interpol_ng.h 26.3 KB
Newer Older
1
2
3
4
5
/*
 *  Copyright (C) 2020 Max-Planck-Society
 *  Author: Martin Reinecke
 */

Martin Reinecke's avatar
Martin Reinecke committed
6
7
8
#ifndef MRUTIL_INTERPOL_NG_H
#define MRUTIL_INTERPOL_NG_H

9
10
#define SIMD_INTERPOL

11
12
#include <vector>
#include <complex>
Martin Reinecke's avatar
Martin Reinecke committed
13
#include <cmath>
14
15
16
17
#include "mr_util/math/constants.h"
#include "mr_util/math/gl_integrator.h"
#include "mr_util/math/es_kernel.h"
#include "mr_util/infra/mav.h"
18
#include "mr_util/infra/simd.h"
19
20
21
22
23
24
#include "mr_util/sharp/sharp.h"
#include "mr_util/sharp/sharp_almhelpers.h"
#include "mr_util/sharp/sharp_geomhelpers.h"
#include "alm.h"
#include "mr_util/math/fft.h"
#include "mr_util/bindings/pybind_utils.h"
Martin Reinecke's avatar
Martin Reinecke committed
25

Martin Reinecke's avatar
Martin Reinecke committed
26
namespace mr {
27

Martin Reinecke's avatar
Martin Reinecke committed
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
namespace detail_fft {

using std::vector;

template<typename T, typename T0> aligned_array<T> alloc_tmp_conv
  (const fmav_info &info, size_t axis, size_t len)
  {
  auto othersize = info.size()/info.shape(axis);
  constexpr auto vlen = native_simd<T0>::size();
  auto tmpsize = len*((othersize>=vlen) ? vlen : 1);
  return aligned_array<T>(tmpsize);
  }

template<typename Tplan, typename T, typename T0, typename Exec>
MRUTIL_NOINLINE void general_convolve(const fmav<T> &in, fmav<T> &out,
  const size_t axis, const vector<T0> &kernel, size_t nthreads,
Martin Reinecke's avatar
Martin Reinecke committed
44
  const Exec &exec)
Martin Reinecke's avatar
Martin Reinecke committed
45
46
47
48
49
50
51
52
53
54
55
56
57
  {
  std::shared_ptr<Tplan> plan1, plan2;

  size_t l_in=in.shape(axis), l_out=out.shape(axis);
  size_t l_min=std::min(l_in, l_out), l_max=std::max(l_in, l_out);
  MR_assert(kernel.size()==l_min/2+1, "bad kernel size");
  plan1 = get_plan<Tplan>(l_in);
  plan2 = get_plan<Tplan>(l_out);

  execParallel(
    util::thread_count(nthreads, in, axis, native_simd<T0>::size()),
    [&](Scheduler &sched) {
      constexpr auto vlen = native_simd<T0>::size();
Martin Reinecke's avatar
Martin Reinecke committed
58
      auto storage = alloc_tmp_conv<T,T0>(in, axis, l_max);
Martin Reinecke's avatar
Martin Reinecke committed
59
60
61
62
63
64
65
66
67
68
69
70
71
      multi_iter<vlen> it(in, out, axis, sched.num_threads(), sched.thread_num());
#ifndef MRUTIL_NO_SIMD
      if (vlen>1)
        while (it.remaining()>=vlen)
          {
          it.advance(vlen);
          auto tdatav = reinterpret_cast<add_vec_t<T> *>(storage.data());
          exec(it, in, out, tdatav, *plan1, *plan2, kernel);
          }
#endif
      while (it.remaining()>0)
        {
        it.advance(1);
Martin Reinecke's avatar
Martin Reinecke committed
72
        auto buf = reinterpret_cast<T *>(storage.data());
Martin Reinecke's avatar
Martin Reinecke committed
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
        exec(it, in, out, buf, *plan1, *plan2, kernel);
        }
    });  // end of parallel region
  }

struct ExecConvR1
  {
  template <typename T0, typename T, size_t vlen> void operator() (
    const multi_iter<vlen> &it, const fmav<T0> &in, fmav<T0> &out,
    T * buf, const pocketfft_r<T0> &plan1, const pocketfft_r<T0> &plan2,
    const vector<T0> &kernel) const
    {
    size_t l_in = plan1.length(),
           l_out = plan2.length(),
           l_min = std::min(l_in, l_out);
    copy_input(it, in, buf);
    plan1.exec(buf, T0(1), true);
Martin Reinecke's avatar
Martin Reinecke committed
90
    for (size_t i=0; i<l_min; ++i) buf[i]*=kernel[(i+1)/2];
Martin Reinecke's avatar
Martin Reinecke committed
91
92
93
94
95
96
97
98
99
100
101
102
    for (size_t i=l_in; i<l_out; ++i) buf[i] = T(0);
    plan2.exec(buf, T0(1), false);
    copy_output(it, buf, out);
    }
  };

