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
From e13e3ec56acfb62543bc1912f1310d00eefba5c3 Mon Sep 17 00:00:00 2001
From: Martin Liska <mli...@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
