[PATCH] Fix pr67963

[PATCH] Fix pr67963

[Yulia Koval]

[-- Attachment #1: Type: text/plain, Size: 733 bytes --]

Hi,

This patch fixes the issue:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963

  gcc/config/i386/i386.c (ix86_option_override_internal) Disable
    80387 mask if lakemont target is set.

diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 4c25c9e..db722aa 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
  break;
       }

+  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
+    {
+      opts->x_target_flags &= ~MASK_80387;
+      opts_set->x_target_flags |= MASK_80387;
+    }
+
   if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
     error ("Intel MPX does not support x32");

Ok for trunk?

Yulia

[-- Attachment #2: patch --]
[-- Type: application/octet-stream, Size: 521 bytes --]

diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 4c25c9e..db722aa 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
 	break;
       }
 
+  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
+    {
+      opts->x_target_flags &= ~MASK_80387;
+      opts_set->x_target_flags |= MASK_80387;
+    }
+
   if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
     error ("Intel MPX does not support x32");
 

Re: [PATCH] Fix pr67963

[H.J. Lu]

On Wed, Oct 14, 2015 at 8:08 AM, Yulia Koval <vaalfreja@gmail.com> wrote:
> Hi,
>
> This patch fixes the issue:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>
>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>     80387 mask if lakemont target is set.
>
> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
> index 4c25c9e..db722aa 100644
> --- a/gcc/config/i386/i386.c
> +++ b/gcc/config/i386/i386.c
> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>   break;
>        }
>
> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
> +    {
> +      opts->x_target_flags &= ~MASK_80387;
> +      opts_set->x_target_flags |= MASK_80387;
> +    }
> +
>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>      error ("Intel MPX does not support x32");
>
> Ok for trunk?

We should add a bit to "struct pta" to indicate availability of
80387 ISA and turn it off for lakemount if 90387 ISA hasn't be
turned on explicitly.

We also need some testcases.

-- 
H.J.

Re: [PATCH] Fix pr67963

[H.J. Lu]

On Wed, Oct 14, 2015 at 8:15 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Wed, Oct 14, 2015 at 8:08 AM, Yulia Koval <vaalfreja@gmail.com> wrote:
>> Hi,
>>
>> This patch fixes the issue:
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>
>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>     80387 mask if lakemont target is set.
>>
>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>> index 4c25c9e..db722aa 100644
>> --- a/gcc/config/i386/i386.c
>> +++ b/gcc/config/i386/i386.c
>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>   break;
>>        }
>>
>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>> +    {
>> +      opts->x_target_flags &= ~MASK_80387;
>> +      opts_set->x_target_flags |= MASK_80387;
>> +    }
>> +
>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>      error ("Intel MPX does not support x32");
>>
>> Ok for trunk?
>
> We should add a bit to "struct pta" to indicate availability of
> 80387 ISA and turn it off for lakemount if 90387 ISA hasn't be
> turned on explicitly.

Something like

diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index a2314e7..1cea58e 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -4348,6 +4348,7 @@ ix86_option_override_internal (bool main_args_p,
       const enum processor_type processor;
       const enum attr_cpu schedule;
       const unsigned HOST_WIDE_INT flags;
+      const unsigned HOST_WIDE_INT mask;
     }
   const processor_alias_table[] =
     {

> We also need some testcases.
>
> --
> H.J.

-- 
H.J.

Re: [PATCH] Fix pr67963

[Uros Bizjak]

On Wed, Oct 14, 2015 at 5:17 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Wed, Oct 14, 2015 at 8:15 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Wed, Oct 14, 2015 at 8:08 AM, Yulia Koval <vaalfreja@gmail.com> wrote:
>>> Hi,
>>>
>>> This patch fixes the issue:
>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>>
>>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>>     80387 mask if lakemont target is set.
>>>
>>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>>> index 4c25c9e..db722aa 100644
>>> --- a/gcc/config/i386/i386.c
>>> +++ b/gcc/config/i386/i386.c
>>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>>   break;
>>>        }
>>>
>>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>>> +    {
>>> +      opts->x_target_flags &= ~MASK_80387;
>>> +      opts_set->x_target_flags |= MASK_80387;
>>> +    }
>>> +
>>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>>      error ("Intel MPX does not support x32");
>>>
>>> Ok for trunk?
>>
>> We should add a bit to "struct pta" to indicate availability of
>> 80387 ISA and turn it off for lakemount if 90387 ISA hasn't be
>> turned on explicitly.
>
> Something like

Hm, I don't think introducing another flag is a good idea.

