From: Cupertino Miranda <cupertino.miranda@oracle.com>
To: gcc-patches@gcc.gnu.org
Cc: jose.marchesi@oracle.com, elena.zannoni@oracle.com,
david.faust@oracle.com,
Cupertino Miranda <cupertino.miranda@oracle.com>
Subject: [PATCH] bpf: Forces __buildin_memcmp not to generate a call upto 1024 bytes.
Date: Mon, 13 Nov 2023 22:37:23 +0000 [thread overview]
Message-ID: <20231113223723.11760-1-cupertino.miranda@oracle.com> (raw)
This patch forces __builtin_memcmp calls upto data sizes of 1024 to
become inline in caller.
This is a requirement by BPF and it mimics the default behaviour of the
clang BPF implementation.
gcc/ChangeLog:
* config/bpf/bpf.cc (bpf_use_by_pieces_infrastructure_p): Added
function to bypass default behaviour.
* config/bpf/bpf.h (COMPARE_MAX_PIECES): Defined to 1024 bytes.
---
gcc/config/bpf/bpf.cc | 16 ++++++++++++++++
gcc/config/bpf/bpf.h | 5 +++++
2 files changed, 21 insertions(+)
diff --git a/gcc/config/bpf/bpf.cc b/gcc/config/bpf/bpf.cc
index a0956a069729..764a3e487cb6 100644
--- a/gcc/config/bpf/bpf.cc
+++ b/gcc/config/bpf/bpf.cc
@@ -1115,6 +1115,22 @@ bpf_small_register_classes_for_mode_p (machine_mode mode)
#define TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P \
bpf_small_register_classes_for_mode_p
+static bool
+bpf_use_by_pieces_infrastructure_p (unsigned HOST_WIDE_INT size,
+ unsigned int align ATTRIBUTE_UNUSED,
+ enum by_pieces_operation op,
+ bool speed_p)
+{
+ if (op != COMPARE_BY_PIECES)
+ return default_use_by_pieces_infrastructure_p (size, align, op, speed_p);
+
+ return size <= COMPARE_MAX_PIECES;
+}
+
+#undef TARGET_USE_BY_PIECES_INFRASTRUCTURE_P
+#define TARGET_USE_BY_PIECES_INFRASTRUCTURE_P \
+ bpf_use_by_pieces_infrastructure_p
+
/* Finally, build the GCC target. */
struct gcc_target targetm = TARGET_INITIALIZER;
diff --git a/gcc/config/bpf/bpf.h b/gcc/config/bpf/bpf.h
index 82702aa7b6ba..1f177ec4c4ef 100644
--- a/gcc/config/bpf/bpf.h
+++ b/gcc/config/bpf/bpf.h
@@ -489,6 +489,11 @@ enum reg_class
locations. */
#define MOVE_MAX 8
+/* Allow upto 1024 bytes moves to occur using by_pieces
+ infrastructure. This mimics clang behaviour when using
+ __builtin_memcmp. */
+#define COMPARE_MAX_PIECES 1024
+
/* An alias for the machine mode for pointers. */
#define Pmode DImode
--
2.30.2
next reply other threads:[~2023-11-13 22:37 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-11-13 22:37 Cupertino Miranda [this message]
2023-11-24 16:54 ` Jose E. Marchesi
2023-11-28 12:46 ` Cupertino Miranda
2023-11-28 12:54 ` Cupertino Miranda
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=20231113223723.11760-1-cupertino.miranda@oracle.com \
--to=cupertino.miranda@oracle.com \
--cc=david.faust@oracle.com \
--cc=elena.zannoni@oracle.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jose.marchesi@oracle.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).