pattern-2.c
index d32cb7585..e64f0115a 100644
--- a/gcc/testsuite/gcc.dg/vect/bb-slp-pattern-2.c
+++ b/gcc/testsuite/gcc.dg/vect/bb-slp-pattern-2.c
@@ -13,13 +13,13 @@ foo (short * __restrict__ a, int * __restrict__ b, int stride)
for (i = 0; i < N/stride; i++, a += stride, b += stride)
{
a[0] = b[0] ? 1 : 7;
- a[1] = b[1] ? 2 : 0;
+ a[1] = b[1] ? 2 : 7;
a[2] = b[2] ? 3 : 0;
- a[3] = b[3] ? 4 : 0;
+ a[3] = b[3] ? 4 : 7;
a[4] = b[4] ? 5 : 0;
a[5] = b[5] ? 6 : 0;
a[6] = b[6] ? 7 : 0;
- a[7] = b[7] ? 8 : 0;
+ a[7] = b[7] ? 8 : 7;
}
}
diff --git a/gcc/tree-ssa-phiopt.c b/gcc/tree-ssa-phiopt.c
index 591b6435f..85587e8d1 100644
--- a/gcc/tree-ssa-phiopt.c
+++ b/gcc/tree-ssa-phiopt.c
@@ -753,7 +753,9 @@ conditional_replacement (basic_block cond_bb, basic_block middle_bb,
gimple_stmt_iterator gsi;
edge true_edge, false_edge;
tree new_var, new_var2;
- bool neg;
+ bool neg = false;
+ int shift = 0;
+ tree nonzero_arg;
/* FIXME: Gimplification of complex type is too hard for now. */
/* We aren't prepared to handle vectors either (and it is a question
@@ -764,14 +766,22 @@ conditional_replacement (basic_block cond_bb, basic_block middle_bb,
|| POINTER_TYPE_P (TREE_TYPE (arg1))))
return false;
- /* The PHI arguments have the constants 0 and 1, or 0 and -1, then
- convert it to the conditional. */
- if ((integer_zerop (arg0) && integer_onep (arg1))
- || (integer_zerop (arg1) && integer_onep (arg0)))
- neg = false;
- else if ((integer_zerop (arg0) && integer_all_onesp (arg1))
- || (integer_zerop (arg1) && integer_all_onesp (arg0)))
+ /* The PHI arguments have the constants 0 and 1, or 0 and -1 or
+ 0 and (1 << cst), then convert it to the conditional. */
+ if (integer_zerop (arg0))
+ nonzero_arg = arg1;
+ else if (integer_zerop (arg1))
+ nonzero_arg = arg0;
+ else
+ return false;
+ if (integer_all_onesp (nonzero_arg))
neg = true;
+ else if (integer_pow2p (nonzero_arg))
+ {
+ shift = tree_log2 (nonzero_arg);
+ if (shift && POINTER_TYPE_P (TREE_TYPE (nonzero_arg)))
+ return false;
+ }
else
return false;
@@ -783,12 +793,12 @@ conditional_replacement (basic_block cond_bb, basic_block middle_bb,
falls through into BB.
There is a single PHI node at the join point (BB) and its arguments
- are constants (0, 1) or (0, -1).
+ are constants (0, 1) or (0, -1) or (0, (1 << shift)).
So, given the condition COND, and the two PHI arguments, we can
rewrite this PHI into non-branching code:
- dest = (COND) or dest = COND'
+ dest = (COND) or dest = COND' or dest = (COND) << shift
We use the condition as-is if the argument associated with the
true edge has the value one or the argument associated with the
@@ -823,6 +833,14 @@ conditional_replacement (basic_block cond_bb, basic_block middle_bb,
cond = fold_build1_loc (gimple_location (stmt),
NEGATE_EXPR, TREE_TYPE (cond), cond);
}
+ else if (shift)
+ {
+ cond = fold_convert_loc (gimple_location (stmt),
+ TREE_TYPE (result), cond);
+ cond = fold_build2_loc (gimple_location (stmt),
+ LSHIFT_EXPR, TREE_TYPE (cond), cond,
+ build_int_cst (integer_type_node, shift));
+ }
/* Insert our new statements at the end of conditional block before the
COND_STMT. */
--
2.27.0.windows.1
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。