Matriplex Namespace Reference

Classes
struct	CholeskyInverter
struct	CholeskyInverter< T, 3, N >
struct	CholeskyInverterSym
struct	CholeskyInverterSym< T, 3, N >
struct	CramerInverter
struct	CramerInverter< T, 2, N >
struct	CramerInverter< T, 3, N >
struct	CramerInverterSym
struct	CramerInverterSym< T, 2, N >
struct	CramerInverterSym< T, 3, N >
class	MatriplexVector
struct	MultiplyCls
struct	MultiplyCls< T, 3, N >
struct	MultiplyCls< T, 6, N >
struct	SymMultiplyCls
struct	SymMultiplyCls< T, 3, N >
struct	SymMultiplyCls< T, 6, N >

Typedefs
typedef int	idx_t
template<typename T , idx_t D1, idx_t D2, idx_t N>
using	MPlex = Matriplex< T, D1, D2, N >
template<typename T , idx_t D, idx_t N>
using	MPlexSym = MatriplexSym< T, D, N >
template<class MP >
using	MPlexVec = MatriplexVector< MP >

Functions
template<typename T , idx_t D, idx_t N>
class	__attribute__ ((aligned(32))) MatriplexSym
template<typename T , idx_t D1, idx_t D2, idx_t N>
class	__attribute__ ((aligned(32))) Matriplex
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	abs (const MPlex< T, D1, D2, N > &a)
void	align_check (const char *pref, void *adr)
void *	aligned_alloc64 (std::size_t size)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	cos (const MPlex< T, D1, D2, N > &a)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	hypot (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D, idx_t N>
void	invertCholesky (MPlexVec< MPlex< T, D, D, N >> &A, int n_to_process=0)
template<typename T , idx_t D, idx_t N>
void	invertCholesky (MPlex< T, D, D, N > &A)
template<typename T , idx_t D, idx_t N>
void	invertCholeskySym (MPlexVec< MPlexSym< T, D, N >> &A, int n_to_process=0)
template<typename T , idx_t D, idx_t N>
void	invertCholeskySym (MPlexSym< T, D, N > &A)
template<typename T , idx_t D, idx_t N>
void	invertCramer (MPlexVec< MPlex< T, D, D, N >> &A, int n_to_process=0)
template<typename T , idx_t D, idx_t N>
void	invertCramer (MPlex< T, D, D, N > &A, double *determ=nullptr)
template<typename T , idx_t D, idx_t N>
void	invertCramerSym (MPlexVec< MPlexSym< T, D, N >> &A, int n_to_process=0)
template<typename T , idx_t D, idx_t N>
void	invertCramerSym (MPlexSym< T, D, N > &A, double *determ=nullptr)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	max (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	min (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
void	min_max (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b, MPlex< T, D1, D2, N > &min, MPlex< T, D1, D2, N > &max)
template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>
void	multiply (const MPlexVec< MPlex< T, D1, D2, N >> &A, const MPlexVec< MPlex< T, D2, D3, N >> &B, MPlexVec< MPlex< T, D1, D3, N >> &C, int n_to_process=0)
template<typename T , idx_t D, idx_t N>
void	multiply (const MPlexVec< MPlexSym< T, D, N >> &A, const MPlexVec< MPlexSym< T, D, N >> &B, MPlexVec< MPlex< T, D, D, N >> &C, int n_to_process=0)
template<typename T , idx_t D, idx_t N>
void	multiply (const MPlexSym< T, D, N > &A, const MPlexSym< T, D, N > &B, MPlex< T, D, D, N > &C)
template<typename T , idx_t D, idx_t N>
void	multiply (const MPlex< T, D, D, N > &A, const MPlex< T, D, D, N > &B, MPlex< T, D, D, N > &C)
template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>
void	multiply3in (MPlexVec< MPlex< T, D1, D2, N >> &A, MPlexVec< MPlex< T, D2, D3, N >> &B, MPlexVec< MPlex< T, D1, D3, N >> &C, int n_to_process=0)
template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>
void	multiplyGeneral (const MPlexVec< MPlex< T, D1, D2, N >> &A, const MPlexVec< MPlex< T, D2, D3, N >> &B, MPlexVec< MPlex< T, D1, D3, N >> &C, int n_to_process=0)
template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>
void	multiplyGeneral (const MPlex< T, D1, D2, N > &A, const MPlex< T, D2, D3, N > &B, MPlex< T, D1, D3, N > &C)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator* (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator* (const MPlex< T, D1, D2, N > &a, T b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator* (T a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator+ (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator+ (const MPlex< T, D1, D2, N > &a, T b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator+ (T a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator- (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator- (const MPlex< T, D1, D2, N > &a, T b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator- (T a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator/ (const MPlex< T, D1, D2, N > &a, const MPlex< T, D1, D2, N > &b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator/ (const MPlex< T, D1, D2, N > &a, T b)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	operator/ (T a, const MPlex< T, D1, D2, N > &b)
constexpr std::size_t	round_up_align64 (std::size_t size)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	sin (const MPlex< T, D1, D2, N > &a)
template<typename T , idx_t D1, idx_t D2, idx_t N>
void	sincos (const MPlex< T, D1, D2, N > &a, MPlex< T, D1, D2, N > &s, MPlex< T, D1, D2, N > &c)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	sqr (const MPlex< T, D1, D2, N > &a)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	sqrt (const MPlex< T, D1, D2, N > &a)
template<typename T , idx_t D1, idx_t D2, idx_t N>
MPlex< T, D1, D2, N >	tan (const MPlex< T, D1, D2, N > &a)

Variables
const idx_t	gSymOffsets [7][36]

Typedef Documentation

idx_t

typedef int Matriplex::idx_t

Definition at line 98 of file MatriplexCommon.h.