Better move -m80387 (together with -msof-float and -mhard-float) to
ISA flags and add corresponding PTA_80387 define.

Uros.

Re: [PATCH] Fix pr67963

[Uros Bizjak]

On Wed, Oct 14, 2015 at 5:08 PM, Yulia Koval <vaalfreja@gmail.com> wrote:
> Hi,
>
> This patch fixes the issue:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>
>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>     80387 mask if lakemont target is set.
>
> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
> index 4c25c9e..db722aa 100644
> --- a/gcc/config/i386/i386.c
> +++ b/gcc/config/i386/i386.c
> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>   break;
>        }
>
> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
> +    {
> +      opts->x_target_flags &= ~MASK_80387;
> +      opts_set->x_target_flags |= MASK_80387;
> +    }
> +
>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>      error ("Intel MPX does not support x32");
>
> Ok for trunk?

The problem is in TARGET_SUBTARGET{32,64}_DEFAULT, that will override
set target flags, unless relevant bit of opts_set->x_target_flags is
set. However, we can't just set x_target_flag, because
__attribute__((target("arch=...."))) won't work. Unfortunately, my
proposed patch in the PR violates this last requirement.

This can probably be solved by adding local "x_target_flags_mask" and
use it after TARGET_SUBTARGET processing.

Uros.

Re: [PATCH] Fix pr67963

[Uros Bizjak]

[-- Attachment #1: Type: text/plain, Size: 1976 bytes --]

On Thu, Oct 15, 2015 at 1:45 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Wed, Oct 14, 2015 at 5:08 PM, Yulia Koval <vaalfreja@gmail.com> wrote:
>> Hi,
>>
>> This patch fixes the issue:
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>
>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>     80387 mask if lakemont target is set.
>>
>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>> index 4c25c9e..db722aa 100644
>> --- a/gcc/config/i386/i386.c
>> +++ b/gcc/config/i386/i386.c
>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>   break;
>>        }
>>
>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>> +    {
>> +      opts->x_target_flags &= ~MASK_80387;
>> +      opts_set->x_target_flags |= MASK_80387;
>> +    }
>> +
>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>      error ("Intel MPX does not support x32");
>>
>> Ok for trunk?
>
> The problem is in TARGET_SUBTARGET{32,64}_DEFAULT, that will override
> set target flags, unless relevant bit of opts_set->x_target_flags is
> set. However, we can't just set x_target_flag, because
> __attribute__((target("arch=...."))) won't work. Unfortunately, my
> proposed patch in the PR violates this last requirement.
>
> This can probably be solved by adding local "x_target_flags_mask" and
> use it after TARGET_SUBTARGET processing.

Attached is a final patch I plan to commit to mainline soon.

2015-10-15  Uros Bizjak  <ubizjak@gmail.com>

    PR target/67963
    * config/i386/i386.c (ix86_option_override_internal): Add PTA_NO_80387.
    Add PTA_NO_80387 to "lakemont".  Disable MASK_80387 target flag
    after target flags were initialized to target defaults.

testsuite/ChangeLog:

2015-10-15  Uros Bizjak  <ubizjak@gmail.com>

    PR target/67963
    * gcc.target/i386/pr67963-1.c: New test.
    * gcc.target/i386/pr67963-2.c: Ditto.

Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.

Uros.

[-- Attachment #2: p.diff.txt --]
[-- Type: text/plain, Size: 3040 bytes --]

Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c	(revision 228843)
+++ config/i386/i386.c	(working copy)
@@ -4247,6 +4247,7 @@ ix86_option_override_internal (bool main_args_p,
 {
   int i;
   unsigned int ix86_arch_mask;
+  int ix86_target_flags_mask = 0;
   const bool ix86_tune_specified = (opts->x_ix86_tune_string != NULL);
   const char *prefix;
   const char *suffix;
@@ -4311,6 +4312,7 @@ ix86_option_override_internal (bool main_args_p,
 #define PTA_PCOMMIT		(HOST_WIDE_INT_1 << 56)
 #define PTA_MWAITX		(HOST_WIDE_INT_1 << 57)
 #define PTA_CLZERO		(HOST_WIDE_INT_1 << 58)
+#define PTA_NO_80387		(HOST_WIDE_INT_1 << 59)
 
