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>
169 lines
6.1 KiB
C++
169 lines
6.1 KiB
C++
// SPDX-FileCopyrightText: 2024 Filip Leonarski, Paul Scherrer Institute <filip.leonarski@psi.ch>
|
|
// SPDX-License-Identifier: GPL-3.0-only
|
|
|
|
#include "MXAnalysisAfterFPGA.h"
|
|
|
|
#include <span>
|
|
|
|
#include "spot_finding/DetModuleSpotFinder_cpu.h"
|
|
#include "../common/CUDAWrapper.h"
|
|
#include "../common/JFJochException.h"
|
|
#include "spot_finding/SpotUtils.h"
|
|
#include "bragg_prediction/BraggPredictionFactory.h"
|
|
|
|
double stddev(const std::vector<float> &v) {
|
|
if (v.size() <= 1)
|
|
return 0.0;
|
|
|
|
double mean = 0.0f;
|
|
|
|
for (const auto &i: v)
|
|
mean += i;
|
|
mean /= v.size();
|
|
|
|
double stddev = 0.0f;
|
|
for (const auto &i: v)
|
|
stddev += (i - mean) * (i - mean);
|
|
|
|
return sqrt(stddev / (v.size() - 1));
|
|
}
|
|
|
|
|
|
MXAnalysisAfterFPGA::MXAnalysisAfterFPGA(const DiffractionExperiment &in_experiment,
|
|
const AzimuthalIntegrationMapping &in_integration,
|
|
IndexAndRefine &indexer)
|
|
: experiment(in_experiment),
|
|
integration(in_integration),
|
|
indexer(indexer),
|
|
prediction(CreateBraggPrediction(experiment.IsRotationIndexing())),
|
|
bragg_engine(std::make_unique<BraggIntegrationEngineCPU>(in_experiment)) {
|
|
if (experiment.IsSpotFindingEnabled())
|
|
find_spots = true;
|
|
|
|
if (experiment.GetAzimuthalIntegrationSettings().IsForceCPUinFPGAWorkflow())
|
|
cpu_azint = std::make_unique<AzIntEngineCPU>(integration);
|
|
}
|
|
|
|
void MXAnalysisAfterFPGA::RunAzimuthalIntegration(const void *image, AzimuthalIntegrationProfile &profile) {
|
|
if (!cpu_azint)
|
|
return;
|
|
|
|
const size_t npixel = experiment.GetPixelsNum();
|
|
|
|
auto run = [&](auto tag) {
|
|
using pixel_t = decltype(tag);
|
|
cpu_azint->RunAzint(std::span<const pixel_t>(static_cast<const pixel_t *>(image), npixel), profile);
|
|
};
|
|
|
|
switch (experiment.GetByteDepthImage()) {
|
|
case 1:
|
|
experiment.IsPixelSigned() ? run(int8_t{}) : run(uint8_t{});
|
|
break;
|
|
case 2:
|
|
experiment.IsPixelSigned() ? run(int16_t{}) : run(uint16_t{});
|
|
break;
|
|
case 4:
|
|
experiment.IsPixelSigned() ? run(int32_t{}) : run(uint32_t{});
|
|
break;
|
|
default:
|
|
throw JFJochException(JFJochExceptionCategory::InputParameterInvalid, "Pixel depth unsupported");
|
|
}
|
|
}
|
|
|
|
void MXAnalysisAfterFPGA::ReadFromFPGA(const DeviceOutput *output, const SpotFindingSettings &settings, size_t module_number) {
|
|
if (state == State::Disabled || !find_spots || !settings.enable) {
|
|
state = State::Disabled;
|
|
} else {
|
|
const auto t0 = std::chrono::steady_clock::now();
|
|
StrongPixelSet strong_pixel_set;
|
|
strong_pixel_set.ReadFPGAOutput(experiment, *output);
|
|
strong_pixel_set.FindSpots(experiment, settings, spots, module_number);
|
|
const auto t1 = std::chrono::steady_clock::now();
