add cdl_option CYGDBG_KERNEL_INSTRUMENT_MLQ

add cdl_option CYGDBG_KERNEL_INSTRUMENT_MLQ

[Pieter-Jan Busschaert]

Hello,


In kernel\current\src\sched\mlqueue.cxx, several calls are made to
CYG_INSTRUMENT_MLQ().
In kernel\current\include\instrmnt.h, this call is either defined as
nothing or as an actual instrumentation instruction, based on the
setting of CYGDBG_KERNEL_INSTRUMENT_MLQ.
However, this setting can not be modified in ecos.ecc because the
cdl_option is missing in kernel\current\cdl\instrument.cdl
This patch fixes this:


--- packages/kernel/current/cdl/instrument.cdl    Mon Sep 28 17:50:00 2009
+++ packages/kernel/current/cdl/instrument.cdl    Mon Sep 28 17:50:00 2009
@@ -246,6 +246,19 @@
         the code related to SMP operations."
 }

+cdl_option CYGDBG_KERNEL_INSTRUMENT_MLQ {
+    display       "Instrument the Multi-level Queue scheduler"
+    default_value 1
+    active_if     CYGSEM_KERNEL_SCHED_MLQUEUE
+    description   "
+        It is possible to perform selective instrumentation at
+        run-time. It is also possible to disable instrumentation
+        in various kernel components at compile-time, thus
+        reducing the code size overheads. This option controls
+        whether or not instrumentation support is compiled into
+        the code for the Multi-level Queue scheduler."
+}
+
 cdl_option CYGDBG_KERNEL_INSTRUMENT_USER {
     display       "Support application-level instrumentation"
     default_value 1

Re: add cdl_option CYGDBG_KERNEL_INSTRUMENT_MLQ

[John Dallaway]

Hi Pieter-Jan

Pieter-Jan Busschaert wrote:

> In kernel\current\src\sched\mlqueue.cxx, several calls are made to
> CYG_INSTRUMENT_MLQ().
> In kernel\current\include\instrmnt.h, this call is either defined as
> nothing or as an actual instrumentation instruction, based on the
> setting of CYGDBG_KERNEL_INSTRUMENT_MLQ.
> However, this setting can not be modified in ecos.ecc because the
> cdl_option is missing in kernel\current\cdl\instrument.cdl
> This patch fixes this:

[ snip ]

This seems perfectly reasonable. Is the resulting MLQ instrumentation
data sensible? If so, I will add the CDL option.

John Dallaway

Re: add cdl_option CYGDBG_KERNEL_INSTRUMENT_MLQ

[Pieter-Jan Busschaert]

Hi,

2009/10/1 John Dallaway <john@dallaway.org.uk>:
> Pieter-Jan Busschaert wrote:
>
>> In kernel\current\src\sched\mlqueue.cxx, several calls are made to
>> CYG_INSTRUMENT_MLQ().
>
> This seems perfectly reasonable. Is the resulting MLQ instrumentation
> data sensible? If so, I will add the CDL option.


I am now analyzing this for the mutex3 test in the kernel (priority
inheritance). At first sight the instrumentation data does look
sensible. If I would see something odd, I will send a patch for that,
but I don't expect that right now.


Kind regards,

Pieter-Jan

Re: add cdl_option CYGDBG_KERNEL_INSTRUMENT_MLQ

[John Dallaway]

Hi Pieter-Jan

Pieter-Jan Busschaert wrote:

> 2009/10/1 John Dallaway <john@dallaway.org.uk>:
>> Pieter-Jan Busschaert wrote:
>>
>>> In kernel\current\src\sched\mlqueue.cxx, several calls are made to
>>> CYG_INSTRUMENT_MLQ().
>> This seems perfectly reasonable. Is the resulting MLQ instrumentation
>> data sensible? If so, I will add the CDL option.
> 
> I am now analyzing this for the mutex3 test in the kernel (priority
> inheritance). At first sight the instrumentation data does look
> sensible. If I would see something odd, I will send a patch for that,
> but I don't expect that right now.

I've now added the CDL option.

Thank you for your report...

John Dallaway