Moved device vendor and type checks to a common header

3 files changed, 12 insertions, 11 deletions

diff --git a/include/internal/clpp11.h b/include/internal/clpp11.h
index e70f9000..b834d8b4 100644
--- a/include/internal/clpp11.h
+++ b/include/internal/clpp11.h
@@ -207,6 +207,12 @@ class Device {
     return true;
   }
 
+  // Query for a specific type of device or brand
+  bool IsCPU() const { return Type() == "CPU"; }
+  bool IsGPU() const { return Type() == "GPU"; }
+  bool IsAMD() const { return Vendor() == "AMD" || Vendor() == "Advanced Micro Devices, Inc."; }
+  bool IsARM() const { return Vendor() == "ARM"; }
+
   // Accessor to the private data-member
   const cl_device_id& operator()() const { return device_; }
  private:
diff --git a/include/internal/tuning.h b/include/internal/tuning.h
index 8fc79aff..a44f79d6 100644
--- a/include/internal/tuning.h
+++ b/include/internal/tuning.h
@@ -61,9 +61,9 @@ void Tuner(int argc, char* argv[]) {
       printf("* Unsupported precision, skipping this tuning run\n\n");
       return;
     }
-    isAMD = device.Vendor() == "AMD" || device.Vendor() == "Advanced Micro Devices, Inc.";
-    isARM = device.Vendor() == "ARM";
-    isGPU = device.Type() == "GPU";
+    isAMD = device.IsAMD();
+    isARM = device.IsARM();
+    isGPU = device.IsGPU();
   }
 
   // Creates input buffers with random data
diff --git a/src/routine.cc b/src/routine.cc
index 1acd814c..ca283b52 100644
--- a/src/routine.cc
+++ b/src/routine.cc
@@ -88,25 +88,20 @@ StatusCode Routine<T>::SetUp() {
   // Adds the name of the routine as a define
   defines += "#define ROUTINE_"+routine_name_+"\n";
 
-  // Determines whether this is a specific device
-  const auto isAMD = device_.Vendor() == "AMD" || device_.Vendor() == "Advanced Micro Devices, Inc.";
-  const auto isARM = device_.Vendor() == "ARM";
-  const auto isGPU = device_.Type() == "GPU";
-
   // For specific devices, use the non-IEE754 compilant OpenCL mad() instruction. This can improve
   // performance, but might result in a reduced accuracy.
-  if (isAMD && isGPU) {
+  if (device_.IsAMD() && device_.IsGPU()) {
     defines += "#define USE_CL_MAD 1\n";
   }
 
   // For specific devices, use staggered/shuffled workgroup indices.
-  if (isAMD && isGPU) {
+  if (device_.IsAMD() && device_.IsGPU()) {
     defines += "#define USE_STAGGERED_INDICES 1\n";
   }
 
   // For specific devices add a global synchronisation barrier to the GEMM kernel to optimize
   // performance through better cache behaviour
-  if (isARM && isGPU) {
+  if (device_.IsARM() && device_.IsGPU()) {
     defines += "#define GLOBAL_MEM_FENCE 1\n";
   }