template<typename T> void convolve_1d(const fmav<T> &in,
  fmav<T> &out, size_t axis, const vector<T> &kernel, size_t nthreads=1)
  {
  MR_assert(axis<in.ndim(), "bad axis number");
  MR_assert(in.ndim()==out.ndim(), "dimensionality mismatch");
  if (in.data()==out.data())
Martin Reinecke's avatar
Martin Reinecke committed
103
    MR_assert(in.stride()==out.stride(), "strides mismatch");
Martin Reinecke's avatar
Martin Reinecke committed
104
105
106
  for (size_t i=0; i<in.ndim(); ++i)
    if (i!=axis)
      MR_assert(in.shape(i)==out.shape(i), "shape mismatch");
Martin Reinecke's avatar
Martin Reinecke committed
107
108
  MR_assert(!((in.shape(axis)&1) || (out.shape(axis)&1)),
    "input and output axis lengths must be even");
Martin Reinecke's avatar
Martin Reinecke committed
109
110
111
112
113
114
  if (in.size()==0) return;
  general_convolve<pocketfft_r<T>>(in, out, axis, kernel, nthreads,
    ExecConvR1());
  }

}
Martin Reinecke's avatar
Martin Reinecke committed
115
116
117

using detail_fft::convolve_1d;

Martin Reinecke's avatar
Martin Reinecke committed
118
namespace detail_interpol_ng {
119

Martin Reinecke's avatar
Martin Reinecke committed
120
using namespace std;
121
122
123
124

template<typename T> class Interpolator
  {
  protected:
125
    bool adjoint;
Martin Reinecke's avatar
Martin Reinecke committed
126
    size_t lmax, kmax, nphi0, ntheta0, nphi, ntheta;
Martin Reinecke's avatar
Martin Reinecke committed
127
    int nthreads;
128
    T ofactor;
Martin Reinecke's avatar
fix    
Martin Reinecke committed
129
    size_t supp;
130
    ES_Kernel kernel;
131
    size_t ncomp;
132
133
134
#ifdef SIMD_INTERPOL
    mav<native_simd<T>,4> scube;
#endif
135
    mav<T,4> cube; // the data cube (theta, phi, 2*mbeam+1, TGC)
136

137
    void correct(mav<T,2> &arr, int spin)
138
      {
139
      T sfct = (spin&1) ? -1 : 1;
Martin Reinecke's avatar
Martin Reinecke committed
140
141
142
143
      mav<T,2> tmp({nphi,nphi0});
      // copy and extend to second half
      for (size_t j=0; j<nphi0; ++j)
        tmp.v(0,j) = arr(0,j);
Martin Reinecke's avatar
Martin Reinecke committed
144
      for (size_t i=1, i2=nphi0-1; i+1<ntheta0; ++i,--i2)
Martin Reinecke's avatar
Martin Reinecke committed
145
        for (size_t j=0,j2=nphi0/2; j<nphi0; ++j,++j2)
146
          {
Martin Reinecke's avatar
Martin Reinecke committed
147
          if (j2>=nphi0) j2-=nphi0;
Martin Reinecke's avatar
Martin Reinecke committed
148
149
          tmp.v(i,j2) = arr(i,j2);
          tmp.v(i2,j) = sfct*tmp(i,j2);
150
          }
Martin Reinecke's avatar
Martin Reinecke committed
151
152
      for (size_t j=0; j<nphi0; ++j)
        tmp.v(ntheta0-1,j) = arr(ntheta0-1,j);
Martin Reinecke's avatar
Martin Reinecke committed
153
      auto fct = kernel.correction_factors(nphi, nphi0/2+1, nthreads);
154
155
      vector<T> k2(fct.size());
      for (size_t i=0; i<fct.size(); ++i) k2[i] = T(fct[i]/nphi0);
Martin Reinecke's avatar
Martin Reinecke committed
156
157
      fmav<T> ftmp(tmp);
      fmav<T> ftmp0(tmp.template subarray<2>({0,0},{nphi0, nphi0}));
158
      convolve_1d(ftmp0, ftmp, 0, k2, nthreads);
Martin Reinecke's avatar
Martin Reinecke committed
159
      fmav<T> ftmp2(tmp.template subarray<2>({0,0},{ntheta, nphi0}));
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
160
      fmav<T> farr(arr);
161
      convolve_1d(ftmp2, farr, 1, k2, nthreads);
162
      }
163
164
    void decorrect(mav<T,2> &arr, int spin)
      {
165
      T sfct = (spin&1) ? -1 : 1;
Martin Reinecke's avatar
Martin Reinecke committed
166
167
      mav<T,2> tmp({nphi,nphi0});
      auto fct = kernel.correction_factors(nphi, nphi0/2+1, nthreads);
168
169
      vector<T> k2(fct.size());
      for (size_t i=0; i<fct.size(); ++i) k2[i] = T(fct[i]/nphi0);
Martin Reinecke's avatar
Martin Reinecke committed
170
171
      fmav<T> farr(arr);
      fmav<T> ftmp2(tmp.template subarray<2>({0,0},{ntheta, nphi0}));
172
      convolve_1d(farr, ftmp2, 1, k2, nthreads);
173
      // extend to second half
Martin Reinecke's avatar
Martin Reinecke committed
174
      for (size_t i=1, i2=nphi-1; i+1<ntheta; ++i,--i2)
Martin Reinecke's avatar
Martin Reinecke committed
175
        for (size_t j=0,j2=nphi0/2; j<nphi0; ++j,++j2)
176
          {
Martin Reinecke's avatar
Martin Reinecke committed
177
          if (j2>=nphi0) j2-=nphi0;
178
179
          tmp.v(i2,j) = sfct*tmp(i,j2);
          }
Martin Reinecke's avatar
Martin Reinecke committed
180
181
      fmav<T> ftmp(tmp);
      fmav<T> ftmp0(tmp.template subarray<2>({0,0},{nphi0, nphi0}));
182
      convolve_1d(ftmp, ftmp0, 0, k2, nthreads);
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
183
      for (size_t j=0; j<nphi0; ++j)
184
        arr.v(0,j) = T(0.5)*tmp(0,j);
Martin Reinecke's avatar
Martin Reinecke committed
185
      for (size_t i=1; i+1<ntheta0; ++i)
186
        for (size_t j=0; j<nphi0; ++j)
Martin Reinecke's avatar
Martin Reinecke committed
187
188
          arr.v(i,j) = tmp(i,j);
      for (size_t j=0; j<nphi0; ++j)
189
        arr.v(ntheta0-1,j) = T(0.5)*tmp(ntheta0-1,j);
190
      }
191

