From: "Martin Liška" <mliska@suse.cz>
To: Richard Biener <richard.guenther@gmail.com>
Cc: Jakub Jelinek <jakub@redhat.com>, Michael Matz <matz@suse.de>,
GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] Optimize macro: make it more predictable
Date: Mon, 13 Sep 2021 15:52:42 +0200 [thread overview]
Message-ID: <6179ac10-8d2c-d0d5-21f4-0b6678fc8592@suse.cz> (raw)
In-Reply-To: <CAFiYyc0=UMqMsocd+8fTWR==5kw7Cg_222_Z23MAzFqDNNyLeQ@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 493 bytes --]
On 8/27/21 11:05, Richard Biener wrote:
> So with ignoring darktable which seems completely insane the cases
> will likely continue
> to work as intended if we change from the current scheme to appending
> as proposed.
All right, I'm addressing the flag_complex_method in a separate sub-thread.
There's slightly updated version of the patch where I modifed the documentation bits.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
Ready to be installed?
Thanks,
Martin
[-- Attachment #2: 0001-Append-target-optimize-attr-to-the-current-cmdline.patch --]
[-- Type: text/x-patch, Size: 7121 bytes --]
From e13e3ec56acfb62543bc1912f1310d00eefba5c3 Mon Sep 17 00:00:00 2001
From: Martin Liska <mliska@suse.cz>
Date: Wed, 2 Jun 2021 08:44:37 +0200
Subject: [PATCH] Append target/optimize attr to the current cmdline.
gcc/c-family/ChangeLog:
* c-common.c (parse_optimize_options): Combine optimize
options with what was provided on the command line.
gcc/ChangeLog:
* toplev.c (toplev::main): Save decoded optimization options.
* toplev.h (save_opt_decoded_options): New.
* doc/extend.texi: Be more clear about optimize and target
attributes.
gcc/testsuite/ChangeLog:
* gcc.target/i386/avx512er-vrsqrt28ps-3.c: Disable fast math.
* gcc.target/i386/avx512er-vrsqrt28ps-5.c: Likewise.
* gcc.target/i386/attr-optimize.c: New test.
---
gcc/c-family/c-common.c | 17 +++++++++++--
gcc/doc/extend.texi | 8 +++++--
gcc/testsuite/gcc.target/i386/attr-optimize.c | 24 +++++++++++++++++++
.../gcc.target/i386/avx512er-vrsqrt28ps-3.c | 2 +-
.../gcc.target/i386/avx512er-vrsqrt28ps-5.c | 2 +-
gcc/toplev.c | 8 +++++++
gcc/toplev.h | 1 +
7 files changed, 56 insertions(+), 6 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/i386/attr-optimize.c
diff --git a/gcc/c-family/c-common.c b/gcc/c-family/c-common.c
index 017e41537ac..09038e3175f 100644
--- a/gcc/c-family/c-common.c
+++ b/gcc/c-family/c-common.c
@@ -5904,9 +5904,22 @@ parse_optimize_options (tree args, bool attr_p)
j++;
}
decoded_options_count = j;
- /* And apply them. */
+
+ /* Merge the decoded options with save_decoded_options. */
+ unsigned save_opt_count = save_opt_decoded_options.length ();
+ unsigned merged_decoded_options_count
+ = save_opt_count + decoded_options_count;
+ cl_decoded_option *merged_decoded_options
+ = XNEWVEC (cl_decoded_option, merged_decoded_options_count);
+
+ for (unsigned i = 0; i < save_opt_count; ++i)
+ merged_decoded_options[i] = save_opt_decoded_options[i];
+ for (unsigned i = 0; i < decoded_options_count; ++i)
+ merged_decoded_options[save_opt_count + i] = decoded_options[i];
+
+ /* And apply them. */
decode_options (&global_options, &global_options_set,
- decoded_options, decoded_options_count,
+ merged_decoded_options, merged_decoded_options_count,
input_location, global_dc, NULL);
free (decoded_options);
diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
index 7fb22ed8063..1cb7e33ca29 100644
--- a/gcc/doc/extend.texi
+++ b/gcc/doc/extend.texi
@@ -3639,7 +3639,10 @@ take function pointer arguments.
@cindex @code{optimize} function attribute
The @code{optimize} attribute is used to specify that a function is to
be compiled with different optimization options than specified on the
-command line. Valid arguments are constant non-negative integers and
+command line. The optimize attribute arguments of a function behave
+behave as if appended to the command-line.
+
+Valid arguments are constant non-negative integers and
strings. Each numeric argument specifies an optimization @var{level}.
Each @var{string} argument consists of one or more comma-separated
substrings. Each substring that begins with the letter @code{O} refers
@@ -3843,7 +3846,8 @@ This attribute prevents stack protection code for the function.
Multiple target back ends implement the @code{target} attribute
to specify that a function is to
be compiled with different target options than specified on the
-command line. One or more strings can be provided as arguments.
+command line. The original target command-line options are ignored.
+One or more strings can be provided as arguments.
