// license:GPL-2.0+
// copyright-holders:Couriersud
/*
* nld_ms_direct.h
*
*/
#ifndef NLD_MS_DIRECT_H_
#define NLD_MS_DIRECT_H_
#include <algorithm>
#include "solver/nld_solver.h"
#include "solver/nld_matrix_solver.h"
#include "solver/vector_base.h"
/* Disabling dynamic allocation gives a ~10% boost in performance
* This flag has been added to support continuous storage for arrays
* going forward in case we implement cuda solvers in the future.
*/
#define NL_USE_DYNAMIC_ALLOCATION (0)
#define TEST_PARALLEL (0)
#if TEST_PARALLEL
#include <thread>
#include <atomic>
#endif
namespace netlist
{
namespace devices
{
//#define nl_ext_double _float128 // slow, very slow
//#define nl_ext_double long double // slightly slower
#define nl_ext_double nl_double
#if TEST_PARALLEL
#define MAXTHR 10
static const int num_thr = 1;
struct thr_intf
{
virtual void do_work(const int id, void *param) = 0;
};
struct ti_t
{
volatile std::atomic<int> lo;
thr_intf *intf;
void *params;
// int block[29]; /* make it 256 bytes */
};
static ti_t ti[MAXTHR];
static std::thread thr[MAXTHR];
int thr_init = 0;
static void thr_process_proc(int id)
{
while (true)
{
while (ti[id].lo.load() == 0)
;
if (ti[id].lo.load() == 2)
return;
ti[id].intf->do_work(id, ti[id].params);
ti[id].lo.store(0);
}
}
static void thr_process(int id, thr_intf *intf, void *params)
{
ti[id].intf = intf;
ti[id].params = params;
ti[id].lo.store(1);
}
static void thr_wait()
{
int c=1;
while (c > 0)
{
c=0;
for (int i=0; i<num_thr; i++)
c += ti[i].lo.load();
}
}
static void thr_initialize()
{
thr_init++;
if (thr_init == 1)
{
for (int i=0; i<num_thr; i++)
{
ti[i].lo = 0;
thr[i] = std::thread(thr_process_proc, i);
}
}
}
static void thr_dispose()
{
thr_init--;
if (thr_init == 0)
{
for (int i=0; i<num_thr; i++)
ti[i].lo = 2;
for (int i=0; i<num_thr; i++)
thr[i].join();
}
}
#endif
template <unsigned m_N, unsigned storage_N>
#if TEST_PARALLEL
class matrix_solver_direct_t: public matrix_solver_t, public thr_intf
#else
class matrix_solver_direct_t: public matrix_solver_t
#endif
{
friend class matrix_solver_t;
public:
matrix_solver_direct_t(netlist_t &anetlist, const pstring &name, const solver_parameters_t *params, const int size);
matrix_solver_direct_t(netlist_t &anetlist, const pstring &name, const eSortType sort, const solver_parameters_t *params, const int size);
virtual ~matrix_solver_direct_t();
virtual void vsetup(analog_net_t::list_t &nets) override;
virtual void reset() override { matrix_solver_t::reset(); }
protected:
virtual int vsolve_non_dynamic(const bool newton_raphson) override;
int solve_non_dynamic(const bool newton_raphson);
inline unsigned N() const { if (m_N == 0) return m_dim; else return m_N; }
void LE_solve();
template <typename T>
void LE_back_subst(T * RESTRICT x);
#if TEST_PARALLEL
int x_i[10];
int x_start[10];
int x_stop[10];
virtual void do_work(const int id, void *param) override;
#endif
#if (NL_USE_DYNAMIC_ALLOCATION)
template <typename T1, typename T2>
inline nl_ext_double &A(const T1 &r, const T2 &c) { return m_A[r * m_pitch + c]; }
template <typename T1>
inline nl_ext_double &RHS(const T1 &r) { return m_A[r * m_pitch + N()]; }
#else
template <typename T1, typename T2>
inline nl_ext_double &A(const T1 &r, const T2 &c) { return m_A[r][c]; }
template <typename T1>