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
192
193
194
195
196
197
198
199
200
201
202
    vector<size_t> getIdx(const mav<T,2> &ptg) const
      {
      vector<size_t> idx(ptg.shape(0));
      constexpr size_t cellsize=16;
      size_t nct = ntheta/cellsize+1,
             ncp = nphi/cellsize+1;
      vector<vector<size_t>> mapper(nct*ncp);
      for (size_t i=0; i<ptg.shape(0); ++i)
        {
        size_t itheta=min(nct-1,size_t(ptg(i,0)/pi*nct)),
               iphi=min(ncp-1,size_t(ptg(i,1)/(2*pi)*ncp));
203
//        MR_assert((itheta<nct)&&(iphi<ncp), "oops");
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
204
205
206
207
208
209
210
211
212
        mapper[itheta*ncp+iphi].push_back(i);
        }
      size_t cnt=0;
      for (const auto &vec: mapper)
        for (auto i:vec)
          idx[cnt++] = i;
      return idx;
      }

213
  public:
214
215
    Interpolator(const vector<Alm<complex<T>>> &slm,
                 const vector<Alm<complex<T>>> &blm,
216
                 bool separate, T epsilon, T ofmin, int nthreads_)
217
      : adjoint(false),
218
219
        lmax(slm.at(0).Lmax()),
        kmax(blm.at(0).Mmax()),
Martin Reinecke's avatar
Martin Reinecke committed
220
221
        nphi0(2*good_size_real(lmax+1)),
        ntheta0(nphi0/2+1),
Martin Reinecke's avatar
Martin Reinecke committed
222
        nphi(std::max<size_t>(20,2*good_size_real(size_t((2*lmax+1)*ofmin/2.)))),
Martin Reinecke's avatar
Martin Reinecke committed
223
        ntheta(nphi/2+1),
Martin Reinecke's avatar
Martin Reinecke committed
224
        nthreads(nthreads_),
225
        ofactor(T(nphi)/(2*lmax+1)),
Martin Reinecke's avatar
fix    
Martin Reinecke committed
226
        supp(ES_Kernel::get_supp(epsilon, ofactor)),
Martin Reinecke's avatar
Martin Reinecke committed
227
        kernel(supp, ofactor, nthreads),
228
        ncomp(separate ? slm.size() : 1),
229
230
231
232
233
234
#ifdef SIMD_INTERPOL
        scube({ntheta+2*supp, nphi+2*supp, ncomp, (2*kmax+1+native_simd<T>::size()-1)/native_simd<T>::size()}),
        cube(reinterpret_cast<T *>(scube.vdata()),{ntheta+2*supp, nphi+2*supp, ncomp, ((2*kmax+1+native_simd<T>::size()-1)/native_simd<T>::size())*native_simd<T>::size()},true)
#else
        cube({ntheta+2*supp, nphi+2*supp, ncomp, 2*kmax+1})
#endif
235
      {
236
      MR_assert((ncomp==1)||(ncomp==3), "currently only 1 or 3 components allowed");
237
238
239
240
241
242
243
244
245
      MR_assert(slm.size()==blm.size(), "inconsistent slm and blm vectors");
      for (size_t i=0; i<slm.size(); ++i)
        {
        MR_assert(slm[i].Lmax()==lmax, "inconsistent Sky lmax");
        MR_assert(slm[i].Mmax()==lmax, "Sky lmax must be equal to Sky mmax");
        MR_assert(blm[i].Lmax()==lmax, "Sky and beam lmax must be equal");
        MR_assert(blm[i].Mmax()==kmax, "Inconcistent beam mmax");
        }

246
247
      MR_assert((supp<=ntheta) && (supp<=nphi), "support too large!");
      Alm<complex<T>> a1(lmax, lmax), a2(lmax,lmax);
Martin Reinecke's avatar
Martin Reinecke committed
248
      auto ginfo = sharp_make_cc_geom_info(ntheta0,nphi0,0.,cube.stride(1),cube.stride(0));
249
250
      auto ainfo = sharp_make_triangular_alm_info(lmax,lmax,1);