MPlex

template<typename T , idx_t D1, idx_t D2, idx_t N>

using Matriplex::MPlex = typedef Matriplex<T, D1, D2, N>

Definition at line 327 of file Matriplex.h.

MPlexSym

template<typename T , idx_t D, idx_t N>

using Matriplex::MPlexSym = typedef MatriplexSym<T, D, N>

Definition at line 279 of file MatriplexSym.h.

MPlexVec

template<class MP >

using Matriplex::MPlexVec = typedef MatriplexVector<MP>

Definition at line 48 of file MatriplexVector.h.

Function Documentation

__attribute__() [1/2]

template<typename T , idx_t D, idx_t N>

class Matriplex::__attribute__

(

(aligned(32))

)

no. of matrix rows

no. of matrix columns

no of elements: lower triangle

size of the whole matriplex

Definition at line 25 of file MatriplexSym.h.

References a, PVValHelper::add(), ASSUME_ALIGNED, b, ALPAKA_ACCELERATOR_NAMESPACE::brokenline::constexpr(), filterCSVwithJSON::copy, ztail::d, gSymOffsets, mps_fire::i, recoMuon::in, dqmiolumiharvest::j, dqmdumpme::k, visualization-live-secondInstance_cfg::m, N, dqmiodumpmetadata::n, hgchebackDigitizer_cfi::noise, unpackBuffers-CaloStage1::offsets, operator=(), operator[](), AlCaHLTBitMon_ParallelJobs::p, alignCSCRings::s, pfClustersFromCombinedCaloHF_cfi::scale, TrackRefitter_38T_cff::src, electronEcalRecHitIsolationLcone_cfi::subtract, submitPVValidationJobs::t, createJobs::tmp, mitigatedMETSequence_cff::U, interactiveExample::ui, findQualityFiles::v, and geometryCSVtoXML::xx.

                                                           {
  public:
    typedef T value_type;

    static constexpr int kRows = D;
    static constexpr int kCols = D;
    static constexpr int kSize = (D + 1) * D / 2;
    static constexpr int kTotSize = N * kSize;

    T fArray[kTotSize];

    MatriplexSym() {}
    MatriplexSym(T v) { setVal(v); }

    idx_t plexSize() const { return N; }

    void setVal(T v) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] = v;
      }
    }

    void add(const MatriplexSym& v) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] += v.fArray[i];
      }
    }

    void scale(T scale) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] *= scale;
      }
    }

    T operator[](idx_t xx) const { return fArray[xx]; }
    T& operator[](idx_t xx) { return fArray[xx]; }

    const idx_t* offsets() const { return gSymOffsets[D]; }
    idx_t off(idx_t i) const { return gSymOffsets[D][i]; }

    const T& constAt(idx_t n, idx_t i, idx_t j) const { return fArray[off(i * D + j) * N + n]; }

    T& At(idx_t n, idx_t i, idx_t j) { return fArray[off(i * D + j) * N + n]; }

    T& operator()(idx_t n, idx_t i, idx_t j) { return At(n, i, j); }
    const T& operator()(idx_t n, idx_t i, idx_t j) const { return constAt(n, i, j); }

    MatriplexSym& operator=(const MatriplexSym& m) {
      memcpy(fArray, m.fArray, sizeof(T) * kTotSize);
      return *this;
    }

    MatriplexSym(const MatriplexSym& m) = default;

    void copySlot(idx_t n, const MatriplexSym& m) {
      for (idx_t i = n; i < kTotSize; i += N) {
        fArray[i] = m.fArray[i];
      }
    }

    void copyIn(idx_t n, const T* arr) {
      for (idx_t i = n; i < kTotSize; i += N) {
        fArray[i] = *(arr++);
      }
    }

    void copyIn(idx_t n, const MatriplexSym& m, idx_t in) {
      for (idx_t i = n; i < kTotSize; i += N, in += N) {
        fArray[i] = m[in];
      }
    }

    void copy(idx_t n, idx_t in) {
      for (idx_t i = n; i < kTotSize; i += N, in += N) {
        fArray[i] = fArray[in];
      }
    }

#if defined(AVX512_INTRINSICS)

    template <typename U>
    void slurpIn(const T* arr, __m512i& vi, const U&, const int N_proc = N) {
      //_mm512_prefetch_i32gather_ps(vi, arr, 1, _MM_HINT_T0);

      const __m512 src = {0};
      const __mmask16 k = N_proc == N ? -1 : (1 << N_proc) - 1;

      for (int i = 0; i < kSize; ++i, ++arr) {
        //_mm512_prefetch_i32gather_ps(vi, arr+2, 1, _MM_HINT_NTA);

        __m512 reg = _mm512_mask_i32gather_ps(src, k, vi, arr, sizeof(U));
        _mm512_mask_store_ps(&fArray[i * N], k, reg);
      }
    }

    // Experimental methods, slurpIn() seems to be at least as fast.
    // See comments in mkFit/MkFitter.cc MkFitter::addBestHit().

    void ChewIn(const char* arr, int off, int vi[N], const char* tmp, __m512i& ui) {
      // This is a hack ... we know sizeof(Hit) = 64 = cache line = vector width.

      for (int i = 0; i < N; ++i) {
        __m512 reg = _mm512_load_ps(arr + vi[i]);
        _mm512_store_ps((void*)(tmp + 64 * i), reg);
      }

      for (int i = 0; i < kSize; ++i) {
        __m512 reg = _mm512_i32gather_ps(ui, tmp + off + i * sizeof(T), 1);
        _mm512_store_ps(&fArray[i * N], reg);
      }
    }

    void Contaginate(const char* arr, int vi[N], const char* tmp) {
      // This is a hack ... we know sizeof(Hit) = 64 = cache line = vector width.

