Date: Mon, 27 May 2013 13:47:47 +0200 From: =?UTF-8?Q?V=C3=A1clav_Zeman?= <vhaisman@gmail.com> To: Lee Thomas <lee_thomas@aslantools.com> Cc: freebsd-hackers@freebsd.org Subject: Re: Performance improvement to strnlen(). Message-ID: <CAKw7uVi3HoCg8qtTn-Sy9VCnDoVTnGJcxAcn5C_he1k8E7DqDw@mail.gmail.com> In-Reply-To: <a88fc3c3c7c4da882af20eed5a61aee1@lthomas.net> References: <2adc4d8e7c55e92d935f61efb4c9d723@lthomas.net> <CAKw7uVi2oofoy=yJQiHR927oGfy74UuYFq68r=C=gUVXBXZuKw@mail.gmail.com> <a88fc3c3c7c4da882af20eed5a61aee1@lthomas.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 27 May 2013 12:20, Lee Thomas wrote: > On 2013-05-27 04:37, V=C3=A1clav Zeman wrote: >> >> On 26 May 2013 21:01, Lee Thomas wrote: >>> >>> On 2013-05-26 08:00, V=C3=A1clav Zeman wrote: >>>> >>>> >>>> On 05/25/2013 10:27 PM, Lee Thomas wrote: >>>>> >>>>> >>>>> + lp =3D (const unsigned long *)((uintptr_t)str & ~LONGPTR_MASK= ); >>>>> + va =3D (*lp - mask01); >>>>> + vb =3D ((~*lp) & mask80); >>>> >>>> >>>> I do not think that this correct C. This is type punning violating the >>>> rules of the language. >>> >>> >>> >>> Hello V=C3=A1clav, >>> >>> The aliasing here is safe, because there are no writes through either o= f >>> the >>> pointers, and the reads are correctly aligned. >> >> I disagree. IANALL but AFAIK, this is simply not allowed by the >> language =3D> UB =3D> even though it seems to work in this instance, you >> are just lucky the UB is actually doing what you expect. >> >> -- >> VZ > > > Hello V=C3=A1clav, > > I am not an expert in C either, so you may be right that this is technica= lly > illegal. However, I copied this code from strlen.c, which has had it, and > still has it, for 4.5 years, and I can't see any way any alias analysis d= one > by the compiler could invalidate this code. In addition, there are many > places in the kernel, and in other codebases I've worked on, where this k= ind > of type conversion is done. See for instance > /sys/amd64/amd64/vm_macdep.c:200, where we compute the base of a thread's > stackframe from a pointer to an unrelated type of 'struct pcb', and then > write to it. > > I am willing to uglify the code in the way you suggest if that is the > general concensus, but I think the code as it stands is both safe and mor= e > legible. You could always put the three lines into a macro to keep the strnlen() more readable. -- VZ
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAKw7uVi3HoCg8qtTn-Sy9VCnDoVTnGJcxAcn5C_he1k8E7DqDw>