clblast::RunKernel, cl::Kernel: unify variants with/without waitForEvents, support empty LWS

2 files changed, 17 insertions, 28 deletions

diff --git a/src/routines/common.cpp b/src/routines/common.cpp
index 21e16954..3969cf9f 100644
--- a/src/routines/common.cpp
+++ b/src/routines/common.cpp
@@ -24,21 +24,23 @@ StatusCode RunKernel(Kernel &kernel, Queue &queue, const Device &device,
                      std::vector<size_t> global, const std::vector<size_t> &local,
                      EventPointer event, const std::vector<Event> &waitForEvents) {
 
-  // Tests for validity of the local thread sizes
-  if (local.size() > device.MaxWorkItemDimensions()) {
-    return StatusCode::kInvalidLocalNumDimensions; 
-  }
-  const auto max_work_item_sizes = device.MaxWorkItemSizes();
-  for (auto i=size_t{0}; i<local.size(); ++i) {
-    if (local[i] > max_work_item_sizes[i]) { return StatusCode::kInvalidLocalThreadsDim; }
-  }
-  auto local_size = size_t{1};
-  for (auto &item: local) { local_size *= item; }
-  if (local_size > device.MaxWorkGroupSize()) { return StatusCode::kInvalidLocalThreadsTotal; }
+  if (!local.empty()) {
+    // Tests for validity of the local thread sizes
+    if (local.size() > device.MaxWorkItemDimensions()) {
+      return StatusCode::kInvalidLocalNumDimensions;
+    }
+    const auto max_work_item_sizes = device.MaxWorkItemSizes();
+    for (auto i=size_t{0}; i<local.size(); ++i) {
+      if (local[i] > max_work_item_sizes[i]) { return StatusCode::kInvalidLocalThreadsDim; }
+    }
+    auto local_size = size_t{1};
+    for (auto &item: local) { local_size *= item; }
+    if (local_size > device.MaxWorkGroupSize()) { return StatusCode::kInvalidLocalThreadsTotal; }
 
-  // Make sure the global thread sizes are at least equal to the local sizes
-  for (auto i=size_t{0}; i<global.size(); ++i) {
-    if (global[i] < local[i]) { global[i] = local[i]; }
+    // Make sure the global thread sizes are at least equal to the local sizes
+    for (auto i=size_t{0}; i<global.size(); ++i) {
+      if (global[i] < local[i]) { global[i] = local[i]; }
+    }
   }
 
   // Tests for local memory usage
@@ -69,13 +71,5 @@ StatusCode RunKernel(Kernel &kernel, Queue &queue, const Device &device,
   return StatusCode::kSuccess;
 }
 
-// As above, but without an event waiting list
-StatusCode RunKernel(Kernel &kernel, Queue &queue, const Device &device,
-                     std::vector<size_t> global, const std::vector<size_t> &local,
-                     EventPointer event) {
-  auto emptyWaitingList = std::vector<Event>();
-  return RunKernel(kernel, queue, device, global, local, event, emptyWaitingList);
-}
-
 // =================================================================================================
 } // namespace clblast
diff --git a/src/routines/common.hpp b/src/routines/common.hpp
index d0bbc707..9d8849c3 100644
--- a/src/routines/common.hpp
+++ b/src/routines/common.hpp
@@ -29,12 +29,7 @@ namespace clblast {
 // Enqueues a kernel, waits for completion, and checks for errors
 StatusCode RunKernel(Kernel &kernel, Queue &queue, const Device &device,
                      std::vector<size_t> global, const std::vector<size_t> &local,
-                     EventPointer event, const std::vector<Event> &waitForEvents);
-
-// As above, but without an event waiting list
-StatusCode RunKernel(Kernel &kernel, Queue &queue, const Device &device,
-                     std::vector<size_t> global, const std::vector<size_t> &local,
-                     EventPointer event);
+                     EventPointer event, const std::vector<Event> &waitForEvents = {});
 
 // =================================================================================================