inline nl_ext_double &RHS(const T1 &r) { return m_A[r][N()]; }
#endif
nl_double m_last_RHS[storage_N]; // right hand side - contains currents
private:
static const std::size_t m_pitch = (((storage_N + 1) + 7) / 8) * 8;
//static const std::size_t m_pitch = (((storage_N + 1) + 15) / 16) * 16;
//static const std::size_t m_pitch = (((storage_N + 1) + 31) / 32) * 32;
#if (NL_USE_DYNAMIC_ALLOCATION)
nl_ext_double * RESTRICT m_A;
#else
nl_ext_double m_A[storage_N][m_pitch];
#endif
//nl_ext_double m_RHSx[storage_N];
const unsigned m_dim;
};
// ----------------------------------------------------------------------------------------
// matrix_solver_direct
// ----------------------------------------------------------------------------------------
template <unsigned m_N, unsigned storage_N>
matrix_solver_direct_t<m_N, storage_N>::~matrix_solver_direct_t()
{
#if (NL_USE_DYNAMIC_ALLOCATION)
pfree_array(m_A);
#endif
#if TEST_PARALLEL
thr_dispose();
#endif
}
template <unsigned m_N, unsigned storage_N>
void matrix_solver_direct_t<m_N, storage_N>::vsetup(analog_net_t::list_t &nets)
{
if (m_dim < nets.size())
log().fatal("Dimension {1} less than {2}", m_dim, nets.size());
matrix_solver_t::setup_base(nets);
/* add RHS element */
for (unsigned k = 0; k < N(); k++)
{
terms_t * t = m_terms[k];
if (!plib::container::contains(t->m_nzrd, N()))
t->m_nzrd.push_back(N());
}
netlist().save(*this, m_last_RHS, "m_last_RHS");
for (unsigned k = 0; k < N(); k++)
netlist().save(*this, RHS(k), plib::pfmt("RHS.{1}")(k));
}
#if TEST_PARALLEL
template <unsigned m_N, unsigned storage_N>
void matrix_solver_direct_t<m_N, storage_N>::do_work(const int id, void *param)
{
const int i = x_i[id];
/* FIXME: Singular matrix? */
const nl_double f = 1.0 / A(i,i);
const unsigned * RESTRICT const p = m_terms[i]->m_nzrd.data();
const unsigned e = m_terms[i]->m_nzrd.size();
/* Eliminate column i from row j */
const unsigned * RESTRICT const pb = m_terms[i]->m_nzbd.data();
const unsigned sj = x_start[id];
const unsigned se = x_stop[id];
for (unsigned jb = sj; jb < se; jb++)
{
const unsigned j = pb[jb];
const nl_double f1 = - A(j,i) * f;
for (unsigned k = 0; k < e; k++)
A(j,p[k]) += A(i,p[k]) * f1;
}
}
#endif
template <unsigned m_N, unsigned storage_N>
void matrix_solver_direct_t<m_N, storage_N>::LE_solve()
{
const unsigned kN = N();
for (unsigned i = 0; i < kN; i++) {
// FIXME: use a parameter to enable pivoting? m_pivot
if (!TEST_PARALLEL && m_params.m_pivot)
{
/* Find the row with the largest first value */
unsigned maxrow = i;
for (unsigned j = i + 1; j < kN; j++)
{
//if (std::abs(m_A[j][i]) > std::abs(m_A[maxrow][i]))
if (A(j,i) * A(j,i) > A(maxrow,i) * A(maxrow,i))
maxrow = j;
}
if (maxrow != i)
{
/* Swap the maxrow and ith row */
for (unsigned k = 0; k < kN + 1; k++) {
std::swap(A(i,k), A(maxrow,k));
}
//std::swap(RHS(i), RHS(maxrow));
}
/* FIXME: Singular matrix? */
const nl_double f = 1.0 / A(i,i);
/* Eliminate column i from row j */
for (unsigned j = i + 1; j < kN; j++)
{
const nl_double f1 = - A(j,i) * f;
if (f1 != NL_FCONST(0.0))
{
const nl_double * RESTRICT pi = &A(i,i+1);
nl_double * RESTRICT pj = &A(j,i+1);
#if 1
vec_add_mult_scalar(kN-i,pi,f1,pj);
#else
vec_add_mult_scalar(kN-i-1,pj,f1,pi);
//for (unsigned k = i+1; k < kN; k++)
// pj[k] = pj[k] + pi[k] * f1;
//for (unsigned k = i+1; k < kN; k++)
