r177035 - Add support for the 'endian' attribute for OpenCL.

Joey Gouly joey.gouly at arm.com
Mon Sep 9 05:57:32 PDT 2013


LGTM.

Sorry, thought you were just going to commit this from our IRC discussion :)

-----Original Message-----
From: aaron.ballman at gmail.com [mailto:aaron.ballman at gmail.com] On Behalf Of Aaron Ballman
Sent: 09 September 2013 13:47
To: Joey Gouly
Cc: llvm cfe; Richard Smith; Eli Friedman
Subject: Re: r177035 - Add support for the 'endian' attribute for OpenCL.

Ping

On Wed, Sep 4, 2013 at 11:13 AM, Aaron Ballman <aaron at aaronballman.com> wrote:
> Based on talking with Joey in IRC, I've removed the attribute from the
> tablegen so that it can go back to being an ignored attribute.  This
> does mean we lose diagnostics about incorrect endian modes, etc.
> However, I think it's an improvement over silently ignoring the
> attribute.  If/when we go to add support for the endian attribute in
> the future, it will be trivial to resurrect the previous
> functionality.
>
> ~Aaron
>
> On Wed, Sep 4, 2013 at 9:14 AM, Joey Gouly <joey.gouly at arm.com> wrote:
>> Me and Aaron talked about this over IRC.
>>
>> The plan for now is to make this produce an "attribute ignored" warning, since full support isn't there yet.
>>
>> -----Original Message-----
>> From: aaron.ballman at gmail.com [mailto:aaron.ballman at gmail.com] On Behalf Of Aaron Ballman
>> Sent: 04 September 2013 13:27
>> To: Joey Gouly
>> Cc: llvm cfe
>> Subject: Re: r177035 - Add support for the 'endian' attribute for OpenCL.
>>
>> Ping?
>>
>> On Sun, Sep 1, 2013 at 1:51 PM, Aaron Ballman <aaron at aaronballman.com> wrote:
>>> I realize this is kind of ancient history, but I'm confused -- this
>>> doesn't attach the attribute to anything in handleEndianAttr.  It does
>>> some semantic checking, but makes no Attr object, and doesn't attach
>>> anything to the decl.  What's more, it seems to have no semantic
>>> changes elsewhere either (no modifications to codegen, etc).
>>>
>>> Am I missing something that makes this functionality complete?  If
>>> not, is anyone actively working on it?
>>>
>>> ~Aaron
>>>
>>> On Thu, Mar 14, 2013 at 5:54 AM, Joey Gouly <joey.gouly at arm.com> wrote:
>>>> Author: joey
>>>> Date: Thu Mar 14 04:54:43 2013
>>>> New Revision: 177035
>>>>
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=177035&view=rev
>>>> Log:
>>>> Add support for the 'endian' attribute for OpenCL.
>>>>
>>>> Added:
>>>>     cfe/trunk/test/SemaOpenCL/endian-attr.cl
>>>> Modified:
>>>>     cfe/trunk/include/clang/Basic/Attr.td
>>>>     cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td
>>>>     cfe/trunk/lib/Sema/SemaDeclAttr.cpp
>>>>
>>>> Modified: cfe/trunk/include/clang/Basic/Attr.td
>>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/Attr.td?rev=177035&r1=177034&r2=177035&view=diff
>>>> ==============================================================================
>>>> --- cfe/trunk/include/clang/Basic/Attr.td (original)
>>>> +++ cfe/trunk/include/clang/Basic/Attr.td Thu Mar 14 04:54:43 2013
>>>> @@ -632,6 +632,11 @@ def ReqdWorkGroupSize : InheritableAttr
>>>>                UnsignedArgument<"ZDim">];
>>>>  }
>>>>
>>>> +def Endian : InheritableAttr {
>>>> +  let Spellings = [GNU<"endian">];
>>>> +  let Args = [IdentifierArgument<"platform">];
>>>> +}
>>>> +
>>>>  def WorkGroupSizeHint :  InheritableAttr {
>>>>    let Spellings = [GNU<"work_group_size_hint">];
>>>>    let Args = [UnsignedArgument<"XDim">,
>>>>
>>>> Modified: cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td
>>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?rev=177035&r1=177034&r2=177035&view=diff
