#include <macrotaskq.h>
◆ StoragePolicy
| Enumerator |
|---|
| StoreFunction | store a madness function in the cloud – can have a large memory impact
|
| StorePointerToFunction | store the pointer to the function in the cloud, the actual function lives in the universe and its coefficients can be copied to the subworlds (e.g. by macrotaskq) when needed. The task itself is responsible for handling data movement
|
| StoreFunctionViaPointer | coefficients to the subworlds when the task is started. This is the default policy.
store a pointer to the function in the cloud, but macrotaskq will move the
|
◆ check_consistency()
| bool madness::MacroTaskInfo::check_consistency |
( |
| ) |
const |
|
inline |
◆ get_all_preset_names()
| static std::vector< std::string > madness::MacroTaskInfo::get_all_preset_names |
( |
| ) |
|
|
inlinestatic |
◆ get_all_presets()
◆ policy_to_string()
◆ preset()
return some preset policies
- "default": StoreFunctionViaPointer, cloud rank-replicated, initial functions node-replicated
- "small_memory": StoreFunctionViaPointer, cloud rank-replicated, initial functions distributed
- "large_memory": StoreFunction, cloud rank-replicated, initial functions distributed the user can also set the policies manually note: the policies are checked for consistency when the MacroTaskQ is created
References cloud_distribution_policy, madness::Distributed, MADNESS_EXCEPTION, madness::name(), madness::NodeReplicated, ptr_target_distribution_policy, madness::RankReplicated, storage_policy, StoreFunction, and StoreFunctionViaPointer.
Referenced by madness::SCF::apply_potential(), get_all_presets(), test_2d_partitioning(), test_deferred(), and test_twice().
◆ to_cloud_storage_policy()
◆ to_json()
| nlohmann::json madness::MacroTaskInfo::to_json |
( |
| ) |
const |
|
inline |
◆ operator<< [1/2]
◆ operator<< [2/2]
◆ to_string
◆ cloud_distribution_policy
◆ ptr_target_distribution_policy
◆ storage_policy
The documentation for this struct was generated from the following file: