[PATCH] Let RTEMS provide clock()

[PATCH] Let RTEMS provide clock()

[Sebastian Huber]

Newlib uses _times_r() in clock().  The problem is that the _times_r()
clock frequency is defined by sysconf(_SC_CLK_TCK).  The clock frequency
of clock() is the constant CLOCKS_PER_SEC.

FreeBSD uses getrusage() for clock().  Since RTEMS has only one process,
the implementation can be simplified.

Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
---
 newlib/configure.host | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/newlib/configure.host b/newlib/configure.host
index 3d967a11b..ba2d8c6c7 100644
--- a/newlib/configure.host
+++ b/newlib/configure.host
@@ -605,7 +605,7 @@ case "${host}" in
 	default_newlib_io_long_long="yes"
 	default_newlib_io_c99_formats="yes"
 	newlib_cflags="${newlib_cflags} -D_COMPILING_NEWLIB"
-	newlib_cflags="${newlib_cflags} -DMALLOC_PROVIDED -DEXIT_PROVIDED -DSIGNAL_PROVIDED -DREENTRANT_SYSCALLS_PROVIDED -DHAVE_NANOSLEEP -DHAVE_BLKSIZE -DHAVE_FCNTL -DHAVE_ASSERT_FUNC"
+	newlib_cflags="${newlib_cflags} -DCLOCK_PROVIDED -DMALLOC_PROVIDED -DEXIT_PROVIDED -DSIGNAL_PROVIDED -DREENTRANT_SYSCALLS_PROVIDED -DHAVE_NANOSLEEP -DHAVE_BLKSIZE -DHAVE_FCNTL -DHAVE_ASSERT_FUNC"
         # turn off unsupported items in posix directory 
 	newlib_cflags="${newlib_cflags} -D_NO_GETLOGIN -D_NO_GETPWENT -D_NO_GETUT -D_NO_GETPASS -D_NO_SIGSET -D_NO_WORDEXP -D_NO_POPEN -D_NO_POSIX_SPAWN"
         # turn off using cli/sti in i386 setjmp/longjmp
-- 
2.12.3

Re: [PATCH] Let RTEMS provide clock()

[Sebastian Huber]

On 06/09/17 10:41, Corinna Vinschen wrote:

> I just don't quite understand.  Is sysconf(_SC_CLK_TCK) different
> from CLOCKS_PER_SEC on RTEMS?

Yes, the CLOCKS_PER_SEC is a POSIX defined constant and the 
sysconf(_SC_CLK_TCK) returns the actual system tick frequency. Which is 
usually 100 or 1000 ticks per second.

-- 
Sebastian Huber, embedded brains GmbH

Address : Dornierstr. 4, D-82178 Puchheim, Germany
Phone   : +49 89 189 47 41-16
Fax     : +49 89 189 47 41-09
E-Mail  : sebastian.huber@embedded-brains.de
PGP     : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.

Re: [PATCH] Let RTEMS provide clock()

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1781 bytes --]

Hi Sebastian,

On Sep  6 08:58, Sebastian Huber wrote:
> Newlib uses _times_r() in clock().  The problem is that the _times_r()
> clock frequency is defined by sysconf(_SC_CLK_TCK).  The clock frequency
> of clock() is the constant CLOCKS_PER_SEC.
> 
> FreeBSD uses getrusage() for clock().  Since RTEMS has only one process,
> the implementation can be simplified.
> 
> Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
> ---
>  newlib/configure.host | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/newlib/configure.host b/newlib/configure.host
> index 3d967a11b..ba2d8c6c7 100644
> --- a/newlib/configure.host
> +++ b/newlib/configure.host
> @@ -605,7 +605,7 @@ case "${host}" in
>  	default_newlib_io_long_long="yes"
>  	default_newlib_io_c99_formats="yes"
>  	newlib_cflags="${newlib_cflags} -D_COMPILING_NEWLIB"
> -	newlib_cflags="${newlib_cflags} -DMALLOC_PROVIDED -DEXIT_PROVIDED -DSIGNAL_PROVIDED -DREENTRANT_SYSCALLS_PROVIDED -DHAVE_NANOSLEEP -DHAVE_BLKSIZE -DHAVE_FCNTL -DHAVE_ASSERT_FUNC"
> +	newlib_cflags="${newlib_cflags} -DCLOCK_PROVIDED -DMALLOC_PROVIDED -DEXIT_PROVIDED -DSIGNAL_PROVIDED -DREENTRANT_SYSCALLS_PROVIDED -DHAVE_NANOSLEEP -DHAVE_BLKSIZE -DHAVE_FCNTL -DHAVE_ASSERT_FUNC"
>          # turn off unsupported items in posix directory 
>  	newlib_cflags="${newlib_cflags} -D_NO_GETLOGIN -D_NO_GETPWENT -D_NO_GETUT -D_NO_GETPASS -D_NO_SIGSET -D_NO_WORDEXP -D_NO_POPEN -D_NO_POSIX_SPAWN"
>          # turn off using cli/sti in i386 setjmp/longjmp
> -- 
> 2.12.3

I see what you're doing and it's ok to push.