 #define PTA_CORE2 \
   (PTA_64BIT | PTA_MMX | PTA_SSE | PTA_SSE2 | PTA_SSE3 | PTA_SSSE3 \
@@ -4355,7 +4357,7 @@ ix86_option_override_internal (bool main_args_p,
       {"i486", PROCESSOR_I486, CPU_NONE, 0},
       {"i586", PROCESSOR_PENTIUM, CPU_PENTIUM, 0},
       {"pentium", PROCESSOR_PENTIUM, CPU_PENTIUM, 0},
-      {"lakemont", PROCESSOR_LAKEMONT, CPU_PENTIUM, 0},
+      {"lakemont", PROCESSOR_LAKEMONT, CPU_PENTIUM, PTA_NO_80387},
       {"pentium-mmx", PROCESSOR_PENTIUM, CPU_PENTIUM, PTA_MMX},
       {"winchip-c6", PROCESSOR_I486, CPU_NONE, PTA_MMX},
       {"winchip2", PROCESSOR_I486, CPU_NONE, PTA_MMX | PTA_3DNOW | PTA_PRFCHW},
@@ -4935,7 +4937,9 @@ ix86_option_override_internal (bool main_args_p,
 	if (processor_alias_table[i].flags & PTA_MWAITX
 	    && !(opts->x_ix86_isa_flags_explicit & OPTION_MASK_ISA_MWAITX))
 	  opts->x_ix86_isa_flags |= OPTION_MASK_ISA_MWAITX;
-
+	if (processor_alias_table[i].flags & PTA_NO_80387
+	    && !(opts_set->x_target_flags & MASK_80387))
+	  ix86_target_flags_mask |= MASK_80387;
 	break;
       }
 
@@ -5126,6 +5130,9 @@ ix86_option_override_internal (bool main_args_p,
         opts->x_target_flags |= MASK_NO_RED_ZONE;
     }
 
+  /* Disable specific processor dependant features after target defaults.  */
+  opts->x_target_flags &= ~ix86_target_flags_mask;
+
   /* Keep nonleaf frame pointers.  */
   if (opts->x_flag_omit_frame_pointer)
     opts->x_target_flags &= ~MASK_OMIT_LEAF_FRAME_POINTER;
Index: testsuite/gcc.target/i386/pr67963-1.c
===================================================================
--- testsuite/gcc.target/i386/pr67963-1.c	(revision 0)
+++ testsuite/gcc.target/i386/pr67963-1.c	(working copy)
@@ -0,0 +1,9 @@
+/* { dg-do compile { target ia32 } } */
+/* { dg-options "-O2 -march=lakemont" } */
+
+float foo (void)
+{
+  return 0.0f;
+}
+
+/* { dg-final { scan-assembler-not "fldz" } } */
Index: testsuite/gcc.target/i386/pr67963-2.c
===================================================================
--- testsuite/gcc.target/i386/pr67963-2.c	(revision 0)
+++ testsuite/gcc.target/i386/pr67963-2.c	(working copy)
@@ -0,0 +1,11 @@
+/* { dg-do compile { target ia32 } } */
+/* { dg-options "-O2 -march=pentium" } */
+
+float
+__attribute__((target("arch=lakemont")))
+foo (void)
+{
+  return 0.0f;
+}
+
+/* { dg-final { scan-assembler-not "fldz" } } */

Re: [PATCH] Fix pr67963

[H.J. Lu]

On Thu, Oct 15, 2015 at 9:57 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Thu, Oct 15, 2015 at 1:45 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>> On Wed, Oct 14, 2015 at 5:08 PM, Yulia Koval <vaalfreja@gmail.com> wrote:
>>> Hi,
>>>
>>> This patch fixes the issue:
>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>>
>>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>>     80387 mask if lakemont target is set.
>>>
>>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>>> index 4c25c9e..db722aa 100644
>>> --- a/gcc/config/i386/i386.c
>>> +++ b/gcc/config/i386/i386.c
>>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>>   break;
>>>        }
>>>
>>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>>> +    {
>>> +      opts->x_target_flags &= ~MASK_80387;
>>> +      opts_set->x_target_flags |= MASK_80387;
>>> +    }
>>> +
>>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>>      error ("Intel MPX does not support x32");
>>>
>>> Ok for trunk?
>>
>> The problem is in TARGET_SUBTARGET{32,64}_DEFAULT, that will override
>> set target flags, unless relevant bit of opts_set->x_target_flags is
>> set. However, we can't just set x_target_flag, because
>> __attribute__((target("arch=...."))) won't work. Unfortunately, my
>> proposed patch in the PR violates this last requirement.
>>
>> This can probably be solved by adding local "x_target_flags_mask" and
>> use it after TARGET_SUBTARGET processing.
>
> Attached is a final patch I plan to commit to mainline soon.
>
> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>
>     PR target/67963
>     * config/i386/i386.c (ix86_option_override_internal): Add PTA_NO_80387.
>     Add PTA_NO_80387 to "lakemont".  Disable MASK_80387 target flag
>     after target flags were initialized to target defaults.
>
> testsuite/ChangeLog:
>
> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>
>     PR target/67963
>     * gcc.target/i386/pr67963-1.c: New test.
>     * gcc.target/i386/pr67963-2.c: Ditto.
>
> Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
>
> Uros.