251
      vector<T>lnorm(lmax+1);
252
      for (size_t i=0; i<=lmax; ++i)
253
        lnorm[i]=T(std::sqrt(4*pi/(2*i+1.)));
254

255
      for (size_t icomp=0; icomp<ncomp; ++icomp)
256
257
258
259
        {
        for (size_t m=0; m<=lmax; ++m)
          for (size_t l=m; l<=lmax; ++l)
            {
260
            if (separate)
261
              a1(l,m) = slm[icomp](l,m)*blm[icomp](l,0).real()*lnorm[l];
262
263
            else
              {
264
265
              a1(l,m) = 0;
              for (size_t j=0; j<slm.size(); ++j)
266
                a1(l,m) += slm[j](l,m)*blm[j](l,0).real()*lnorm[l];
267
268
              }
            }
269
        auto m1 = cube.template subarray<2>({supp,supp,icomp,0},{ntheta,nphi,0,0});
270
271
272
273
274
275
276
277
278
279
280
281
282
283
        sharp_alm2map(a1.Alms().data(), m1.vdata(), *ginfo, *ainfo, 0, nthreads);
        correct(m1,0);

        for (size_t k=1; k<=kmax; ++k)
          {
          for (size_t m=0; m<=lmax; ++m)
            for (size_t l=m; l<=lmax; ++l)
              {
              if (l<k)
                a1(l,m)=a2(l,m)=0.;
              else
                {
                if (separate)
                  {
284
                  auto tmp = blm[icomp](l,k)*(-2*lnorm[l]);
285
286
287
288
289
290
291
292
                  a1(l,m) = slm[icomp](l,m)*tmp.real();
                  a2(l,m) = slm[icomp](l,m)*tmp.imag();
                  }
                else
                  {
                  a1(l,m) = a2(l,m) = 0;
                  for (size_t j=0; j<slm.size(); ++j)
                    {
293
                    auto tmp = blm[j](l,k)*(-2*lnorm[l]);
294
295
296
297
                    a1(l,m) += slm[j](l,m)*tmp.real();
                    a2(l,m) += slm[j](l,m)*tmp.imag();
                    }
                  }
Martin Reinecke's avatar
fixes    
Martin Reinecke committed
298
                }
299
              }
300
301
          auto m1 = cube.template subarray<2>({supp,supp,icomp,2*k-1},{ntheta,nphi,0,0});
          auto m2 = cube.template subarray<2>({supp,supp,icomp,2*k  },{ntheta,nphi,0,0});
Martin Reinecke's avatar
fixes    
Martin Reinecke committed
302
303
304
305
          sharp_alm2map_spin(k, a1.Alms().data(), a2.Alms().data(), m1.vdata(),
            m2.vdata(), *ginfo, *ainfo, 0, nthreads);
          correct(m1,k);
          correct(m2,k);
306
          }
307
        }
308

309
310
311
      // fill border regions
      for (size_t i=0; i<supp; ++i)
        for (size_t j=0, j2=nphi/2; j<nphi; ++j,++j2)
312
          for (size_t k=0; k<cube.shape(3); ++k)
313
            {
314
            T fct = (((k+1)/2)&1) ? -1 : 1;
315
            if (j2>=nphi) j2-=nphi;
316
            for (size_t l=0; l<cube.shape(2); ++l)
317
              {
318
319
              cube.v(supp-1-i,j2+supp,l,k) = fct*cube(supp+1+i,j+supp,l,k);
              cube.v(supp+ntheta+i,j2+supp,l,k) = fct*cube(supp+ntheta-2-i,j+supp,l,k);
320
              }
321
322
323
            }
      for (size_t i=0; i<ntheta+2*supp; ++i)
        for (size_t j=0; j<supp; ++j)
324
325
          for (size_t k=0; k<cube.shape(3); ++k)
            for (size_t l=0; l<cube.shape(2); ++l)
326
            {
327
328
            cube.v(i,j,l,k) = cube(i,j+nphi,l,k);
            cube.v(i,j+nphi+supp,l,k) = cube(i,j+supp,l,k);
329
330
331
            }
      }

332
    Interpolator(size_t lmax_, size_t kmax_, size_t ncomp_, T epsilon, T ofmin, int nthreads_)
333
334
335
336
337
      : adjoint(true),
        lmax(lmax_),
        kmax(kmax_),
        nphi0(2*good_size_real(lmax+1)),
        ntheta0(nphi0/2+1),
Martin Reinecke's avatar
Martin Reinecke committed
338
        nphi(std::max<size_t>(20,2*good_size_real(size_t((2*lmax+1)*ofmin/2.)))),
339
340
        ntheta(nphi/2+1),
        nthreads(nthreads_),
341
        ofactor(T(nphi)/(2*lmax+1)),
342
343
        supp(ES_Kernel::get_supp(epsilon, ofactor)),
        kernel(supp, ofactor, nthreads),
344
        ncomp(ncomp_),
345
346
347
348
349
350
#ifdef SIMD_INTERPOL
        scube({ntheta+2*supp, nphi+2*supp, ncomp, (2*kmax+1+native_simd<T>::size()-1)/native_simd<T>::size()}),
        cube(reinterpret_cast<T *>(scube.vdata()),{ntheta+2*supp, nphi+2*supp, ncomp, ((2*kmax+1+native_simd<T>::size()-1)/native_simd<T>::size())*native_simd<T>::size()},true)
#else
        cube({ntheta+2*supp, nphi+2*supp, ncomp, 2*kmax+1})
#endif
351
      {
352
      MR_assert((ncomp==1)||(ncomp==3), "currently only 1 or 3 components allowed");
353
354
355
356
      MR_assert((supp<=ntheta) && (supp<=nphi), "support too large!");
      cube.apply([](T &v){v=0.;});
      }