I just don't quite understand.  Is sysconf(_SC_CLK_TCK) different
from CLOCKS_PER_SEC on RTEMS?


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
Red Hat

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] Let RTEMS provide clock()

[Joel Sherrill]

On 9/6/2017 3:53 AM, Sebastian Huber wrote:
> On 06/09/17 10:41, Corinna Vinschen wrote:
> 
>> I just don't quite understand.  Is sysconf(_SC_CLK_TCK) different
>> from CLOCKS_PER_SEC on RTEMS?
> 
> Yes, the CLOCKS_PER_SEC is a POSIX defined constant and the
> sysconf(_SC_CLK_TCK) returns the actual system tick frequency. Which is
> usually 100 or 1000 ticks per second.
> 

Making it more complicated, the system tick frequency is
something that the end user can set for each application.
We just need to make sure that the constant CLOCKS_PER_SECOND
is appropriate for use with clock() as defined here:

http://pubs.opengroup.org/onlinepubs/9699919799/functions/clock.html

Sebastian.. I assume that all that matches up per POSIX, right?

--joel

Re: [PATCH] Let RTEMS provide clock()

[Craig Howland]

On 09/06/2017 09:56 AM, Joel Sherrill wrote:
>
>
> On 9/6/2017 3:53 AM, Sebastian Huber wrote:
>> On 06/09/17 10:41, Corinna Vinschen wrote:
>>
>>> I just don't quite understand.  Is sysconf(_SC_CLK_TCK) different
>>> from CLOCKS_PER_SEC on RTEMS?
>>
>> Yes, the CLOCKS_PER_SEC is a POSIX defined constant and the
>> sysconf(_SC_CLK_TCK) returns the actual system tick frequency. Which is
>> usually 100 or 1000 ticks per second.
>>
>
> Making it more complicated, the system tick frequency is
> something that the end user can set for each application.
> We just need to make sure that the constant CLOCKS_PER_SECOND
> is appropriate for use with clock() as defined here:
>
> http://pubs.opengroup.org/onlinepubs/9699919799/functions/clock.html
>
> Sebastian.. I assume that all that matches up per POSIX, right?
>
> --joel
CLOCKS_PER_SEC is not necessarily constant on all systems.  It is only POSIX 
which requires it to be fixed (at 1 million).  The C standard defines it as "an 
expression with type clock_t that is the number per second of the value returned 
by the clock function."  (In addition, POSIX does not require microsecond 
accuracy, so the _times_r() return value could simply be scaled to 
CLOCKS_PER_SEC if so desired as an alternative way of doing it.)

Re: [PATCH] Let RTEMS provide clock()

[Sebastian Huber]

On 06/09/17 16:30, Craig Howland wrote:
> On 09/06/2017 09:56 AM, Joel Sherrill wrote:
>>
>>
>> On 9/6/2017 3:53 AM, Sebastian Huber wrote:
>>> On 06/09/17 10:41, Corinna Vinschen wrote:
>>>
>>>> I just don't quite understand.  Is sysconf(_SC_CLK_TCK) different
>>>> from CLOCKS_PER_SEC on RTEMS?
>>>
>>> Yes, the CLOCKS_PER_SEC is a POSIX defined constant and the
>>> sysconf(_SC_CLK_TCK) returns the actual system tick frequency. Which is
>>> usually 100 or 1000 ticks per second.
>>>
>>
>> Making it more complicated, the system tick frequency is
>> something that the end user can set for each application.
>> We just need to make sure that the constant CLOCKS_PER_SECOND
>> is appropriate for use with clock() as defined here:
>>
>> http://pubs.opengroup.org/onlinepubs/9699919799/functions/clock.html
>>
>> Sebastian.. I assume that all that matches up per POSIX, right?
>>
>> --joel
> CLOCKS_PER_SEC is not necessarily constant on all systems.  It is only 
> POSIX which requires it to be fixed (at 1 million).  The C standard 
> defines it as "an expression with type clock_t that is the number per 
> second of the value returned by the clock function."  (In addition, 
> POSIX does not require microsecond accuracy, so the _times_r() return 
> value could simply be scaled to CLOCKS_PER_SEC if so desired as an 
> alternative way of doing it.)

This CLOCKS_PER_SEC seems to be a real mess. Its 100 or 1000000 on 
Newlib, 128 on FreeBSD, 100 on NetBSD and OpenBSD, 1000000 on glibc. 
XSI-conformant is 1000000:

http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/time.h.html

With respect to times() the document above says:

"To obtain the number of clock ticks per second returned by the 
/times/() 
<http://pubs.opengroup.org/onlinepubs/9699919799/functions/times.html> 
function, applications should call /sysconf/(_SC_CLK_TCK)."

-- 
Sebastian Huber, embedded brains GmbH

Address : Dornierstr. 4, D-82178 Puchheim, Germany
Phone   : +49 89 189 47 41-16
Fax     : +49 89 189 47 41-09
E-Mail  : sebastian.huber@embedded-brains.de
PGP     : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.