Project
Loading...
Searching...
No Matches
GPUTPCCFDecodeZS.h
Go to the documentation of this file.
1// Copyright 2019-2020 CERN and copyright holders of ALICE O2.
2// See https://alice-o2.web.cern.ch/copyright for details of the copyright holders.
3// All rights not expressly granted are reserved.
4//
5// This software is distributed under the terms of the GNU General Public
6// License v3 (GPL Version 3), copied verbatim in the file "COPYING".
7//
8// In applying this license CERN does not waive the privileges and immunities
9// granted to it by virtue of its status as an Intergovernmental Organization
10// or submit itself to any jurisdiction.
11
14
15#ifndef O2_GPU_DECODE_ZS_H
16#define O2_GPU_DECODE_ZS_H
17
18#include "clusterFinderDefs.h"
19#include "GPUGeneralKernels.h"
20#include "GPUConstantMem.h"
21#include "TPCBase/PadPos.h"
25
26namespace o2::gpu
27{
28
29class GPUTPCClusterFinder;
30
32{
33 public:
34 struct GPUSharedMemory /*: public GPUKernelTemplate::GPUSharedMemoryScan64<int32_t, GPUCA_WARP_SIZE>*/ {
35 GPUCA_SHARED_STORAGE(uint32_t ZSPage[o2::tpc::TPCZSHDR::TPC_ZS_PAGE_SIZE / sizeof(uint32_t)]);
36 uint32_t RowClusterOffset[o2::tpc::TPCZSHDR::TPC_MAX_ZS_ROW_IN_ENDPOINT];
41 GPUAtomic(uint32_t) rowOffsetCounter;
42 };
43
44 enum K : int32_t {
46 };
47
49
52 {
53 return processors.tpcClusterer;
54 }
55
56 GPUhdi() constexpr static gpudatatypes::RecoStep GetRecoStep()
57 {
59 }
60
61 template <int32_t iKernel = defaultKernel, typename... Args>
63};
64
66{
67
68 public:
71 {
72 return processors.tpcClusterer;
73 }
74
75 GPUhdi() constexpr static gpudatatypes::RecoStep GetRecoStep()
76 {
78 }
79
88
89 template <class Decoder>
91
92 GPUd() static o2::tpc::PadPos GetPadAndRowFromFEC(processorType& clusterer, int32_t cru, int32_t rawFecChannel, int32_t fecInPartition);
93 GPUd() static void WriteCharge(processorType& clusterer, float charge, o2::tpc::PadPos pos, tpccf::TPCFragmentTime localTime, size_t positionOffset);
95
96 GPUdi() static const uint8_t* ConsumeBytes(const uint8_t*& page, size_t nbytes)
97 {
98 const uint8_t* oldPage = page;
99 page += nbytes;
100 return oldPage;
101 }
102
103 GPUdi() static uint8_t ConsumeByte(const uint8_t*& page)
104 {
105 return *(page++);
106 }
107
108 template <typename T>
109 GPUdi() static const T* ConsumeHeader(const uint8_t*& page)
110 {
111 assert(size_t(page) % alignof(T) == 0);
112 return reinterpret_cast<const T*>(ConsumeBytes(page, sizeof(T)));
113 }
114
115 template <typename T = uint8_t>
116 GPUdi() static const T* Peek(const uint8_t* page, ptrdiff_t offset = 0)
117 {
118 // if ((size_t(page) + offset) % alignof(T) != 0) {
119 // printf("page = %zu, offset = %zu, alignof = %zu\n", size_t(page), offset, alignof(T));
120 // }
121 assert((size_t(page) + offset) % alignof(T) == 0);
122 return reinterpret_cast<const T*>(page + offset);
123 }
124
125 GPUdi() static float ADCToFloat(uint32_t adc, uint32_t decodeMask, float decodeBitsFactor)
126 {
127 return float(adc & decodeMask) * decodeBitsFactor;
128 }
129};
130
132{
133 public:
134 // constants for decoding
135 static inline GPUglobalconstexpr() int32_t DECODE_BITS = tpc::TPCZSHDRV2::TPC_ZS_NBITS_V34;
136 static inline GPUglobalconstexpr() float DECODE_BITS_FACTOR = 1.f / (1 << (DECODE_BITS - 10));
137 static inline GPUglobalconstexpr() uint32_t DECODE_MASK = (1 << DECODE_BITS) - 1;
138
140 // GPUCA_SHARED_STORAGE(uint32_t ZSPage[o2::tpc::TPCZSHDR::TPC_ZS_PAGE_SIZE / sizeof(uint32_t)]);
141 };
142
143 template <int32_t iKernel = defaultKernel, typename... Args>
144 GPUd() static void Thread(int32_t nBlocks, int32_t nThreads, int32_t iBlock, int32_t iThread, GPUSharedMemory& smem, processorType& clusterer, Args... args);
145
146 GPUd() static size_t DecodePage(GPUSharedMemory& smem, DecodeCtx& ctx);
147
148 GPUd() static void GetChannelBitmask(const tpc::zerosupp_link_based::CommonHeader& tbHdr, uint32_t* chan);
149 GPUd() static bool ChannelIsActive(const uint32_t* chan, uint8_t chanIndex);
150
151 GPUd() static void DecodeTB(GPUSharedMemory& smem, DecodeCtx& ctx, const uint8_t* adcData, uint32_t nAdc, const uint32_t* channelMask, int32_t timeBin, int32_t cru, int32_t fecInPartition);
152};
153
155{
156 public:
157 // constants for decoding
158 static inline GPUglobalconstexpr() int32_t DECODE_BITS = o2::tpc::TPCZSHDRV2::TPC_ZS_NBITS_V34;
159 static inline GPUglobalconstexpr() float DECODE_BITS_FACTOR = 1.f / (1 << (DECODE_BITS - 10));
