MATLAB 코드에서 커널 생성

CUDA^® GPU 커널을 생성하는 MATLAB 코드 구조와 패턴

GPU Coder™는 MATLAB^® 코드의 특정 알고리즘 구조와 패턴에 최적화된 CUDA 커널을 생성하고 실행합니다. 생성 코드는 cuFFT, cuSolver, cuBLAS, cuDNN, TensorRT를 포함하여 최적화된 NVIDIA^® CUDA 라이브러리를 호출합니다. 생성된 코드를 소스 코드, 정적 라이브러리 또는 동적 라이브러리 형태로 프로젝트에 통합할 수 있으며, 데스크탑, 서버, 그리고 NVIDIA Jetson, DRIVE 및 기타 플랫폼에 내장된 GPU용으로 컴파일할 수 있습니다. 또한 GPU Coder를 사용하여 수기로 작성한 CUDA 코드를 알고리즘과 생성된 코드에 포함시킬 수 있습니다.

앱

모두 확장

GPU Coder

GPU Coder	MATLAB 코드에서 CUDA 코드 생성
GPU 환경 검사	GPU 코드 생성 환경에 대한 확인과 설정

함수

모두 확장

코드 생성

`coder.gpuConfig`	Create GPU code generation configuration
`codegen`	MATLAB 코드에서 C/C++ 코드 생성
`gpucoder`	GPU Coder 앱 열기
`coder.checkGpuInstall`	GPU 코드 생성 환경 확인

GPU 커널 프라그마

`coder.gpu.kernel`	Pragma that maps `for`-loops to GPU kernels
`coder.gpu.kernelfun`	함수를 GPU 커널에 매핑하는 프라그마
`coder.gpu.nokernel`	Pragma to disable kernel creation for loops
`coder.ceval`	생성 코드에서 C/C++ 함수 호출
`coder.gpu.iterations`	Pragma that provides information to the code generator for making parallelization decisions on variable bound loops

GPU 메모리 프라그마

`coder.gpu.constantMemory`	Pragma that maps a variable to the constant memory on GPU
`coder.gpu.persistentMemory`	Pragma to allocate a variable as persistent memory on the GPU
`cudaMemoryManager`	Query memory usage by shared GPU memory manager for MEX functions (R2024a 이후)

GPU 아토믹 연산

`gpucoder.atomicAdd`	Atomically add value and variable in global or shared memory (R2021b 이후)
`gpucoder.atomicAnd`	Atomically perform bit-wise AND between value and variable in global or shared memory (R2021b 이후)
`gpucoder.atomicCAS`	Atomically compare and swap value of variable in global or shared memory (R2021b 이후)
`gpucoder.atomicDec`	Atomically decrement variable in global or shared memory within upper bound (R2021b 이후)
`gpucoder.atomicExch`	Atomically exchange variable in global or shared memory with value (R2021b 이후)
`gpucoder.atomicInc`	Atomically increment variable in global or shared memory within upper bound (R2021b 이후)
`gpucoder.atomicMax`	Atomically find the maximum between value and variable in global or shared memory (R2021b 이후)
`gpucoder.atomicMin`	Atomically find the minimum between value and variable in global or shared memory (R2021b 이후)
`gpucoder.atomicOr`	Atomically perform bit-wise OR between value and variable in global or shared memory (R2021b 이후)
`gpucoder.atomicSub`	Atomically subtract value from variable in global or shared memory (R2021b 이후)
`gpucoder.atomicXor`	Atomically perform bit-wise XOR between value and variable in global or shared memory (R2021b 이후)

코드 생성을 위한 프로그래밍

`half`	Construct half-precision numeric object
`stencilfun`	Generate CUDA code for stencil functions (R2022b 이후)
`selectdata`	Select slices of arrays and generate CUDA code (R2025a 이후)
`gpucoder.matrixMatrixKernel`	Optimized GPU implementation of functions containing matrix-matrix operations
`gpucoder.batchedMatrixMultiply`	Optimized GPU implementation of batched matrix multiply operation
`gpucoder.stridedMatrixMultiply`	Optimized GPU implementation of strided and batched matrix multiply operation
`gpucoder.batchedMatrixMultiplyAdd`	Optimized GPU implementation of batched matrix multiply with add operation
`gpucoder.stridedMatrixMultiplyAdd`	Optimized GPU implementation of strided, batched matrix multiply with add operation
`gpucoder.sort`	Optimized GPU implementation of the MATLAB sort function
`gpucoder.ctranspose`	Optimized GPU implementation of the MATLAB transpose function
`gpucoder.transpose`	MATLAB 전치 함수의 최적화된 GPU 구현
`gpucoder.reduce`	Optimized GPU implementation for reduction operations