|
|
spot_finding_time_total += (t1 - t0);
|
|
spot_finding_timing_active = true;
|
|
|
|
state = State::Enabled;
|
|
}
|
|
}
|
|
|
|
void MXAnalysisAfterFPGA::ReadFromCPU(DeviceOutput *output, const SpotFindingSettings &settings, size_t module_number) {
|
|
std::unique_lock ul(read_from_cpu_mutex);
|
|
|
|
if (state == State::Disabled || !find_spots || !settings.enable) {
|
|
state = State::Disabled;
|
|
} else {
|
|
const auto t0 = std::chrono::steady_clock::now();
|
|
|
|
state = State::Enabled;
|
|
|
|
std::vector<float> d_map(RAW_MODULE_SIZE);
|
|
|
|
experiment.CalcSpotFinderResolutionMap(d_map.data(), module_number);
|
|
|
|
arr_mean.resize(RAW_MODULE_SIZE);
|
|
arr_sttdev.resize(RAW_MODULE_SIZE);
|
|
arr_valid_count.resize(RAW_MODULE_SIZE);
|
|
arr_strong_pixel.resize(RAW_MODULE_SIZE);
|
|
|
|
if (experiment.GetByteDepthImage() == 2)
|
|
FindSpots(*output,
|
|
settings,
|
|
d_map.data(),
|
|
arr_mean.data(),
|
|
arr_sttdev.data(),
|
|
arr_valid_count.data(),
|
|
arr_strong_pixel.data());
|
|
else if (experiment.GetByteDepthImage() == 4)
|
|
FindSpots<int32_t>(*output,
|
|
settings,
|
|
d_map.data(),
|
|
arr_mean.data(),
|
|
arr_sttdev.data(),
|
|
arr_valid_count.data(),
|
|
arr_strong_pixel.data());
|
|
else if (experiment.GetByteDepthImage() == 1)
|
|
FindSpots<int8_t>(*output,
|
|
settings,
|
|
d_map.data(),
|
|
arr_mean.data(),
|
|
arr_sttdev.data(),
|
|
arr_valid_count.data(),
|
|
arr_strong_pixel.data());
|
|
|
|
StrongPixelSet strong_pixel_set;
|
|
strong_pixel_set.ReadFPGAOutput(experiment, *output);
|
|
strong_pixel_set.FindSpots(experiment, settings, spots, module_number);
|
|
|
|
const auto t1 = std::chrono::steady_clock::now();
|
|
spot_finding_time_total += (t1 - t0);
|
|
spot_finding_timing_active = true;
|
|
}
|
|
}
|
|
|
|
void MXAnalysisAfterFPGA::Process(DataMessage &message, const SpotFindingSettings& spot_finding_settings) {
|
|
if (find_spots && (state == State::Enabled)) {
|
|
const auto t0 = std::chrono::steady_clock::now();
|
|
SpotAnalyze(experiment, spot_finding_settings, spots, message);
|
|
const auto t1 = std::chrono::steady_clock::now();
|
|
spot_finding_time_total += (t1 - t0);
|
|
|
|
if (spot_finding_settings.indexing)
|
|
indexer.ProcessImage(message, spot_finding_settings, *prediction,
|
|
[this, &message](const std::vector<Reflection> &predicted, size_t npredicted, int64_t image_number) {
|
|
return bragg_engine->Run(message.image, predicted, npredicted, image_number);
|
|
});
|
|
}
|
|
|
|
if (spot_finding_timing_active) {
|
|
// total spot-finding time for the whole image
|
|
message.spot_finding_time_s = spot_finding_time_total.count() / 1e6;
|
|
|
|
// report/store ms here
|
|
spot_finding_time_total = std::chrono::duration<double, std::micro>{0.0};
|
|
spot_finding_timing_active = false;
|
|
}
|
|
|
|
spots.clear();
|
|
state = State::Idle;
|
|
}
|