Build Packages / build:portable:linux-x86_64:nocuda (push) Successful in 16m43s
Build Packages / build:jfjoch:ubuntu2404:nocuda (push) Successful in 28m46s
Build Packages / build:jfjoch:rocky8:cuda-sls9 (push) Successful in 32m5s
Build Packages / build:jfjoch:rocky9:cuda (push) Successful in 39m44s
Build Packages / build:portable:macos-arm64:nocuda (push) Successful in 5m7s
Build Packages / Create release (push) Successful in 8s
Build Packages / Build documentation (push) Successful in 49s
Build Packages / build:portable:linux-x86_64:cuda (push) Successful in 22m30s
Build Packages / build:jfjoch:rocky8:nocuda (push) Successful in 31m54s
Build Packages / build:rugnux:linux-aarch64:cuda (push) Successful in 11m29s
Build Packages / build:portable:windows-x86_64:nocuda (push) Successful in 23m51s
Build Packages / build:jfjoch:ubuntu2204:nocuda (push) Successful in 26m7s
Build Packages / build:jfjoch:ubuntu2204:cuda (push) Successful in 29m38s
Build Packages / build:portable:windows-x86_64:cuda (push) Successful in 28m6s
Build Packages / HDF5 consumer tests (DIALS, XDS) (push) Successful in 29m44s
Build Packages / build:jfjoch:rocky9:nocuda (push) Successful in 33m22s
Build Packages / Generate python client (push) Successful in 47s
Build Packages / build:jfjoch:rocky9:cuda-sls9 (push) Successful in 34m6s
Build Packages / Unit tests (push) Successful in 1h2m47s
Build Packages / build:jfjoch:rocky8:cuda (push) Successful in 35m59s
Build Packages / build:jfjoch:ubuntu2404:cuda (push) Successful in 34m8s
Build Packages / Validation Battery (push) Successful in 38m31s
* Rugnux: Refinement several times faster, one protocol on the CPU and the GPU, parts of it on the Apple GPU; its random draws are the same on every platform. * Rugnux: Refinement follows common practice: the free set decides only the restraint weights and the displacement model (isotropic, TLS or anisotropic, by Hamilton's test), everything else is decided on the working set; lower R-free for placed and predicted models. * Rugnux: Restraint dictionaries via `--restraints` (or among the inputs) and more built-in ligands; a model whose space group is a subgroup of the data's is refined in its own group; reflections plus a model select `--mode refine` on their own. * Rugnux: `-C` fixes the crystal system on the axes as typed; `-S` alone fixes the space group only, the cell is still found and reduced; a space group or cell stored in the input file is ignored, with a line in the log. * Rugnux: More reliable multi-sweep merging; `--mode scale` merges several stills runs as one; `_plot.txt` and the report name each input; a second crystal form is reported for stills given a cell. * jfjoch_broker (experimental): Live Rugnux on FPGA detectors (JUNGFRAU) takes the same path as a DECTRIS stream, rotation and serial alike, model step included, with analysis threads per GPU; frames waiting for the analysis can be viewed; no host stage drops a frame - a source faster than the analysis is held at the card. * jfjoch_broker: Summation of more than 256 frames per image is split between the card and the host (`summation_mode` is then `hybrid`), and live Rugnux runs on such collections; pedestal runs take a receiver of their own. * jfjoch_broker (experimental): `stills_downsample` analyses 1 of every N serial images (chosen by a hash; the first 2000 all; the others written without analysis; reset to 1 by `/initialize`); `max_spot_count` left out means 250 for per-image indexing and up to 1000 for live stills; the live status gives the merge before the model work, the fitted resolution and the anomalous peak height of each unexplained blob. * jfjoch_broker: Without a `pcie` entry the broker runs on a mock of the card instead of the HLS simulation; its spot, ROI and azimuthal results are zero then. * Web frontend: Redesigned Rugnux live page; Vite 8, TypeScript 6, Mol* 5.13 and dependency security updates. * Tools: `jfjoch_fpga_stills_test` measures live stills on the cards with a stored dataset; `jfjoch_fpga_test` loses the HLS-simulation option. * Other: FFTW no longer fetched or linked (the CPU FFT indexer keeps the name `FFTW`); PCIe driver now GPL-2.0-or-later and third-party notices complete; several crash fixes. Reviewed-on: #98 Co-authored-by: Filip Leonarski <filip.leonarski@psi.ch>
143 lines
5.8 KiB
C++
143 lines
5.8 KiB
C++
// SPDX-FileCopyrightText: 2026 Filip Leonarski, Paul Scherrer Institute <filip.leonarski@psi.ch>
|
|
// SPDX-License-Identifier: GPL-3.0-only
|
|
|
|
#include <catch2/catch_all.hpp>
|
|
|
|
#include <algorithm>
|
|
#include <cmath>
|
|
#include <complex>
|
|
#include <random>
|
|
#include <vector>
|
|
|
|
#include "../common/CpuFFT.h"
|
|
#include "../image_analysis/structure_refinement/ModelFFT.h"
|
|
|
|
// pocketfft (common/CpuFFT) on the shapes and layouts the call sites use, against a direct DFT in double
|
|
// precision: the forward transforms to float rounding, the backward ones by a round trip (an unnormalised
|
|
// forward + backward pair gives the input times the number of points).
