GPUTracking: place the remaining cluster-finder constants in the constant address space

MSL requires every variable at program scope, class-scope statics included, to
name an address space. GPUglobalconstexpr() expands to constexpr everywhere
except Metal, so this is a no-op for host, CUDA, HIP, OpenCL and cling.

These were missed by the earlier pass: the header they sit in only started
being compiled once the errors ahead of it were gone.

Author: ktf

GPU/GPUTracking/DataCompression/GPUTPCCompressionTrackModel.h

@@ -40,7 +40,7 @@ namespace o2::gpu
 
 struct GPUParam;
 
-constexpr float MaxSinPhi = 0.999f;
+GPUglobalconstexpr() float MaxSinPhi = 0.999f;
 
 class GPUTPCCompressionTrackModel
 {

GPU/GPUTracking/Refit/GPUTrackingRefit.cxx

@@ -38,7 +38,7 @@ using namespace o2::track;
 using namespace o2::base;
 using namespace o2::tpc;
 
-static constexpr int32_t kIGNORE_ENDS = 3;
+static GPUglobalconstexpr() int32_t kIGNORE_ENDS = 3;
 
 #define IgnoreErrors(SNP)                                                                                            \
   if (mIgnoreErrorsOnTrackEnds) {                                                                                    \

GPU/GPUTracking/TPCClusterFinder/GPUTPCCFClusterizer.h

@@ -36,7 +36,7 @@ class MCLabelAccumulator;
 class GPUTPCCFClusterizer : public GPUKernelTemplate
 {
  public:
-  static constexpr size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFClusterizer);
+  static GPUglobalconstexpr() size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFClusterizer);
   struct GPUSharedMemory {
     CfChargePos posBcast[SCRATCH_PAD_WORK_GROUP_SIZE];
     PackedCharge buf[SCRATCH_PAD_WORK_GROUP_SIZE * SCRATCH_PAD_BUILD_N];

GPU/GPUTracking/TPCClusterFinder/GPUTPCCFDecodeZS.h

@@ -132,9 +132,9 @@ class GPUTPCCFDecodeZSLink : public GPUTPCCFDecodeZSLinkBase
 {
  public:
   // constants for decoding
-  static inline constexpr int32_t DECODE_BITS = tpc::TPCZSHDRV2::TPC_ZS_NBITS_V34;
-  static inline constexpr float DECODE_BITS_FACTOR = 1.f / (1 << (DECODE_BITS - 10));
-  static inline constexpr uint32_t DECODE_MASK = (1 << DECODE_BITS) - 1;
+  static inline GPUglobalconstexpr() int32_t DECODE_BITS = tpc::TPCZSHDRV2::TPC_ZS_NBITS_V34;
+  static inline GPUglobalconstexpr() float DECODE_BITS_FACTOR = 1.f / (1 << (DECODE_BITS - 10));
+  static inline GPUglobalconstexpr() uint32_t DECODE_MASK = (1 << DECODE_BITS) - 1;
 
   struct GPUSharedMemory : GPUKernelTemplate::GPUSharedMemoryWarpScan64<uint8_t, GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFDecodeZSLink)> {
     // GPUCA_SHARED_STORAGE(uint32_t ZSPage[o2::tpc::TPCZSHDR::TPC_ZS_PAGE_SIZE / sizeof(uint32_t)]);
@@ -155,11 +155,11 @@ class GPUTPCCFDecodeZSDenseLink : public GPUTPCCFDecodeZSLinkBase
 {
  public:
   // constants for decoding
-  static inline constexpr int32_t DECODE_BITS = o2::tpc::TPCZSHDRV2::TPC_ZS_NBITS_V34;
-  static inline constexpr float DECODE_BITS_FACTOR = 1.f / (1 << (DECODE_BITS - 10));
-  static inline constexpr uint32_t DECODE_MASK = (1 << DECODE_BITS) - 1;
+  static inline GPUglobalconstexpr() int32_t DECODE_BITS = o2::tpc::TPCZSHDRV2::TPC_ZS_NBITS_V34;
+  static inline GPUglobalconstexpr() float DECODE_BITS_FACTOR = 1.f / (1 << (DECODE_BITS - 10));
+  static inline GPUglobalconstexpr() uint32_t DECODE_MASK = (1 << DECODE_BITS) - 1;
 
