#ifndef ROOFIT_BATCHCOMPUTE_ROOBATCHCOMPUTE_H
#define ROOFIT_BATCHCOMPUTE_ROOBATCHCOMPUTE_H

#include "RooSpan.h"
#include "RooVDTHeaders.h"
#include "RunContext.h"
#include "BracketAdapter.h"
#include "DllImport.h" //for R__EXTERN, needed for windows

class RooAbsReal;
class RooListProxy;

/**
 * Namespace for dispatching RooFit computations to various backends.
 *
 * This namespace contains an interface for providing high-performance computation functions for use in RooAbsReal::evaluateSpan(),
 * see RooBatchComputeInterface.
 *
 * Furthermore, several implementations of this interface can be created, which reside in RooBatchCompute::RF_ARCH, where
 * RF_ARCH may be replaced by the architecture that this implementation targets, e.g. SSE, AVX, etc.
 *
 * Using the pointer RooBatchCompute::dispatch, a computation request can be dispatched to the fastest backend that is available
 * on a specific platform.
 */
namespace RooBatchCompute {
  /**
   * \brief The interface which should be implemented to provide optimised computation functions for implementations of RooAbsReal::evaluateSpan().
   *
   * This interface contains the signatures of the compute functions of every PDF that has an optimised implementation available.
   * These are the functions that perform the actual computations in batches.
   *
   * Several implementations of this interface may be provided, e.g. SSE, AVX, AVX2 etc. At run time, the fastest implementation of this interface
   * is selected, and using a virtual call, the computation is dispatched to the best backend.
   *
   * \see RooBatchCompute::dispatch, RooBatchComputeClass, RF_ARCH
   */ 
  class RooBatchComputeInterface {
  public:
    virtual ~RooBatchComputeInterface() = default;
    virtual RooSpan<double> computeArgusBG(const RooAbsReal*, RunContext&, RooSpan<const double> m, RooSpan<const double> m0, RooSpan<const double> c, RooSpan<const double> p) = 0;
    virtual void computeBernstein(size_t batchSize, double * __restrict output, const double * __restrict const xData, double xmin, double xmax, std::vector<double> coef) = 0;
    virtual RooSpan<double> computeBifurGauss(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> mean, RooSpan<const double> sigmaL, RooSpan<const double> sigmaR) = 0;
    virtual RooSpan<double> computeBukin(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> Xp, RooSpan<const double> sigp, RooSpan<const double> xi, RooSpan<const double> rho1, RooSpan<const double> rho2) = 0;
    virtual RooSpan<double> computeBreitWigner(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> mean, RooSpan<const double> width) = 0;
    virtual RooSpan<double> computeCBShape(const RooAbsReal*, RunContext&, RooSpan<const double> m, RooSpan<const double> m0, RooSpan<const double> sigma, RooSpan<const double> alpha, RooSpan<const double> n) = 0;
    virtual void computeChebychev(size_t batchSize, double * __restrict output, const double * __restrict const xData, double xmin, double xmax, std::vector<double> coef) = 0;
    virtual RooSpan<double> computeChiSquare(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> ndof) = 0;
    virtual RooSpan<double> computeDstD0BG(const RooAbsReal*, RunContext&, RooSpan<const double> dm, RooSpan<const double> dm0, RooSpan<const double> C, RooSpan<const double> A, RooSpan<const double> B) = 0;
    virtual RooSpan<double> computeExponential(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> c) = 0;
    virtual RooSpan<double> computeGamma(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> gamma, RooSpan<const double> beta, RooSpan<const double> mu) = 0;
    virtual RooSpan<double> computeGaussian(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> mean, RooSpan<const double> sigma) = 0;
    virtual RooSpan<double> computeJohnson(const RooAbsReal*, RunContext&, RooSpan<const double> mass, RooSpan<const double> mu, RooSpan<const double> lambda, RooSpan<const double> gamma, RooSpan<const double> delta, double massThreshold) = 0;
    virtual RooSpan<double> computeLandau(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> mean, RooSpan<const double> sigma) = 0;
    virtual RooSpan<double> computeLognormal(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> m0, RooSpan<const double> k) = 0;
    virtual RooSpan<double> computeNovosibirsk(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> peak, RooSpan<const double> width, RooSpan<const double> tail) = 0;
    virtual RooSpan<double> computePoisson(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> mean, bool protectNegative, bool noRounding) = 0;
    virtual void computePolynomial(size_t batchSize, double * __restrict output, const double * __restrict const xData, int lowestOrder, std::vector<BracketAdapterWithMask> &coef) = 0;
    virtual RooSpan<double> computeVoigtian(const RooAbsReal*, RunContext&, RooSpan<const double> x, RooSpan<const double> mean, RooSpan<const double> width, RooSpan<const double> sigma) = 0;
  };

  /**
   * This dispatch pointer points to an implementation of the compute library, provided one has been loaded.
   * Using a virtual call, computation requests are dispatched to backends with architecture-specific functions
   * such as SSE, AVX, AVX2, etc.
   *
   * \see RooBatchComputeInterface, RooBatchComputeClass, RF_ARCH
   */
  R__EXTERN RooBatchComputeInterface* dispatch;
}

#endif