This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH 5/6] vfprintf: Introduce printf_positional function


On 05/21/2015 06:55 AM, Carlos O'Donell wrote:

> Oh! You're saying that the performance domain of positional versus
> non-positional is different and should not be lumped together, as
> Florian did not lump them together in his analysis.

Here's the run with both vfprintf-patched (build) and unpatched
(build-master) sources:

==> build/benchtests/bench-sprintf.out <==
  "sprintf": {
   "positional": {
    "duration": 2.4935e+10,
    "iterations": 1.6674e+07,
    "max": 1724.24,
    "min": 1486.34,
    "mean": 1495.44
   },
   "non-positional": {
    "duration": 2.49332e+10,
    "iterations": 2.5191e+07,
    "max": 1204.71,
    "min": 984.675,
    "mean": 989.767
   }
  }
==> build-master/benchtests/bench-sprintf.out <==
  "sprintf": {
   "positional": {
    "duration": 2.49366e+10,
    "iterations": 1.4544e+07,
    "max": 1948.33,
    "min": 1709.47,
    "mean": 1714.57
   },
   "non-positional": {
    "duration": 2.49334e+10,
    "iterations": 2.3758e+07,
    "max": 1277.08,
    "min": 1042.82,
    "mean": 1049.47
   }
  }

Again, the numbers are suspiciously good.  But at least they do not show
a performance regression.

I'll commit the remaining vfprintf patches then.

-- 
Florian Weimer / Red Hat Product Security


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]