Re: [PATCH] final: Improve output for -dp and -fverbose-asm

[Jeff Law <law at redhat dot com>]

On 11/30/2017 09:28 AM, Martin Sebor wrote:
> On 11/30/2017 04:16 AM, Segher Boessenkool wrote:
>> On Wed, Nov 29, 2017 at 08:46:41PM -0700, Martin Sebor wrote:
>>> On 11/29/2017 04:13 PM, Segher Boessenkool wrote:
>>>> This improves the assembler output (for -dp and -fverbose-asm) in
>>>> several ways.  It always prints the insn_cost.  It does not print
>>>> "[length = NN]" but "[c=NN l=NN]", to save space.  It does not add one
>>>> to the instruction alternative number (everything else starts counting
>>>> those at 0, too).  And finally, it tries to keep things lined up in
>>>> columns a bit better.
>>>>
>>>> Tested on powerpc64-linux {-m32,-m64}; is this okay for trunk?
>>>
>>> [c=NN l=NN] will be meaningless to those without some deeper
>>> insight into/experience with what's actually being printed.
>>> It might as well say [NN NN].  But such extreme terseness would
>>
>> Length isn't printed on all targets, fwiw.
>>
>>> seem to run counter to the documented purpose of -fverbose-asm
>>> to "Put extra commentary information in the generated assembly
>>> code to make it more readable."
>>
>> The point is that [length = 12] takes up an awful lot of space.  The
>> output of -fverbose-asm alread suffers from information overload.
> 
> Amount of space vs amount of detail are two different concerns.
> If you feel that the output is overly detailed then adding even
> more to it won't help.  Other than that, I don't think trading
> readability for space savings makes sense in a format whose main
> purpose is to be readable.  If it's line length that's a concern
> then using spaces instead of tabs would make them look shorter.
> You can also remove the brackets and use length=NN to shave off
> a couple of bytes.  Or print length only when it's not the typical
> default.  Or print the hex encoding instead.
Some of those things may make sense as well.  Though we also have to be
careful because some points don't have length information at all.  SIgh.

> 
>>> Looking further in the manual I don't see the [length=NN] bit
>>> mentioned (nor does your patch add a mention of it) so while
>>> the meaning of [length=NN] isn't exactly obvious, using [l=NN]
>>> instead certainly won't make it easier to read.  Does the manual
>>> need updating?
>>
>> Should -fverbose-asm print length (and cost) at all?  They should be
>> printed for -dp (which is for debugging the *compiler*), but are they
>> very useful for -fverbose-asm (which is primarily for debugging the
>> *user program*)?
> 
> I don't have the answers to these questions.  What I can say
> is that using one letter abbreviations for short words is not
> going to make it more readable, on the contrary.  len= would
> be okay.
I think length and costing information are definitely things we want to
include.  Length is less of an issue now than it was in the past, but it
definitely has value.

jeff