load_state

oneMKL - Data Parallel C++ Developer Reference

Download PDF

ID 772045

Date 6/24/2024

Version

Public

A newer version of this document is available. Customers should click here to go to the newest version.

Visible to Intel only — GUID: GUID-6A7DB125-1828-490C-85F9-168213E30026

View Details

load_state

Loads the state of the random number engine from the provided memory buffer or file, then creates new engine object.

Description
API

Description

The load_state function allows you to create a new random number engine object from the binary state of another engine, which was written in a memory buffer or file. You can use different sycl::queue objects for the new engine.

API

Syntax

Load from Memory Interface

namespace oneapi::mkl::rng {
  template<typename Engine>
  Engine load_state (const sycl::queue& queue,
                     const std::uint8_t* mem);
}

Load from File Interface

namespace oneapi::mkl::rng {
  template<typename Engine>
  Engine load_state (const sycl::queue& queue,
                     const std::string& filename); // deprecated since 2024.1 release
}

> Note: The Load from File function is deprecated. Use the Load from Memory function instead.

Include Files

oneapi/mkl/rng.hpp

Input Parameters

Load from Memory Interface

Name	Type	Description
queue	`const sycl::queue& queue`	`sycl::queue` object, which will be used for the newly-created engine.
mem	`const std::uint8_t*`	Memory, where engine's state was stored.

Load from File Interface

Name	Type	Description
queue	`const sycl::queue& queue`	`sycl::queue` object, which will be used for the newly-created engine.
filename	`const std::string&`	Name of the file where engine's state was written.

Output Parameters

Type	Description
Engine	New random number engine object, which would be created from the given `sycl::queue` and loaded engine's state.

The following code illustrates how to store the engine's state to the: engine_state.dat file and load it back:

Code for Save/Load state functionality

   // Creating GPU engine
   oneapi::mkl::rng::default_engine engine(gpu_queue);

   // Checking how much memory is required to save
   auto mem_size = oneapi::mkl::rng::get_state_size(engine);

   // Allocating memory buffer
   std::uint8_t* mem_buf = new std::uint8_t[mem_size];

   // Saving state of engine in the file
   oneapi::mkl::rng::save_state(engine, mem_buf);

   // Generating random numbers from the GPU engine
   oneapi::mkl::rng::generate(oneapi::mkl::rng::uniform{}, engine, n, r1);

   // Loading state for the CPU queue
   auto engine_2 = oneapi::mkl::rng::load_state<oneapi::mkl::rng::default_engine>(cpu_queue, mem_buf);

   // Generating random numbers from the CPU engine
   oneapi::mkl::rng::generate(oneapi::mkl::rng::uniform{}, engine_2, n, r2);

   // Cleaning up memory buffer
   delete[] mem_buf;

Select Your Language

Using Intel.com Search

Quick Links

Recent Searches

Advanced Search

Only search in

oneMKL - Data Parallel C++ Developer Reference

load_state

Description

API