[patch i386]: Avoid for leaf functions stack_alignment_needed

[patch i386]: Avoid for leaf functions stack_alignment_needed

[Kai Tietz]

[-- Attachment #1: Type: text/plain, Size: 502 bytes --]

Hi,

this patch addresses the emit of stack-alignment for leaf functions
without register-saves, additional local stack, and no
tls-descriptor-call.

ChangeLog

2010-06-11  Kai Tietz

        * config/i386/i386.c (ix86_compute_frame_layout): Avoid stack-alignment
        for simple leaf-functions.

Tested for x86_64-pc-mingw32, and i686-pc-cygwin.
Ok for apply?

Regards,
Kai

-- 
|  (\_/) This is Bunny. Copy and paste
| (='.'=) Bunny into your signature to help
| (")_(") him gain world domination

[-- Attachment #2: opt_x64_frame.diff --]
[-- Type: application/octet-stream, Size: 626 bytes --]

Index: gcc/gcc/config/i386/i386.c
===================================================================
--- gcc.orig/gcc/config/i386/i386.c	2010-06-11 18:15:35.948017200 +0200
+++ gcc/gcc/config/i386/i386.c	2010-06-11 19:18:49.973023000 +0200
@@ -8190,6 +8190,11 @@
   frame->padding1 = ((offset + stack_alignment_needed - 1)
 		     & -stack_alignment_needed) - offset;
 
+  if (!size && offset == UNITS_PER_WORD &&
+      current_function_is_leaf && cfun->calls_alloca == 0
+      && !ix86_current_function_calls_tls_descriptor)
+    frame->padding1 = 0;
+
   offset += frame->padding1;
 
   /* Frame pointer points here.  */

Re: [patch i386]: Avoid for leaf functions stack_alignment_needed

[H.J. Lu]

On Fri, Jun 11, 2010 at 10:47 AM, Kai Tietz <ktietz70@googlemail.com> wrote:
> Hi,
>
> this patch addresses the emit of stack-alignment for leaf functions
> without register-saves, additional local stack, and no
> tls-descriptor-call.
>
> ChangeLog
>
> 2010-06-11  Kai Tietz
>
>        * config/i386/i386.c (ix86_compute_frame_layout): Avoid stack-alignment
>        for simple leaf-functions.
>
> Tested for x86_64-pc-mingw32, and i686-pc-cygwin.
> Ok for apply?
>

Is there a testcase?

-- 
H.J.

Re: [patch i386]: Avoid for leaf functions stack_alignment_needed

[Kai Tietz]

[-- Attachment #1: Type: text/plain, Size: 1023 bytes --]

2010/6/11 H.J. Lu <hjl.tools@gmail.com>:
> On Fri, Jun 11, 2010 at 10:47 AM, Kai Tietz <ktietz70@googlemail.com> wrote:
>> Hi,
>>
>> this patch addresses the emit of stack-alignment for leaf functions
>> without register-saves, additional local stack, and no
>> tls-descriptor-call.
>>
>> ChangeLog
>>
>> 2010-06-11  Kai Tietz
>>
>>        * config/i386/i386.c (ix86_compute_frame_layout): Avoid stack-alignment
>>        for simple leaf-functions.
>>
>> Tested for x86_64-pc-mingw32, and i686-pc-cygwin.
>> Ok for apply?
>>
>
> Is there a testcase?
>
> --
> H.J.
>

I added to leaf function tests for this patch

Additional ChangeLog for testsuite

        * gcc.target/x86_64/abi/callabi/leaf-1.c: New test.
        * gcc.target/x86_64/abi/callabi/leaf-2.c: New test.

Tested for x86_64-pc-linux-gnu, and for x86_64-pc-mingw32. Ok for apply?

Regards,
Kai

-- 
|  (\_/) This is Bunny. Copy and paste
| (='.'=) Bunny into your signature to help
| (")_(") him gain world domination

[-- Attachment #2: opt_x64_frame.diff --]
[-- Type: application/octet-stream, Size: 1833 bytes --]

Index: gcc/gcc/config/i386/i386.c
===================================================================
--- gcc.orig/gcc/config/i386/i386.c	2010-06-12 11:01:30.809062300 +0200
+++ gcc/gcc/config/i386/i386.c	2010-06-12 12:06:17.727381300 +0200
@@ -8190,6 +8190,11 @@
   frame->padding1 = ((offset + stack_alignment_needed - 1)
 		     & -stack_alignment_needed) - offset;
 
+  if (!size && offset == ((frame->nregs + 1) * UNITS_PER_WORD)
+      && current_function_is_leaf && cfun->calls_alloca == 0
+      && !ix86_current_function_calls_tls_descriptor)
+    frame->padding1 = 0;
+
   offset += frame->padding1;
 