160 static inline GPUglobalconstexpr() uint32_t DECODE_MASK = (1 << DECODE_BITS) - 1;
161
162 static inline GPUglobalconstexpr() int32_t MaxNLinksPerTimebin = 16;
163
165 // GPUCA_SHARED_STORAGE(uint32_t ZSPage[o2::tpc::TPCZSHDR::TPC_ZS_PAGE_SIZE / sizeof(uint32_t)]);
166 uint16_t samplesPerLinkEnd[MaxNLinksPerTimebin]; // Offset from end of TB link header to first sample not in this link
167 uint8_t linkIds[MaxNLinksPerTimebin];
168 uint8_t rawFECChannels[MaxNLinksPerTimebin * 80];
169 };
170
171 template <int32_t iKernel = defaultKernel, typename... Args>
172 GPUd() static void Thread(int32_t nBlocks, int32_t nThreads, int32_t iBlock, int32_t iThread, GPUSharedMemory& smem, processorType& clusterer, Args... args);
173
174 GPUd() static uint32_t DecodePage(GPUSharedMemory& smem, DecodeCtx& ctx);
175
176 GPUd() static bool ChannelIsActive(const uint8_t* chan, uint16_t chanIndex);
177
178 // Decode a single timebin within an 8kb page.
179 // Returns the number of samples decoded from the page
180 // or negative value to indicate an error (no samples are written in this case)
181 template <bool PayloadExtendsToNextPage>
182 GPUd() static int16_t DecodeTB(GPUSharedMemory& smem, DecodeCtx& ctx, const header::RAWDataHeader* rawDataHeader, int32_t cru, uint16_t nSamplesLeftInPage, const uint8_t* payloadEnd, const uint8_t* nextPage);
183};
184
185} // namespace o2::gpu
186
187#endif
int16_t charge
Definition RawEventData.h:5
unsigned short int uint16_t
std::vector< std::string > header
#define GPUsharedref()
#define GPUglobalconstexpr()
#define GPUd()
#define GPUCA_GET_THREAD_COUNT(...)
zsPage * page
int32_t nSamples
Timeframe settings Average distance in cm of events merged into Randomize distance around average distance of merged events Simulate correct bunch interactions instead of placing only the average number of events A value[n] sets TFlen for[n] collisions in Number of bunches per trainsort Number of bunch trains Instantaneous interaction rate Do not simulate border Total number of collisions to be placed in the interior of all time Do not select random but walk over array of events in stride steps h Event generator settings h configStandalone Use CPU for GPU Force usage of the specified GPU device no CPU fallback Continue processing after an error s
configStandalone Reconstruction rec_tpc multiplier for the combined cluster track error during track following area in cm for the search of for z only used if decrease removal by when above this local occupancy rowx Shift of TPC revert cluster flag edge bit distance to edge exceeds this ZOffset for secondary multiple of radius max P of electron K BB bands crossing max P of electron p BB bands crossing max P of electron d BB bands crossing max P of electron t BB bands crossing min accepted dEdxMaxTPC of the track Additive extra cluster error for Y2 if edge flag set Additive extra cluster error for Y2 if single set Additive extra cluster error for Y2 if or single set Additive extra cluster error for Z2 if or single set Extra errors to add to track when crossing depending on addErrorsCECrossing Inner portion of y range in sector that is not used in searching for extrapolated track candidates factor2YS for track merging factor2K for track merging Margin in cm of Y position when rejecting edge clusters based on uncorrected track Y Max fraction of shared clusters for tracklet Except pad from cluster finding if total number of charges in a fragment is above this Except pad from cluster finding if number of consecutive charges in a fragment is above this Threshold where a timebin is considered disabling the noisy pad check for that pad Treat signal above this minimum as saturated Smoothing factor for the exponential Highly Ionising Particle tail filter Number of timebins to use for the averaging Add additional custom track errors when crossing relative as fraction of Do N fit passes in final fit of< 0:reject at max -n hits") AddOptionRTC(rejectIFCLowRadiusCluster, uint8_t, 1, "", 0, "Reject clusters that get the IFC mask error during refit") AddOptionRTC(dEdxTruncLow, uint8_t, 2, "", 0, "Low truncation threshold, fraction of 128") AddOptionRTC(dEdxTruncHigh, uint8_t, 77, "", 0, "High truncation threshold, fraction of 128") AddOptionRTC(extrapolationTracking, int8_t, 1, "", 0, "Enable Extrapolation Tracking(prolong tracks to adjacent sectors to find short