//A(j,k) += A(i,k) * f1;
RHS(j) += RHS(i) * f1;
#endif
}
}
}
else
{
#if TEST_PARALLEL
const unsigned eb = m_terms[i]->m_nzbd.size();
if (eb > 0)
{
//printf("here %d\n", eb);
unsigned chunks = (eb + num_thr) / (num_thr + 1);
for (int p=0; p < num_thr + 1; p++)
{
x_i[p] = i;
x_start[p] = chunks * p;
x_stop[p] = std::min(chunks*(p+1), eb);
if (p<num_thr && x_start[p] < x_stop[p]) thr_process(p, this, nullptr);
}
if (x_start[num_thr] < x_stop[num_thr])
do_work(num_thr, nullptr);
thr_wait();
}
else if (eb > 0)
{
x_i[0] = i;
x_start[0] = 0;
x_stop[0] = eb;
do_work(0, nullptr);
}
#else
/* FIXME: Singular matrix? */
const nl_double f = 1.0 / A(i,i);
const auto &nzrd = m_terms[i]->m_nzrd;
const auto &nzbd = m_terms[i]->m_nzbd;
for (auto & j : nzbd)
{
const nl_double f1 = -f * A(j,i);
for (auto & k : nzrd)
A(j,k) += A(i,k) * f1;
//RHS(j) += RHS(i) * f1;
#endif
}
}
}
}
template <unsigned m_N, unsigned storage_N>
template <typename T>
void matrix_solver_direct_t<m_N, storage_N>::LE_back_subst(
T * RESTRICT x)
{
const unsigned kN = N();
/* back substitution */
if (m_params.m_pivot)
{
for (int j = kN - 1; j >= 0; j--)
{
T tmp = 0;
for (unsigned k = j+1; k < kN; k++)
tmp += A(j,k) * x[k];
x[j] = (RHS(j) - tmp) / A(j,j);
}
}
else
{
for (int j = kN - 1; j >= 0; j--)
{
T tmp = 0;
const auto *p = m_terms[j]->m_nzrd.data();
const auto e = m_terms[j]->m_nzrd.size() - 1; /* exclude RHS element */
for (unsigned k = 0; k < e; k++)
{
const auto pk = p[k];
tmp += A(j,pk) * x[pk];
}
x[j] = (RHS(j) - tmp) / A(j,j);
}
}
}
template <unsigned m_N, unsigned storage_N>
int matrix_solver_direct_t<m_N, storage_N>::solve_non_dynamic(ATTR_UNUSED const bool newton_raphson)
{
nl_double new_V[storage_N]; // = { 0.0 };
this->LE_solve();
this->LE_back_subst(new_V);
if (newton_raphson)
{
nl_double err = delta(new_V);
store(new_V);
return (err > this->m_params.m_accuracy) ? 2 : 1;
}
else
{
store(new_V);
return 1;
}
}
template <unsigned m_N, unsigned storage_N>
inline int matrix_solver_direct_t<m_N, storage_N>::vsolve_non_dynamic(const bool newton_raphson)
{
build_LE_A<matrix_solver_direct_t>();
build_LE_RHS<matrix_solver_direct_t>();
for (unsigned i=0, iN=N(); i < iN; i++)
m_last_RHS[i] = RHS(i);
this->m_stat_calculations++;
return this->solve_non_dynamic(newton_raphson);
}
template <unsigned m_N, unsigned storage_N>
matrix_solver_direct_t<m_N, storage_N>::matrix_solver_direct_t(netlist_t &anetlist, const pstring &name,
const solver_parameters_t *params, const int size)
: matrix_solver_t(anetlist, name, ASCENDING, params)
, m_dim(size)
{
#if (NL_USE_DYNAMIC_ALLOCATION)
m_A = palloc_array(nl_ext_double, N() * m_pitch);
#endif
for (unsigned k = 0; k < N(); k++)
{
m_last_RHS[k] = 0.0;
}
#if TEST_PARALLEL
thr_initialize();
#endif
}
template <unsigned m_N, unsigned storage_N>
matrix_solver_direct_t<m_N, storage_N>::matrix_solver_direct_t(netlist_t &anetlist, const pstring &name,
const eSortType sort, const solver_parameters_t *params, const int size)
: matrix_solver_t(anetlist, name, sort, params)
, m_dim(size)
{
#if (NL_USE_DYNAMIC_ALLOCATION)
m_A = palloc_array(nl_ext_double, N() * m_pitch);
#endif
for (unsigned k = 0; k < N(); k++)
{
m_last_RHS[k] = 0.0;
}
#if TEST_PARALLEL
thr_initialize();
#endif
}
} //namespace devices
} // namespace netlist
#endif /* NLD_MS_DIRECT_H_ */