      for (int i = 0; i < N; ++i) {
        __m512 reg = _mm512_load_ps(arr + vi[i]);
        _mm512_store_ps((void*)(tmp + 64 * i), reg);
      }
    }

    void Plexify(const char* tmp, __m512i& ui) {
      for (int i = 0; i < kSize; ++i) {
        __m512 reg = _mm512_i32gather_ps(ui, tmp + i * sizeof(T), 1);
        _mm512_store_ps(&fArray[i * N], reg);
      }
    }

#elif defined(AVX2_INTRINSICS)

    template <typename U>
    void slurpIn(const T* arr, __m256i& vi, const U&, const int N_proc = N) {
      const __m256 src = {0};

      __m256i k = _mm256_setr_epi32(0, 1, 2, 3, 4, 5, 6, 7);
      __m256i k_sel = _mm256_set1_epi32(N_proc);
      __m256i k_master = _mm256_cmpgt_epi32(k_sel, k);

      k = k_master;
      for (int i = 0; i < kSize; ++i, ++arr) {
        __m256 reg = _mm256_mask_i32gather_ps(src, arr, vi, (__m256)k, sizeof(U));
        // Restore mask (docs say gather clears it but it doesn't seem to).
        k = k_master;
        _mm256_maskstore_ps(&fArray[i * N], k, reg);
      }
    }

#else

    void slurpIn(const T* arr, int vi[N], const int N_proc = N) {
      // Separate N_proc == N case (gains about 7% in fit test).
      if (N_proc == N) {
        for (int i = 0; i < kSize; ++i) {
          for (int j = 0; j < N; ++j) {
            fArray[i * N + j] = *(arr + i + vi[j]);
          }
        }
      } else {
        for (int i = 0; i < kSize; ++i) {
          for (int j = 0; j < N_proc; ++j) {
            fArray[i * N + j] = *(arr + i + vi[j]);
          }
        }
      }
    }

#endif

    void copyOut(idx_t n, T* arr) const {
      for (idx_t i = n; i < kTotSize; i += N) {
        *(arr++) = fArray[i];
      }
    }

    void setDiagonal3x3(idx_t n, T d) {
      T* p = fArray + n;

      p[0 * N] = d;
      p[1 * N] = 0;
      p[2 * N] = d;
      p[3 * N] = 0;
      p[4 * N] = 0;
      p[5 * N] = d;
    }

    MatriplexSym& subtract(const MatriplexSym& a, const MatriplexSym& b) {
      // Does *this = a - b;

#pragma omp simd
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] = a.fArray[i] - b.fArray[i];
      }

      return *this;
    }

    // ==================================================================
    // Operations specific to Kalman fit in 6 parameter space
    // ==================================================================

    void addNoiseIntoUpperLeft3x3(T noise) {
      T* p = fArray;
      ASSUME_ALIGNED(p, 64);

#pragma omp simd
      for (idx_t n = 0; n < N; ++n) {
        p[0 * N + n] += noise;
        p[2 * N + n] += noise;
        p[5 * N + n] += noise;
      }
    }

    void invertUpperLeft3x3() {
      typedef T TT;

      T* a = fArray;
      ASSUME_ALIGNED(a, 64);

#pragma omp simd
      for (idx_t n = 0; n < N; ++n) {
        const TT c00 = a[2 * N + n] * a[5 * N + n] - a[4 * N + n] * a[4 * N + n];
        const TT c01 = a[4 * N + n] * a[3 * N + n] - a[1 * N + n] * a[5 * N + n];
        const TT c02 = a[1 * N + n] * a[4 * N + n] - a[2 * N + n] * a[3 * N + n];
        const TT c11 = a[5 * N + n] * a[0 * N + n] - a[3 * N + n] * a[3 * N + n];
        const TT c12 = a[3 * N + n] * a[1 * N + n] - a[4 * N + n] * a[0 * N + n];
        const TT c22 = a[0 * N + n] * a[2 * N + n] - a[1 * N + n] * a[1 * N + n];

        // Force determinant calculation in double precision.
        const double det = (double)a[0 * N + n] * c00 + (double)a[1 * N + n] * c01 + (double)a[3 * N + n] * c02;
        const TT s = TT(1) / det;

        a[0 * N + n] = s * c00;
        a[1 * N + n] = s * c01;
        a[2 * N + n] = s * c11;
        a[3 * N + n] = s * c02;
        a[4 * N + n] = s * c12;
        a[5 * N + n] = s * c22;
      }
    }

    Matriplex<T, 1, 1, N> ReduceFixedIJ(idx_t i, idx_t j) const {
      Matriplex<T, 1, 1, N> t;
      for (idx_t n = 0; n < N; ++n) {
        t[n] = constAt(n, i, j);
      }
      return t;
    }
  };

__attribute__() [2/2]

template<typename T , idx_t D1, idx_t D2, idx_t N>

class Matriplex::__attribute__

(

(aligned(32))

)

return no. of matrix rows

return no. of matrix columns

return no of elements: rows*columns

size of the whole matriplex

Definition at line 11 of file Matriplex.h.

References a, abs(), PVValHelper::add(), b, ALPAKA_ACCELERATOR_NAMESPACE::brokenline::constexpr(), filterCSVwithJSON::copy, cos(), hypot(), mps_fire::i, recoMuon::in, dqmiolumiharvest::j, dqmdumpme::k, visualization-live-secondInstance_cfg::m, N, dqmiodumpmetadata::n, operator*=(), operator+=(), operator-(), operator-=(), operator/=(), operator=(), operator[](), pfClustersFromCombinedCaloHF_cfi::scale, sin(), sqr(), sqrt(), TrackRefitter_38T_cff::src, submitPVValidationJobs::t, tan(), createJobs::tmp, mitigatedMETSequence_cff::U, interactiveExample::ui, findQualityFiles::v, and geometryCSVtoXML::xx.