segments)") AddOptionRTC(disableRefitAttachment, uint8_t, 0, "", 0, "Bitmask to disable certain attachment steps during refit(1:attachment, 2:propagation, 4:loop following, 8:mirroring)") AddOptionRTC(rejectionStrategy, uint8_t, o2::gpu::GPUSettings::RejectionStrategyA, "", 0, "Enable rejection of TPC clusters for compression(0=no, 1=strategy A, 2=strategy B)") AddOptionRTC(mergeLoopersAfterburner, uint8_t, 1, "", 0, "Run afterburner for additional looper merging") AddOptionRTC(compressionTypeMask, uint8_t, o2::gpu::GPUSettings::CompressionFull, "", 0, "TPC Compression mode bits(1=truncate charge/width LSB, 2=differences, 4=track-model)") AddOptionRTC(compressionSortOrder, uint8_t, o2::gpu::GPUSettings::SortTime, "", 0, "Sort order of TPC compression(0=time, 1=pad, 2=Z-time-pad, 3=Z-pad-time, 4=no sorting(use incoming order))") AddOptionRTC(sigBitsCharge, uint8_t, 4, "", 0, "Number of significant bits for TPC cluster charge in compression mode 1") AddOptionRTC(sigBitsWidth, uint8_t, 3, "", 0, "Number of significant bits for TPC cluster width in compression mode 1") AddOptionRTC(dropLoopers, uint8_t, 0, "", 0, "Drop looping tracks starting from second loop") AddOptionRTC(mergerInterpolateErrors, uint8_t, 1, "", 0, "Use interpolation instead of extrapolation for chi2 based cluster rejection") AddOptionRTC(mergerInterpolateRejectAlsoOnCurrentPosition, uint8_t, 1, "", 0, "When using mergerInterpolateErrors, reject based on chi2 twice computed with interpolated and current track position starting from NDF > mergerNonInterpolateRejectMinNDF Merge tracks accross the central electrode Enable PID response Add sub threshold clusters in TPC dEdx computation OR mask of TPC flags that will reject the cluster in alternative dEdx Reject edge clusters based on uncorrected track Y during track fit Increase decrease removal by when on this number of pads from the edge h configStandalone Reconstruction rec_trd Max chi2 for TRD tracklets to be matched to a track Chi2 cut for strict matching mode Number of sigmas for ITS TPC track time error estimate Addition to search road around track prolongation along Y in cm Optimal tracklet rphi resolution in maximum chi2 per attached tracklet for TRD tracks take into account the full likelihood of the slope including exponential tails Tracks with less attached tracklets are discarded after the tracking Post trigger Pile up integration time in BCs fits all iterations but the last one in projections Use material lookup table for TPC refit Apply workarounds to avoid cyclic merge should not be needed tpc dyn configStandalone Processing proc_rtc Replace constant variables static by constexpr expressions Compile RTC in deterministic with NO_FAST_MATH flags and GPUCA_DETERMINISTIC_MODE define Use RTC to optimize GPU code h configStandalone Processing proc_rtctech Use a file lock to serialize access to the cache folder int32_t
configStandalone Reconstruction rec_tpc float
Definition of the RAW Data Header.
uint16_t pos
Definition RawData.h:3
definitions to deal with the link based zero suppression format
Definitions of TPC Zero Suppression Data Headers.
int32_t int32_t int32_t processorType & processors
GPUdi() static uint8_t ConsumeByte(const uint8_t *&page)
GPUdi() static const T *Peek(const uint8_t *page
GPUdi() static const T *ConsumeHeader(const uint8_t *&page)
GPUhdi() static processorType *Processor(GPUConstantMem &processors)
GPUhdi() const expr static gpudatatypes
static int32_t int32_t int32_t int32_t int32_t int32_t tpcTimeBinCut
static int32_t int32_t int32_t int32_t int32_t firstHBF
int32_t int32_t int32_t processorType Args args
GPUd() static void Thread(int32_t nBlocks
static int32_t int32_t nThreads
GPUhdi() const expr static gpudatatypes
static int32_t int32_t int32_t iBlock
static int32_t int32_t int32_t int32_t iThread
int32_t int32_t int32_t processorType & clusterer
GLdouble f
Definition glcorearb.h:310
GLintptr offset
Definition glcorearb.h:660
typedef void(APIENTRYP PFNGLCULLFACEPROC)(GLenum mode)
GPUdi() o2
Definition TrackTRD.h:39
a couple of static helper functions to create timestamp values for CCDB queries or override obsolete ...
GPUCA_SHARED_STORAGE(uint32_t ZSPage[o2::tpc::TPCZSHDR::TPC_ZS_PAGE_SIZE/sizeof(uint32_t)])
uint32_t RowClusterOffset[o2::tpc::TPCZSHDR::TPC_MAX_ZS_ROW_IN_ENDPOINT]
GPUAtomic(uint32_t) rowOffsetCounter
coder decode(ctfImage, triggersD, clustersD)
ArrayADC adc