357
    void interpol (const mav<T,2> &ptg, mav<T,2> &res) const
358
      {
359
360
361
362
#ifdef SIMD_INTERPOL
      constexpr size_t vl=native_simd<T>::size();
      MR_assert(scube.stride(3)==1, "bad stride");
#endif
363
      MR_assert(!adjoint, "cannot be called in adjoint mode");
364
365
      MR_assert(ptg.shape(0)==res.shape(0), "dimension mismatch");
      MR_assert(ptg.shape(1)==3, "second dimension must have length 3");
366
      MR_assert(res.shape(1)==ncomp, "# of components mismatch");
367
368
369
      T delta = T(2)/supp;
      T xdtheta = T((ntheta-1)/pi),
        xdphi = T(nphi/(2*pi));
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
370
      auto idx = getIdx(ptg);
Martin Reinecke's avatar
Martin Reinecke committed
371
      execStatic(idx.size(), nthreads, 0, [&](Scheduler &sched)
372
        {
Martin Reinecke's avatar
Martin Reinecke committed
373
374
        vector<T> wt(supp), wp(supp);
        vector<T> psiarr(2*kmax+1);
375
376
377
378
#ifdef SIMD_INTERPOL
        vector<native_simd<T>> psiarr2((2*kmax+1+vl-1)/vl);
        for (auto &v:psiarr2) v=0;
#endif
Martin Reinecke's avatar
Martin Reinecke committed
379
        while (auto rng=sched.getNext()) for(auto ind=rng.lo; ind<rng.hi; ++ind)
380
          {
Martin Reinecke's avatar
Martin Reinecke committed
381
          size_t i=idx[ind];
382
383
          T f0=T(0.5*supp+ptg(i,0)*xdtheta);
          size_t i0 = size_t(f0+T(1));
Martin Reinecke's avatar
Martin Reinecke committed
384
385
          for (size_t t=0; t<supp; ++t)
            wt[t] = kernel((t+i0-f0)*delta - 1);
386
          T f1=T(0.5)*supp+ptg(i,1)*xdphi;
Martin Reinecke's avatar
Martin Reinecke committed
387
388
389
390
          size_t i1 = size_t(f1+1.);
          for (size_t t=0; t<supp; ++t)
            wp[t] = kernel((t+i1-f1)*delta - 1);
          psiarr[0]=1.;
391
392
          double psi=ptg(i,2);
          double cpsi=cos(psi), spsi=sin(psi);
Martin Reinecke's avatar
Martin Reinecke committed
393
394
395
          double cnpsi=cpsi, snpsi=spsi;
          for (size_t l=1; l<=kmax; ++l)
            {
396
397
            psiarr[2*l-1]=T(cnpsi);
            psiarr[2*l]=T(snpsi);
Martin Reinecke's avatar
Martin Reinecke committed
398
399
400
401
            const double tmp = snpsi*cpsi + cnpsi*spsi;
            cnpsi=cnpsi*cpsi - snpsi*spsi;
            snpsi=tmp;
            }
402
403
404
405
#ifdef SIMD_INTERPOL
          memcpy(reinterpret_cast<T *>(psiarr2.data()), psiarr.data(),
            (2*kmax+1)*sizeof(T));
#endif
406
407
          if (ncomp==1)
            {
408
#ifndef SIMD_INTERPOL
409
            T vv=0;
410
411
412
            for (size_t j=0; j<supp; ++j)
              for (size_t k=0; k<supp; ++k)
                for (size_t l=0; l<2*kmax+1; ++l)
413
                  vv += cube(i0+j,i1+k,0,l)*wt[j]*wp[k]*psiarr[l];
414
            res.v(i,0) = vv;
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
#else
            const native_simd<T> *p=&scube(i0,i1,0,0);
            ptrdiff_t d0 = scube.stride(0);
            ptrdiff_t d1 = scube.stride(1);
            size_t nv=scube.shape(3);
            native_simd<T> vv=0.;
            for (size_t j=0; j<supp; ++j, p+=d0)
              {
              auto p1=p;
              T wtj = wt[j];
              for (size_t k=0; k<supp; ++k, p1+=d1)
                {
                auto p2=p1;
                native_simd<T> tvv=0;
                for (size_t l=0; l<nv; ++l,++p2)
                  tvv += *p2*psiarr2[l];
                vv += wtj*wp[k]*tvv;
                }
              }
            res.v(i,0) = reduce(vv, std::plus<>());
#endif
436
437
438
            }
          else // ncomp==3
            {
439
#ifndef SIMD_INTERPOL
440
            T v0=0., v1=0., v2=0.;
441
442
443
444
            for (size_t j=0; j<supp; ++j)
              for (size_t k=0; k<supp; ++k)
                for (size_t l=0; l<2*kmax+1; ++l)
                  {
445
                  auto tmp = wt[j]*wp[k]*psiarr[l];
446
447
448
                  v0 += cube(i0+j,i1+k,0,l)*tmp;
                  v1 += cube(i0+j,i1+k,1,l)*tmp;
                  v2 += cube(i0+j,i1+k,2,l)*tmp;
449
450
451
452
                  }
            res.v(i,0) = v0;
            res.v(i,1) = v1;
            res.v(i,2) = v2;
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
#else
            const native_simd<T> *p=&scube(i0,i1,0,0);
            ptrdiff_t d0 = scube.stride(0);
            ptrdiff_t d1 = scube.stride(1);
            ptrdiff_t d2 = scube.stride(2);
            size_t nv=scube.shape(3);
            native_simd<T> v0=0., v1=0., v2=0.;
            for (size_t j=0; j<supp; ++j, p+=d0)
              {
              auto p1=p;
              T wtj = wt[j];
              for (size_t k=0; k<supp; ++k, p1+=d1)
                {
                auto p2=p1;
                native_simd<T> wtjwpk = wtj*wp[k];
                for (size_t l=0; l<nv; ++l,++p2)
                  {
                  auto tmp = wtjwpk*psiarr2[l];
                  v0 += p2[0]*tmp;
                  v1 += p2[d2]*tmp;
                  v2 += p2[2*d2]*tmp;
                  }
                }
              }
            res.v(i,0) = reduce(v0, std::plus<>());
            res.v(i,1) = reduce(v1, std::plus<>());
            res.v(i,2) = reduce(v2, std::plus<>());
#endif
481
            }
482
          }
Martin Reinecke's avatar
Martin Reinecke committed
483
        });
484
      }
485

486
487
488
    size_t support() const
      { return supp; }

489
    void deinterpol (const mav<T,2> &ptg, const mav<T,2> &data)
490
      {
491
492
493
494
#ifdef SIMD_INTERPOL
      constexpr size_t vl=native_simd<T>::size();
      MR_assert(scube.stride(3)==1, "bad stride");
#endif
495
496
497
      MR_assert(adjoint, "can only be called in adjoint mode");
      MR_assert(ptg.shape(0)==data.shape(0), "dimension mismatch");
      MR_assert(ptg.shape(1)==3, "second dimension must have length 3");
498
      MR_assert(data.shape(1)==ncomp, "# of components mismatch");
499
500
501
      T delta = T(2)/supp;
      T xdtheta = T((ntheta-1)/pi),
        xdphi = T(nphi/(2*pi));
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
502
      auto idx = getIdx(ptg);
503
504
505
506
507
508
509

      constexpr size_t cellsize=16;
      size_t nct = ntheta/cellsize+5,
             ncp = nphi/cellsize+5;
      mav<std::mutex,2> locks({nct,ncp});

      execStatic(idx.size(), nthreads, 0, [&](Scheduler &sched)
510
        {
511
        size_t b_theta=99999999999999, b_phi=9999999999999999;
512
513
        vector<T> wt(supp), wp(supp);
        vector<T> psiarr(2*kmax+1);
514
515
516
517
#ifdef SIMD_INTERPOL
        vector<native_simd<T>> psiarr2((2*kmax+1+vl-1)/vl);
        for (auto &v:psiarr2) v=0;
#endif
518
519
520
        while (auto rng=sched.getNext()) for(auto ind=rng.lo; ind<rng.hi; ++ind)
          {
          size_t i=idx[ind];
521
          T f0=T(0.5)*supp+ptg(i,0)*xdtheta;
522
523
524
          size_t i0 = size_t(f0+1.);
          for (size_t t=0; t<supp; ++t)
            wt[t] = kernel((t+i0-f0)*delta - 1);
525
          T f1=T(0.5)*supp+ptg(i,1)*xdphi;
526
527
528
529
530
531
532
533
534
          size_t i1 = size_t(f1+1.);
          for (size_t t=0; t<supp; ++t)
            wp[t] = kernel((t+i1-f1)*delta - 1);
          psiarr[0]=1.;
          double psi=ptg(i,2);
          double cpsi=cos(psi), spsi=sin(psi);
          double cnpsi=cpsi, snpsi=spsi;
          for (size_t l=1; l<=kmax; ++l)
            {
535
536
            psiarr[2*l-1]=T(cnpsi);
            psiarr[2*l]=T(snpsi);
537
538
539
540
            const double tmp = snpsi*cpsi + cnpsi*spsi;
            cnpsi=cnpsi*cpsi - snpsi*spsi;
            snpsi=tmp;
            }
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
          size_t b_theta_new = i0/cellsize,
                 b_phi_new = i1/cellsize;
          if ((b_theta_new!=b_theta) || (b_phi_new!=b_phi))
            {
            if (b_theta<locks.shape(0))  // unlock
              {
              locks.v(b_theta,b_phi).unlock();
              locks.v(b_theta,b_phi+1).unlock();
              locks.v(b_theta+1,b_phi).unlock();
              locks.v(b_theta+1,b_phi+1).unlock();
              }
            b_theta = b_theta_new;
            b_phi = b_phi_new;
            locks.v(b_theta,b_phi).lock();
            locks.v(b_theta,b_phi+1).lock();
            locks.v(b_theta+1,b_phi).lock();
            locks.v(b_theta+1,b_phi+1).lock();
            }
559
560
561
562
#ifdef SIMD_INTERPOL
          memcpy(reinterpret_cast<T *>(psiarr2.data()), psiarr.data(),
            (2*kmax+1)*sizeof(T));
#endif
563
564
          if (ncomp==1)
            {
565
#ifndef SIMD_INTERPOL
566
            T val = data(i,0);
567
568
569
            for (size_t j=0; j<supp; ++j)
