[v3,2/4] lmbench: memsize: increase delay for slow FPGAs

otherwise memsize bails out and erroneously reports 1 MB

NOK
----
| />/lmbench/bin/arc/memsize 16
|
| 1

OK
---
| />/lmbench/bin/arc/memsize 16
| 2MB OK3MB OK4MB OK5MB OK6MB OK7MB OK8MB OK9MB OK10MB OK11MB OK12MB OK13MB OK14MB OK15MB OK16MB OK
| 16

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
---
 ...c-to-prevent-memsize-from-timingout-.patch | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 package/lmbench/0003-TOO_LONG-100-usec-to-prevent-memsize-from-timingout-.patch

Vineet, All,

On 2021-05-10 11:00 -0700, Vineet Gupta via buildroot spake thusly:
> otherwise memsize bails out and erroneously reports 1 MB
> 
> NOK
> ----
> | />/lmbench/bin/arc/memsize 16
> |
> | 1
> 
> OK
> ---

This triple-dash line is the usual git marker to end the actual commit
log, so everything below was dropped when applying the patch with git-am.
I've fixed that, and applied to master, thanks.

However, I wonder how relevant lmbench is nowadays. Indeed, there has
been zero activity on lmbench source code since June 2009, almost 12
years ago now...

Regards,
Yann E. MORIN.

> | />/lmbench/bin/arc/memsize 16
> | 2MB OK3MB OK4MB OK5MB OK6MB OK7MB OK8MB OK9MB OK10MB OK11MB OK12MB OK13MB OK14MB OK15MB OK16MB OK
> | 16
> 
> Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
> ---
>  ...c-to-prevent-memsize-from-timingout-.patch | 28 +++++++++++++++++++
>  1 file changed, 28 insertions(+)
>  create mode 100644 package/lmbench/0003-TOO_LONG-100-usec-to-prevent-memsize-from-timingout-.patch
> 
> diff --git a/package/lmbench/0003-TOO_LONG-100-usec-to-prevent-memsize-from-timingout-.patch b/package/lmbench/0003-TOO_LONG-100-usec-to-prevent-memsize-from-timingout-.patch
> new file mode 100644
> index 000000000000..6c3e7370e11b
> --- /dev/null
> +++ b/package/lmbench/0003-TOO_LONG-100-usec-to-prevent-memsize-from-timingout-.patch
> @@ -0,0 +1,28 @@
> +From d85b61666aa2728f68b69d02bfb162c432df709f Mon Sep 17 00:00:00 2001
> +From: vgupta <vgupta@4eed8e25-bc12-0410-818a-f70ee6866280>
> +Date: Sat, 3 Mar 2012 10:02:24 +0000
> +Subject: [PATCH] TOO_LONG 100 usec to prevent memsize from timingout @80Mhz
> +
> +git-svn-id: http://sjvm-subversion02.internal.synopsys.com/svn/OSS/ARC_Linux/tests/lmbench-3.0-a9@157318 4eed8e25-bc12-0410-818a-f70ee6866280
> +
> +Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
> +---
> + src/memsize.c | 2 +-
> + 1 file changed, 1 insertion(+), 1 deletion(-)
> +
> +diff --git a/src/memsize.c b/src/memsize.c
> +index eb25a0924872..b333cf8cbd32 100644
> +--- a/src/memsize.c
> ++++ b/src/memsize.c
> +@@ -15,7 +15,7 @@ char	*id = "$Id$\n";
> + #define	CHK(x)	if ((x) == -1) { perror("x"); exit(1); }
> + 
> + #ifndef	TOO_LONG
> +-#define	TOO_LONG	10	/* usecs */
> ++#define	TOO_LONG	100	/* usecs */
> + #endif
> + 
> + int	alarm_triggered = 0;
> +-- 
> +2.25.1
> +
> -- 
> 2.25.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

Hi Yann,

On 5/17/21 1:09 PM, Yann E. MORIN wrote:
>
>> OK
>> ---
> This triple-dash line is the usual git marker to end the actual commit
> log, so everything below was dropped when applying the patch with git-am.
> I've fixed that, and applied to master, thanks.

Yep I'm aware of that and sorry - there's no easy way to "underline" 
stuff in a patch :-)

> However, I wonder how relevant lmbench is nowadays. Indeed, there has
> been zero activity on lmbench source code since June 2009, almost 12
> years ago now...

Depends on who you ask. I think it is still relevant for embedded 
platforms.  Especially when doing new CPU and/or ISA bringup - former we 
tend to do regularly at Synopsys it gives you a great "snapshot" of 
performance of 2 systems: old and new. And this is true starting from 
fork/shell latency to memory latency and bandwidth tests which are still 
pretty smart for something written so long ago.

Thx for merging the patches.

-Vineet

vineet, All,

On 2021-05-17 21:53 +0000, Vineet Gupta spake thusly:
> On 5/17/21 1:09 PM, Yann E. MORIN wrote:
> >
> >> OK
> >> ---
> > This triple-dash line is the usual git marker to end the actual commit
> > log, so everything below was dropped when applying the patch with git-am.
> > I've fixed that, and applied to master, thanks.
> Yep I'm aware of that and sorry - there's no easy way to "underline" 
> stuff in a patch :-)

Oh yes, there _is_. ;-)

> > However, I wonder how relevant lmbench is nowadays. Indeed, there has
> > been zero activity on lmbench source code since June 2009, almost 12
> > years ago now...
> Depends on who you ask. I think it is still relevant for embedded 
> platforms.  Especially when doing new CPU and/or ISA bringup - former we 
> tend to do regularly at Synopsys it gives you a great "snapshot" of 
> performance of 2 systems: old and new. And this is true starting from 
> fork/shell latency to memory latency and bandwidth tests which are still 
> pretty smart for something written so long ago.

Ah, yes, that makes sense, indeed.

Regards,
Yann E. MORIN.

>>>>> "Vineet" == Vineet Gupta via buildroot <buildroot@busybox.net> writes:

 > otherwise memsize bails out and erroneously reports 1 MB
 > NOK
 > ----
 > | />/lmbench/bin/arc/memsize 16
 > |
 > | 1

 > OK
 > ---
 > | />/lmbench/bin/arc/memsize 16
 > | 2MB OK3MB OK4MB OK5MB OK6MB OK7MB OK8MB OK9MB OK10MB OK11MB OK12MB OK13MB OK14MB OK15MB OK16MB OK
 > | 16

 > Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

Committed to 2021.02.x, thanks.