Replace the free functions BraggIntegrate2D/ProfileIntegrate2D with the BraggIntegrationEngine (CPU/GPU) as the live integrator. - IndexAndRefine no longer holds the integrator: ProcessImage takes a per-worker BraggIntegrateFn callback (ProcessImage is called concurrently by the shared IndexAndRefine, so the stateful engine must not be a member). - WithoutFPGA/jfjoch_process: owns a GPU engine when a GPU is present, else CPU, and passes the GPU-resident preprocessed buffer so integration runs on-device. - AfterFPGA: forces CPU and integrates straight off the assembled CompressedImage via a templated per-pixel sampler - only the reflection-disk pixels are read, no whole-image copy (the FPGA host runs up to 36 GB/s). Sampler maps type min/max to INT32_MIN/INT32_MAX on read; special/saturation only, no +/-1 band. - Remove BraggIntegrate2D/ProfileIntegrate2D and their test; keep IntegratorMode. Prediction: buffer up to 20000 candidates but return the 10000 closest to the Ewald sphere (deterministic partial_sort on |dist_ewald|, hkl tiebreak) instead of the GPU atomic-fill order. Serialized output stays <=10000, so the frame transport headroom and its CBOR guard are unchanged. integration_model exposed via OpenAPI (bragg_integration_settings schema, /config/bragg_integration PUT/GET, added to jfjoch_settings and jfjoch_statistics) and the frontend (BraggIntegrationSettings dropdown). Regenerated C++/TS clients and redoc. Validated old-vs-new on all 18 /data/rotation_test crystals: indexing rate and space group bit-identical; ISa/CC identical on 16/18 (one improved, EcwtAL500 ISa 0.0->6.7); new CompressedImage-vs-buffer and GPU-vs-CPU parity tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
138 lines
6.1 KiB
C++
138 lines
6.1 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 <cmath>
|
|
#include <cstdint>
|
|
#include <vector>
|
|
|
|
#include "../common/BraggIntegrationSettings.h"
|
|
#include "../common/CompressedImage.h"
|
|
#include "../common/DetectorSetup.h"
|
|
#include "../common/DiffractionExperiment.h"
|
|
#include "../common/Reflection.h"
|
|
#include "../image_analysis/bragg_integration/BraggIntegrationEngineCPU.h"
|
|
#include "../image_analysis/image_preprocessing/ImagePreprocessorBuffer.h"
|
|
|
|
// The FPGA workflow integrates straight off the assembled CompressedImage (any pixel type) instead of
|
|
// a preprocessed int32 buffer, reading only the reflection disks. These tests pin that the typed
|
|
// sampler produces exactly the same intensities as the equivalent int32 preprocessed buffer - i.e. the
|
|
// zero-copy route is numerically identical to the buffer route, not merely close.
|
|
|
|
namespace {
|
|
|
|
struct Scene {
|
|
std::vector<int32_t> image; // INT32_MIN = masked, INT32_MAX = saturated
|
|
std::vector<Reflection> predicted;
|
|
};
|
|
|
|
Reflection MakeReflection(float x, float y, float d, int hkl) {
|
|
Reflection r{};
|
|
r.h = hkl; r.k = hkl; r.l = hkl;
|
|
r.predicted_x = x;
|
|
r.predicted_y = y;
|
|
r.d = d;
|
|
r.rlp = 1.0f;
|
|
r.partiality = 1.0f;
|
|
return r;
|
|
}
|
|
|
|
// A grid of clean Gaussian spots on a flat background, values kept well inside int16 range so the same
|
|
// scene can be represented as Int16 and Int32 without clipping. A few masked/saturated pixels exercise
|
|
// the sentinel handling.
