[PATCH] [OPENMP] Codegen for threadprivate variables

[Bataev, Alexey]

Hal, no, this is part of codegen patch. I can't test it without codegen. 
If this one would be testable without codegen, we'd included this 
already in our previous patches.

Best regards,
Alexey Bataev
=============
Software Engineer
Intel Compiler Team

23.09.2014 16:50, Hal Finkel пишет:
> ----- Original Message -----
>> From: "Alexey Bataev" <a.bataev at hotmail.com>
>> To: "a bataev" <a.bataev at hotmail.com>, dgregor at apple.com, hfinkel at anl.gov, rjmccall at gmail.com,
>> fraggamuffin at gmail.com, ejstotzer at gmail.com, richard at metafoo.co.uk, cbergstrom at pathscale.com
>> Cc: cfe-commits at cs.uiuc.edu
>> Sent: Tuesday, September 23, 2014 5:47:21 AM
>> Subject: Re: [PATCH] [OPENMP] Codegen for threadprivate variables
>>
> [snip]
>
> To be clear, regarding these changes below, is either of them separately testable? If so, please go ahead and commit them now with a test.
>
>   -Hal
>
>> ================
>> Comment at: lib/Parse/Parser.cpp:627
>> @@ -626,4 +626,3 @@
>>     case tok::annot_pragma_openmp:
>> -    ParseOpenMPDeclarativeDirective();
>> -    return DeclGroupPtrTy();
>> +    return ParseOpenMPDeclarativeDirective();
>>     case tok::annot_pragma_ms_pointers_to_members:
>> ----------------
>> hfinkel wrote:
>>> Is this a separate change?
>> Nope, this is required for codegen, without it the declaration is not
>> passed to codegen consumer.
>>
>> ================
>> Comment at: lib/Serialization/ASTReaderDecl.cpp:2186
>> @@ -2185,3 +2185,3 @@
>>         isa<ObjCImplDecl>(D) ||
>> -      isa<ImportDecl>(D))
>> +      isa<ImportDecl>(D) || isa<OMPThreadPrivateDecl>(D))
>>       return true;
>> ----------------
>> hfinkel wrote:
>>> This is a separate change.
>> Nope, it's a part of codegen, without this change codegen for
>> construct loaded from .pch does not work.
>>
>> http://reviews.llvm.org/D4002
>>
>>
>>