Skip to content

GitLab

Commit 6e02111c authored by Ethan Coon's avatar Ethan Coon
Browse files

working multiple test in legion with multiple partitions

parent acee1e39
Hide whitespace changes
Inline Side-by-side
tests/tests_legion/CanopyHydrology_kern1_multiple.cc
View file @ 6e02111c
......@@ -22,6 +22,7 @@
#include "readers.hh"
#include "CanopyHydrology.hh"
#include "legion.h"
#include "domains.hh"
using namespace Legion;
......@@ -54,12 +55,6 @@ enum FieldIDs {
};
} // namespace
static const int n_months = 12;
static const int n_pfts = 17;
static const int n_max_times = 31 * 24 * 2; // max days per month times hours per
// day * half hour timestep
static const int n_grid_cells = 24;
std::array<double,3> SumMinMaxReduction(const Task *task,
const std::vector<PhysicalRegion> &regions,
......@@ -68,6 +63,7 @@ std::array<double,3> SumMinMaxReduction(const Task *task,
assert(regions.size() == 1);
assert(task->regions.size() == 1);
assert(task->regions[0].privilege_fields.size() == 1);
std::cout << "LOG: Executing SumMinMax Task" << std::endl;
FieldID fid = *(task->regions[0].privilege_fields.begin());
FieldAccessor<READ_ONLY,double,2,coord_t,
......@@ -98,8 +94,14 @@ void InitPhenology(const Task *task,
std::cout << "LOG: Executing InitPhenology task" << std::endl;
const FieldAccessor<WRITE_DISCARD,double,2> elai(regions[0], FieldIDs::ELAI);
const FieldAccessor<WRITE_DISCARD,double,2> esai(regions[0], FieldIDs::ESAI);
ELM::Utils::read_phenology("../links/surfacedataWBW.nc", n_months, n_pfts, 0, elai, esai);
ELM::Utils::read_phenology("../links/surfacedataBRW.nc", n_months, n_pfts, n_months, elai, esai);
Rect<2> my_bounds = Domain(runtime->get_index_space_domain(regions[0].get_logical_region().get_index_space()));
coord_t n_grid_cells = my_bounds.hi[0] - my_bounds.lo[0] + 1;
coord_t n_pfts = my_bounds.hi[1] - my_bounds.lo[1] + 1;
assert(n_grid_cells == 24); // hard coded as two reads of 2x 12 increments
ELM::Utils::read_phenology("../links/surfacedataWBW.nc", 12, n_pfts, 0, elai, esai);
ELM::Utils::read_phenology("../links/surfacedataBRW.nc", 12, n_pfts, 12, elai, esai);
}
int InitForcing(const Task *task,
......@@ -111,17 +113,20 @@ int InitForcing(const Task *task,
assert(task->regions[0].privilege_fields.size() == 4); // rain, snow, temp, irrig
std::cout << "LOG: Executing InitForcing task" << std::endl;
Rect<2> my_bounds = Domain(runtime->get_index_space_domain(regions[0].get_logical_region().get_index_space()));
coord_t n_times_max = my_bounds.hi[0] - my_bounds.lo[0] + 1;
coord_t n_grid_cells = my_bounds.hi[1] - my_bounds.lo[1] + 1;
// init rain, snow, and air temp through reader
const FieldAccessor<WRITE_DISCARD,double,2> rain(regions[0], FieldIDs::FORC_RAIN);
const FieldAccessor<WRITE_DISCARD,double,2> snow(regions[0], FieldIDs::FORC_SNOW);
const FieldAccessor<WRITE_DISCARD,double,2> air_temp(regions[0], FieldIDs::FORC_AIR_TEMP);
int n_times = ELM::Utils::read_forcing("../links/forcing", n_max_times, 0, n_grid_cells,
int n_times = ELM::Utils::read_forcing("../links/forcing", n_times_max, 0, n_grid_cells,
rain, snow, air_temp);
// init irrig to zero
const FieldAccessor<WRITE_DISCARD,double,2> irrig(regions[0], FieldIDs::FORC_IRRIG);
for (size_t t=0; t!=n_max_times; ++t) {
for (size_t t=0; t!=n_times_max; ++t) {
for (size_t g=0; g!=n_grid_cells; ++g) {
irrig[t][g] = 0.;
}
......@@ -137,6 +142,7 @@ void CanopyHydrology_Interception_task(const Task *task,
{
assert(regions.size() == 3);
assert(task->regions.size() == 3);
std::cout << "LOG: Executing Interception task" << std::endl;
// process args / parameters
int lcv_time;
......@@ -172,11 +178,16 @@ void CanopyHydrology_Interception_task(const Task *task,
const AffineAccessorRW qflx_rain_grnd(regions[2], FieldIDs::QFLX_RAIN_GRND);
const AffineAccessorRW h2ocan(regions[2], FieldIDs::H2O_CAN);
LogicalRegion lr = regions[2].get_logical_region();
IndexSpaceT<2> is(lr.get_index_space());
Rect<2> bounds = Domain(runtime->get_index_space_domain(is));
std::cout << "LOG: With bounds: " << bounds.lo << "," << bounds.hi << std::endl;
int n_irrig_steps_left = 0.; // NOTE: still not physical quite sure what to do with this one.