              for (size_t k=0; k<supp; ++k)
                for (size_t l=0; l<2*kmax+1; ++l)
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
                  cube.v(i0+j,i1+k,0,l) += val*wt[j]*wp[k]*psiarr[l];
#else
            native_simd<T> val = data(i,0);
            native_simd<T> *p=&scube.v(i0,i1,0,0);
            ptrdiff_t d0 = scube.stride(0);
            ptrdiff_t d1 = scube.stride(1);
            size_t nv=scube.shape(3);
            for (size_t j=0; j<supp; ++j, p+=d0)
              {
              auto p1=p;
              T wtj = wt[j];
              for (size_t k=0; k<supp; ++k, p1+=d1)
                {
                auto p2=p1;
                native_simd<T> tv = wtj*wp[k]*val;
                for (size_t l=0; l<nv; ++l,++p2)
                  *p2 += tv*psiarr2[l];
                }
              }
#endif
590
591
592
            }
          else // ncomp==3
            {
593
#ifndef SIMD_INTERPOL
594
            T v0=data(i,0), v1=data(i,1), v2=data(i,2);
595
596
597
            for (size_t j=0; j<supp; ++j)
              for (size_t k=0; k<supp; ++k)
                {
598
                T t0 = wt[j]*wp[k];
599
600
                for (size_t l=0; l<2*kmax+1; ++l)
                  {
601
                  T tmp = t0*psiarr[l];
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
                  cube.v(i0+j,i1+k,0,l) += v0*tmp;
                  cube.v(i0+j,i1+k,1,l) += v1*tmp;
                  cube.v(i0+j,i1+k,2,l) += v2*tmp;
                  }
                }
#else
            native_simd<T> v0=data(i,0), v1=data(i,1), v2=data(i,2);
            native_simd<T> *p=&scube.v(i0,i1,0,0);
            ptrdiff_t d0 = scube.stride(0);
            ptrdiff_t d1 = scube.stride(1);
            ptrdiff_t d2 = scube.stride(2);
            size_t nv=scube.shape(3);
            for (size_t j=0; j<supp; ++j, p+=d0)
              {
              auto p1=p;
              T wtj = wt[j];
              for (size_t k=0; k<supp; ++k, p1+=d1)
                {
                auto p2=p1;
                native_simd<T> wtjwpk = wtj*wp[k];
                for (size_t l=0; l<nv; ++l,++p2)
                  {
                  auto tmp = wtjwpk*psiarr2[l];
                  p2[0] += v0*tmp;
                  p2[d2] += v1*tmp;
                  p2[2*d2] += v2*tmp;
628
629
                  }
                }
630
631
              }
#endif
632
            }
633
          }
634
635
636
637
638
639
640
        if (b_theta<locks.shape(0))  // unlock
          {
          locks.v(b_theta,b_phi).unlock();
          locks.v(b_theta,b_phi+1).unlock();
          locks.v(b_theta+1,b_phi).unlock();
          locks.v(b_theta+1,b_phi+1).unlock();
          }
641
642
        });
      }
643
    void getSlm (const vector<Alm<complex<T>>> &blm, vector<Alm<complex<T>>> &slm)
644
645
      {
      MR_assert(adjoint, "can only be called in adjoint mode");
646
647
      MR_assert((blm.size()==ncomp) || (ncomp==1), "incorrect number of beam a_lm sets");
      MR_assert((slm.size()==ncomp) || (ncomp==1), "incorrect number of sky a_lm sets");
648
649
650
651
652
      Alm<complex<T>> a1(lmax, lmax), a2(lmax,lmax);
      auto ginfo = sharp_make_cc_geom_info(ntheta0,nphi0,0.,cube.stride(1),cube.stride(0));
      auto ainfo = sharp_make_triangular_alm_info(lmax,lmax,1);