|
|
|
|
namespace {
|
|
|
|
std::vector<float> Random(size_t n, unsigned seed) {
|
|
std::mt19937 rng(seed);
|
|
std::normal_distribution<float> g(0.0f, 1.0f);
|
|
std::vector<float> v(n);
|
|
for (auto &x : v)
|
|
x = g(rng);
|
|
return v;
|
|
}
|
|
|
|
// Largest difference relative to the largest value of the reference.
|
|
template <typename T, typename U>
|
|
double RelativeError(const std::vector<T> &a, const std::vector<U> &ref) {
|
|
REQUIRE(a.size() == ref.size());
|
|
double diff = 0, scale = 0;
|
|
for (size_t i = 0; i < a.size(); i++) {
|
|
diff = std::max(diff, std::abs(std::complex<double>(a[i]) - std::complex<double>(ref[i])));
|
|
scale = std::max(scale, std::abs(std::complex<double>(ref[i])));
|
|
}
|
|
return diff / scale;
|
|
}
|
|
|
|
// Forward DFT of a row-major array along every axis, computed directly, in double precision.
|
|
std::vector<std::complex<double>> DirectDFT(const std::vector<size_t> &shape, const std::vector<float> &in) {
|
|
std::vector<std::complex<double>> a(in.begin(), in.end());
|
|
size_t stride = a.size();
|
|
for (size_t n : shape) {
|
|
stride /= n;
|
|
std::vector<std::complex<double>> line(n);
|
|
for (size_t base = 0; base < a.size(); base++) {
|
|
if ((base / stride) % n != 0)
|
|
continue; // not the first element of a line along this axis
|
|
for (size_t k = 0; k < n; k++) {
|
|
std::complex<double> s = 0;
|
|
for (size_t j = 0; j < n; j++)
|
|
s += a[base + j * stride] * std::polar(1.0, -2.0 * M_PI * static_cast<double>(j * k % n) / n);
|
|
line[k] = s;
|
|
}
|
|
for (size_t k = 0; k < n; k++)
|
|
a[base + k * stride] = line[k];
|
|
}
|
|
}
|
|
return a;
|
|
}
|
|
|
|
// The half (n/2 + 1 values) of the last axis that a real transform keeps.
|
|
std::vector<std::complex<double>> HalfLastAxis(const std::vector<size_t> &shape,
|
|
const std::vector<std::complex<double>> &full) {
|
|
const size_t n = shape.back(), nc = n / 2 + 1;
|
|
std::vector<std::complex<double>> half;
|
|
for (size_t i = 0; i < full.size(); i++)
|
|
if (i % n < nc)
|
|
half.push_back(full[i]);
|
|
return half;
|
|
}
|
|
|
|
std::vector<float> Scaled(const std::vector<float> &v, double s) {
|
|
std::vector<float> r(v.size());
|
|
for (size_t i = 0; i < v.size(); i++)
|
|
r[i] = static_cast<float>(v[i] * s);
|
|
return r;
|
|
}
|
|
|
|
} // namespace
|
|
|
|
TEST_CASE("CpuFFT_BatchedR2CMatchesDirectDFT", "[fft]") {
|
|
// The FFT indexer's histograms: one transform per direction; 3142 = 2 x 1571 is a real length.