                                                        {
  public:
    typedef T value_type;

    static constexpr int kRows = D1;
    static constexpr int kCols = D2;
    static constexpr int kSize = D1 * D2;
    static constexpr int kTotSize = N * kSize;

    T fArray[kTotSize];

    Matriplex() {}
    Matriplex(T v) { setVal(v); }

    idx_t plexSize() const { return N; }

    void setVal(T v) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] = v;
      }
    }

    void add(const Matriplex& v) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] += v.fArray[i];
      }
    }

    void scale(T scale) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] *= scale;
      }
    }

    T operator[](idx_t xx) const { return fArray[xx]; }
    T& operator[](idx_t xx) { return fArray[xx]; }

    const T& constAt(idx_t n, idx_t i, idx_t j) const { return fArray[(i * D2 + j) * N + n]; }

    T& At(idx_t n, idx_t i, idx_t j) { return fArray[(i * D2 + j) * N + n]; }

    T& operator()(idx_t n, idx_t i, idx_t j) { return fArray[(i * D2 + j) * N + n]; }
    const T& operator()(idx_t n, idx_t i, idx_t j) const { return fArray[(i * D2 + j) * N + n]; }

    Matriplex& operator=(T t) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = t;
      return *this;
    }

    Matriplex& operator+=(T t) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] += t;
      return *this;
    }

    Matriplex& operator-=(T t) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] -= t;
      return *this;
    }

    Matriplex& operator*=(T t) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] *= t;
      return *this;
    }

    Matriplex& operator/=(T t) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] /= t;
      return *this;
    }

    Matriplex& operator+=(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] += a.fArray[i];
      return *this;
    }

    Matriplex& operator-=(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] -= a.fArray[i];
      return *this;
    }

    Matriplex& operator*=(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] *= a.fArray[i];
      return *this;
    }

    Matriplex& operator/=(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] /= a.fArray[i];
      return *this;
    }

    Matriplex operator-() {
      Matriplex t;
      for (idx_t i = 0; i < kTotSize; ++i)
        t.fArray[i] = -fArray[i];
      return t;
    }

    Matriplex& abs(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::abs(a.fArray[i]);
      return *this;
    }
    Matriplex& abs() {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::abs(fArray[i]);
      return *this;
    }

    Matriplex& sqrt(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::sqrt(a.fArray[i]);
      return *this;
    }
    Matriplex& sqrt() {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::sqrt(fArray[i]);
      return *this;
    }

    Matriplex& sqr(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = a.fArray[i] * a.fArray[i];
      return *this;
    }
    Matriplex& sqr() {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = fArray[i] * fArray[i];
      return *this;
    }

    Matriplex& hypot(const Matriplex& a, const Matriplex& b) {
      for (idx_t i = 0; i < kTotSize; ++i) {
        fArray[i] = a.fArray[i] * a.fArray[i] + b.fArray[i] * b.fArray[i];
      }
      return sqrt();
    }

    Matriplex& sin(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::sin(a.fArray[i]);
      return *this;
    }
    Matriplex& sin() {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::sin(fArray[i]);
      return *this;
    }

    Matriplex& cos(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::cos(a.fArray[i]);
      return *this;
    }
    Matriplex& cos() {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::cos(fArray[i]);
      return *this;
    }

    Matriplex& tan(const Matriplex& a) {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::tan(a.fArray[i]);
      return *this;
    }
    Matriplex& tan() {
      for (idx_t i = 0; i < kTotSize; ++i)
        fArray[i] = std::tan(fArray[i]);
      return *this;
    }

    //---------------------------------------------------------

    void copySlot(idx_t n, const Matriplex& m) {
      for (idx_t i = n; i < kTotSize; i += N) {
        fArray[i] = m.fArray[i];
      }
    }

    void copyIn(idx_t n, const T* arr) {
      for (idx_t i = n; i < kTotSize; i += N) {
        fArray[i] = *(arr++);
      }
    }

    void copyIn(idx_t n, const Matriplex& m, idx_t in) {
      for (idx_t i = n; i < kTotSize; i += N, in += N) {
        fArray[i] = m[in];
      }
    }

    void copy(idx_t n, idx_t in) {
      for (idx_t i = n; i < kTotSize; i += N, in += N) {
        fArray[i] = fArray[in];
      }
    }

#if defined(AVX512_INTRINSICS)

    template <typename U>
    void slurpIn(const T* arr, __m512i& vi, const U&, const int N_proc = N) {
      //_mm512_prefetch_i32gather_ps(vi, arr, 1, _MM_HINT_T0);

      const __m512 src = {0};
      const __mmask16 k = N_proc == N ? -1 : (1 << N_proc) - 1;

      for (int i = 0; i < kSize; ++i, ++arr) {
        //_mm512_prefetch_i32gather_ps(vi, arr+2, 1, _MM_HINT_NTA);

        __m512 reg = _mm512_mask_i32gather_ps(src, k, vi, arr, sizeof(U));
        _mm512_mask_store_ps(&fArray[i * N], k, reg);
      }
    }

    // Experimental methods, slurpIn() seems to be at least as fast.
    // See comments in mkFit/MkFitter.cc MkFitter::addBestHit().
    void ChewIn(const char* arr, int off, int vi[N], const char* tmp, __m512i& ui) {
      // This is a hack ... we know sizeof(Hit) = 64 = cache line = vector width.

      for (int i = 0; i < N; ++i) {
        __m512 reg = _mm512_load_ps(arr + vi[i]);
        _mm512_store_ps((void*)(tmp + 64 * i), reg);
      }

      for (int i = 0; i < kSize; ++i) {
        __m512 reg = _mm512_i32gather_ps(ui, tmp + off + i * sizeof(T), 1);
        _mm512_store_ps(&fArray[i * N], reg);
      }
    }

    void Contaginate(const char* arr, int vi[N], const char* tmp) {
      // This is a hack ... we know sizeof(Hit) = 64 = cache line = vector width.