Do we support -O2 -march=lakemont with

__attribute__((target("arch=silvermont")))


-- 
H.J.

Re: [PATCH] Fix pr67963

[Uros Bizjak]

On Thu, Oct 15, 2015 at 6:59 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Thu, Oct 15, 2015 at 9:57 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
>> On Thu, Oct 15, 2015 at 1:45 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>> On Wed, Oct 14, 2015 at 5:08 PM, Yulia Koval <vaalfreja@gmail.com> wrote:
>>>> Hi,
>>>>
>>>> This patch fixes the issue:
>>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>>>
>>>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>>>     80387 mask if lakemont target is set.
>>>>
>>>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>>>> index 4c25c9e..db722aa 100644
>>>> --- a/gcc/config/i386/i386.c
>>>> +++ b/gcc/config/i386/i386.c
>>>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>>>   break;
>>>>        }
>>>>
>>>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>>>> +    {
>>>> +      opts->x_target_flags &= ~MASK_80387;
>>>> +      opts_set->x_target_flags |= MASK_80387;
>>>> +    }
>>>> +
>>>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>>>      error ("Intel MPX does not support x32");
>>>>
>>>> Ok for trunk?
>>>
>>> The problem is in TARGET_SUBTARGET{32,64}_DEFAULT, that will override
>>> set target flags, unless relevant bit of opts_set->x_target_flags is
>>> set. However, we can't just set x_target_flag, because
>>> __attribute__((target("arch=...."))) won't work. Unfortunately, my
>>> proposed patch in the PR violates this last requirement.
>>>
>>> This can probably be solved by adding local "x_target_flags_mask" and
>>> use it after TARGET_SUBTARGET processing.
>>
>> Attached is a final patch I plan to commit to mainline soon.
>>
>> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>>
>>     PR target/67963
>>     * config/i386/i386.c (ix86_option_override_internal): Add PTA_NO_80387.
>>     Add PTA_NO_80387 to "lakemont".  Disable MASK_80387 target flag
>>     after target flags were initialized to target defaults.
>>
>> testsuite/ChangeLog:
>>
>> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>>
>>     PR target/67963
>>     * gcc.target/i386/pr67963-1.c: New test.
>>     * gcc.target/i386/pr67963-2.c: Ditto.
>>
>> Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
>>
>> Uros.
>
> Do we support -O2 -march=lakemont with
>
> __attribute__((target("arch=silvermont")))

Hm, no.

Uros.

Re: [PATCH] Fix pr67963

[H.J. Lu]

On Thu, Oct 15, 2015 at 12:16 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Thu, Oct 15, 2015 at 6:59 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Thu, Oct 15, 2015 at 9:57 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>> On Thu, Oct 15, 2015 at 1:45 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>>> On Wed, Oct 14, 2015 at 5:08 PM, Yulia Koval <vaalfreja@gmail.com> wrote:
>>>>> Hi,
>>>>>
>>>>> This patch fixes the issue:
>>>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>>>>
>>>>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>>>>     80387 mask if lakemont target is set.
>>>>>
>>>>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>>>>> index 4c25c9e..db722aa 100644
>>>>> --- a/gcc/config/i386/i386.c
>>>>> +++ b/gcc/config/i386/i386.c
>>>>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>>>>   break;
>>>>>        }
>>>>>
>>>>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>>>>> +    {
>>>>> +      opts->x_target_flags &= ~MASK_80387;
>>>>> +      opts_set->x_target_flags |= MASK_80387;
>>>>> +    }
>>>>> +
>>>>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>>>>      error ("Intel MPX does not support x32");
>>>>>
>>>>> Ok for trunk?
>>>>
>>>> The problem is in TARGET_SUBTARGET{32,64}_DEFAULT, that will override
>>>> set target flags, unless relevant bit of opts_set->x_target_flags is
>>>> set. However, we can't just set x_target_flag, because
>>>> __attribute__((target("arch=...."))) won't work. Unfortunately, my
>>>> proposed patch in the PR violates this last requirement.
>>>>
>>>> This can probably be solved by adding local "x_target_flags_mask" and
>>>> use it after TARGET_SUBTARGET processing.
>>>
>>> Attached is a final patch I plan to commit to mainline soon.
>>>
>>> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>>>
>>>     PR target/67963
>>>     * config/i386/i386.c (ix86_option_override_internal): Add PTA_NO_80387.
>>>     Add PTA_NO_80387 to "lakemont".  Disable MASK_80387 target flag
>>>     after target flags were initialized to target defaults.
>>>
>>> testsuite/ChangeLog:
>>>
>>> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>>>
>>>     PR target/67963
>>>     * gcc.target/i386/pr67963-1.c: New test.
>>>     * gcc.target/i386/pr67963-2.c: Ditto.
>>>
>>> Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
>>>
>>> Uros.
>>
>> Do we support -O2 -march=lakemont with
>>
>> __attribute__((target("arch=silvermont")))
>
> Hm, no.
>