코드 구성 설정

모두 확장

GPU 코드

GPU 코드 생성	Control GPU code generation
GPU 장치 ID	CUDA device selection
최소 Compute Capability	Minimum compute capability for code generation
사용자 지정 Compute Capability	Virtual GPU architecture
Malloc 모드	GPU memory allocation
스택 제한	Stack limit per GPU thread
커널당 최대 블록 수	Maximum number of blocks created during a kernel launch
안전한 빌드	Error checking in the generated code
커널 이름 접두사	Custom kernel name prefixes
컴파일러 플래그	Pass additional flags to GPU compiler
cuBLAS 사용	Replace math function calls with `cuBLAS` library calls
cuSOLVER 사용	Replace math function calls with `cuSOLVER` library calls
cuFFT 사용	Replace `fft` function calls with `cuFFT` library calls
GPU 메모리 관리자 활성화	Use GPU memory manager

객체

모두 확장

코드 구성

`coder.GpuCodeConfig`	Configuration parameters for CUDA code generation from MATLAB code
`coder.MexCodeConfig`	Configuration parameters for MEX function generation from MATLAB code
`coder.CodeConfig`	MATLAB 코드에서 C/C++ 코드를 생성하기 위한 구성 파라미터
`coder.EmbeddedCodeConfig`	Configuration parameters for C/C++ code generation from MATLAB code with Embedded Coder
`coder.gpuEnvConfig`	Configuration object for checking the GPU code generation environment

도움말 항목

Configure GPU Code Generation
Configure the code generator using configuration objects or the GPU Coder app.
Kernels from Element-Wise Loops
Create kernels from MATLAB functions containing scalarized, element-wise math operations.
Generate GPU Kernels for Reduction Operations
Create kernels from MATLAB functions containing reduction operations.
Kernels from Library Calls

Target GPU optimized math libraries such as cuBLAS, cuSOLVER, and cuFFT.
- Generate GPU Code That Uses the NVIDIA cuBLAS Library
- cuSOLVER Example
- FFT 예제
Support for GPU Arrays
Generate CUDA code that uses GPU arrays.
Use Dynamically Allocated C++ Arrays in Generated Function Interfaces
Understand and use dynamically allocated arrays from the generated CUDA C++ function interfaces.
Call Custom CUDA Kernels from the Generated Code
Integrate custom CUDA kernels with MATLAB code intended for code generation.
Call Custom CUDA Device Functions from Generated Code
Integrate custom GPU device functions with MATLAB code intended for code generation.
Design Patterns
Create kernels for MATLAB functions containing computational design patterns.
Reduce GPU Memory Allocations By Using GPU Memory Manager
Avoid repetitive memory allocations by creating and reusing memory pools for generated CUDA applications.
What Is Half Precision?
Introduction to the half-precision data type in MATLAB and Simulink^®.
Half Precision Code Generation Support
C/C++ and GPU code generation support for functions that support half-precision inputs.

추천 예제

$Simulate Diffraction Patterns Using CUDA FFT Libraries$

Simulate Diffraction Patterns Using CUDA FFT Libraries

Use GPU Coder™ to leverage the CUDA® Fast Fourier Transform library (cuFFT) to compute two-dimensional FFT on a NVIDIA® GPU. The two-dimensional Fourier transform is used in optics to calculate far-field diffraction patterns. When a monochromatic light source passes through a small aperture, such as in Young's double-slit experiment, you can observe these diffraction patterns. This example also shows how to pass GPU inputs to an entry-point function when generating CUDA MEX, source code, static libraries, dynamic libraries, and executables. By using this functionality, the performance of the generated code is improved by minimizing the number of cudaMemcpy calls in the generated code.

스크립트 열기

QR Decomposition on NVIDIA GPU Using cuSOLVER Libraries

Create a standalone CUDA® executable that leverages the NVIDIA® cuSOLVER library. The example uses a curve fitting application that mimics automatic lane tracking on a road to illustrate:

라이브 스크립트 열기

Stencil Processing on GPU

Generate CUDA® kernels for stencil type operations by implementing "Game of Life" by John H. Conway.

스크립트 열기

Benchmark Solving a Linear System by Using GPU Coder

Benchmark solving a linear system by generating CUDA® code. Use matrix left division, also known as mldivide or the backslash operator (\), to solve the system of linear equations A*x = b for x (that is, compute x = A\b).

라이브 스크립트 열기

안개 보정

이 예제는 GPU 코드 생성을 위해 영상 처리 함수를 사용하는 방법을 보여줍니다. 이 예제에서는 안개 낀 이미지를 입력으로 받아서 안개가 제거된 이미지를 생성합니다. 이 예제는 전형적인 안개 보정 알고리즘입니다. 이 예제에서는 conv2, im2gray, imhist 함수를 사용합니다.

라이브 스크립트 열기

Generate GPU Code That Computes Stereo Disparity

Generate a CUDA® MEX function that computes the stereo disparity of two images. You generate code from MATLAB® functions that compute the stereo disparity. The generated CUDA MEX functions compute the stereo disparity using integer or half-precision values.

라이브 스크립트 열기

Accelerating Radar Signal Processing Using GPU

Compare the performance of a conventional radar signal processing chain implemented in interpreted MATLAB and on a graphical processing unit (GPU).

(Radar Toolbox)

R2024a 이후