      for (int i = 0; i < N; ++i) {
        __m512 reg = _mm512_load_ps(arr + vi[i]);
        _mm512_store_ps((void*)(tmp + 64 * i), reg);
      }
    }

    void Plexify(const char* tmp, __m512i& ui) {
      for (int i = 0; i < kSize; ++i) {
        __m512 reg = _mm512_i32gather_ps(ui, tmp + i * sizeof(T), 1);
        _mm512_store_ps(&fArray[i * N], reg);
      }
    }

#elif defined(AVX2_INTRINSICS)

    template <typename U>
    void slurpIn(const T* arr, __m256i& vi, const U&, const int N_proc = N) {
      // Casts to float* needed to "support" also T=HitOnTrack.
      // Note that sizeof(float) == sizeof(HitOnTrack) == 4.

      const __m256 src = {0};

      __m256i k = _mm256_setr_epi32(0, 1, 2, 3, 4, 5, 6, 7);
      __m256i k_sel = _mm256_set1_epi32(N_proc);
      __m256i k_master = _mm256_cmpgt_epi32(k_sel, k);

      k = k_master;
      for (int i = 0; i < kSize; ++i, ++arr) {
        __m256 reg = _mm256_mask_i32gather_ps(src, (float*)arr, vi, (__m256)k, sizeof(U));
        // Restore mask (docs say gather clears it but it doesn't seem to).
        k = k_master;
        _mm256_maskstore_ps((float*)&fArray[i * N], k, reg);
      }
    }

#else

    void slurpIn(const T* arr, int vi[N], const int N_proc = N) {
      // Separate N_proc == N case (gains about 7% in fit test).
      if (N_proc == N) {
        for (int i = 0; i < kSize; ++i) {
          for (int j = 0; j < N; ++j) {
            fArray[i * N + j] = *(arr + i + vi[j]);
          }
        }
      } else {
        for (int i = 0; i < kSize; ++i) {
          for (int j = 0; j < N_proc; ++j) {
            fArray[i * N + j] = *(arr + i + vi[j]);
          }
        }
      }
    }

#endif

    void copyOut(idx_t n, T* arr) const {
      for (idx_t i = n; i < kTotSize; i += N) {
        *(arr++) = fArray[i];
      }
    }

    Matriplex<T, 1, 1, N> ReduceFixedIJ(idx_t i, idx_t j) const {
      Matriplex<T, 1, 1, N> t;
      for (idx_t n = 0; n < N; ++n) {
        t[n] = constAt(n, i, j);
      }
      return t;
    }
  };

abs()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::abs

(

const MPlex< T, D1, D2, N > &

a

)

Definition at line 418 of file Matriplex.h.

References a, and submitPVValidationJobs::t.

Referenced by __attribute__().

                                                        {
    MPlex<T, D1, D2, N> t;
    return t.abs(a);
  }

align_check()

void Matriplex::align_check	(	const char *	pref,
		void *	adr
	)

Definition at line 4 of file MatriplexCommon.cc.

Referenced by mkfit::MkFitter::checkAlignment().

                                                {
    printf("%s 0x%llx  -  modulo 64 = %lld\n", pref, (long long unsigned)adr, (long long)adr % 64);
  }

aligned_alloc64()

void* Matriplex::aligned_alloc64 ( std::size_t  size )

inline

Definition at line 13 of file Memory.h.

References aligned_alloc(), and round_up_align64().

Referenced by mkfit::Pool< mkfit::MkFitter >::create(), and Matriplex::MatriplexVector< MP >::MatriplexVector().

{ return std::aligned_alloc(64, round_up_align64(size)); }

cos()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::cos

(

const MPlex< T, D1, D2, N > &

a

)

Definition at line 448 of file Matriplex.h.

References a, and submitPVValidationJobs::t.

Referenced by __attribute__(), and sincos().

                                                        {
    MPlex<T, D1, D2, N> t;
    return t.cos(a);
  }

hypot()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::hypot	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 436 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