>>>> ==============================================================================
>>>> --- cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td (original)
>>>> +++ cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td Thu Mar 14 04:54:43 2013
>>>> @@ -2150,6 +2150,8 @@ def warn_attribute_protected_visibility
>>>>    Warning<"target does not support 'protected' visibility; using 'default'">,
>>>>    InGroup<DiagGroup<"unsupported-visibility">>;
>>>>  def err_mismatched_visibility: Error<"visibility does not match previous declaration">;
>>>> +def warn_attribute_unknown_endian : Warning<"unknown endian '%0'">,
>>>> +  InGroup<IgnoredAttributes>;
>>>>  def note_previous_attribute : Note<"previous attribute is here">;
>>>>  def err_unknown_machine_mode : Error<"unknown machine mode %0">;
>>>>  def err_unsupported_machine_mode : Error<"unsupported machine mode %0">;
>>>>
>>>> Modified: cfe/trunk/lib/Sema/SemaDeclAttr.cpp
>>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaDeclAttr.cpp?rev=177035&r1=177034&r2=177035&view=diff
>>>> ==============================================================================
>>>> --- cfe/trunk/lib/Sema/SemaDeclAttr.cpp (original)
>>>> +++ cfe/trunk/lib/Sema/SemaDeclAttr.cpp Thu Mar 14 04:54:43 2013
>>>> @@ -2799,6 +2799,15 @@ static void handleVecTypeHint(Sema &S, D
>>>>                                                 ParmType, Attr.getLoc()));
>>>>  }
>>>>
>>>> +static void handleEndianAttr(Sema &S, Decl *D, const AttributeList &Attr) {
>>>> +  if (!dyn_cast<VarDecl>(D))
>>>> +    S.Diag(Attr.getLoc(), diag::warn_attribute_wrong_decl_type) << "endian"
>>>> +                                                                << 9;
>>>> +  StringRef EndianType = Attr.getParameterName()->getName();
>>>> +  if (EndianType != "host" && EndianType != "device")
>>>> +    S.Diag(Attr.getLoc(), diag::warn_attribute_unknown_endian) << EndianType;
>>>> +}
>>>> +
>>>>  SectionAttr *Sema::mergeSectionAttr(Decl *D, SourceRange Range,
>>>>                                      StringRef Name,
>>>>                                      unsigned AttrSpellingListIndex) {
>>>> @@ -4783,6 +4792,10 @@ static void ProcessInheritableDeclAttr(S
>>>>    case AttributeList::AT_VecTypeHint:
>>>>      handleVecTypeHint(S, D, Attr); break;
>>>>
>>>> +  case AttributeList::AT_Endian:
>>>> +    handleEndianAttr(S, D, Attr);
>>>> +    break;
>>>> +
>>>>    case AttributeList::AT_InitPriority:
>>>>        handleInitPriorityAttr(S, D, Attr); break;
>>>>
>>>>
>>>> Added: cfe/trunk/test/SemaOpenCL/endian-attr.cl
>>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/SemaOpenCL/endian-attr.cl?rev=177035&view=auto
>>>> ==============================================================================
>>>> --- cfe/trunk/test/SemaOpenCL/endian-attr.cl (added)
>>>> +++ cfe/trunk/test/SemaOpenCL/endian-attr.cl Thu Mar 14 04:54:43 2013
>>>> @@ -0,0 +1,9 @@
>>>> +// RUN: %clang_cc1 -verify %s
>>>> +
>>>> +constant long a __attribute__((endian(host))) = 100;
>>>> +
>>>> +constant long b __attribute__((endian(device))) = 100;
>>>> +
>>>> +constant long c __attribute__((endian(none))) = 100; // expected-warning {{unknown endian 'none'}}
>>>> +
>>>> +void func() __attribute__((endian(host))); // expected-warning {{endian attribute only applies to variables}}
>>>>
>>>>
>>>> _______________________________________________
>>>> cfe-commits mailing list
>>>> cfe-commits at cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>
>>
>>
>>








More information about the cfe-commits mailing list