On 10-12-2014 11:49, Fernando de Oliveira wrote:
> On 10-12-2014 11:22, William Harrington wrote:
>> On Wed, December 10, 2014 11:33, Armin K. wrote:
>>> On 12/10/2014 12:28 PM, Armin K. wrote:
>>>> On 12/09/2014 08:33 PM, [email protected] wrote:
>>>>> Author: fernando
>>
>>>>> Modified:
>>>>>    trunk/BOOK/xsoft/other/thunderbird.xml
>>>>>
>>>>> Modified: trunk/BOOK/xsoft/other/thunderbird.xml
>>>>> ==============================================================================
>>>>> --- trunk/BOOK/xsoft/other/thunderbird.xml        Tue Dec  9 11:19:25
>>>>> 2014      (r15176)
>>>>> +++ trunk/BOOK/xsoft/other/thunderbird.xml        Tue Dec  9 11:33:19
>>>>> 2014      (r15177)
>>>>> @@ -11,7 +11,7 @@
>>>>>    <!ENTITY thunderbird-md5sum
>>>>> "3781dfb541412c7f6b530a654b834ce5">
>>>>>    <!ENTITY thunderbird-size              "164 MB">
>>>>>    <!ENTITY thunderbird-buildsize         "3.6 GB (68 MB installed)">
>>>>> -  <!ENTITY thunderbird-time              "30 SBU (using
>>>>> parallelism=4)">
>>>>> +  <!ENTITY thunderbird-time              "16 SBU (using
>>>>> parallelism=4)">
>>>> As you can see, this largely depends on what's running on the system.
>>>> Why bother upgrading it this much? I only check it each major release.
>>>> Seems like too much work.
>>>>
>>>
>>> Don't get me wrong though, I have nothing against you doing this, but
>>> doesn't it make it little less true for someone? I mean, obviously the
>>> former value was from a system under (average) load and the second one
>>> is from a sort of base system with no additional software running. Why
>>> not add these two, then divide them by two to get a value that's mostly
>>> true for base system and systems under (average) load (ie, a wm, desktop
>>> or simply just X).
> 
> The problem has been discussed before, by Ken, Bruce, Igor and I. In one
> of the discussions, we talked about the time taken for libxul.so to
> link. In the case you mention, the difference in time is exactly that.
> First it took extra 14 SBU for linking. Depending on the hardware, it
> will take less or more. I got in my host 18 SBU, the day I got 30SBU in
> the VM.
> 
> 
> We have discussed about theses different values. I thinks it is not
> satisfactory, when about half (I've even got 2/3) of the time is for the
> linking.
> 
> 
> I believe you missed the discussions.

I'm sorry. Please ignore previous message.

>>>
>>> --
>>> Note: My last name is not Krejzi.
>>
>> Not to mention, having an SBU when using multiple CPU's or cores when the
>> rest of the book has SBU's based with one CPU or one core, and when the
>> LFS book SBU is measured with one core/cpu. May as well put in every book
>> an SBU based on multiple jobs when the system has so many CPU's or cores.
>> I suggest leaving the SBU as if someone is using one CPU or one core.
>>
>> Here's one major issue with the SBU and then stating the SBU when using 4
>> CPU's or cores,
>>
>> When the first SBU is measured it is using one CPU or one core. When using
>> multiple jobs, the SBU is not going to be accurate, especially when
>> someone moves the build to a different machine.
>>
>> I would suggest to leave the SBU with one CPU or core. Adding a note with
>> machine specifics could be used for SBU's when using multiple CPU's or
>> cores.
>>
>> Sincerely,
>>
>> William Harrington
> 
> The way I am doing has been introduced by Bruce. Later, there has been a
> discussion about it, I think in the dev list, and its under "SBU values
> in BLFS", at:
> 
> http://www.linuxfromscratch.org/blfs/view/svn/introduction/conventions.html
> 
> There are many packages with that, either in the build SBU or tests SBU.
> 
> I believe you missed all those.

I'm sorry. Please ignore previous message.

-- 
[]s,
Fernando
-- 
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to