From: Philipp Tomsich <philipp.tomsich@vrull.eu>
To: gcc-patches@gcc.gnu.org
Cc: Christoph Muellner <christoph.muellner@vrull.eu>,
Tamar Christina <tamar.christina@arm.com>,
Jiang-Ning Liu <jiangning.liu@amperecomputing.com>,
Richard Biener <rguenther@suse.de>,
Jeff Law <jlaw@ventanamicro.com>,
Philipp Tomsich <philipp.tomsich@vrull.eu>
Subject: [PATCH] ifcombine: fold two bit tests with different polarity
Date: Thu, 10 Nov 2022 00:08:15 +0100 [thread overview]
Message-ID: <20221109230815.3240583-1-philipp.tomsich@vrull.eu> (raw)
Our ifcombine pass combines 2 single-bit tests into a single test of
the form "(a & T) == T", requiring the same polarity (i.e., tests for
bit set/cleared) for both bit-tests. However some applications test
against two bits expecting one set and the other cleared.
This adds support for the case "(a & T) == C" where T is a constant
with 2 bits set and C has only one of those bits set.
gcc/ChangeLog:
* tree-ssa-ifcombine.cc (ifcombine_ifandif): Add support for
combining two bit-tests that test for bits of different
polarity.
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/ssa-ifcombine-15.c: New test.
Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
---
.../gcc.dg/tree-ssa/ssa-ifcombine-15.c | 14 +++++
gcc/tree-ssa-ifcombine.cc | 56 +++++++++++++++++++
2 files changed, 70 insertions(+)
create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/ssa-ifcombine-15.c
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ssa-ifcombine-15.c b/gcc/testsuite/gcc.dg/tree-ssa/ssa-ifcombine-15.c
new file mode 100644
index 00000000000..081faa39628
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/ssa-ifcombine-15.c
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O -fdump-tree-ifcombine-details-blocks" } */
+
+void sink();
+
+void reversed(unsigned char *a)
+{
+ if (*a & 0x60)
+ if (!(*a & 0x02))
+ g();
+}
+
+/* { dg-final { scan-tree-dump "optimizing double bit test" } } */
+
diff --git a/gcc/tree-ssa-ifcombine.cc b/gcc/tree-ssa-ifcombine.cc
index cd6331f84db..ea49cc2bff1 100644
--- a/gcc/tree-ssa-ifcombine.cc
+++ b/gcc/tree-ssa-ifcombine.cc
@@ -498,6 +498,62 @@ ifcombine_ifandif (basic_block inner_cond_bb, bool inner_inv,
return true;
}
+ /* See if we test polarity-reversed single bits of the same name in
+ both tests. In that case remove the outer test, merging both
+ else edges, and change the inner one to test for
+ name & (bit1 | bit2) == (bit2). */
+ else if ((recognize_single_bit_test (inner_cond, &name1, &bit1, !inner_inv)
+ && recognize_single_bit_test (outer_cond, &name2, &bit2, outer_inv)
+ && name1 == name2)
+ || (recognize_single_bit_test (inner_cond, &name2, &bit2, inner_inv)
+ && recognize_single_bit_test (outer_cond, &name1, &bit1, !outer_inv)
+ && name1 == name2))
+ {
+ tree t, t2, t3;
+
+ /* Do it. */
+ gsi = gsi_for_stmt (inner_cond);
+ t = fold_build2 (LSHIFT_EXPR, TREE_TYPE (name1),
+ build_int_cst (TREE_TYPE (name1), 1), bit1);
+ t2 = fold_build2 (LSHIFT_EXPR, TREE_TYPE (name1),
+ build_int_cst (TREE_TYPE (name1), 1), bit2);
+ t = fold_build2 (BIT_IOR_EXPR, TREE_TYPE (name1), t, t2);
+ t = force_gimple_operand_gsi (&gsi, t, true, NULL_TREE,
+ true, GSI_SAME_STMT);
+ t3 = fold_build2 (BIT_AND_EXPR, TREE_TYPE (name1), name1, t);
+ t3 = force_gimple_operand_gsi (&gsi, t3, true, NULL_TREE,
+ true, GSI_SAME_STMT);
+ t = fold_build2 (result_inv ? NE_EXPR : EQ_EXPR,
+ boolean_type_node, t2, t3);
+ t = canonicalize_cond_expr_cond (t);
+ if (!t)
+ return false;
+ gimple_cond_set_condition_from_tree (inner_cond, t);
+ update_stmt (inner_cond);
+
+ /* Leave CFG optimization to cfg_cleanup. */
+ gimple_cond_set_condition_from_tree (outer_cond,
+ outer_inv ? boolean_false_node : boolean_true_node);
+ update_stmt (outer_cond);
+
+ update_profile_after_ifcombine (inner_cond_bb, outer_cond_bb);
+
+ if (dump_file)
+ {
+ fprintf (dump_file, "optimizing double bit test to ");
+ print_generic_expr (dump_file, name1);
+ fprintf (dump_file, " & T == C\nwith temporary T = (1 << ");
+ print_generic_expr (dump_file, bit1);
+ fprintf (dump_file, ") | (1 << ");
+ print_generic_expr (dump_file, bit2);
+ fprintf (dump_file, ")\nand temporary C = (1 << ");
+ print_generic_expr (dump_file, bit2);
+ fprintf (dump_file, ")\n");
+ }
+
+ return true;
+ }
+
/* See if we have two bit tests of the same name in both tests.
In that case remove the outer test and change the inner one to
test for name & (bits1 | bits2) != 0. */
--
2.34.1
next reply other threads:[~2022-11-09 23:08 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-09 23:08 Philipp Tomsich [this message]
2022-11-10 7:02 ` Richard Biener
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=20221109230815.3240583-1-philipp.tomsich@vrull.eu \
--to=philipp.tomsich@vrull.eu \
--cc=christoph.muellner@vrull.eu \
--cc=gcc-patches@gcc.gnu.org \
--cc=jiangning.liu@amperecomputing.com \
--cc=jlaw@ventanamicro.com \
--cc=rguenther@suse.de \
--cc=tamar.christina@arm.com \
/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).