Re: [PATCH 1/2] bswap.h: drop unaligned loads

[Thomas Guyot <tguyot@gmail.com>]

On 2020-09-25 05:05, Carlo Arenas wrote:
> On Thu, Sep 24, 2020 at 6:15 PM brian m. carlson
> <sandals@crustytoothpaste.net> wrote:
>> On 2020-09-24 at 19:21:11, Jeff King wrote:
>>>   [stock]
>>>   Benchmark #1: t/helper/test-tool sha1 <foo.rand
>>>     Time (mean ± σ):      6.638 s ±  0.081 s    [User: 6.269 s, System: 0.368 s]
>>>     Range (min … max):    6.550 s …  6.841 s    10 runs
> 
> slightly offtopic but what generates this nicely formatted output?

It turns out I may have finally reproduced my diff speed regression at
home - just today - and I read this :) I may use it to test, nice timing!

About my unrelated regression, TL;DR, have anyone benchmarked under a
VM? (more context below...)

>> I cannot speak for s390, since I have never owned one
> 
> I happen to be lucky enough to have access to one (RHEL 8.2/z15, gcc
> 8.3.1) and seems (third consecutive run):
> 
> stock: user: 7.555s, system: 1.191s
> -DNO_UNALIGNED_LOADS: user: 7.561s, system: 1.189s

So also somewhat offtopic too, but I've been trying lately to reproduce
a speed regression with "git diff --no-ext-diff --quiet" (used by git
prompt in contrib) on some large repos between 2.9.5 and 2.16.2. Overall
the diff was over 3x as slow where I tested, and I measured a couple
timings between mmap of the same two large files between the two
version, the delta was approx 10x! (no other syscalls so it's all
computing and - if my theory proves right - I guess a lot of pagefaults
or similar handling control back to the host).

I prepped a VM where I will do more testing (I ran one preliminary test
and there seemed to be a visible difference, but I didn't have a proper
repo to test with yet). The point being it might be worth comparing the
two on VM's as well.

Regards,

Thomas