pyCommonUtils.cpp File Reference

#include "shambase/DistributedData.hpp"
#include "shamalgs/primitives/compute_histogram.hpp"
#include "shamalgs/primitives/upper_bound.hpp"
#include "shambackends/DeviceBuffer.hpp"
#include "shambackends/MemPerfInfos.hpp"
#include "shambackends/kernel_call.hpp"
#include "shambindings/pybind11_stl.hpp"
#include "shambindings/pybindaliases.hpp"
#include "shambindings/pytypealias.hpp"
#include "shamcomm/logs.hpp"
#include "shamrock/solvergraph/Field.hpp"
#include "shamsys/NodeInstance.hpp"
#include <pybind11/cast.h>
#include <pybind11/complex.h>
#include <shambackends/sycl.hpp>
#include <vector>

Include dependency graph for pyCommonUtils.cpp:

Go to the source code of this file.

Namespaces
namespace	sham
	namespace for backends this one is named only sham since shambackends is too long to write

Functions
template<typename T>
DeviceBuffer< T > &	sham::operator+= (DeviceBuffer< T > &lhs, const DeviceBuffer< T > &rhs)
template<typename T>
DeviceBuffer< T > &	sham::operator/= (DeviceBuffer< T > &lhs, const DeviceBuffer< T > &rhs)
m	def ("compute_histogram", [](std::vector< f64 > bin_edges, shamrock::solvergraph::Field< f64 > &x_field, shamrock::solvergraph::Field< f64 > &y_field, bool do_average) -> std::vector< f64 > { auto dev_sched=shamsys::instance::get_compute_scheduler_ptr();u32 nx=bin_edges.size() - 1;std::vector< f64 > bin_edge_inf(nx);std::vector< f64 > bin_edge_sup(nx);for(size_t i=0;i< nx;i++) { bin_edge_inf[i]=bin_edges[i];bin_edge_sup[i]=bin_edges[i+1];} sham::DeviceBuffer< f64 > ret(bin_edge_inf.size(), dev_sched);ret.fill(0);sham::DeviceBuffer< f64 > bin_inf(bin_edge_inf.size(), dev_sched);sham::DeviceBuffer< f64 > bin_sup(bin_edge_inf.size(), dev_sched);bin_inf.copy_from_stdvec(bin_edge_inf);bin_sup.copy_from_stdvec(bin_edge_sup);shambase::DistributedData< u32 > obj_cnts=x_field.get_obj_cnts();obj_cnts.for_each([&](u64 id_patch, const unsigned int &obj_cnt) { ret+=shamalgs::primitives::compute_histogram< f64 >(dev_sched, bin_inf, bin_sup, obj_cnt, [](const f64 &bin_edge_inf, const f64 &bin_edge_sup, const f64 &x_val, const f64 &y_val, bool &has_value) { has_value=x_val >=bin_edge_inf &&x_val< bin_edge_sup;return has_value ? y_val :0;}, x_field.get_buf(id_patch), y_field.get_buf(id_patch));});shamalgs::collective::reduce_buffer_in_place_sum(ret, MPI_COMM_WORLD);if(do_average) { sham::DeviceBuffer< f64 > norm(bin_edge_inf.size(), dev_sched);norm.fill(0);obj_cnts.for_each([&](u64 id_patch, const unsigned int &obj_cnt) { sham::DeviceBuffer< f64 > unit_buf(obj_cnt, dev_sched);unit_buf.fill(1);norm+=shamalgs::primitives::compute_histogram< f64 >(dev_sched, bin_inf, bin_sup, obj_cnt, [](const f64 &bin_edge_inf, const f64 &bin_edge_sup, const f64 &x_val, const f64 &y_val, bool &has_value) { has_value=x_val >=bin_edge_inf &&x_val< bin_edge_sup;return has_value ? y_val :0;}, x_field.get_buf(id_patch), unit_buf);});shamalgs::collective::reduce_buffer_in_place_sum(norm, MPI_COMM_WORLD);ret/=norm;} return ret.copy_to_stdvec();}, py::kw_only{}, py::arg("bin_edges"), py::arg("x_field"), py::arg("y_field"), py::arg("do_average")=false)
m	def ("compute_histogram_convolve_x", [](std::vector< f64 > bin_edges, shamrock::solvergraph::Field< f64 > &x_field, shamrock::solvergraph::Field< f64 > &y_field, shamrock::solvergraph::Field< f64 > &size_field, bool do_average) -> std::vector< f64 > { auto dev_sched=shamsys::instance::get_compute_scheduler_ptr();u32 nx=bin_edges.size() - 1;std::vector< f64 > bin_edge_inf(nx);std::vector< f64 > bin_edge_sup(nx);for(size_t i=0;i< nx;i++) { bin_edge_inf[i]=bin_edges[i];bin_edge_sup[i]=bin_edges[i+1];} sham::DeviceBuffer< f64 > ret(bin_edge_inf.size(), dev_sched);ret.fill(0);sham::DeviceBuffer< f64 > bin_inf(bin_edge_inf.size(), dev_sched);sham::DeviceBuffer< f64 > bin_sup(bin_edge_inf.size(), dev_sched);bin_inf.copy_from_stdvec(bin_edge_inf);bin_sup.copy_from_stdvec(bin_edge_sup);shambase::DistributedData< u32 > obj_cnts=x_field.get_obj_cnts();obj_cnts.for_each([&](u64 id_patch, const unsigned int &obj_cnt) { ret+=shamalgs::primitives::compute_histogram< f64 >(dev_sched, bin_inf, bin_sup, obj_cnt, [](const f64 &bin_edge_inf, const f64 &bin_edge_sup, const f64 &x_val, const f64 &y_val, const f64 &size_val, bool &has_value) { has_value=x_val >=bin_edge_inf - size_val &&x_val< bin_edge_sup+size_val;return has_value ? y_val :0;}, x_field.get_buf(id_patch), y_field.get_buf(id_patch), size_field.get_buf(id_patch));});shamalgs::collective::reduce_buffer_in_place_sum(ret, MPI_COMM_WORLD);if(do_average) { sham::DeviceBuffer< f64 > norm(bin_edge_inf.size(), dev_sched);norm.fill(0);obj_cnts.for_each([&](u64 id_patch, const unsigned int &obj_cnt) { sham::DeviceBuffer< f64 > unit_buf(obj_cnt, dev_sched);unit_buf.fill(1);norm+=shamalgs::primitives::compute_histogram< f64 >(dev_sched, bin_inf, bin_sup, obj_cnt, [](const f64 &bin_edge_inf, const f64 &bin_edge_sup, const f64 &x_val, const f64 &y_val, const f64 &size_val, bool &has_value) { has_value=x_val >=bin_edge_inf - size_val &&x_val< bin_edge_sup+size_val;return has_value ? y_val :0;}, x_field.get_buf(id_patch), unit_buf, size_field.get_buf(id_patch));});shamalgs::collective::reduce_buffer_in_place_sum(norm, MPI_COMM_WORLD);ret/=norm;} return ret.copy_to_stdvec();}, py::kw_only{}, py::arg("bin_edges"), py::arg("x_field"), py::arg("y_field"), py::arg("size_field"), py::arg("do_average")=false)
m	def ("compute_histogram_2d", [](std::vector< f64 > bin_edges_x, std::vector< f64 > bin_edges_y, shamrock::solvergraph::Field< f64 > &x_field, shamrock::solvergraph::Field< f64 > &y_field) -> std::vector< u64 > { auto dev_sched=shamsys::instance::get_compute_scheduler_ptr();u32 nx=bin_edges_x.size() - 1;u32 ny=bin_edges_y.size() - 1;sham::DeviceBuffer< u64 > ret(nx *ny, dev_sched);ret.fill(0);shambase::DistributedData< u32 > obj_cnts=x_field.get_obj_cnts();sham::DeviceBuffer< f64 > binsx(bin_edges_x.size(), dev_sched);sham::DeviceBuffer< f64 > binsy(bin_edges_y.size(), dev_sched);binsx.copy_from_stdvec(bin_edges_x);binsy.copy_from_stdvec(bin_edges_y);obj_cnts.for_each([&](u64 id_patch, const unsigned int &obj_cnt) { sham::kernel_call(dev_sched->get_queue(), sham::MultiRef{ binsx, binsy, x_field.get_buf(id_patch), y_field.get_buf(id_patch)}, sham::MultiRef{ret}, obj_cnt, [nx, ny](u32 id, const f64 *__restrict x_bins, const f64 *__restrict y_bins, const f64 *__restrict x_field, const f64 *__restrict y_field, u64 *__restrict pic) { auto get_pic_coord=[&](u32 ix, u32 iy) { return ix+iy *nx;};f64 x_val=x_field[id];f64 y_val=y_field[id];bool is_in_x_range=x_bins[0]<=x_val &&x_val<=x_bins[nx];bool is_in_y_range=y_bins[0]<=y_val &&y_val<=y_bins[ny];if(!(is_in_x_range &&is_in_y_range)) { return;} u32 ix=shamalgs::primitives::binary_search_upper_bound(x_bins, 0, nx+1, x_val);u32 iy=shamalgs::primitives::binary_search_upper_bound(y_bins, 0, ny+1, y_val);if(ix >=nx||iy >=ny) { return;} using atomic_ref_T=sycl::atomic_ref< u64, sycl::memory_order_relaxed, sycl::memory_scope_device, sycl::access::address_space::global_space >;atomic_ref_T pic_ref(pic[get_pic_coord(ix, iy)]);pic_ref++;});});shamalgs::collective::reduce_buffer_in_place_sum(ret, MPI_COMM_WORLD);return ret.copy_to_stdvec();}, py::kw_only{}, py::arg("bin_edges_x"), py::arg("bin_edges_y"), py::arg("x_field"), py::arg("y_field"))

Variables
	ON_PYTHON_INIT

Detailed Description

Author

Timothée David–Cléris (tim.s.nosp@m.hamr.nosp@m.ock@p.nosp@m.roto.nosp@m.n.me)

Definition in file pyCommonUtils.cpp.

Variable Documentation

ON_PYTHON_INIT

ON_PYTHON_INIT

Initial value:

{
    auto &m = root_module

Definition at line 68 of file pyCommonUtils.cpp.