[clang] 3590ede - [clang][Interp] Support vec_step

[Timm Bäder via cfe-commits]

Author: Timm Bäder
Date: 2024-04-29T18:21:24+02:00
New Revision: 3590ede848d67cf7d49d43e1830f90b053e3e76f

URL: https://github.com/llvm/llvm-project/commit/3590ede848d67cf7d49d43e1830f90b053e3e76f
DIFF: https://github.com/llvm/llvm-project/commit/3590ede848d67cf7d49d43e1830f90b053e3e76f.diff

LOG: [clang][Interp] Support vec_step

Added: 
    clang/test/AST/Interp/opencl.cl

Modified: 
    clang/lib/AST/Interp/ByteCodeExprGen.cpp
    clang/test/SemaOpenCL/vec_step.cl

Removed: 
    


################################################################################
diff  --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 3ceccfdebfca66..17f95e7f3cac0d 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -1358,6 +1358,20 @@ bool ByteCodeExprGen<Emitter>::VisitUnaryExprOrTypeTraitExpr(
     assert(E->getTypeOfArgument()->isSizelessVectorType());
   }
 
+  if (Kind == UETT_VecStep) {
+    if (const auto *VT = E->getTypeOfArgument()->getAs<VectorType>()) {
+      unsigned N = VT->getNumElements();
+
+      // The vec_step built-in functions that take a 3-component
+      // vector return 4. (OpenCL 1.1 spec 6.11.12)
+      if (N == 3)
+        N = 4;
+
+      return this->emitConst(N, E);
+    }
+    return this->emitConst(1, E);
+  }
+
   return false;
 }
 

diff  --git a/clang/test/AST/Interp/opencl.cl b/clang/test/AST/Interp/opencl.cl
new file mode 100644
index 00000000000000..b9ba4f8b9b555a
--- /dev/null
+++ b/clang/test/AST/Interp/opencl.cl
@@ -0,0 +1,32 @@
+// RUN: %clang_cc1 -fsyntax-only -verify=ref,both %s
+// RUN: %clang_cc1 -fsyntax-only -verify=expected,both %s -fexperimental-new-constant-interpreter
+
+// both-no-diagnostics
+
+typedef int int2 __attribute__((ext_vector_type(2)));
+typedef int int3 __attribute__((ext_vector_type(3)));
+typedef int int4 __attribute__((ext_vector_type(4)));
+typedef int int8 __attribute__((ext_vector_type(8)));
+typedef int int16 __attribute__((ext_vector_type(16)));
+
+void foo(int3 arg1, int8 arg2) {
+  int4 auto1;
+  int16 *auto2;
+  int auto3;
+  int2 auto4;
+  struct S *incomplete1;
+
+  int res1[vec_step(arg1) == 4 ? 1 : -1];
+  int res2[vec_step(arg2) == 8 ? 1 : -1];
+  int res3[vec_step(auto1) == 4 ? 1 : -1];
+  int res4[vec_step(*auto2) == 16 ? 1 : -1];
+  int res5[vec_step(auto3) == 1 ? 1 : -1];
+  int res6[vec_step(auto4) == 2 ? 1 : -1];
+  int res7[vec_step(int2) == 2 ? 1 : -1];
+  int res8[vec_step(int3) == 4 ? 1 : -1];
+  int res9[vec_step(int4) == 4 ? 1 : -1];
+  int res10[vec_step(int8) == 8 ? 1 : -1];
+  int res11[vec_step(int16) == 16 ? 1 : -1];
+  int res12[vec_step(void) == 1 ? 1 : -1];
+}
+

diff  --git a/clang/test/SemaOpenCL/vec_step.cl b/clang/test/SemaOpenCL/vec_step.cl
index afb6dc94d92e5e..c116f09b351fd2 100644
--- a/clang/test/SemaOpenCL/vec_step.cl
+++ b/clang/test/SemaOpenCL/vec_step.cl
@@ -1,4 +1,4 @@
-// RUN: %clang_cc1 -fsyntax-only -pedantic -verify %s
+// RUN: %clang_cc1 -fsyntax-only -pedantic -verify %s -fexperimental-new-constant-interpreter
 
 typedef int int2 __attribute__((ext_vector_type(2)));
 typedef int int3 __attribute__((ext_vector_type(3)));