Referenced by __attribute__(), mkfit::MkFinder::addBestHit(), RegressionHelper::applyCombinationRegression(), objects.METAnalyzer.METAnalyzer::applyDeltaMet(), mkfit::MkFinder::bkFitFitTracks(), mkfit::MkFinder::bkFitFitTracksBH(), PFEGammaAlgo::calculateEleMVA(), PhotonEnergyCalibratorRun2::calibrate(), ElectronEnergyCalibratorRun2::calibrate(), mkfit::TrackBase::canReachRadius(), MuonGEMBaseHarvestor::computeEfficiency(), DQMGenericClient::computeEfficiency(), JetReCalibrator.Type1METCorrector::correct(), mkfit::TrackBase::d0BeamSpot(), L1JetRecoTreeProducer::doPFMetNoMu(), L1JetRecoTreeProducer::doPUPPIMetNoMu(), HLTRegionalEcalResonanceFilter::doSelection(), pat::PATMuonProducer::embedHighLevel(), pat::PATElectronProducer::embedHighLevel(), trklet::TrackletEventProcessor::event(), Phase2TrackerMonitorDigi::fillITPixelDigiHistos(), LHETablesProducer::fillLHEObjectTable(), Phase2TrackerMonitorDigi::fillOTDigiHistos(), Phase2TrackerValidateDigi::fillSimHitInfo(), mkfit::MkFinder::findCandidates(), GEMEfficiencyAnalyzer::findCSCSegmentCosmics(), GenParticles2HepMCConverter::FourVector(), Point::GetSigmaDeltaMu(), mkfit::kalmanOperation(), mkfit::kalmanPropagateAndComputeChi2(), mkfit::kalmanPropagateAndUpdate(), objects.METAnalyzer::makeGenTkMet(), mkfit::TrackBase::maxReachRadius(), LowPtElectronModifier::modifyObject(), l1tpf::ParametricResolution::operator()(), reco::parser::hypot_f::operator()(), mkfit::MkFitter::printPt(), DeepMETProducer::produce(), DeepMETSonicProducer::produce(), L1TPFMetNoMuProducer::produce(), PseudoTopProducer::produce(), PATTracksToPackedCandidates::produce(), EvtPlaneProducer::produce(), L1FPGATrackProducer::produce(), mkfit::mini_propagators::InitialState::propagate_to_r(), mkfit::mini_propagators::InitialStatePlex::propagate_to_r(), mkfit::propagateHelixToPlaneMPlex(), mkfit::propagateHelixToZMPlex(), mkfit::MkBase::propagateTracksToHitR(), mkfit::MkBase::propagateTracksToPCAZ(), trklet::L1TStub::r(), mkfit::TrackBase::rAtZ(), EcalUncalibRecHitWorkerMultiFit::run(), pf2pat::IPCutPFCandidateSelectorDefinition::select(), MultiTrackSelector::select(), HIMultiTrackSelector::select(), mkfit::MkFinder::selectHitIndices(), mkfit::MkFinder::selectHitIndicesV2(), pat::LeptonUpdater< T >::setDZ(), EGEtScaleSysModifier::setEcalEnergy(), ElectronEnergyCalibrator::setEcalEnergy(), PhotonEnergyCalibrator::setEnergyAndSystVarations(), JetReCalibrator::setFakeRawMETOnOldMiniAODs(), BJetId::setNNVectorVar(), pat::MET::shiftedP4(), pat::MET::shiftedP4_74x(), XHistogram::splitSegment(), DD4hep_XHistogram::splitSegment(), objects.METAnalyzer::sumXY(), reco::ForwardProton::t(), and mkfit::TrackBase::zAtR().

                                                                                        {
    MPlex<T, D1, D2, N> t;
    return t.hypot(a, b);
  }

invertCholesky() [1/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCholesky	(	MPlexVec< MPlex< T, D, D, N >> &	A,
		int	n_to_process = 0
	)

Definition at line 126 of file MatriplexVector.h.

References A, mps_fire::i, invertCholesky(), and np.

                                                                            {
    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      invertCholesky(A[i]);
    }
  }

invertCholesky() [2/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCholesky

(

MPlex< T, D, D, N > &

A

)

Definition at line 784 of file Matriplex.h.

References A, and Matriplex::CholeskyInverter< T, D, N >::invert().

Referenced by invertCholesky().

                                            {
    CholeskyInverter<T, D, N>::invert(A);
  }

invertCholeskySym() [1/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCholeskySym	(	MPlexVec< MPlexSym< T, D, N >> &	A,
		int	n_to_process = 0
	)

Definition at line 144 of file MatriplexVector.h.

References A, mps_fire::i, invertCholeskySym(), and np.

                                                                               {
    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      invertCholeskySym(A[i]);
    }
  }

invertCholeskySym() [2/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCholeskySym

(

MPlexSym< T, D, N > &

A

)

Definition at line 471 of file MatriplexSym.h.

References A, and Matriplex::CholeskyInverterSym< T, D, N >::invert().

Referenced by invertCholeskySym().

                                               {
    CholeskyInverterSym<T, D, N>::invert(A);
  }

invertCramer() [1/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCramer	(	MPlexVec< MPlex< T, D, D, N >> &	A,
		int	n_to_process = 0
	)

Definition at line 117 of file MatriplexVector.h.

References A, mps_fire::i, invertCramer(), and np.

                                                                          {
    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      invertCramer(A[i]);
    }
  }

invertCramer() [2/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCramer	(	MPlex< T, D, D, N > &	A,
		double *	determ = nullptr
	)

Definition at line 730 of file Matriplex.h.

References A, and Matriplex::CramerInverter< T, D, N >::invert().

Referenced by mkfit::conformalFitMPlex(), and invertCramer().

                                                                    {
    CramerInverter<T, D, N>::invert(A, determ);
  }

invertCramerSym() [1/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCramerSym	(	MPlexVec< MPlexSym< T, D, N >> &	A,
		int	n_to_process = 0
	)

Definition at line 135 of file MatriplexVector.h.

References A, mps_fire::i, invertCramerSym(), and np.

                                                                             {
    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      invertCramerSym(A[i]);
    }
  }

invertCramerSym() [2/2]

template<typename T , idx_t D, idx_t N>

void Matriplex::invertCramerSym	(	MPlexSym< T, D, N > &	A,
		double *	determ = nullptr
	)

Definition at line 420 of file MatriplexSym.h.

References A, and Matriplex::CramerInverterSym< T, D, N >::invert().

Referenced by invertCramerSym(), mkfit::kalmanOperation(), mkfit::kalmanOperationEndcap(), and mkfit::kalmanOperationPlane().

                                                                       {
    CramerInverterSym<T, D, N>::invert(A, determ);
  }

max()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::max	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 488 of file Matriplex.h.

References a, b, mps_fire::i, and submitPVValidationJobs::t.

Referenced by min_max().

                                                                                      {
    MPlex<T, D1, D2, N> t;
    for (idx_t i = 0; i < a.kTotSize; ++i) {
      t.fArray[i] = std::max(a.fArray[i], b.fArray[i]);
    }
    return t;
  }

min()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::min	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 479 of file Matriplex.h.

References a, b, mps_fire::i, and submitPVValidationJobs::t.

Referenced by min_max().

