[PATCH] D41915: [lldCOFF] Print detailed timing information with /VERBOSE

Zachary Turner via llvm-commits llvm-commits at lists.llvm.org
Wed Jan 17 21:44:38 PST 2018


I know what you mean, but this is not actually a strict tree, because the
nodes do not strictly add up to the sum of the children.  Maybe someone
else can chime in, but I feel it is very awkward for me looking at a set of
numbers to see the total at the top.  If you look at MSVC's own /TIME
option, or other profiler outputs, or even something like an accounting
spreadsheet / ledger or just a simple math problem, the total is always at
the bottom.


On Wed, Jan 17, 2018 at 8:28 PM Rui Ueyama <ruiu at google.com> wrote:

> Well, I don't think that's inconsistent, but showing only the total at
> bottom seems inconsistent, because timers are organized as a tree. When we
> show some tree, usually the root is not shown in a special way. Think of
> the file tree in Windows Explorer for example. The root is at the beginning
> just like other sub-trees. It's also different from what we usually do with
> pen and paper because it's a tree.
>
>  47
> +23
>   17
>   6
> ----
>  70
>
> The following seems much more natural to me.
>
> 70
>  |- 47
>  |- 23
>      |- 17
>      |- 6
>
>
> On Wed, Jan 17, 2018 at 5:55 PM, Zachary Turner <zturner at google.com>
> wrote:
>
>> Putting the total at the top is inconsistent with every other profiling
>> output i can find from similar tools, and also with normal pen and paper
>> math which would be written
>>
>> ```
>> 47
>> +23
>> —-
>> 70
>> ```
>>
>> Can we leave it at the bottom?
>>
>> On Wed, Jan 17, 2018 at 1:07 PM Rui Ueyama via Phabricator <
>> reviews at reviews.llvm.org> wrote:
>>
>>> ruiu added a comment.
>>>
>>> It seems it is still showing the total timer at the end. Can you fix it?
>>>
>>>
>>> Repository:
>>>   rLLD LLVM Linker
>>>
>>> https://reviews.llvm.org/D41915
>>>
>>>
>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180118/11b16b27/attachment.html>


More information about the llvm-commits mailing list