Developer Guide

oneAPI GPU Optimization Guide

Download PDF
ID 771772
Date 3/31/2023
Public

A newer version of this document is available.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

Document Table of Contents
Document Table of Contents x
oneAPI GPU Optimization Guide
oneAPI GPU Optimization Guide x
Introduction Getting Started Parallelization Intel® Xe GPU Architecture General-Purpose Computing on GPU Media Graphics Computing on GPU References Terms and Conditions
General-Purpose Computing on GPU x
Execution Model Overview Thread Mapping and GPU Occupancy Kernels Using Libraries for GPU Offload Host/Device Memory, Buffer and USM Host/Device Coordination Using Multiple Heterogeneous Devices Compilation OpenMP Offloading Tuning Guide Multi-GPU, Multi-Stack and Multi-C-Slice Architecture and Programming Level Zero Performance Profiling and Analysis Configuring GPU Device
Kernels x
Sub-Groups and SIMD Vectorization Removing Conditional Checks Registerization and Avoiding Register Spills Small Register Mode vs. Large Register Mode Shared Local Memory Pointer Aliasing and the Restrict Directive Synchronization among Threads in a Kernel Considerations for Selecting Work-Group Size Reduction Kernel Launch Executing Multiple Kernels on the Device at the Same Time Submitting Kernels to Multiple Queues Avoiding Redundant Queue Constructions Programming Intel® XMX Using SYCL Joint Matrix Extension Doing I/O in the Kernel
Synchronization among Threads in a Kernel x
Atomic Operations Local Barriers vs Global Atomics
Using Libraries for GPU Offload x
Using Performance Libraries Using Standard Library Functions in SYCL Kernels Efficiently Implementing Fourier Correlation Using oneAPI Math Kernel Library (oneMKL)
Host/Device Memory, Buffer and USM x
Unified Shared Memory Allocations Performance Impact of USM and Buffers Optimizing Memory Movement Between Host and Accelerator Avoiding Moving Data Back and Forth between Host and Device Avoiding Declaring Buffers in a Loop Buffer Accessor Modes
Host/Device Coordination x
Asynchronous and Overlapping Data Transfers Between Host and Device
Compilation x
Just-In-Time Compilation Ahead-Of-Time Compilation Specialization Constants Accuracy Versus Performance Tradeoffs in Floating-Point Computations
OpenMP Offloading Tuning Guide x
OpenMP Directives OpenMP Execution Model Terminology Compiling and Running an OpenMP Application Offloading oneMKL Computations onto the GPU Tools for Analyzing Performance of OpenMP Applications OpenMP Offload Best Practices
OpenMP Offload Best Practices x
Using More GPU Resources Minimizing Data Transfers and Memory Allocations Making Better Use of OpenMP Constructs Memory Allocation Fortran Example Clauses: is_device_ptr, use_device_ptr, has_device_addr, use_device_addr Prefetching
Multi-GPU, Multi-Stack and Multi-C-Slice Architecture and Programming x
Multi-Stack GPU Architecture GPU Memory System Implicit Scaling Explicit Scaling Intel® MPI for GPU Clusters Intra-Device and Inter-Device Data Transfers for MPI+OpenMP Programs Terminology
Implicit Scaling x
Introduction Work Scheduling and Memory Distribution Simple Examples Programming Principles
Explicit Scaling x
Explicit Scaling on Multi-GPU, Multi-Stack, Multi-C-Slice in SYCL Explicit Scaling on Multi-GPU, Multi-Stack and Multi-C-Slice in OpenMP Explicit Scaling Summary
Level Zero x
Immediate Command Lists
Performance Profiling and Analysis x
Using the Timers Intel® VTuneTM Profiler Intel® Advisor Intel® Intercept Layer for OpenCLTM Applications Intel® Profiling Tools Interfaces for GPU
Intel® VTuneTM Profiler x
Hardware-assisted Stall Sampling
Intel® Advisor x
Identify Regions to Offload to GPU with Offload Modeling Run a GPU Roofline Analysis Optimize Memory-bound Applications with GPU Roofline
Media Graphics Computing on GPU x
Optimizing Media Pipelines Performance Analysis with Intel® Graphics Performance Analyzers
Optimizing Media Pipelines x
Media Engine Hardware Media API Options for Hardware Acceleration Media Pipeline Parallelism Media Pipeline Inter-operation and Memory Sharing SYCL-Blur Example
oneAPI GPU Optimization Guide

Synchronization among Threads in a Kernel

Level Two Title