Do we issue an error or silently ignore
__attribute__((target("arch=silvermont")))?
If we don't support it, should we support

-O2 -march=silvermont

__attribute__((target("arch=lakemont")))


-- 
H.J.

Re: [PATCH] Fix pr67963

[Uros Bizjak]

[-- Attachment #1: Type: text/plain, Size: 3059 bytes --]

On Thu, Oct 15, 2015 at 9:21 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Thu, Oct 15, 2015 at 12:16 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>> On Thu, Oct 15, 2015 at 6:59 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> On Thu, Oct 15, 2015 at 9:57 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>>> On Thu, Oct 15, 2015 at 1:45 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>>>> On Wed, Oct 14, 2015 at 5:08 PM, Yulia Koval <vaalfreja@gmail.com> wrote:
>>>>>> Hi,
>>>>>>
>>>>>> This patch fixes the issue:
>>>>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67963
>>>>>>
>>>>>>   gcc/config/i386/i386.c (ix86_option_override_internal) Disable
>>>>>>     80387 mask if lakemont target is set.
>>>>>>
>>>>>> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
>>>>>> index 4c25c9e..db722aa 100644
>>>>>> --- a/gcc/config/i386/i386.c
>>>>>> +++ b/gcc/config/i386/i386.c
>>>>>> @@ -4943,6 +4943,12 @@ ix86_option_override_internal (bool main_args_p,
>>>>>>   break;
>>>>>>        }
>>>>>>
>>>>>> +  if (!strcmp (opts->x_ix86_arch_string, "lakemont"))
>>>>>> +    {
>>>>>> +      opts->x_target_flags &= ~MASK_80387;
>>>>>> +      opts_set->x_target_flags |= MASK_80387;
>>>>>> +    }
>>>>>> +
>>>>>>    if (TARGET_X32 && (opts->x_ix86_isa_flags & OPTION_MASK_ISA_MPX))
>>>>>>      error ("Intel MPX does not support x32");
>>>>>>
>>>>>> Ok for trunk?
>>>>>
>>>>> The problem is in TARGET_SUBTARGET{32,64}_DEFAULT, that will override
>>>>> set target flags, unless relevant bit of opts_set->x_target_flags is
>>>>> set. However, we can't just set x_target_flag, because
>>>>> __attribute__((target("arch=...."))) won't work. Unfortunately, my
>>>>> proposed patch in the PR violates this last requirement.
>>>>>
>>>>> This can probably be solved by adding local "x_target_flags_mask" and
>>>>> use it after TARGET_SUBTARGET processing.
>>>>
>>>> Attached is a final patch I plan to commit to mainline soon.
>>>>
>>>> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>>>>
>>>>     PR target/67963
>>>>     * config/i386/i386.c (ix86_option_override_internal): Add PTA_NO_80387.
>>>>     Add PTA_NO_80387 to "lakemont".  Disable MASK_80387 target flag
>>>>     after target flags were initialized to target defaults.
>>>>
>>>> testsuite/ChangeLog:
>>>>
>>>> 2015-10-15  Uros Bizjak  <ubizjak@gmail.com>
>>>>
>>>>     PR target/67963
>>>>     * gcc.target/i386/pr67963-1.c: New test.
>>>>     * gcc.target/i386/pr67963-2.c: Ditto.
>>>>
>>>> Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
>>>>
>>>> Uros.
>>>
>>> Do we support -O2 -march=lakemont with
>>>
>>> __attribute__((target("arch=silvermont")))
>>
>> Hm, no.
>>
>
> Do we issue an error or silently ignore
> __attribute__((target("arch=silvermont")))?
> If we don't support it, should we support
>
> -O2 -march=silvermont
>
> __attribute__((target("arch=lakemont")))

Actually, we have to re-initialize:

  opts->x_target_flags
    |= (TARGET_DEFAULT | TARGET_SUBTARGET_DEFAULT) & ~opts_set->x_target_flags;

just before TARGET_SUBTARGET{32,64}_DEFAULT processing, and it will work.

