[Mesa-dev] [PATCH 3/5] nir: Avoid structure initalization expressions.

[Jason Ekstrand]

On Wed, Apr 13, 2016 at 11:53 AM, Jose Fonseca <jfonseca at vmware.com> wrote:

> On 13/04/16 19:35, Jason Ekstrand wrote:
>
>>
>>
>> On Wed, Apr 13, 2016 at 10:19 AM, Emil Velikov <emil.l.velikov at gmail.com
>> <mailto:emil.l.velikov at gmail.com>> wrote:
>>
>>     From: Jose Fonseca <jfonseca at vmware.com <mailto:jfonseca at vmware.com>>
>>
>>     Not supported by MSVC, and completely unnecessary -- inline functions
>>     work just as well.
>>
>>     NIR_SRC_INIT/NIR_DEST_INIT could and probably should be replaced by
>> the
>>     inline functions.
>>
>>     Cc: Jason Ekstrand <jason at jlekstrand.net <mailto:jason at jlekstrand.net
>> >>
>>
>>
>>     ---
>>
>>     Jason, I believe you (used to) have an alternative in the vulkan
>> branch.
>>     This approach looks more reasonable imho, although I don't feel
>> strongly
>>     either way.
>>
>>     -Emil
>>     ---
>>       src/compiler/nir/nir.h | 18 ++++++++++++++++--
>>       1 file changed, 16 insertions(+), 2 deletions(-)
>>
>>     diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
>>     index ebac750..0adadc2 100644
>>     --- a/src/compiler/nir/nir.h
>>     +++ b/src/compiler/nir/nir.h
>>     @@ -503,7 +503,14 @@ typedef struct nir_src {
>>          bool is_ssa;
>>       } nir_src;
>>
>>     -#define NIR_SRC_INIT (nir_src) { { NULL } }
>>     +static inline nir_src
>>     +nir_src_init(void)
>>     +{
>>     +   nir_src src = { { NULL } };
>>     +   return src;
>>     +}
>>     +
>>     +#define NIR_SRC_INIT nir_src_init()
>>
>>
>> Is this a problem in C or only in C++?
>>
>
> IIRC, the problem was with C.
>
> If it's only a problem in C++,
>> another option (and I've written this patch before) is to do
>>
>> #ifdef _cplusplus
>> #define NIR_SRC_INIT nir_src()
>> #else
>> #define NIR_SRC_INIT { { NULL } }
>> #endif
>>
>> and the same for NIR_ALU_SRC_INIT, NIR_DST_INIT, etc.
>>
>
> But is there any thing to be gained by using { { NULL } } as opposed to
> the inline?  Is the generated code better, or error messages cleaner or
> something like that?
>

I don't know.  What I'd really prefer is

#define NIR_SRC_INIT ((nir_src) { { NULL } })

which is type-safe but doesn't work in C++.


> Generally speaking inlines tend to have much better characteristics that
> macros (e.g, in this case, type safety).
>

I guess I'm ok with the inline.  I think it's ugly and nasty but the
compiler should get rid of it just fine.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160413/8b7da4aa/attachment.html>