      // move stuff from border regions onto the main grid
653
      for (size_t i=0; i<cube.shape(0); ++i)
654
        for (size_t j=0; j<supp; ++j)
655
656
          for (size_t k=0; k<cube.shape(3); ++k)
            for (size_t l=0; l<cube.shape(2); ++l)
657
              {
658
659
              cube.v(i,j+nphi,l,k) += cube(i,j,l,k);
              cube.v(i,j+supp,l,k) += cube(i,j+nphi+supp,l,k);
660
              }
661
662
      for (size_t i=0; i<supp; ++i)
        for (size_t j=0, j2=nphi/2; j<nphi; ++j,++j2)
663
          for (size_t k=0; k<cube.shape(3); ++k)
664
            {
665
            T fct = (((k+1)/2)&1) ? -1 : 1;
666
            if (j2>=nphi) j2-=nphi;
667
            for (size_t l=0; l<cube.shape(2); ++l)
668
              {
669
670
              cube.v(supp+1+i,j+supp,l,k) += fct*cube(supp-1-i,j2+supp,l,k);
              cube.v(supp+ntheta-2-i, j+supp,l,k) += fct*cube(supp+ntheta+i,j2+supp,l,k);
671
              }
672
            }
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
673
674
675

      // special treatment for poles
      for (size_t j=0,j2=nphi/2; j<nphi/2; ++j,++j2)
676
677
        for (size_t k=0; k<cube.shape(3); ++k)
          for (size_t l=0; l<cube.shape(2); ++l)
678
            {
679
            T fct = (((k+1)/2)&1) ? -1 : 1;
680
            if (j2>=nphi) j2-=nphi;
681
682
683
684
685
686
            T tval = (cube(supp,j+supp,l,k) + fct*cube(supp,j2+supp,l,k));
            cube.v(supp,j+supp,l,k) = tval;
            cube.v(supp,j2+supp,l,k) = fct*tval;
            tval = (cube(supp+ntheta-1,j+supp,l,k) + fct*cube(supp+ntheta-1,j2+supp,l,k));
            cube.v(supp+ntheta-1,j+supp,l,k) = tval;
            cube.v(supp+ntheta-1,j2+supp,l,k) = fct*tval;
687
            }
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
688

689
      vector<T>lnorm(lmax+1);
690
      for (size_t i=0; i<=lmax; ++i)
691
        lnorm[i]=T(std::sqrt(4*pi/(2*i+1.)));
692

693
694
695
      for (size_t j=0; j<blm.size(); ++j)
        slm[j].SetToZero();

696
      for (size_t icomp=0; icomp<ncomp; ++icomp)
697
        {
698
        bool separate = ncomp>1;
699
        {
700
        auto m1 = cube.template subarray<2>({supp,supp,icomp,0},{ntheta,nphi,0,0});
701
702
        decorrect(m1,0);
        sharp_alm2map_adjoint(a1.Alms().vdata(), m1.data(), *ginfo, *ainfo, 0, nthreads);
703
704
        for (size_t m=0; m<=lmax; ++m)
          for (size_t l=m; l<=lmax; ++l)
705
            if (separate)
706
              slm[icomp](l,m) += conj(a1(l,m))*blm[icomp](l,0).real()*lnorm[l];
707
708
            else
              for (size_t j=0; j<blm.size(); ++j)
709
                slm[j](l,m) += conj(a1(l,m))*blm[j](l,0).real()*lnorm[l];
710
711
712
        }
        for (size_t k=1; k<=kmax; ++k)
          {
713
714
          auto m1 = cube.template subarray<2>({supp,supp,icomp,2*k-1},{ntheta,nphi,0,0});
          auto m2 = cube.template subarray<2>({supp,supp,icomp,2*k  },{ntheta,nphi,0,0});
715
716
717
718
719
720
721
722
          decorrect(m1,k);
          decorrect(m2,k);

          sharp_alm2map_spin_adjoint(k, a1.Alms().vdata(), a2.Alms().vdata(), m1.data(),
            m2.data(), *ginfo, *ainfo, 0, nthreads);
          for (size_t m=0; m<=lmax; ++m)
            for (size_t l=m; l<=lmax; ++l)
              if (l>=k)
723
724
                {
                if (separate)
725
                  {
726
                  auto tmp = conj(blm[icomp](l,k))*(-2*lnorm[l]);
727
728
                  slm[icomp](l,m) += conj(a1(l,m))*tmp.real();
                  slm[icomp](l,m) -= conj(a2(l,m))*tmp.imag();
729
                  }
730
731
732
                else
                  for (size_t j=0; j<blm.size(); ++j)
                    {
733
                    auto tmp = conj(blm[j](l,k))*(-2*lnorm[l]);
734
735
736
737
                    slm[j](l,m) += conj(a1(l,m))*tmp.real();
                    slm[j](l,m) -= conj(a2(l,m))*tmp.imag();
                    }
                }
738
          }
739
740
        }
      }
741
742
  };

743
744
745
double epsilon_guess(size_t support, double ofactor)
  { return std::sqrt(12.)*std::exp(-(support*ofactor)); }

Martin Reinecke's avatar
Martin Reinecke committed
746
}
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
747

Martin Reinecke's avatar
Martin Reinecke committed
748
using detail_interpol_ng::Interpolator;
749
using detail_interpol_ng::epsilon_guess;
750

Martin Reinecke's avatar
Martin Reinecke committed
751
}
752

Martin Reinecke's avatar
Martin Reinecke committed
753
#endif