                                                                                      {
    MPlex<T, D1, D2, N> t;
    for (idx_t i = 0; i < a.kTotSize; ++i) {
      t.fArray[i] = std::min(a.fArray[i], b.fArray[i]);
    }
    return t;
  }

min_max()

template<typename T , idx_t D1, idx_t D2, idx_t N>

void Matriplex::min_max	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b,
		MPlex< T, D1, D2, N > &	min,
		MPlex< T, D1, D2, N > &	max
	)

Definition at line 468 of file Matriplex.h.

References a, b, mps_fire::i, max(), and min().

Referenced by mkfit::MkFinder::selectHitIndicesV2().

                                         {
    for (idx_t i = 0; i < a.kTotSize; ++i) {
      min.fArray[i] = std::min(a.fArray[i], b.fArray[i]);
      max.fArray[i] = std::max(a.fArray[i], b.fArray[i]);
    }
  }

multiply() [1/4]

template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>

void Matriplex::multiply	(	const MPlexVec< MPlex< T, D1, D2, N >> &	A,
		const MPlexVec< MPlex< T, D2, D3, N >> &	B,
		MPlexVec< MPlex< T, D1, D3, N >> &	C,
		int	n_to_process = 0
	)

Definition at line 53 of file MatriplexVector.h.

References A, cms::cuda::assert(), B, correctionTermsCaloMet_cff::C, mps_fire::i, multiply(), and np.

                                      {
    assert(A.size() == B.size());
    assert(A.size() == C.size());

    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      multiply(A[i], B[i], C[i]);
    }
  }

multiply() [2/4]

template<typename T , idx_t D, idx_t N>

void Matriplex::multiply	(	const MPlexVec< MPlexSym< T, D, N >> &	A,
		const MPlexVec< MPlexSym< T, D, N >> &	B,
		MPlexVec< MPlex< T, D, D, N >> &	C,
		int	n_to_process = 0
	)

Definition at line 100 of file MatriplexVector.h.

References A, cms::cuda::assert(), B, correctionTermsCaloMet_cff::C, mps_fire::i, multiply(), and np.

                                      {
    assert(A.size() == B.size());
    assert(A.size() == C.size());

    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      multiply(A[i], B[i], C[i]);
    }
  }

multiply() [3/4]

template<typename T , idx_t D, idx_t N>

void Matriplex::multiply	(	const MPlexSym< T, D, N > &	A,
		const MPlexSym< T, D, N > &	B,
		MPlex< T, D, D, N > &	C
	)

Definition at line 347 of file MatriplexSym.h.

References A, B, correctionTermsCaloMet_cff::C, and Matriplex::SymMultiplyCls< T, D, N >::multiply().

                                                                                              {
    SymMultiplyCls<T, D, N>::multiply(A, B, C);
  }

multiply() [4/4]

template<typename T , idx_t D, idx_t N>

void Matriplex::multiply	(	const MPlex< T, D, D, N > &	A,
		const MPlex< T, D, D, N > &	B,
		MPlex< T, D, D, N > &	C
	)

Definition at line 646 of file Matriplex.h.

References A, B, correctionTermsCaloMet_cff::C, Matriplex::MultiplyCls< T, D, N >::multiply(), and N.

Referenced by multiply(), multiply3in(), and AlignmentExtendedCorrelationsEntry::operator*=().

                                                                                              {
#ifdef DEBUG
    printf("Multipl %d %d\n", D, N);
#endif

    MultiplyCls<T, D, N>::multiply(A, B, C);
  }

multiply3in()

template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>

void Matriplex::multiply3in	(	MPlexVec< MPlex< T, D1, D2, N >> &	A,
		MPlexVec< MPlex< T, D2, D3, N >> &	B,
		MPlexVec< MPlex< T, D1, D3, N >> &	C,
		int	n_to_process = 0
	)

Definition at line 83 of file MatriplexVector.h.

References A, cms::cuda::assert(), B, correctionTermsCaloMet_cff::C, mps_fire::i, multiply(), and np.

                                         {
    assert(A.size() == B.size());
    assert(A.size() == C.size());

    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      multiply(A[i], B[i], C[i]);
      multiply(B[i], C[i], A[i]);
      multiply(C[i], A[i], B[i]);
    }
  }

multiplyGeneral() [1/2]

template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>

void Matriplex::multiplyGeneral	(	const MPlexVec< MPlex< T, D1, D2, N >> &	A,
		const MPlexVec< MPlex< T, D2, D3, N >> &	B,
		MPlexVec< MPlex< T, D1, D3, N >> &	C,
		int	n_to_process = 0
	)

Definition at line 68 of file MatriplexVector.h.

References A, cms::cuda::assert(), B, correctionTermsCaloMet_cff::C, mps_fire::i, multiplyGeneral(), and np.

                                             {
    assert(A.size() == B.size());
    assert(A.size() == C.size());

    const int np = n_to_process ? n_to_process : A.size();

    for (int i = 0; i < np; ++i) {
      multiplyGeneral(A[i], B[i], C[i]);
    }
  }

multiplyGeneral() [2/2]

template<typename T , idx_t D1, idx_t D2, idx_t D3, idx_t N>

void Matriplex::multiplyGeneral	(	const MPlex< T, D1, D2, N > &	A,
		const MPlex< T, D2, D3, N > &	B,
		MPlex< T, D1, D3, N > &	C
	)

Definition at line 501 of file Matriplex.h.

References A, B, correctionTermsCaloMet_cff::C, trklet::D3, mps_fire::i, dqmiolumiharvest::j, dqmdumpme::k, N, and dqmiodumpmetadata::n.

Referenced by multiplyGeneral().

