lkml.org 
[lkml]   [2015]   [Nov]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH v2] string_helpers: fix precision loss for some inputs
Date
On Wed, Nov 04 2015, James Bottomley <James.Bottomley@HansenPartnership.com> wrote:

> On Wed, 2015-11-04 at 00:26 +0100, Rasmus Villemoes wrote:
>> On Tue, Nov 03 2015, James Bottomley <James.Bottomley@HansenPartnership.com> wrote:
>> >> Please spell it U32_MAX
>> >
>> > Why? there's no reason not to use the arithmetic UINT_MAX here. Either
>> > works, of course but UINT_MAX is standard.
>>
>> We're dealing with explicitly sized integers
>
> An integer is explicitly sized: it's 32 bits. That's why UINT_MAX is
> a universal constant.

In the Linux universe, yes. It's kind of amusing how you try to argue
based on the UINT_MAX name being (defined in a) standard while at the same
time very much rely on sizeof(int) having a value which is not specified
by the standard. I repeat:

>> U32_MAX is the natural name for the appropriate constant.

(and it's defined right next to UINT_MAX in kernel.h, so it's not like
you'd have to introduce that macro).

>> Also, you could do > U32_MAX instead of >= U32_MAX, but that's unlikely
>> to make any difference (well, except it might generate slightly better
>> code, since it would allow gcc to just test the upper half for being 0,
>> which might be cheaper on some architectures than comparing to a
>> literal).
>
> Heh if we're going to be that concerned about the code generation, then
> we should just tell gcc exactly how to do it instead of hoping it can
> work it out for itself, so
>
> while (blk_size >> 32) {
> ...

Nah, that would still require the compiler to be able to transform that
to the other form, which apparently it isn't. On x86_64, the simplest
is to load U32_MAX once into a register and then do r/r comparisons, but
when it's possible to directly test the upper half (e.g. when the 64 bit
value is represented in a pair of 32 bit registers) that's much
simpler. gcc generates good code for 'blk_size > U32_MAX' on both x86_64
and x86_32, but ends up doing an extra cmp on x86_32 for >=, and ends up
doing mov,shift,test inside the loop on x86_64 for 'blk_size >> 32'.

Rasmus


\
 
 \ /
  Last update: 2015-11-04 10:21    [W:0.090 / U:0.256 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site