Re: help wanted reproducing reported performance regression in groff 1.24

[G. Branden Robinson]

Hi Morten,

At 2026-03-17T15:47:07+0100, Morten Bo Johansen wrote:
> On 2026-03-17 G. Branden Robinson wrote:
> 
> > Which of these standard C++ libraries is groff linked with on your
> > system?
> >
> > For me:
> >
> > $ ldd $(type -p groff)
> >         linux-vdso.so.1 (0x00007ffe7df89000)
> >         libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 
> > (0x00007fa7659a9000)
> >         libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fa765865000)
> >         libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 
> > (0x00007fa76584b000)
> >         libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fa765677000)
> >         /lib64/ld-linux-x86-64.so.2 (0x00007fa765bc2000)
> 
> $ ldd $(type -p groff)
>         linux-vdso.so.1 (0x00007f8911ba9000)
>         libstdc++.so.6 => /usr/lib/libstdc++.so.6 (0x00007f8911800000)
>         libm.so.6 => /usr/lib/libm.so.6 (0x00007f89116e2000)
>         libgcc_s.so.1 => /usr/lib/libgcc_s.so.1 (0x00007f8911b26000)
>         libc.so.6 => /usr/lib/libc.so.6 (0x00007f89114f1000)
>         /lib64/ld-linux-x86-64.so.2 => /usr/lib64/ld-linux-x86-64.so.2 
> (0x00007f8911bab000)

Okay.  You're using GCC's C++ runtime.  Just like me.

> > My glibc and libstdc++ are geriatric compared to yours.
> 
> You use Debian, right? Arch is a pretty bleeding edge distro.

Yup.

> >> > Can you re-run your test scenario with the `-Z` option added to
> >> > groff's command line?  I'm curious to know how cutting grotty out of
> >> > the execution pipeline affects the numbers.
> >> 
> >> I used this script:
> >
> > Hmm, that's dimich's, not mine.
> 
> Oh, sorry. Here are the times with your script. No problem there,
> then:
> 
> generating up to 20 copies of "/home/mojo/tmp/bash.1" using device 'utf8'
> find data in 68145a.*.dat
> n  1.23.0  1.24.0
> 1  0.36  0.38
> 2  0.74  0.74
> 3  1.27  1.32
> 4  1.59  1.52
> 5  1.94  2.00
> 6  2.47  2.52
> 7  3.11  3.01
> 8  3.58  3.38
> 9  3.96  4.02
> 10  4.66  4.64
> 11  5.39  5.64
> 12  6.12  6.16
> 13  7.17  7.12
> 14  8.05  7.96
> 15  9.13  9.17
> 16  10.34  9.85
> 17  11.07  10.99
> 18  12.01  12.30
> 19  13.43  13.24
> 20  14.74  14.50

I'm glad you can reproduce my results.

But this makes the issue even more suspicious.

Are the Arch guys using a bad tactic for measuring performance?

Regards,
Branden

signature.asc
Description: PGP signature