                                                                                                           {
    for (idx_t i = 0; i < D1; ++i) {
      for (idx_t j = 0; j < D3; ++j) {
        const idx_t ijo = N * (i * D3 + j);

#pragma omp simd
        for (idx_t n = 0; n < N; ++n) {
          C.fArray[ijo + n] = 0;
        }

        for (idx_t k = 0; k < D2; ++k) {
          const idx_t iko = N * (i * D2 + k);
          const idx_t kjo = N * (k * D3 + j);

#pragma omp simd
          for (idx_t n = 0; n < N; ++n) {
            C.fArray[ijo + n] += A.fArray[iko + n] * B.fArray[kjo + n];
          }
        }
      }
    }
  }

operator*() [1/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator*	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 348 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                                            {
    MPlex<T, D1, D2, N> t = a;
    t *= b;
    return t;
  }

operator*() [2/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator*	(	const MPlex< T, D1, D2, N > &	a,
		T	b
	)

Definition at line 376 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t *= b;
    return t;
  }

operator*() [3/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator*	(	T	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 404 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t *= b;
    return t;
  }

operator+() [1/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator+	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 334 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                                            {
    MPlex<T, D1, D2, N> t = a;
    t += b;
    return t;
  }

operator+() [2/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator+	(	const MPlex< T, D1, D2, N > &	a,
		T	b
	)

Definition at line 362 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t += b;
    return t;
  }

operator+() [3/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator+	(	T	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 390 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t += b;
    return t;
  }

operator-() [1/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator-	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 341 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

Referenced by __attribute__().

                                                                                            {
    MPlex<T, D1, D2, N> t = a;
    t -= b;
    return t;
  }

operator-() [2/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator-	(	const MPlex< T, D1, D2, N > &	a,
		T	b
	)

Definition at line 369 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t -= b;
    return t;
  }

operator-() [3/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator-	(	T	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 397 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t -= b;
    return t;
  }

operator/() [1/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator/	(	const MPlex< T, D1, D2, N > &	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 355 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                                            {
    MPlex<T, D1, D2, N> t = a;
    t /= b;
    return t;
  }

operator/() [2/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator/	(	const MPlex< T, D1, D2, N > &	a,
		T	b
	)

Definition at line 383 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t /= b;
    return t;
  }

operator/() [3/3]

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::operator/	(	T	a,
		const MPlex< T, D1, D2, N > &	b
	)

Definition at line 411 of file Matriplex.h.

References a, b, and submitPVValidationJobs::t.

                                                                   {
    MPlex<T, D1, D2, N> t = a;
    t /= b;
    return t;
  }

round_up_align64()

constexpr std::size_t Matriplex::round_up_align64

(

std::size_t

size

)

Definition at line 8 of file Memory.h.

References ALPAKA_ACCELERATOR_NAMESPACE::brokenline::constexpr(), and ALPAKA_ACCELERATOR_NAMESPACE::pixelClustering::pixelStatus::mask.

Referenced by aligned_alloc64().

                                                       {
    constexpr std::size_t mask = 64 - 1;
    return size & mask ? (size & ~mask) + 64 : size;
  }

sin()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::sin

(

const MPlex< T, D1, D2, N > &

a

)

Definition at line 442 of file Matriplex.h.

References a, and submitPVValidationJobs::t.

Referenced by __attribute__(), and sincos().

                                                        {
    MPlex<T, D1, D2, N> t;
    return t.sin(a);
  }

sincos()

template<typename T , idx_t D1, idx_t D2, idx_t N>

void Matriplex::sincos	(	const MPlex< T, D1, D2, N > &	a,
		MPlex< T, D1, D2, N > &	s,
		MPlex< T, D1, D2, N > &	c
	)

Definition at line 454 of file Matriplex.h.

References a, HltBtagPostValidation_cff::c, cos(), mps_fire::i, alignCSCRings::s, and sin().

                                                                                            {
    for (idx_t i = 0; i < a.kTotSize; ++i) {
      s.fArray[i] = std::sin(a.fArray[i]);
      c.fArray[i] = std::cos(a.fArray[i]);
    }
  }

sqr()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::sqr

(

const MPlex< T, D1, D2, N > &

a

)

Definition at line 430 of file Matriplex.h.

References a, and submitPVValidationJobs::t.

Referenced by __attribute__().

                                                        {
    MPlex<T, D1, D2, N> t;
    return t.sqr(a);
  }

sqrt()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::sqrt

(

const MPlex< T, D1, D2, N > &

a

)

Definition at line 424 of file Matriplex.h.

References a, and submitPVValidationJobs::t.

Referenced by __attribute__(), Matriplex::CholeskyInverter< T, 3, N >::invert(), and mkfit::MkFinder::selectHitIndicesV2().

                                                         {
    MPlex<T, D1, D2, N> t;
    return t.sqrt(a);
  }

tan()

template<typename T , idx_t D1, idx_t D2, idx_t N>

MPlex<T, D1, D2, N> Matriplex::tan

(

const MPlex< T, D1, D2, N > &

a

)

Definition at line 462 of file Matriplex.h.

References a, and submitPVValidationJobs::t.

Referenced by __attribute__().

                                                        {
    MPlex<T, D1, D2, N> t;
    return t.tan(a);
  }

Variable Documentation

gSymOffsets

const idx_t Matriplex::gSymOffsets[7][36]

Initial value:

= {{},
                                    {},
                                    {0, 1, 1, 2},
                                    {0, 1, 3, 1, 2, 4, 3, 4, 5},  
                                    {},
                                    {},
                                    {0, 1, 3, 6, 10, 15, 1,  2,  4,  7,  11, 16, 3,  4,  5,  8,  12, 17,
                                     6, 7, 8, 9, 13, 18, 10, 11, 12, 13, 14, 19, 15, 16, 17, 18, 19, 20}}

Definition at line 13 of file MatriplexSym.h.

Referenced by __attribute__().