On Mon, Feb 1, 2016 at 8:15 AM, Nathan Sidwell <nat...@acm.org> wrote:
> On 02/01/16 10:32, Jakub Jelinek wrote:
>>
>> On Mon, Feb 01, 2016 at 09:15:05AM -0500, Nathan Sidwell wrote:
>>>
>>> On 01/29/16 10:18, Jakub Jelinek wrote:
>>>>
>>>> On Thu, Jan 28, 2016 at 10:38:51AM -0500, Nathan Sidwell wrote:
>>>>>
>>>>> This patch adds default compute dimension handling. Users rarely
>>>>> specify
>>>>> compute dimensions, expecting the toolchain to DTRT. More savvy users
>>>>> would
>>>>> like to specify global defaults. This patch permits both.
>>>>
>>>>
>>>> Isn't it better to be able to override the defaults on the library side?
>>>> I mean, when when somebody is compiling the code, often he doesn't know
>>>> the
>>>> exact properties of the hw it will be run on, if he does, I think it is
>>>> better to specify them explicitly in the code.
>>>
>>>
>>> I realized that it's actually not possible to markup the code in this
>>> way,
>>> as an 'intermediate' user. One can exercise complete control by saying
>>> exactly the axis/axes over which a loop is to be partitioned, and then
>>> specify the geometry. But one cannot use the 'auto' feature and have the
>>> compiler choose an axis without also relying on the compiler choosing a
>>> size
>>> for that axis. As I already said, IMHO being able to specify a
>>> compile-time size is useful.
>>
>>
>> Ok, I won't fight against it. But please make sure it can be overridden
>> on
>> the library side too.
>
>
> Absolutely, thanks!
>
This breaks bootstrap on x86:
../../src-trunk/gcc/omp-low.c: In function ‘void
oacc_parse_default_dims(const char*)’:
../../src-trunk/gcc/omp-low.c:20288:47: warning: comparison between
signed and unsigned integer expressions [-Wsign-compare]
if (errno || val <= 0 || (unsigned)val != val)
^
--
H.J.