Uros.

[-- Attachment #2: p.diff.txt --]
[-- Type: text/plain, Size: 3402 bytes --]

Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c	(revision 228843)
+++ config/i386/i386.c	(working copy)
@@ -4247,6 +4247,7 @@ ix86_option_override_internal (bool main_args_p,
 {
   int i;
   unsigned int ix86_arch_mask;
+  int ix86_target_flags_mask = 0;
   const bool ix86_tune_specified = (opts->x_ix86_tune_string != NULL);
   const char *prefix;
   const char *suffix;
@@ -4311,6 +4312,7 @@ ix86_option_override_internal (bool main_args_p,
 #define PTA_PCOMMIT		(HOST_WIDE_INT_1 << 56)
 #define PTA_MWAITX		(HOST_WIDE_INT_1 << 57)
 #define PTA_CLZERO		(HOST_WIDE_INT_1 << 58)
+#define PTA_NO_80387		(HOST_WIDE_INT_1 << 59)
 
 #define PTA_CORE2 \
   (PTA_64BIT | PTA_MMX | PTA_SSE | PTA_SSE2 | PTA_SSE3 | PTA_SSSE3 \
@@ -4355,7 +4357,7 @@ ix86_option_override_internal (bool main_args_p,
       {"i486", PROCESSOR_I486, CPU_NONE, 0},
       {"i586", PROCESSOR_PENTIUM, CPU_PENTIUM, 0},
       {"pentium", PROCESSOR_PENTIUM, CPU_PENTIUM, 0},
-      {"lakemont", PROCESSOR_LAKEMONT, CPU_PENTIUM, 0},
+      {"lakemont", PROCESSOR_LAKEMONT, CPU_PENTIUM, PTA_NO_80387},
       {"pentium-mmx", PROCESSOR_PENTIUM, CPU_PENTIUM, PTA_MMX},
       {"winchip-c6", PROCESSOR_I486, CPU_NONE, PTA_MMX},
       {"winchip2", PROCESSOR_I486, CPU_NONE, PTA_MMX | PTA_3DNOW | PTA_PRFCHW},
@@ -4935,7 +4937,9 @@ ix86_option_override_internal (bool main_args_p,
 	if (processor_alias_table[i].flags & PTA_MWAITX
 	    && !(opts->x_ix86_isa_flags_explicit & OPTION_MASK_ISA_MWAITX))
 	  opts->x_ix86_isa_flags |= OPTION_MASK_ISA_MWAITX;
-
+	if (processor_alias_table[i].flags & PTA_NO_80387
+	    && !(opts_set->x_target_flags & MASK_80387))
+	  ix86_target_flags_mask |= MASK_80387;
 	break;
       }
 
@@ -5094,6 +5098,9 @@ ix86_option_override_internal (bool main_args_p,
   if (!opts_set->x_ix86_branch_cost)
     opts->x_ix86_branch_cost = ix86_tune_cost->branch_cost;
 
+  opts->x_target_flags
+    |= (TARGET_DEFAULT | TARGET_SUBTARGET_DEFAULT) & ~opts_set->x_target_flags;
+
   if (TARGET_64BIT_P (opts->x_ix86_isa_flags))
     {
       opts->x_target_flags
@@ -5126,6 +5133,9 @@ ix86_option_override_internal (bool main_args_p,
         opts->x_target_flags |= MASK_NO_RED_ZONE;
     }
 
+  /* Disable specific processor dependant features after target defaults.  */
+  opts->x_target_flags &= ~ix86_target_flags_mask;
+
   /* Keep nonleaf frame pointers.  */
   if (opts->x_flag_omit_frame_pointer)
     opts->x_target_flags &= ~MASK_OMIT_LEAF_FRAME_POINTER;
Index: testsuite/gcc.target/i386/pr67963-1.c
===================================================================
--- testsuite/gcc.target/i386/pr67963-1.c	(revision 0)
+++ testsuite/gcc.target/i386/pr67963-1.c	(working copy)
@@ -0,0 +1,9 @@
+/* { dg-do compile { target ia32 } } */
+/* { dg-options "-O2 -march=lakemont" } */
+
+float foo (void)
+{
+  return 0.0f;
+}
+
+/* { dg-final { scan-assembler-not "fldz" } } */
Index: testsuite/gcc.target/i386/pr67963-2.c
===================================================================
--- testsuite/gcc.target/i386/pr67963-2.c	(revision 0)
+++ testsuite/gcc.target/i386/pr67963-2.c	(working copy)
@@ -0,0 +1,11 @@
+/* { dg-do compile { target ia32 } } */
+/* { dg-options "-O2 -march=pentium" } */
+
+float
+__attribute__((target("arch=lakemont")))
+foo (void)
+{
+  return 0.0f;
+}
+
+/* { dg-final { scan-assembler-not "fldz" } } */