|
|
for (size_t n : {3142, 3150, 1024}) {
|
|
const size_t howmany = 5;
|
|
const std::vector<float> in = Random(howmany * n, n);
|
|
std::vector<std::complex<float>> ours(howmany * (n / 2 + 1));
|
|
cpu_fft::R2C({howmany, n}, {1}, in.data(), ours.data());
|
|
// DirectDFT over the batch axis too would mix the lines: transform each line on its own.
|
|
std::vector<std::complex<double>> ref;
|
|
for (size_t i = 0; i < howmany; i++) {
|
|
const std::vector<float> line(in.begin() + i * n, in.begin() + (i + 1) * n);
|
|
const auto r = HalfLastAxis({n}, DirectDFT({n}, line));
|
|
ref.insert(ref.end(), r.begin(), r.end());
|
|
}
|
|
CHECK(RelativeError(ours, ref) < 1e-5);
|
|
}
|
|
}
|
|
|
|
TEST_CASE("CpuFFT_2DRoundTrip", "[fft]") {
|
|
// The beam-centre search: a 2D r2c and c2r on a 2-3-5-7-smooth padded size.
|
|
const std::vector<size_t> shape = {210, 336};
|
|
const std::vector<float> in = Random(shape[0] * shape[1], 1);
|
|
std::vector<std::complex<float>> ours(shape[0] * (shape[1] / 2 + 1));
|
|
cpu_fft::R2C(shape, {0, 1}, in.data(), ours.data());
|
|
CHECK(RelativeError(ours, HalfLastAxis(shape, DirectDFT(shape, in))) < 1e-5);
|
|
|
|
std::vector<float> back(in.size());
|
|
cpu_fft::C2R(shape, {0, 1}, ours.data(), back.data());
|
|
CHECK(RelativeError(back, Scaled(in, static_cast<double>(in.size()))) < 1e-5);
|
|
}
|
|
|
|
TEST_CASE("CpuFFT_3DRoundTrip", "[fft]") {
|
|
// The model maps (u fastest, w halved) and the tNCS Patterson (row-major, last axis halved) are both a
|
|
// row-major 3D transform halving the last axis; an odd length and a factor 7 included.
|
|
const std::vector<size_t> shape = {70, 45, 54};
|
|
const std::vector<float> in = Random(shape[0] * shape[1] * shape[2], 2);
|
|
std::vector<std::complex<float>> ours(shape[0] * shape[1] * (shape[2] / 2 + 1));
|
|
cpu_fft::R2C(shape, {0, 1, 2}, in.data(), ours.data());
|
|
CHECK(RelativeError(ours, HalfLastAxis(shape, DirectDFT(shape, in))) < 1e-5);
|
|
|
|
std::vector<float> back(in.size());
|
|
cpu_fft::C2R(shape, {0, 1, 2}, ours.data(), back.data());
|
|
CHECK(RelativeError(back, Scaled(in, static_cast<double>(in.size()))) < 1e-5);
|
|
}
|
|
|
|
TEST_CASE("CpuFFT_MapToFPhiInParallelMatchesMapToFPhi", "[fft]") {
|
|
gemmi::Grid<float> map;
|
|
map.unit_cell.set(60, 50, 40, 90, 90, 90);
|
|
map.spacegroup = gemmi::find_spacegroup_by_name("P1");
|
|
map.set_size(60, 50, 40);
|
|
map.data = Random(map.data.size(), 3);
|
|
const int reach = 12;
|
|
const gemmi::FPhiGrid<float> ref = CropHalfL(MapToFPhi(map), reach);
|
|
for (size_t nthreads : {1, 4}) {
|
|
const gemmi::FPhiGrid<float> ours = MapToFPhiInParallel(map, reach, nthreads);
|
|
CHECK(RelativeError(ours.data, ref.data) < 1e-5);
|
|
}
|
|
}
|