|
|
Scene BuildScene(size_t width, size_t height) {
|
|
Scene s;
|
|
s.image.assign(width * height, 12);
|
|
const int margin = 45, spacing = 60;
|
|
int hkl = 1;
|
|
for (int gy = 0; margin + gy * spacing < static_cast<int>(height) - margin; ++gy)
|
|
for (int gx = 0; margin + gx * spacing < static_cast<int>(width) - margin; ++gx) {
|
|
const float cx = static_cast<float>(margin + gx * spacing) + 0.3f;
|
|
const float cy = static_cast<float>(margin + gy * spacing) - 0.2f;
|
|
const double amp = 150.0 + 40.0 * ((gx * 7 + gy * 13) % 20); // <= ~950, safe for int16
|
|
const double sigma = 1.3;
|
|
for (int dy = -6; dy <= 6; ++dy)
|
|
for (int dx = -6; dx <= 6; ++dx) {
|
|
const int x = static_cast<int>(std::lround(cx)) + dx;
|
|
const int y = static_cast<int>(std::lround(cy)) + dy;
|
|
if (x < 0 || y < 0 || x >= static_cast<int>(width) || y >= static_cast<int>(height)) continue;
|
|
const double ex = x - cx, ey = y - cy;
|
|
s.image[y * width + x] += static_cast<int32_t>(std::lround(amp * std::exp(-(ex * ex + ey * ey) / (2.0 * sigma * sigma))));
|
|
}
|
|
const float d = 1.4f + 0.12f * static_cast<float>((gx + gy) % 12);
|
|
s.predicted.push_back(MakeReflection(cx, cy, d, hkl++));
|
|
}
|
|
for (int k = 0; k < 20; ++k) {
|
|
const size_t idx = (static_cast<size_t>(k) * 2654435761u) % s.image.size();
|
|
s.image[idx] = (k % 2) ? INT32_MIN : INT32_MAX;
|
|
}
|
|
return s;
|
|
}
|
|
|
|
DiffractionExperiment MakeExperiment(IntegratorMode mode) {
|
|
DiffractionExperiment experiment(DetJF(2));
|
|
experiment.DetectorDistance_mm(100.0f).IncidentEnergy_keV(WVL_1A_IN_KEV).BeamX_pxl(400.0f).BeamY_pxl(400.0f);
|
|
BraggIntegrationSettings settings;
|
|
settings.Integrator(mode);
|
|
experiment.ImportBraggIntegrationSettings(settings);
|
|
return experiment;
|
|
}
|
|
|
|
void RequireIdentical(const std::vector<Reflection> &a, const std::vector<Reflection> &b) {
|
|
REQUIRE(a.size() == b.size());
|
|
REQUIRE(a.size() > 40);
|
|
for (size_t i = 0; i < a.size(); ++i) {
|
|
INFO("reflection " << i << " hkl " << a[i].h);
|
|
CHECK(a[i].h == b[i].h);
|
|
CHECK(a[i].I == b[i].I);
|
|
CHECK(a[i].sigma == b[i].sigma);
|
|
CHECK(a[i].bkg == b[i].bkg);
|
|
}
|
|
}
|
|
|
|
} // namespace
|
|
|
|
TEST_CASE("BraggIntegrationEngineCPU_CompressedImageMatchesBuffer", "[Integration]") {
|
|
for (const auto mode : {IntegratorMode::BoxSum, IntegratorMode::ProfileGaussian, IntegratorMode::ProfileEmpirical}) {
|
|
const DiffractionExperiment experiment = MakeExperiment(mode);
|
|
const size_t W = experiment.GetXPixelsNum(), H = experiment.GetYPixelsNum();
|
|
const size_t npixel = experiment.GetPixelsNum();
|
|
const Scene scene = BuildScene(W, H);
|
|
REQUIRE(scene.image.size() == npixel);
|
|
|
|
BraggIntegrationEngineCPU engine(experiment);
|
|
|
|
// Route A: the preprocessed int32 buffer.
|
|
ImagePreprocessorBuffer buffer(npixel);
|
|
for (size_t i = 0; i < npixel; ++i) buffer[i] = scene.image[i];
|
|
const auto from_buffer = engine.Run(buffer, scene.predicted, scene.predicted.size(), 7);
|
|
|
|
SECTION("Int32 CompressedImage") {
|
|
const CompressedImage image(scene.image, W, H);
|
|
const auto from_image = engine.Run(image, scene.predicted, scene.predicted.size(), 7);
|
|
RequireIdentical(from_buffer, from_image);
|
|
}
|
|
|
|
SECTION("Int16 CompressedImage") {
|
|
// Same scene as int16 (masked -> INT16_MIN, saturated -> INT16_MAX) with a matching int32
|
|
// buffer built by the exact sampler mapping; the two must integrate identically.
|
|
std::vector<int16_t> img16(npixel);
|
|
std::vector<int32_t> buf32(npixel);
|
|
for (size_t i = 0; i < npixel; ++i) {
|
|
const int32_t v = scene.image[i];
|
|
if (v == INT32_MIN) { img16[i] = INT16_MIN; buf32[i] = INT32_MIN; }
|
|
else if (v == INT32_MAX) { img16[i] = INT16_MAX; buf32[i] = INT32_MAX; }
|
|
else { img16[i] = static_cast<int16_t>(v); buf32[i] = v; }
|
|
}
|
|
ImagePreprocessorBuffer buffer16(npixel);
|
|
for (size_t i = 0; i < npixel; ++i) buffer16[i] = buf32[i];
|
|
const auto ref16 = engine.Run(buffer16, scene.predicted, scene.predicted.size(), 7);
|
|
|
|
const CompressedImage image(img16, W, H);
|
|
const auto from_image = engine.Run(image, scene.predicted, scene.predicted.size(), 7);
|
|
RequireIdentical(ref16, from_image);
|
|
}
|
|
}
|
|
}
|