Re: [PATCH] Fix pr67963

[Uros Bizjak]

On Thu, Oct 15, 2015 at 9:30 PM, Uros Bizjak <ubizjak@gmail.com> wrote:

>>>> Do we support -O2 -march=lakemont with
>>>>
>>>> __attribute__((target("arch=silvermont")))
>>>
>>> Hm, no.
>>>
>>
>> Do we issue an error or silently ignore
>> __attribute__((target("arch=silvermont")))?
>> If we don't support it, should we support
>>
>> -O2 -march=silvermont
>>
>> __attribute__((target("arch=lakemont")))
>
> Actually, we have to re-initialize:
>
>   opts->x_target_flags
>     |= (TARGET_DEFAULT | TARGET_SUBTARGET_DEFAULT) & ~opts_set->x_target_flags;
>
> just before TARGET_SUBTARGET{32,64}_DEFAULT processing, and it will work.

No, this won't work. The value of MASK_NO_FANCY_MATH depend on
MASK_80387setting, and once fancy math bit is set, it couldn't be
cleared for march != lakemont.

It looks just like we want to error out when lakemont is enabled with -m80387.

Uros.

Re: [PATCH] Fix pr67963

[Uros Bizjak]

[-- Attachment #1: Type: text/plain, Size: 1067 bytes --]

On Fri, Oct 16, 2015 at 8:43 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Thu, Oct 15, 2015 at 9:30 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>
>>>>> Do we support -O2 -march=lakemont with
>>>>>
>>>>> __attribute__((target("arch=silvermont")))
>>>>
>>>> Hm, no.
>>>>
>>>
>>> Do we issue an error or silently ignore
>>> __attribute__((target("arch=silvermont")))?
>>> If we don't support it, should we support
>>>
>>> -O2 -march=silvermont
>>>
>>> __attribute__((target("arch=lakemont")))
>>
>> Actually, we have to re-initialize:
>>
>>   opts->x_target_flags
>>     |= (TARGET_DEFAULT | TARGET_SUBTARGET_DEFAULT) & ~opts_set->x_target_flags;
>>
>> just before TARGET_SUBTARGET{32,64}_DEFAULT processing, and it will work.
>
> No, this won't work. The value of MASK_NO_FANCY_MATH depend on
> MASK_80387setting, and once fancy math bit is set, it couldn't be
> cleared for march != lakemont.
>
> It looks just like we want to error out when lakemont is enabled with -m80387.

Like in the attached patch, that also slightly improves existing error
reporting.

Uros.

[-- Attachment #2: p.diff.txt --]
[-- Type: text/plain, Size: 1778 bytes --]

Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c	(revision 228863)
+++ config/i386/i386.c	(working copy)
@@ -4949,16 +4949,30 @@ ix86_option_override_internal (bool main_args_p,
     {
       /* Verify that x87/MMX/SSE/AVX is off for -miamcu.  */
       if (TARGET_80387_P (opts->x_target_flags))
-	sorry ("X87 FPU isn%'t supported in Intel MCU psABI");
-      else if ((opts->x_ix86_isa_flags & (OPTION_MASK_ISA_MMX
-					  | OPTION_MASK_ISA_SSE
-					  | OPTION_MASK_ISA_AVX)))
-	sorry ("%s isn%'t supported in Intel MCU psABI",
-	       TARGET_MMX_P (opts->x_ix86_isa_flags)
-	       ? "MMX"
-	       : TARGET_SSE_P (opts->x_ix86_isa_flags) ? "SSE" : "AVX");
+	sorry ("X87 FPU is not supported in Intel MCU psABI");
+      if ((opts->x_ix86_isa_flags & (OPTION_MASK_ISA_MMX
+				     | OPTION_MASK_ISA_SSE
+				     | OPTION_MASK_ISA_AVX)))
+	sorry ("%s is not supported in Intel MCU psABI",
+	       TARGET_AVX_P (opts->x_ix86_isa_flags)
+	       ? "AVX"
+	       : TARGET_SSE_P (opts->x_ix86_isa_flags) ? "SSE" : "MMX");
     }
 