   /* Frame pointer points here.  */
Index: gcc/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-1.c
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ gcc/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-1.c	2010-06-12 11:42:45.859627000 +0200
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mabi=sysv" } */
+
+__attribute__ ((ms_abi))
+int foo (void)
+{
+  return 0;
+}
+
+/* { dg-final { scan-assembler-not "%rsp" } } */
+
Index: gcc/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-2.c
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ gcc/gcc/testsuite/gcc.target/x86_64/abi/callabi/leaf-2.c	2010-06-12 12:01:27.105758700 +0200
@@ -0,0 +1,25 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mabi=sysv" } */
+
+extern int glb1, gbl2, gbl3;
+
+__attribute__ ((ms_abi))
+int foo (void)
+{
+  int r = 1;
+  int i, j, k;
+  for (i = 0; i < glb1; i++)
+  {
+     r *= (i + 1);
+     for (j = gbl2; j > 0; --j)
+       {
+	 for (k = 0; k < gbl3; k++)
+	   r += (i + k * j);
+       }
+  }
+
+  return r;
+}
+
+/* { dg-final { scan-assembler-not "%rsp" } } */
+

Re: [patch i386]: Avoid for leaf functions stack_alignment_needed

[Kai Tietz]

PING

2010/6/12 Kai Tietz <ktietz70@googlemail.com>:
> 2010/6/11 H.J. Lu <hjl.tools@gmail.com>:
>> On Fri, Jun 11, 2010 at 10:47 AM, Kai Tietz <ktietz70@googlemail.com> wrote:
>>> Hi,
>>>
>>> this patch addresses the emit of stack-alignment for leaf functions
>>> without register-saves, additional local stack, and no
>>> tls-descriptor-call.
>>>
>>> ChangeLog
>>>
>>> 2010-06-11  Kai Tietz
>>>
>>>        * config/i386/i386.c (ix86_compute_frame_layout): Avoid stack-alignment
>>>        for simple leaf-functions.
>>>
>>> Tested for x86_64-pc-mingw32, and i686-pc-cygwin.
>>> Ok for apply?
>>>
>>
>> Is there a testcase?
>>
>> --
>> H.J.
>>
>
> I added to leaf function tests for this patch
>
> Additional ChangeLog for testsuite
>
>        * gcc.target/x86_64/abi/callabi/leaf-1.c: New test.
>        * gcc.target/x86_64/abi/callabi/leaf-2.c: New test.
>
> Tested for x86_64-pc-linux-gnu, and for x86_64-pc-mingw32. Ok for apply?
>
> Regards,
> Kai
>
> --
> |  (\_/) This is Bunny. Copy and paste
> | (='.'=) Bunny into your signature to help
> | (")_(") him gain world domination
>



-- 
|  (\_/) This is Bunny. Copy and paste
| (='.'=) Bunny into your signature to help
| (")_(") him gain world domination

Re: [patch i386]: Avoid for leaf functions stack_alignment_needed

[Kai Tietz]

2010/6/16 Kai Tietz <ktietz70@googlemail.com>:
> PING
>
> 2010/6/12 Kai Tietz <ktietz70@googlemail.com>:
>> 2010/6/11 H.J. Lu <hjl.tools@gmail.com>:
>>> On Fri, Jun 11, 2010 at 10:47 AM, Kai Tietz <ktietz70@googlemail.com> wrote:
>>>> Hi,
>>>>
>>>> this patch addresses the emit of stack-alignment for leaf functions
>>>> without register-saves, additional local stack, and no
>>>> tls-descriptor-call.
>>>>
>>>> ChangeLog
>>>>
>>>> 2010-06-11  Kai Tietz
>>>>
>>>>        * config/i386/i386.c (ix86_compute_frame_layout): Avoid stack-alignment
>>>>        for simple leaf-functions.
>>>>
>>>> Tested for x86_64-pc-mingw32, and i686-pc-cygwin.
>>>> Ok for apply?
>>>>
>>>
>>> Is there a testcase?
>>>
>>> --
>>> H.J.
>>>
>>
>> I added to leaf function tests for this patch
>>
>> Additional ChangeLog for testsuite
>>
>>        * gcc.target/x86_64/abi/callabi/leaf-1.c: New test.
>>        * gcc.target/x86_64/abi/callabi/leaf-2.c: New test.
>>
>> Tested for x86_64-pc-linux-gnu, and for x86_64-pc-mingw32. Ok for apply?
>>
>> Regards,
>> Kai
>>
>> --
>> |  (\_/) This is Bunny. Copy and paste
>> | (='.'=) Bunny into your signature to help
>> | (")_(") him gain world domination
>>
>
>
>
> --
> |  (\_/) This is Bunny. Copy and paste
> | (='.'=) Bunny into your signature to help
> | (")_(") him gain world domination
>

Committed adjusted patch at revision 161064 (and testsuite part at
161063). Patch approved by Jan on IRC.

Thanks,
Kai

-- 
|  (\_/) This is Bunny. Copy and paste
| (='.'=) Bunny into your signature to help
| (")_(") him gain world domination