for (size_t g = 0; g != n_grid_cells; ++g) {
for (size_t p = 0; p != n_pfts; ++p) {
for (size_t g = bounds.lo[0]; g != bounds.hi[0]+1; ++g) {
for (size_t p = bounds.lo[1]; p != bounds.hi[1]+1; ++p) {
ELM::CanopyHydrology_Interception(dtime,
forc_rain[lcv_time][g], forc_snow[lcv_time][g], forc_irrig[lcv_time][g],
ltype, ctype, urbpoi, do_capsnow,
......@@ -197,105 +208,59 @@ void top_level_task(const Task *task,
{
std::cout << "LOG: Executing Top Level Task" << std::endl;
const int n_pfts = 17;
const int n_times_max = 31 * 24 * 2; // max days per month times hours per
// day * half hour timestep
const int n_grid_cells = 24;
const int n_parts = 4;
// -----------------------------------------------------------------------------
// SETUP Phase
// -----------------------------------------------------------------------------
//
// Create index spaces
// Create data
//
// create a domain and index space for PFT-state
const Rect<2> pft_state_domain(Point<2>(0,0), Point<2>(n_grid_cells-1, n_pfts-1));
IndexSpace pft_state_is = runtime->create_index_space(ctx, pft_state_domain);
std::cout << "LOG: Created index space for PFTs: " << pft_state_is.get_id() << std::endl;
// create a domain and index space for forcing data
const Rect<2> forcing_domain(Point<2>(0,0), Point<2>(n_max_times-1,n_grid_cells-1));
IndexSpace forcing_is = runtime->create_index_space(ctx, forcing_domain);
std::cout << "LOG: Created index space for forcing data: " << forcing_is.get_id() << std::endl;
// grid cell x pft data for phenology
auto phenology_fs_ids = std::vector<unsigned>{ FieldIDs::ELAI, FieldIDs::ESAI };
Data2D phenology(n_grid_cells, n_pfts, n_parts, "phenology", phenology_fs_ids,
ctx, runtime);
//
// Create field spaces
//
// phenology field space
FieldSpace phenology_fs = runtime->create_field_space(ctx);
auto phenology_fs_ids = std::vector<FieldIDs::FieldIDs>{ FieldIDs::ELAI, FieldIDs::ESAI };
printf("Created field space for phenology: %x\n", phenology_fs.get_id());
{
FieldAllocator allocator = runtime->create_field_allocator(ctx, phenology_fs);
for (auto id : phenology_fs_ids) allocator.allocate_field(sizeof(double), id);
}
// forcing field space
FieldSpace forcing_fs = runtime->create_field_space(ctx);
auto forcing_fs_ids = std::vector<FieldIDs::FieldIDs>{
// ntimes x grid cells forcing data
auto forcing_fs_ids = std::vector<unsigned>{
FieldIDs::FORC_RAIN, FieldIDs::FORC_SNOW, FieldIDs::FORC_AIR_TEMP, FieldIDs::FORC_IRRIG};
printf("Created field space for forcing: %x\n", forcing_fs.get_id());
{
FieldAllocator allocator = runtime->create_field_allocator(ctx, forcing_fs);
for (auto id : forcing_fs_ids) allocator.allocate_field(sizeof(double), id);
}
// grid-cell flux data field space
FieldSpace flux_fs = runtime->create_field_space(ctx);
auto flux_fs_ids = std::vector<FieldIDs::FieldIDs>{
Data2D_Transposed forcing(n_grid_cells, n_times_max, n_parts, "forcing", forcing_fs_ids,
ctx, runtime);
// grid cell x pft water state and flux outputs
auto flux_fs_ids = std::vector<unsigned>{
FieldIDs::QFLX_PREC_INTR, FieldIDs::QFLX_IRRIG, FieldIDs::QFLX_PREC_GRND,
FieldIDs::QFLX_SNWCP_LIQ, FieldIDs::QFLX_SNWCP_ICE,
FieldIDs::QFLX_SNOW_GRND_PATCH, FieldIDs::QFLX_RAIN_GRND,
FieldIDs::H2O_CAN};
printf("Created field space for PFT-level fluxes: %x\n", flux_fs.get_id());
{
FieldAllocator allocator = runtime->create_field_allocator(ctx, flux_fs);
for (auto id : flux_fs_ids) allocator.allocate_field(sizeof(double), id);
}
//
// Physical Regions
//
// Logical region is the cross product of IndexSpace and FieldSpace --
// create logical regions, physical regions, and the block til memory is
// available.