+  if (ix86_arch == PROCESSOR_LAKEMONT)
+    {
+      /* Verify that x87/MMX/SSE/AVX is off for -march=lakemont.  */
+      if (TARGET_80387_P (opts->x_target_flags))
+	error ("X87 FPU is not supported on Lakemont CPU");
+      if ((opts->x_ix86_isa_flags & (OPTION_MASK_ISA_MMX
+				     | OPTION_MASK_ISA_SSE
+				     | OPTION_MASK_ISA_AVX)))
+	error ("%s is not supported on Lakemont CPU",
+	       TARGET_AVX_P (opts->x_ix86_isa_flags)
+	       ? "AVX"
+	       : TARGET_SSE_P (opts->x_ix86_isa_flags) ? "SSE" : "MMX");
+    }
+
   if (!strcmp (opts->x_ix86_arch_string, "generic"))
     error ("generic CPU can be used only for %stune=%s %s",
 	   prefix, suffix, sw);

Re: [PATCH] Fix pr67963

[H.J. Lu]

On Fri, Oct 16, 2015 at 2:35 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Fri, Oct 16, 2015 at 8:43 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
>> On Thu, Oct 15, 2015 at 9:30 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>
>>>>>> Do we support -O2 -march=lakemont with
>>>>>>
>>>>>> __attribute__((target("arch=silvermont")))
>>>>>
>>>>> Hm, no.
>>>>>
>>>>
>>>> Do we issue an error or silently ignore
>>>> __attribute__((target("arch=silvermont")))?
>>>> If we don't support it, should we support
>>>>
>>>> -O2 -march=silvermont
>>>>
>>>> __attribute__((target("arch=lakemont")))
>>>
>>> Actually, we have to re-initialize:
>>>
>>>   opts->x_target_flags
>>>     |= (TARGET_DEFAULT | TARGET_SUBTARGET_DEFAULT) & ~opts_set->x_target_flags;
>>>
>>> just before TARGET_SUBTARGET{32,64}_DEFAULT processing, and it will work.
>>
>> No, this won't work. The value of MASK_NO_FANCY_MATH depend on
>> MASK_80387setting, and once fancy math bit is set, it couldn't be
>> cleared for march != lakemont.
>>
>> It looks just like we want to error out when lakemont is enabled with -m80387.
>
> Like in the attached patch, that also slightly improves existing error
> reporting.
>

We should use a bit instead of checking PROCESSOR_LAKEMONT
so that we don't need to check another PROCESSOR_XXX for
a new IA MCU processor.

Thanks.


-- 
H.J.

Re: [PATCH] Fix pr67963

[H.J. Lu]

On Fri, Oct 16, 2015 at 3:31 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Fri, Oct 16, 2015 at 2:35 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
>> On Fri, Oct 16, 2015 at 8:43 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>> On Thu, Oct 15, 2015 at 9:30 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>>>
>>>>>>> Do we support -O2 -march=lakemont with
>>>>>>>
>>>>>>> __attribute__((target("arch=silvermont")))
>>>>>>
>>>>>> Hm, no.
>>>>>>
>>>>>
>>>>> Do we issue an error or silently ignore
>>>>> __attribute__((target("arch=silvermont")))?
>>>>> If we don't support it, should we support
>>>>>
>>>>> -O2 -march=silvermont
>>>>>
>>>>> __attribute__((target("arch=lakemont")))
>>>>
>>>> Actually, we have to re-initialize:
>>>>
>>>>   opts->x_target_flags
>>>>     |= (TARGET_DEFAULT | TARGET_SUBTARGET_DEFAULT) & ~opts_set->x_target_flags;
>>>>
>>>> just before TARGET_SUBTARGET{32,64}_DEFAULT processing, and it will work.
>>>
>>> No, this won't work. The value of MASK_NO_FANCY_MATH depend on
>>> MASK_80387setting, and once fancy math bit is set, it couldn't be
>>> cleared for march != lakemont.
>>>
>>> It looks just like we want to error out when lakemont is enabled with -m80387.
>>
>> Like in the attached patch, that also slightly improves existing error
>> reporting.
>>
>
> We should use a bit instead of checking PROCESSOR_LAKEMONT
> so that we don't need to check another PROCESSOR_XXX for
> a new IA MCU processor.
>

Another related bug:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67985

We may be able to fix both in a patch.

-- 
H.J.