]> git.ipfire.org Git - thirdparty/gcc.git/commitdiff
Make option mvzeroupper independent of optimization level.
authorliuhongt <hongtao.liu@intel.com>
Mon, 26 Jun 2023 01:50:25 +0000 (09:50 +0800)
committerliuhongt <hongtao.liu@intel.com>
Wed, 28 Jun 2023 01:54:03 +0000 (09:54 +0800)
pass_insert_vzeroupper is under condition

TARGET_AVX && TARGET_VZEROUPPER
&& flag_expensive_optimizations && !optimize_size

But the document of mvzeroupper doesn't mention the insertion
required -O2 and above, it may confuse users when they explicitly
use -Os -mvzeroupper.

------------
mvzeroupper
Target Mask(VZEROUPPER) Save
Generate vzeroupper instruction before a transfer of control flow out of
the function.
------------

The patch moves flag_expensive_optimizations && !optimize_size to
ix86_option_override_internal. It makes -mvzeroupper independent of
optimization level, but still keeps the behavior of architecture
tuning(emit_vzeroupper) unchanged.

gcc/ChangeLog:

* config/i386/i386-features.cc (pass_insert_vzeroupper:gate):
Move flag_expensive_optimizations && !optimize_size to ..
* config/i386/i386-options.cc (ix86_option_override_internal):
.. this, it makes -mvzeroupper independent of optimization
level, but still keeps the behavior of architecture
tuning(emit_vzeroupper) unchanged.

gcc/testsuite/ChangeLog:

* gcc.target/i386/avx-vzeroupper-29.c: New testcase.
* gcc.target/i386/avx-vzeroupper-12.c: Adjust testcase.
* gcc.target/i386/avx-vzeroupper-7.c: Ditto.
* gcc.target/i386/avx-vzeroupper-9.c: Ditto.

gcc/config/i386/i386-features.cc
gcc/config/i386/i386-options.cc
gcc/testsuite/gcc.target/i386/avx-vzeroupper-12.c
gcc/testsuite/gcc.target/i386/avx-vzeroupper-29.c [new file with mode: 0644]
gcc/testsuite/gcc.target/i386/avx-vzeroupper-7.c
gcc/testsuite/gcc.target/i386/avx-vzeroupper-9.c

index 6fe41c3c24f3429905ead6e89d6a5dbc931ab201..6a2444eb6b68cc190436974423872d5bdd53e116 100644 (file)
@@ -1875,8 +1875,7 @@ public:
   /* opt_pass methods: */
   virtual bool gate (function *)
     {
-      return TARGET_AVX && TARGET_VZEROUPPER
-       && flag_expensive_optimizations && !optimize_size;
+      return TARGET_AVX && TARGET_VZEROUPPER;
     }
 
   virtual unsigned int execute (function *)
index 099cec4b6102c5294fcc9606f8484642d5b0ca4e..70ddcd284c4b6b1310fcd69548b0dca237ec35a4 100644 (file)
@@ -2676,7 +2676,9 @@ ix86_option_override_internal (bool main_args_p,
     sorry ("%<-mcall-ms2sysv-xlogues%> isn%'t currently supported with SEH");
 
   if (!(opts_set->x_target_flags & MASK_VZEROUPPER)
-      && TARGET_EMIT_VZEROUPPER)
+      && TARGET_EMIT_VZEROUPPER
+      && flag_expensive_optimizations
+      && !optimize_size)
     opts->x_target_flags |= MASK_VZEROUPPER;
   if (!(opts_set->x_target_flags & MASK_STV))
     opts->x_target_flags |= MASK_STV;
index e694d4048bdbd0be746fdafed0fcb77c1b9e3714..5a40e87832c8bcb0c7f05e973be62bd4e5059e08 100644 (file)
@@ -16,5 +16,6 @@ foo ()
   _mm256_zeroupper ();
 }
 
-/* { dg-final { scan-assembler-times "avx_vzeroupper" 4 } } */
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 4 { target ia32 } } } */
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 5 { target { ! ia32 } } } } */
 /* { dg-final { scan-assembler-times "\\*avx_vzeroall" 1 } } */
diff --git a/gcc/testsuite/gcc.target/i386/avx-vzeroupper-29.c b/gcc/testsuite/gcc.target/i386/avx-vzeroupper-29.c
new file mode 100644 (file)
index 0000000..4af6377
--- /dev/null
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O0 -mavx -mtune=generic -mvzeroupper -dp" } */
+
+#include <immintrin.h>
+
+extern __m256 x, y;
+
+void
+foo ()
+{
+  x = y;
+}
+
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 1 } } */
index ab6d68779b380b34faca7beec60055a7a31897b4..75fe58897830e3a92fd5e9b0ece2ff5df6ba57d3 100644 (file)
@@ -12,4 +12,5 @@ foo ()
   _mm256_zeroupper ();
 }
 
-/* { dg-final { scan-assembler-times "avx_vzeroupper" 1 } } */
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 1 { target ia32 } } } */
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 2 { target { ! ia32 } } } } */
index 974e1626a6d83c371a1340f8b1d5f63564908c63..fa0a6dfcaacb608423fd63a6e5a097834c658273 100644 (file)
@@ -15,4 +15,5 @@ foo ()
   _mm256_zeroupper ();
 }
 
-/* { dg-final { scan-assembler-times "avx_vzeroupper" 4 } } */
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 4 { target ia32 } } } */
+/* { dg-final { scan-assembler-times "avx_vzeroupper" 5 { target { ! ia32 } } } } */