LogicalRegion phenology_lr = runtime->create_logical_region(ctx, pft_state_is, phenology_fs);
RegionRequirement phenology_req(phenology_lr, READ_WRITE, EXCLUSIVE, phenology_lr);
for (auto id : phenology_fs_ids) phenology_req.add_field(id);
InlineLauncher phenology_region_launcher(phenology_req);
PhysicalRegion phenology_region = runtime->map_region(ctx, phenology_region_launcher);
LogicalRegion forcing_lr = runtime->create_logical_region(ctx, forcing_is, forcing_fs);
RegionRequirement forcing_req(forcing_lr, READ_WRITE, EXCLUSIVE, forcing_lr);
for (auto id : forcing_fs_ids) forcing_req.add_field(id);
InlineLauncher forcing_region_launcher(forcing_req);
PhysicalRegion forcing_region = runtime->map_region(ctx, forcing_region_launcher);
LogicalRegion flux_lr = runtime->create_logical_region(ctx, pft_state_is, flux_fs);
RegionRequirement flux_req(flux_lr, READ_WRITE, EXCLUSIVE, flux_lr);
for (auto id : flux_fs_ids) flux_req.add_field(id);
InlineLauncher flux_region_launcher(flux_req);
PhysicalRegion flux_region = runtime->map_region(ctx, flux_region_launcher);
Data2D flux(n_grid_cells, n_pfts, n_parts, "flux", flux_fs_ids, ctx, runtime);
// -----------------------------------------------------------------------------
// END SETUP
// -----------------------------------------------------------------------------
// create a color space for indexed launching. This is what a Data1D
// color_space would look like.
auto color_space = Rect<1>(Point<1>(0), Point<1>(n_parts-1));
// -----------------------------------------------------------------------------
// Initialization Phase
// -----------------------------------------------------------------------------
// launch task to read phenology
std::cout << "LOG: Launching Init Phenology" << std::endl;
// phenology_region.wait_until_valid(); // actually likely not needed, implicit wait on accessor
TaskLauncher phenology_launcher(TaskIDs::INIT_PHENOLOGY, TaskArgument(NULL, 0));
phenology_launcher.add_region_requirement(
RegionRequirement(phenology_lr, WRITE_DISCARD, EXCLUSIVE, phenology_lr));
RegionRequirement(phenology.logical_region, WRITE_DISCARD, EXCLUSIVE,
phenology.logical_region));
for (auto id : phenology_fs_ids) phenology_launcher.add_field(0,id);
runtime->execute_task(ctx, phenology_launcher);
// launch task to read forcing
std::cout << "LOG: Launching Init Forcing" << std::endl;
// forcing_region.wait_until_valid(); // actually likely not needed, implicit wait on accessor
TaskLauncher forcing_launcher(TaskIDs::INIT_FORCING, TaskArgument(NULL, 0));
forcing_launcher.add_region_requirement(
RegionRequirement(forcing_lr, WRITE_DISCARD, EXCLUSIVE, forcing_lr));
RegionRequirement(forcing.logical_region, WRITE_DISCARD, EXCLUSIVE,
forcing.logical_region));
for (auto id : forcing_fs_ids) forcing_launcher.add_field(0,id);
auto forcing_future = runtime->execute_task(ctx, forcing_launcher);
int n_times = forcing_future.get_result<int>();
......@@ -317,37 +282,46 @@ void top_level_task(const Task *task,
soln_file << "Time\t Total Canopy Water\t Min Water\t Max Water" << std::endl;
soln_file << std::setprecision(16) << 0 << "\t" << 0.0 << "\t" << 0.0 << "\t" << 0.0 << std::endl;
// DEBUG HACK --ETC
//n_times = 12;
// END DEBUG HACK
std::vector<Future> futures;
