GPU: Fix typo in variable name, fix a comment, fix a debug message

Author: davidrohr

GPU/GPUTracking/Base/GPUReconstructionCPU.cxx

@@ -68,7 +68,7 @@ inline void GPUReconstructionCPUBackend::runKernelBackendInternal(const krnlSetu
   int32_t nThreads = getNKernelHostThreads(false);
   if (nThreads > 1) {
     if (mProcessingSettings.debugLevel >= 5) {
-      printf("Running %d Threads\n", nThreads);
+      printf("Running %d Threads\n", mThreading->activeThreads->max_concurrency());
     }
     tbb::this_task_arena::isolate([&] {
       mThreading->activeThreads->execute([&] {
@@ -91,10 +91,10 @@ inline void GPUReconstructionCPUBackend::runKernelBackendInternal(const krnlSetu
 template <>
 inline void GPUReconstructionCPUBackend::runKernelBackendInternal<GPUMemClean16, 0>(const krnlSetupTime& _xyz, void* const& ptr, uint64_t const& size)
 {
-  int32_t nnThreads = std::max<int32_t>(1, std::min<int32_t>(size / (16 * 1024 * 1024), getNKernelHostThreads(true)));
-  if (nnThreads > 1) {
-    tbb::parallel_for(0, nnThreads, [&](int iThread) {
-      size_t threadSize = size / nnThreads;
+  int32_t nThreads = std::max<int32_t>(1, std::min<int32_t>(size / (16 * 1024 * 1024), getNKernelHostThreads(true)));
+  if (nThreads > 1) {
+    tbb::parallel_for(0, nThreads, [&](int iThread) {
+      size_t threadSize = size / nThreads;
       if (threadSize % 4096) {
         threadSize += 4096 - threadSize % 4096;
       }

GPU/GPUTracking/cmake/kernel_helpers.cmake

@@ -76,7 +76,6 @@ function(o2_gpu_add_kernel kernel_name kernel_files)
   set_property(TARGET O2_GPU_KERNELS APPEND PROPERTY O2_GPU_KERNEL_NAMES "${kernel_name}")
   set_property(TARGET O2_GPU_KERNELS APPEND PROPERTY O2_GPU_KERNEL_INCLUDES "${TMP_KERNEL_CLASS_FILE}")
   set_property(TARGET O2_GPU_KERNELS APPEND PROPERTY O2_GPU_KERNEL_FILES "${TMP_KERNEL_CLASS_FILE}.cxx")
-  # add_custom_command OUTPUT option does not support target-dependend generator expressions, thus this workaround
 
   set(O2_GPU_KERNEL_TEMPLATE_FILES "GPUConstantMem.h")
   if (GPUCA_BUILD_DEBUG)
@@ -102,6 +101,7 @@ function(o2_gpu_add_kernel kernel_name kernel_files)
   list(TRANSFORM O2_GPU_KERNEL_TEMPLATE_FILES PREPEND "#include \"")
   list(JOIN O2_GPU_KERNEL_TEMPLATE_FILES "\n" O2_GPU_KERNEL_TEMPLATE_FILES)
 
+  # add_custom_command OUTPUT option does not support target-dependend generator expressions, thus this workaround to create CUDA and HIP files
   string(REPLACE ", " "_" TMP_FILENAME "${kernel_name}")
   if(CUDA_ENABLED)
     set(TMP_FILENAMEA "${O2_GPU_KERNEL_WRAPPER_FOLDER}/krnl_${TMP_FILENAME}.cu")