This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] aarch64: thunderx2 memcpy implementation cleanup and streamlining
I agree, I see no point in just removing the branches names.
Also, for the branching rewriting you need to show if it has
any performance implications with at least the benchtests results
for memcpy.
On 20/03/2019 11:52, Patrick McGehearty wrote:
> I'll note that the named labels are helpful
> for understanding the intent of the code
> while the numeric labels make it more difficult
> to find all the sources for a particular branch target.
>
> Bottom line: numeric labels make assembly code
> more difficult to comprehend and maintain if
> you are not completely familiar with its logic.
>
> I've done considerable work in hand optimization
> of assembly language code (multiple architectures)
> over the years and generally find numeric labels
> to be troublesome, especially for larger routines.
>
>
> On 3/20/2019 9:31 AM, Anton Youdkevitch wrote:
>> Replaced named labels with the numeric ones to
>> reduce clutter.
>> Rewrote branching scheme in load and merge chunk
>> to be aligned with the most probable case.
>>
>> ChangeLog:
>>
>> * sysdeps/aarch64/multiarch/memcpy_thunderx2.S: cleanup
>>
>