for (int i=0; i!=n_times; ++i) {
auto args = std::make_tuple(i, dtime, ltype, ctype, urbpoi, do_capsnow, dewmx, frac_veg_nosno);
TaskLauncher interception_launcher(TaskIDs::CANOPY_HYDROLOGY_INTERCEPTION,
TaskArgument(&args, sizeof(args)));
ArgumentMap arg_map;
IndexLauncher interception_launcher(TaskIDs::CANOPY_HYDROLOGY_INTERCEPTION,
color_space, TaskArgument(&args, sizeof(args)), arg_map);
// -- permissions on forcing
interception_launcher.add_region_requirement(
RegionRequirement(forcing_lr, READ_ONLY, EXCLUSIVE, forcing_lr));
RegionRequirement(forcing.logical_partition, forcing.projection_id,
READ_ONLY, EXCLUSIVE, forcing.logical_region));
interception_launcher.add_field(0, FieldIDs::FORC_RAIN);
interception_launcher.add_field(0, FieldIDs::FORC_SNOW);
interception_launcher.add_field(0, FieldIDs::FORC_IRRIG);
// -- permissions on phenology
interception_launcher.add_region_requirement(
RegionRequirement(phenology_lr, READ_ONLY, EXCLUSIVE, phenology_lr));
RegionRequirement(phenology.logical_partition, phenology.projection_id,
READ_ONLY, EXCLUSIVE, phenology.logical_region));
interception_launcher.add_field(1, FieldIDs::ELAI);
interception_launcher.add_field(1, FieldIDs::ESAI);
// -- permissions on output
interception_launcher.add_region_requirement(
RegionRequirement(flux_lr, READ_WRITE, EXCLUSIVE, flux_lr));
RegionRequirement(flux.logical_partition, flux.projection_id,
READ_WRITE, EXCLUSIVE, flux.logical_region));
for (auto id : flux_fs_ids) interception_launcher.add_field(2, id);
// -- launch the interception
runtime->execute_task(ctx, interception_launcher);
runtime->execute_index_space(ctx, interception_launcher);
// launch accumulator for h2ocan
// NOTE: need to somehow make this a reduction or something? Shouldn't be on the full region!
TaskLauncher accumlate_launcher(TaskIDs::UTIL_SUM_MIN_MAX_REDUCTION, TaskArgument());
accumlate_launcher.add_region_requirement(
RegionRequirement(flux_lr, READ_ONLY, EXCLUSIVE, flux_lr));
RegionRequirement(flux.logical_region, READ_ONLY, EXCLUSIVE, flux.logical_region));
accumlate_launcher.add_field(0, FieldIDs::H2O_CAN);
futures.push_back(runtime->execute_task(ctx, accumlate_launcher));
}
......@@ -363,21 +337,9 @@ void top_level_task(const Task *task,
<< "\t" << sum_min_max[1]
<< "\t" << sum_min_max[2] << std::endl;
}
// clean up resources
soln_file.close();
runtime->destroy_logical_region(ctx, phenology_lr);
runtime->destroy_logical_region(ctx, forcing_lr);
runtime->destroy_logical_region(ctx, flux_lr);
runtime->destroy_field_space(ctx, phenology_fs);
runtime->destroy_field_space(ctx, forcing_fs);
runtime->destroy_field_space(ctx, flux_fs);
runtime->destroy_index_space(ctx, pft_state_is);
runtime->destroy_index_space(ctx, forcing_is);
}
// Main just calls top level task
int main(int argc, char **argv)
{
......@@ -417,6 +379,12 @@ int main(int argc, char **argv)
registrar.set_leaf();
Runtime::preregister_task_variant<std::array<double,3>,SumMinMaxReduction>(registrar, "sum_min_max_reduction");
}
Runtime::preregister_projection_functor(Data2D::projection_id,
new Data2D::LocalProjectionFunction());
Runtime::preregister_projection_functor(Data2D_Transposed::projection_id,
new Data2D_Transposed::LocalProjectionFunction());
return Runtime::start(argc, argv);
}
......