-  static inline constexpr int32_t MaxNLinksPerTimebin = 16;
+  static inline GPUglobalconstexpr() int32_t MaxNLinksPerTimebin = 16;
 
   struct GPUSharedMemory : GPUKernelTemplate::GPUSharedMemoryWarpScan64<uint8_t, GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFDecodeZSDenseLink)> {
     // GPUCA_SHARED_STORAGE(uint32_t ZSPage[o2::tpc::TPCZSHDR::TPC_ZS_PAGE_SIZE / sizeof(uint32_t)]);

GPU/GPUTracking/TPCClusterFinder/GPUTPCCFDeconvolution.h

@@ -29,7 +29,7 @@ namespace o2::gpu
 class GPUTPCCFDeconvolution : public GPUKernelTemplate
 {
  public:
-  static constexpr size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFDeconvolution);
+  static GPUglobalconstexpr() size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFDeconvolution);
   struct GPUSharedMemory : public GPUKernelTemplate::GPUSharedMemoryScan64<int16_t, SCRATCH_PAD_WORK_GROUP_SIZE> {
     CfChargePos posBcast1[SCRATCH_PAD_WORK_GROUP_SIZE];
     uint8_t aboveThresholdBcast[SCRATCH_PAD_WORK_GROUP_SIZE];

GPU/GPUTracking/TPCClusterFinder/GPUTPCCFNoiseSuppression.h

@@ -34,7 +34,7 @@ class GPUTPCCFNoiseSuppression : public GPUKernelTemplate
     noiseSuppression = 0,
     updatePeaks = 1,
   };
-  static constexpr size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFNoiseSuppression_noiseSuppression);
+  static GPUglobalconstexpr() size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFNoiseSuppression_noiseSuppression);
   static_assert(GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFNoiseSuppression_noiseSuppression) == GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFNoiseSuppression_updatePeaks));
 
   struct GPUSharedMemory {

GPU/GPUTracking/TPCClusterFinder/GPUTPCCFPeakFinder.h

@@ -30,7 +30,7 @@ struct CfChargePos;
 class GPUTPCCFPeakFinder : public GPUKernelTemplate
 {
  public:
-  static constexpr size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFPeakFinder);
+  static GPUglobalconstexpr() size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFPeakFinder);
   struct GPUSharedMemory : public GPUKernelTemplate::GPUSharedMemoryScan64<int16_t, SCRATCH_PAD_WORK_GROUP_SIZE> {
     CfChargePos posBcast[SCRATCH_PAD_WORK_GROUP_SIZE];
     PackedCharge buf[SCRATCH_PAD_WORK_GROUP_SIZE * SCRATCH_PAD_SEARCH_N];

GPU/GPUTracking/TPCClusterFinder/GPUTPCNNClusterizerKernels.h

@@ -37,7 +37,7 @@ class GPUTPCNNClusterizerKernels : public GPUKernelTemplate
 {
  public:
   // Must all have same number of threads, since they use a common SCRATCH_PAD_WORK_GROUP_SIZE below
-  static constexpr size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFClusterizer);
+  static GPUglobalconstexpr() size_t SCRATCH_PAD_WORK_GROUP_SIZE = GPUCA_GET_THREAD_COUNT(GPUCA_LB_GPUTPCCFClusterizer);
   using GPUSharedMemory = GPUTPCCFClusterizer::GPUSharedMemory;
 
   GPUhdi() constexpr static gpudatatypes::RecoStep GetRecoStep()

GPU/GPUTracking/TRDTracking/GPUTRDGeometry.h

@@ -74,7 +74,7 @@ class GPUTRDGeometry : private o2::trd::GeometryFlat
   GPUd() int32_t GetRowMax(int32_t layer, int32_t stack, int32_t sector) const { return getRowMax(layer, stack, sector); }
   GPUd() bool ChamberInGeometry(int32_t det) const { return chamberInGeometry(det); }
 
-  static constexpr int32_t kNstack = o2::trd::constants::NSTACK;
+  static GPUglobalconstexpr() int32_t kNstack = o2::trd::constants::NSTACK;
 };
 } // namespace o2::gpu