Re: [kvm-ppc-devel] [PATCH] [3/3] kvmppc: make tlb tracing

Re: [kvm-ppc-devel] [PATCH] [3/3] kvmppc: make tlb tracing

[Arnd Bergmann]

On Monday 14 April 2008, ehrhardt@linux.vnet.ibm.com wrote:
> @@ -58,6 +58,14 @@ config KVM_POWERPC_440_TRACE_INSTRUCTION
>           channel.
>           If unsure, say N.
>  
> +config KVM_POWERPC_440_TRACE_TLB
> +       bool "ppc440 guest tlb activity tracing"
> +       depends on KVM && 44x && KVM_POWERPC_440
> +       select RELAY
> +       ---help---
> +         Adds the complete tracing of the tlb activities via a relayfs channel.
> +         If unsure, say N.
> +

I think you're taking the configurability a little too far, do you really
want users to be able to select the three new tracing options separately?
Choosing between all stats and no stats at all sounds sufficient to me.

	Arnd <><

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
kvm-ppc-devel mailing list
kvm-ppc-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-ppc-devel

Re: [kvm-ppc-devel] [PATCH] [3/3] kvmppc: make tlb tracing

[Christian Ehrhardt]

Arnd Bergmann wrote:
> On Monday 14 April 2008, ehrhardt@linux.vnet.ibm.com wrote:
>> @@ -58,6 +58,14 @@ config KVM_POWERPC_440_TRACE_INSTRUCTION
>>           channel.
>>           If unsure, say N.
>>  
>> +config KVM_POWERPC_440_TRACE_TLB
>> +       bool "ppc440 guest tlb activity tracing"
>> +       depends on KVM && 44x && KVM_POWERPC_440
>> +       select RELAY
>> +       ---help---
>> +         Adds the complete tracing of the tlb activities via a relayfs channel.
>> +         If unsure, say N.
>> +
> 
> I think you're taking the configurability a little too far, do you really
> want users to be able to select the three new tracing options separately?
> Choosing between all stats and no stats at all sounds sufficient to me.
> 
> 	Arnd <><

hmmm ... I intended it exaclty to be that selectable to save runtime overhead because e.g. the relay code add overhead all the time.
We might think of enabling them at runtime, but since all are developer tools I would prefer the .config way.
E.g. If I want to low overhead profile instructions so I would only enable KVM_POWERPC_440_INSTRUCTION_STAT and none
of the others. Actually that's the only reason the kvm_stat version exists, because the instruction tracing has the same info + much more just with more overhead.
btw - I already think of removing the kvm_stat variant, because the relay based tracing revealed to be not as overhead bound as I thought.

-- 

Grüsse / regards, 
Christian Ehrhardt
IBM Linux Technology Center, Open Virtualization

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
kvm-ppc-devel mailing list
kvm-ppc-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-ppc-devel