tests/tests_legion/domains.hh 0 → 100644
View file @ 6e02111c
#ifndef DOMAINS_HH_
#define DOMAINS_HH_
#include "legion.h"
using namespace Legion;
namespace ProjectionIDs {
enum ProjectionIDs {
IDENTITY,
TO_1D_FROM_2D,
TO_1D_FROM_2D_TRANSPOSED
};
} // namespace
struct Data1D {
Data1D(coord_t n_grid_cells_, coord_t n_parts_,
const std::vector<unsigned>& field_ids_,
const std::string& name_,
Context ctx_, Runtime* runtime_) :
name(name_),
n_grid_cells(n_grid_cells_),
domain(Point<1>(0), Point<1>(n_grid_cells_-1)),
field_ids(field_ids_),
n_parts(n_parts_),
color_domain(Point<1>(0), Point<1>(n_parts_-1)),
ctx(ctx_),
runtime(runtime_)
{
index_space = runtime->create_index_space<1, coord_t>(ctx, domain);
auto block = Point<1>(n_grid_cells / n_parts);
index_partition = runtime->create_partition_by_blockify<1,coord_t>(ctx, index_space, block);
runtime->attach_name(index_partition, (std::string("index_partition_")+name).c_str());
field_space = runtime->create_field_space(ctx);
{
FieldAllocator allocator = runtime->create_field_allocator(ctx, field_space);
for (auto fid : field_ids) allocator.allocate_field(sizeof(double), fid);
}
runtime->attach_name(field_space, (std::string("field_space_")+name).c_str());
logical_region = runtime->create_logical_region(ctx, index_space, field_space);
runtime->attach_name(logical_region, (std::string("logical_region_")+name).c_str());
logical_partition = runtime->get_logical_partition(ctx, logical_region, index_partition);
runtime->attach_name(logical_partition, (std::string("logical_partition_")+name).c_str());
}
~Data1D() {
runtime->destroy_logical_partition(ctx, logical_partition);
runtime->destroy_logical_region(ctx, logical_region);
runtime->destroy_field_space(ctx, field_space);
runtime->destroy_index_partition(ctx, index_partition);
runtime->destroy_index_space(ctx, index_space);
}
std::string name;
coord_t n_grid_cells;
const Rect<1> domain;
IndexSpaceT<1,coord_t> index_space;
IndexPartitionT<1, coord_t> index_partition;
std::vector<unsigned> field_ids;
FieldSpace field_space;
LogicalRegion logical_region;
LogicalPartition logical_partition;
coord_t n_parts;
const Rect<1> color_domain;
const static unsigned projection_id = ProjectionIDs::IDENTITY;
Context ctx;
Runtime* runtime;
};
struct Data2D {
Data2D(coord_t n_grid_cells_, coord_t n_second_, coord_t n_parts_,
const std::string& name_,
const std::vector<unsigned>& field_ids_,
Context ctx_, Runtime* runtime_) :
name(name_),
n_grid_cells(n_grid_cells_),
n_second(n_second_),
domain(Point<2>(0,0), Point<2>(n_grid_cells_-1, n_second_-1)),
field_ids(field_ids_),
n_parts(n_parts_),
color_domain(Point<2>(0,0), Point<2>(n_parts_-1,0)),
ctx(ctx_),
runtime(runtime_)
{
index_space = runtime->create_index_space<2,coord_t>(ctx, domain);
auto block = Point<2>(n_grid_cells / n_parts, n_second);
index_partition = runtime->create_partition_by_blockify<2,coord_t>(ctx, index_space, block);
runtime->attach_name(index_partition, (std::string("index_partition_")+name).c_str());
field_space = runtime->create_field_space(ctx);
{
FieldAllocator allocator = runtime->create_field_allocator(ctx, field_space);
for (auto fid : field_ids) allocator.allocate_field(sizeof(double), fid);
}
runtime->attach_name(field_space, (std::string("field_space_")+name).c_str());
logical_region = runtime->create_logical_region(ctx, index_space, field_space);
runtime->attach_name(logical_region, (std::string("logical_region_")+name).c_str());
logical_partition = runtime->get_logical_partition(ctx, logical_region, index_partition);
runtime->attach_name(logical_partition, (std::string("logical_partition_")+name).c_str());
}
~Data2D() {
runtime->destroy_logical_partition(ctx, logical_partition);
runtime->destroy_logical_region(ctx, logical_region);
runtime->destroy_field_space(ctx, field_space);
runtime->destroy_index_partition(ctx, index_partition);
runtime->destroy_index_space(ctx, index_space);
}
std::string name;
coord_t n_grid_cells, n_second;
const Rect<2> domain;
