Fix usage of cl_intel_required_subgroup_size for non-intel driver

PiperOrigin-RevId: 640434035
tensorflow · Jun 5, 2024 · 025efe7 · 025efe7
1 parent e6971ed
commit 025efe7
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/tensorflow/lite/delegates/gpu/common/tasks/conv_generic.cc b/tensorflow/lite/delegates/gpu/common/tasks/conv_generic.cc
@@ -488,7 +488,7 @@ std::string ConvGeneric::GenerateConv(const GpuInfo& gpu_info,
          std::to_string(work_group_size_.y) + ", " +
          std::to_string(work_group_size_.z) + ")))\n";
   }
-  if (use_simd_broadcast && gpu_info.IsIntel() && gpu_info.IsApiOpenCl() &&
+  if (use_simd_broadcast && gpu_info.IsApiOpenCl() &&
       gpu_info.SupportsExtension("cl_intel_required_subgroup_size")) {
     c += "__attribute__((intel_reqd_sub_group_size(" +
          std::to_string(simd_size) + ")))\n";