public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
* An error of gcc-ss-20030303
@ 2003-04-23 10:36 Jie Zhang
  2003-04-23 22:27 ` tm_gccmail
  2003-04-23 22:43 ` Richard Henderson
  0 siblings, 2 replies; 4+ messages in thread
From: Jie Zhang @ 2003-04-23 10:36 UTC (permalink / raw)
  To: gcc

The following code is copied from spinlock.c in uClibc (after 
preprocessing and 'inline' after static removed):

static void __pthread_release(int * spinlock)
{
  asm ("" : : : "memory");
  *spinlock = 0;
  __asm __volatile ("" : "=m" (*spinlock) : "0" (*spinlock));
}

When compiled using gcc-3.0.2, no error is reported. But gcc-3.2 and 
gcc-ss-20030303 report the follwing error:

t.c: In function `__pthread_release':
t.c:5: inconsistent operand constraints in an `asm'

I dump the rtl using -dr and find that gcc-3.0.2 uses a pseudo register 
to hold the address of 'spinlock' and the output operand and input 
operand of __asm__ use the same register. But gcc-ss-20030303 uses two 
pseudo registers for the output operand and input operand of __asm__ 
respectively, both pseudo registers hold the same value -- the address 
of 'spinlock'. Then when checking the constraints, gcc reports the error.

Is it a bug of gcc-3.2.x and gcc-3.3? Or the above __asm__ usage is 
wrong in gcc-3.2.x and gcc-3.3? I think gcc-3.0.2 is right?

- Jie


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: An error of gcc-ss-20030303
  2003-04-23 10:36 An error of gcc-ss-20030303 Jie Zhang
@ 2003-04-23 22:27 ` tm_gccmail
  2003-04-23 22:43 ` Richard Henderson
  1 sibling, 0 replies; 4+ messages in thread
From: tm_gccmail @ 2003-04-23 22:27 UTC (permalink / raw)
  To: Jie Zhang; +Cc: gcc

On Wed, 23 Apr 2003, Jie Zhang wrote:

> The following code is copied from spinlock.c in uClibc (after 
> preprocessing and 'inline' after static removed):
> 
> static void __pthread_release(int * spinlock)
> {
>   asm ("" : : : "memory");
>   *spinlock = 0;
>   __asm __volatile ("" : "=m" (*spinlock) : "0" (*spinlock));
> }
> 

Why not use a pointer to a volatile int?

Toshi


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: An error of gcc-ss-20030303
  2003-04-23 10:36 An error of gcc-ss-20030303 Jie Zhang
  2003-04-23 22:27 ` tm_gccmail
@ 2003-04-23 22:43 ` Richard Henderson
  2003-04-24  5:42   ` Jie Zhang
  1 sibling, 1 reply; 4+ messages in thread
From: Richard Henderson @ 2003-04-23 22:43 UTC (permalink / raw)
  To: Jie Zhang; +Cc: gcc

On Wed, Apr 23, 2003 at 03:41:45PM +0800, Jie Zhang wrote:
>  __asm __volatile ("" : "=m" (*spinlock) : "0" (*spinlock));

Don't use matching constraints on memories.


r~

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: An error of gcc-ss-20030303
  2003-04-23 22:43 ` Richard Henderson
@ 2003-04-24  5:42   ` Jie Zhang
  0 siblings, 0 replies; 4+ messages in thread
From: Jie Zhang @ 2003-04-24  5:42 UTC (permalink / raw)
  To: gcc

Richard Henderson wrote:

> On Wed, Apr 23, 2003 at 03:41:45PM +0800, Jie Zhang wrote:
>  
>
>> __asm __volatile ("" : "=m" (*spinlock) : "0" (*spinlock));
>>   
>
>
> Don't use matching constraints on memories.
>  
>
The problem is solved by changing this line into

__asm __volatile ("" : "+m" (*spinlock));

I should have read the document more carefully :-P
Thanks all!

- Jie

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2003-04-24  2:54 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-04-23 10:36 An error of gcc-ss-20030303 Jie Zhang
2003-04-23 22:27 ` tm_gccmail
2003-04-23 22:43 ` Richard Henderson
2003-04-24  5:42   ` Jie Zhang

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).