IndexSpaceT<2,coord_t> index_space;
IndexPartitionT<2, coord_t> index_partition;
std::vector<unsigned> field_ids;
FieldSpace field_space;
LogicalRegion logical_region;
LogicalPartition logical_partition;
coord_t n_parts;
const Rect<2> color_domain;
Context ctx;
Runtime* runtime;
class LocalProjectionFunction : public ProjectionFunctor{
public:
LocalProjectionFunction() {}
LocalProjectionFunction(Runtime *rt) {}
virtual ~LocalProjectionFunction() {}
public:
virtual LogicalRegion project(const Mappable *mappable, unsigned index,
LogicalRegion upper_bound,
const DomainPoint &point) override {
assert(false);
}
virtual LogicalRegion project(const Mappable *mappable, unsigned index,
LogicalPartition upper_bound,
const DomainPoint &point) override {
Point<2,coord_t> color; color[0] = point[0]; color[1] = 0;
return runtime->get_logical_subregion_by_color(upper_bound, color);
}
virtual bool is_exclusive(void) const override { return true; }
virtual unsigned get_depth(void) const override { return 0; }
};
const static unsigned projection_id = ProjectionIDs::TO_1D_FROM_2D;
};
struct Data2D_Transposed {
Data2D_Transposed(coord_t n_grid_cells_, coord_t n_second_, coord_t n_parts_,
const std::string& name_,
const std::vector<unsigned>& field_ids_,
Context ctx_, Runtime* runtime_) :
name(name_),
n_grid_cells(n_grid_cells_),
n_second(n_second_),
domain(Point<2>(0,0), Point<2>(n_second_-1, n_grid_cells_-1)),
field_ids(field_ids_),
n_parts(n_parts_),
color_domain(Point<2>(0,0), Point<2>(0,n_parts_-1)),
ctx(ctx_),
runtime(runtime_)
{
index_space = runtime->create_index_space<2,coord_t>(ctx, domain);
auto block = Point<2>(n_second, n_grid_cells / n_parts);
index_partition = runtime->create_partition_by_blockify<2,coord_t>(ctx, index_space, block);
runtime->attach_name(index_partition, (std::string("index_partition_")+name).c_str());
field_space = runtime->create_field_space(ctx);
{
FieldAllocator allocator = runtime->create_field_allocator(ctx, field_space);
for (auto fid : field_ids) allocator.allocate_field(sizeof(double), fid);
}
runtime->attach_name(field_space, (std::string("field_space_")+name).c_str());
logical_region = runtime->create_logical_region(ctx, index_space, field_space);
runtime->attach_name(logical_region, (std::string("logical_region_")+name).c_str());
logical_partition = runtime->get_logical_partition(ctx, logical_region, index_partition);
runtime->attach_name(logical_partition, (std::string("logical_partition_")+name).c_str());
}
~Data2D_Transposed() {
runtime->destroy_logical_partition(ctx, logical_partition);
runtime->destroy_logical_region(ctx, logical_region);
runtime->destroy_field_space(ctx, field_space);
runtime->destroy_index_partition(ctx, index_partition);
runtime->destroy_index_space(ctx, index_space);
}
std::string name;
coord_t n_grid_cells, n_second;
const Rect<2> domain;
IndexSpaceT<2,coord_t> index_space;
IndexPartitionT<2, coord_t> index_partition;
std::vector<unsigned> field_ids;
FieldSpace field_space;
LogicalRegion logical_region;
LogicalPartition logical_partition;
coord_t n_parts;
const Rect<2> color_domain;
static const unsigned projection_id = ProjectionIDs::TO_1D_FROM_2D_TRANSPOSED;
Context ctx;
Runtime* runtime;
class LocalProjectionFunction : public ProjectionFunctor{
public:
LocalProjectionFunction() {}
LocalProjectionFunction(Runtime *rt) {}
virtual ~LocalProjectionFunction() {}
public:
virtual LogicalRegion project(const Mappable *mappable, unsigned index,
LogicalRegion upper_bound,
const DomainPoint &point) override {
assert(false);
}
virtual LogicalRegion project(const Mappable *mappable, unsigned index,
LogicalPartition upper_bound,
const DomainPoint &point) override {
Point<2,coord_t> color; color[0] = 0; color[1] = point[0];
return runtime->get_logical_subregion_by_color(upper_bound, color);
}
virtual bool is_exclusive(void) const override { return true; }
virtual unsigned get_depth(void) const override { return 0; }
};
};
#endif
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment