|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
|
#ifndef CUDADAtaFormats_HGCal_HGCRecHitGPUProduct_H
#define CUDADAtaFormats_HGCal_HGCRecHitGPUProduct_H
#include <cassert>
#include <numeric>
#include "HeterogeneousCore/CUDAUtilities/interface/device_unique_ptr.h"
#include "CUDADataFormats/HGCal/interface/HGCRecHitSoA.h"
#include "CUDADataFormats/HGCal/interface/ConstHGCRecHitSoA.h"
#include "CUDADataFormats/HGCal/interface/HGCUncalibRecHitSoA.h"
class HGCRecHitGPUProduct {
public:
HGCRecHitGPUProduct() = default;
explicit HGCRecHitGPUProduct(uint32_t nhits, const cudaStream_t &stream) : nhits_(nhits) {
size_tot_ = std::accumulate(sizes_.begin(), sizes_.end(), 0); //this might be done at compile time
pad_ = ((nhits - 1) / 32 + 1) * 32; //align to warp boundary (assumption: warpSize = 32)
mem_ = cms::cuda::make_device_unique<std::byte[]>(pad_ * size_tot_, stream);
}
~HGCRecHitGPUProduct() = default;
HGCRecHitGPUProduct(const HGCRecHitGPUProduct &) = delete;
HGCRecHitGPUProduct &operator=(const HGCRecHitGPUProduct &) = delete;
HGCRecHitGPUProduct(HGCRecHitGPUProduct &&) = default;
HGCRecHitGPUProduct &operator=(HGCRecHitGPUProduct &&) = default;
HGCRecHitSoA get() {
HGCRecHitSoA soa;
soa.energy_ = reinterpret_cast<float *>(mem_.get());
soa.time_ = soa.energy_ + pad_;
soa.timeError_ = soa.time_ + pad_;
soa.id_ = reinterpret_cast<uint32_t *>(soa.timeError_ + pad_);
soa.flagBits_ = soa.id_ + pad_;
soa.son_ = reinterpret_cast<uint8_t *>(soa.flagBits_ + pad_);
soa.nbytes_ = size_tot_;
soa.nhits_ = nhits_;
soa.pad_ = pad_;
return soa;
}
ConstHGCRecHitSoA get() const {
ConstHGCRecHitSoA soa;
soa.energy_ = reinterpret_cast<float const *>(mem_.get());
soa.time_ = soa.energy_ + pad_;
soa.timeError_ = soa.time_ + pad_;
soa.id_ = reinterpret_cast<uint32_t const *>(soa.timeError_ + pad_);
soa.flagBits_ = soa.id_ + pad_;
soa.son_ = reinterpret_cast<uint8_t const *>(soa.flagBits_ + pad_);
return soa;
}
//number of hits stored in the SoA
uint32_t nHits() const { return nhits_; }
//pad of memory block (used for warp alignment, slighlty larger than 'nhits_')
uint32_t pad() const { return pad_; }
//number of bytes of the SoA
uint32_t nBytes() const { return size_tot_; }
private:
cms::cuda::device::unique_ptr<std::byte[]> mem_;
static constexpr std::array<int, memory::npointers::ntypes_hgcrechits_soa> sizes_ = {
{memory::npointers::float_hgcrechits_soa * sizeof(float),
memory::npointers::uint32_hgcrechits_soa * sizeof(uint32_t),
memory::npointers::uint8_hgcrechits_soa * sizeof(uint8_t)}};
uint32_t pad_;
uint32_t nhits_;
uint32_t size_tot_;
};
#endif //CUDADAtaFormats_HGCal_HGCRecHitGPUProduct_H
|