Each string consists of one or more comma-separated suffixes to
the @code{-m} prefix jointly forming the name of a machine-dependent
option. @xref{Submodel Options,,Machine-Dependent Options}.
diff --git a/gcc/testsuite/gcc.target/i386/attr-optimize.c b/gcc/testsuite/gcc.target/i386/attr-optimize.c
new file mode 100644
index 00000000000..f5db028f1fd
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/attr-optimize.c
@@ -0,0 +1,24 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-O1 -ftree-slp-vectorize -march=znver1 -fdump-tree-optimized" } */
+
+/* Use -O2, but -ftree-slp-vectorize option should be preserved and used. */
+#pragma GCC optimize "-O2"
+
+typedef struct {
+ long n[4];
+} secp256k1_fe;
+
+void *a;
+int c;
+static void
+fn1(secp256k1_fe *p1, int p2)
+{
+ p1->n[0] = p1->n[1] = p2;
+}
+void
+fn2()
+{
+ fn1(a, !c);
+}
+
+/* { dg-final { scan-tree-dump { MEM <vector\(2\) long int> \[[^]]*\] = } "optimized" } } */
diff --git a/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-3.c b/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-3.c
index 1ba8172d6e3..40aefb50844 100644
--- a/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-3.c
+++ b/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-3.c
@@ -8,7 +8,7 @@
#define MAX 1000
#define EPS 0.00001
-__attribute__ ((noinline, optimize (1)))
+__attribute__ ((noinline, optimize (1, "-fno-fast-math")))
void static
compute_rsqrt_ref (float *a, float *r)
{
diff --git a/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-5.c b/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-5.c
index e067a81e562..498f4d50aa5 100644
--- a/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-5.c
+++ b/gcc/testsuite/gcc.target/i386/avx512er-vrsqrt28ps-5.c
@@ -8,7 +8,7 @@
#define MAX 1000
#define EPS 0.00001
-__attribute__ ((noinline, optimize (1)))
+__attribute__ ((noinline, optimize (1, "-fno-fast-math")))
void static
compute_sqrt_ref (float *a, float *r)
{
diff --git a/gcc/toplev.c b/gcc/toplev.c
index 14d1335e79e..538ffdbd31a 100644
--- a/gcc/toplev.c
+++ b/gcc/toplev.c
@@ -121,6 +121,9 @@ static bool no_backend;
struct cl_decoded_option *save_decoded_options;
unsigned int save_decoded_options_count;
+/* Vector of saved Optimization decoded command line options. */
+auto_vec<cl_decoded_option> save_opt_decoded_options;
+
/* Used to enable -fvar-tracking, -fweb and -frename-registers according
to optimize in process_options (). */
#define AUTODETECT_VALUE 2
@@ -2342,6 +2345,11 @@ toplev::main (int argc, char **argv)
&save_decoded_options,
&save_decoded_options_count);
+ /* Save Optimization decoded options. */
+ for (unsigned i = 0; i < save_decoded_options_count; ++i)
+ if (cl_options[save_decoded_options[i].opt_index].flags & CL_OPTIMIZATION)
+ save_opt_decoded_options.safe_push (save_decoded_options[i]);
+
/* Perform language-specific options initialization. */
lang_hooks.init_options (save_decoded_options_count, save_decoded_options);
diff --git a/gcc/toplev.h b/gcc/toplev.h
index f543554b15f..c44c5ff926a 100644
--- a/gcc/toplev.h
+++ b/gcc/toplev.h
@@ -23,6 +23,7 @@ along with GCC; see the file COPYING3. If not see
/* Decoded options, and number of such options. */
extern struct cl_decoded_option *save_decoded_options;
extern unsigned int save_decoded_options_count;
+extern auto_vec<cl_decoded_option> save_opt_decoded_options;
class timer;
--
2.33.0
next prev parent reply other threads:[~2021-09-13 13:52 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-23 11:47 Martin Liška
2020-11-03 13:27 ` Richard Biener
2020-11-03 13:34 ` Jakub Jelinek
2020-11-03 13:40 ` Richard Biener
2020-11-09 10:35 ` Martin Liška
2020-11-26 13:56 ` Martin Liška
2020-12-07 11:03 ` Martin Liška
2021-01-11 13:10 ` Martin Liška
2020-11-09 10:27 ` Martin Liška
2020-11-06 17:34 ` Jeff Law
2020-11-09 10:36 ` Martin Liška
2021-07-01 13:13 ` Martin Liška
2021-08-10 15:52 ` Martin Liška
2021-08-24 11:06 ` Martin Liška
2021-08-24 12:13 ` Richard Biener
2021-08-24 13:04 ` Martin Liška
2021-08-26 11:04 ` Richard Biener
2021-08-26 12:39 ` Martin Liška
2021-08-26 13:20 ` Richard Biener
2021-08-27 8:35 ` Martin Liška
2021-08-27 9:05 ` Richard Biener
2021-09-13 13:52 ` Martin Liška [this message]
2021-09-19 5:46 ` Jeff Law
2021-09-06 11:37 ` [PATCH] flag_complex_method: support optimize attribute Martin Liška
2021-09-06 11:46 ` Jakub Jelinek
2021-09-06 12:16 ` Richard Biener
2021-09-06 12:24 ` Jakub Jelinek
2021-09-07 9:42 ` Martin Liška
2021-09-13 13:32 ` Martin Liška
2021-09-19 14:45 ` Jeff Law
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=6179ac10-8d2c-d0d5-21f4-0b6678fc8592@suse.cz \
--to=mliska@suse.cz \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=matz@suse.de \
--cc=richard.guenther@gmail.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).