Hi Egor, On Tue, Mar 21, 2017 at 6:20 AM, Egor Churaev via cfe-commits <cfe-commits@lists.llvm.org> wrote: > Author: echuraev > Date: Tue Mar 21 08:20:57 2017 > New Revision: 298369 > > URL: http://llvm.org/viewvc/llvm-project?rev=298369&view=rev > Log: > [OpenCL] Added diagnostic for checking length of vector > > Reviewers: Anastasia, cfe-commits > > Reviewed By: Anastasia > > Subscribers: bader, yaxunl > > Differential Revision: https://reviews.llvm.org/D30937 > > Added: > cfe/trunk/test/SemaOpenCL/vector_swizzle_length.cl > Modified: > cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td > cfe/trunk/lib/Sema/SemaExprMember.cpp > > Modified: cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?rev=298369&r1=298368&r2=298369&view=diff > ============================================================================== > --- cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td (original) > +++ cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td Tue Mar 21 08:20:57 > 2017 > @@ -8236,6 +8236,8 @@ def err_opencl_ptrptr_kernel_param : Err > def err_kernel_arg_address_space : Error< > "pointer arguments to kernel functions must reside in '__global', " > "'__constant' or '__local' address space">; > +def err_opencl_ext_vector_component_invalid_length : Error< > + "vector component access has invalid length %0. Supported: > 1,2,3,4,8,16.">; > def err_opencl_function_variable : Error< > "%select{non-kernel function|function scope}0 variable cannot be declared > in %1 address space">; > def err_static_function_scope : Error< > > Modified: cfe/trunk/lib/Sema/SemaExprMember.cpp > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaExprMember.cpp?rev=298369&r1=298368&r2=298369&view=diff > ============================================================================== > --- cfe/trunk/lib/Sema/SemaExprMember.cpp (original) > +++ cfe/trunk/lib/Sema/SemaExprMember.cpp Tue Mar 21 08:20:57 2017 > @@ -284,6 +284,14 @@ IsRGBA(char c) { > } > } > > +// OpenCL v1.1, s6.1.7 > +// The component swizzle length must be in accordance with the acceptable > +// vector sizes. > +static bool IsValidOpenCLComponentSwizzleLength(unsigned len) > +{ > + return (len >= 1 && len <= 4) || len == 8 || len == 16; > +} > + > /// Check an ext-vector component access expression. > /// > /// VK should be set in advance to the value kind of the base > @@ -376,6 +384,19 @@ CheckExtVectorComponent(Sema &S, QualTyp > } > } > > + if (!HalvingSwizzle) { > + unsigned SwizzleLength = CompName->getLength(); > + > + if (HexSwizzle) > + SwizzleLength--; > + > + if (IsValidOpenCLComponentSwizzleLength(SwizzleLength) == false) { > + S.Diag(OpLoc, diag::err_opencl_ext_vector_component_invalid_length) > + << SwizzleLength << SourceRange(CompLoc); > + return QualType(); > + } > + } > + > // The component accessor looks fine - now we need to compute the actual > type. > // The vector type is implied by the component accessor. For example, > // vec4.b is a float, vec4.xy is a vec2, vec4.rgb is a vec3, etc. > > Added: cfe/trunk/test/SemaOpenCL/vector_swizzle_length.cl > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/test/SemaOpenCL/vector_swizzle_length.cl?rev=298369&view=auto > ============================================================================== > --- cfe/trunk/test/SemaOpenCL/vector_swizzle_length.cl (added) > +++ cfe/trunk/test/SemaOpenCL/vector_swizzle_length.cl Tue Mar 21 08:20:57 > 2017 > @@ -0,0 +1,10 @@ > +// RUN: %clang_cc1 %s -verify -pedantic -fsyntax-only > + > +typedef float float8 __attribute__((ext_vector_type(8))); > + > +void foo() { > + float8 f2 = (float8)(0, 0, 0, 0, 0, 0, 0, 0); > + > + f2.s01234; // expected-error {{vector component access has invalid > length 5. Supported: 1,2,3,4,8,16}} > + f2.xyzxy; // expected-error {{vector component access has invalid length > 5. Supported: 1,2,3,4,8,16}} > +}
Sorry for the necromancy here, but I wonder if we should only make this if LangOpts.OpenCL in on. Is there an initial intent for not to? The rationale is that we have given users support for ext_vector_type without OpenCL mode with arbitrary vectors lengths not defined in "6.1.2 Built-In Vector Data Types", that said it makes sense to support the component notation for those. I'm happy to fix it, I just need to know if this covers some background I'm not aware of. Thanks, -- Bruno Cardoso Lopes http://www.brunocardoso.cc _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits