On 09/18/2018 04:02 PM, Jim Jagielski wrote:
>
>
>> On Sep 18, 2018, at 9:53 AM, Jim Jagielski <[email protected]
>> <mailto:[email protected]>> wrote:
>>
>>
>>
>>> On Sep 18, 2018, at 9:33 AM, Ruediger Pluem <[email protected]
>>> <mailto:[email protected]>> wrote:
>>>
>>>
>>>
>>> On 09/17/2018 05:50 PM, [email protected] <mailto:[email protected]> wrote:
>>>> Author: jim
>>>> Date: Mon Sep 17 15:50:19 2018
>>>> New Revision: 1841078
>>>>
>>>> URL: http://svn.apache.org/viewvc?rev=1841078&view=rev
>>>> Log:
>>>> Add in Atomics for 64bit ints
>>>>
>>>> Added:
>>>> apr/apr/trunk/atomic/unix/builtins64.c (with props)
>>>> apr/apr/trunk/atomic/unix/mutex64.c (with props)
>>>> apr/apr/trunk/atomic/win32/apr_atomic64.c (with props)
>>>> Modified:
>>>> apr/apr/trunk/CHANGES
>>>> apr/apr/trunk/apr.dsp
>>>> apr/apr/trunk/include/apr_atomic.h
>>>> apr/apr/trunk/include/arch/unix/apr_arch_atomic.h
>>>> apr/apr/trunk/test/testatomic.c
>>>>
>>>
>>> Hm, don't we miss some configure magic here?
>>
>>
>> How so? All configure does is see if __sync_val_compare_and_swap, et.al.
>> exists
>
> If it does, then it adds the 64bit versions. This is for Unix. For other
> platforms, either we know they do have 64bit
> atomics (eg: Windows) or else we use the portable versions of 64bit atomics.
Ahh, now I get it. Thanks. If the existing test for the compiler atomics
succeeds then 32bit *and* 64bit atomics are
offered by the compiler on a 64 bit platform, correct?
Regards
Rüdiger
