CUDA tutorial/fr

Introduction

Dans ce tutoriel, nous présentons le composant de calcul hautement parallèle qu'est le processeur graphique (ou GPU pour graphics processing unit) et nous abordons le modèle CUDA avec quelques-unes de ses librairies numériques utilisées en calcul de haute performance.

Prérequis

Ce tutoriel montre comment utiliser CUDA pour accélérer des programmes en C ou en C++; une bonne connaissance d'un de ces langages vous permettra d’en tirer le meilleur profit. Si CUDA sert aussi aux programmes en Fortran, nous nous limiterons ici à CUDA pour C/C++ et utiliserons le terme CUDA C. Il s'agit essentiellement de produire des fonctions en C/C++ pouvant être exécutées par les CPUs et les GPUs.

Objectifs d'apprentissage

• Comprendre l'architecture d'un GPU
• Comprendre le déroulement d'un programme CUDA
• Comprendre et gérer les différents types de mémoires GPU
• Écrire et compiler un exemple de code CUDA

Qu'est-ce qu'un GPU?

Un GPU (pour graphics processing unit) est un processeur monopuce capable d'effectuer des calculs mathématiques rapidement pour produire des rendus d'images. Depuis quelques années, la puissance du GPU sert aussi à accélérer l'exécution de calculs intensifs dans plusieurs domaines de la recherche scientifique de pointe.

Qu'est-ce que CUDA?

CUDA est l'abréviation de compute unified device architecture. Il s'agit d'un environnement logiciel flexible et d'un modèle de programmation pour le traitement de calculs parallèles intensifs permettant l'accès aux instructions et à la mémoire d'un GPU.

Architecture du GPU

Un GPU se conmpose  :

• d'une mémoire globale
  • semblable à la mémoire du CPU
  • accessible par un CPU et un GPU
• des multiprocesseurs en continu (SMs pour streaming multiprocessors)
  • chaque SM est formé de plusieurs processeurs en continu (SPs pour streaming processors)
  • ces SMs effectuent les calculs
  • chaque SM est doté d'une unité de contrôle, de registres, de pipelines d'exécution, etc. qui lui sont propres

Modèle de programmation

Voyons d'abord quelques termes importants :

• Hôte : désigne le CPU et la mémoire principale
• Composant : désigne le GPU et sa mémoire

Modèle de programmation

Voyons d'abord quelques termes importants : Hôte : désigne le CPU et sa mémoire (mémoire hôte) Composant : désigne le GPU et sa mémoire (mémoire du composant)

Le modèle CUDA est un modèle hétérogène où à la fois le CPU et le GPU sont utilisés. Le code CUDA peut gérer les deux types de mémoires, la mémoire hôte CPU et la mémoire du composant GPU. Le code exécute aussi les fonctions du GPU appelées kernels (noyaux). Ces fonctions sont exécutées en parallèle par plusieurs fils GPU. Le processus comporte cinq étapes :

1. Déclaration et allocation de la mémoire hôte CPU et de la mémoire du composant GPU.
2. Initialisation de la mémoire hôte CPU.
3. Transfert des données de la mémoire hôte CPU à la mémoire du composant GPU.
4. Exécution des fonctions GPU (kernels).
5. Retour des données à la mémoire hôte CPU.

Modèle d'exécution

Le modèle CUDA est un modèle hétérogène qui utilise à la fois le CPU et le GPU. Le code CUDA peut gérer les deux types de mémoires, la mémoire principale du CPU et la mémoire du composant GPU le code exécute aussi les fonctions du GPU appelées noyaux (kernels). Ces fonctions sont exécutées en parallèle par plusieurs fils GPU. Le processus comporte cinq étapes :

1. déclaration et allocation de la mémoire principale et de la mémoire du composant
2. initialisation de la mémoire principale
3. transfert des données de la mémoire principale à la mémoire du composant
4. exécution des fonctions GPU (noyaux)
5. retour des données à la mémoire principale

Block-threading model

CUDA block-threading model where threads are organized into blocks while blocks are further organized into grid.

Given a very large number of threads - in order to achieve massive parallelism one has to use all the threads possible - in a CUDA kernel, one needs to organize them somehow. In CUDA, all the threads are structured in threading blocks, the blocks are further organized into grids, as shown in the accompanying figure. In distributing the threads we must make sure that the following conditions are satisfied:

• threads within a block cooperate via the shared memory
• threads in different blocks can not cooperate

In this model the threads within a block work on the same set of instructions (but perhaps with different data sets) and exchange data between each other via shared memory. Threads in other blocks do the same thing (see the figure).

Threads within a block intercommunicate via shared memory.

Each thread uses IDs to decide what data to work on:

• Block IDs: 1D or 2D (blockIdx.x, blockIdx.y)
• Thread IDs: 1D, 2D, or 3D (threadIdx.x, threadIdx.y, threadIdx.z)

Such a model simplifies memory addressing when processing multi-dimensional data.

Thread scheduling

Usually a streaming microprocessor (SM) executes one threading block at a time. The code is executed in groups of 32 threads (called warps). A hardware scheduller is free to assign blocks to any SM at any time. Furthermore, when an SM gets the block assigned to it, it does not mean that this particular block will be executed non-stop. In fact, the scheduler can postpone/suspend execution of such blocks under certain conditions when e.g. data becomes unavailable (indeed, it is quite time-consuming to read data from the global GPU memory). When it happens, the scheduler executes another threading block which is ready for execution. This is a so-called zero-overhead scheduling which makes the execution more streamlined so that SMs are not idle.

Types of GPU memories

There are several types of memories available for CUDA operations:

• Global memory
  • off-chip, good for I/O, but relatively slow
• Shared memory
  • on-chip, good for thread collaboration, very fast
• Registers and Local Memory
  • thread work space , very fast
• Constant memory

A few basic CUDA operations

CUDA memory allocation

• cudaMalloc((void**)&array, size)
  • Allocates object in the device memory. Requires address of a pointer of allocated array and size.
• cudaFree(array)
  • Deallocates object from the memory. Requires just a pointer to the array.

Data transfer

• cudaMemcpy(array_dest, array_orig, size, direction)
  • Copy the data from either device to host or host to device. Requires pointers to the arrays, size and the direction type (cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, cudaMemcpyDeviceToDevice, etc.)
• cudaMemcpyAsync
  • Same as cudaMemcpy, but transfers the data asynchronously which means it doesn't block the execution of other processes.

A simple CUDA C program

The following example shows how to add two numbers on the GPU using CUDA. Note that this is just an exercise, it's very simple, so it will not scale at all.

__global__   void add (int *a, int *b, int *c){

	*c = *a + *b;
}
int main(void){
	int a, b, c;
	int *dev_a, *dev_b, *dev_c;
	int size = sizeof(int);

//  allocate device copies of a,b, c
cudaMalloc ( (void**) &dev_a, size);
cudaMalloc ( (void**) &dev_b, size);
cudaMalloc ( (void**) &dev_c, size);

a=2; b=7;
//  copy inputs to device
cudaMemcpy (dev_a, &a, size, cudaMemcpyHostToDevice);
cudaMemcpy (dev_b, &b, size, cudaMemcpyHostToDevice);

// launch add() kernel on GPU, passing parameters
add <<< 1, 1 >>> (dev_a, dev_b, dev_c);

// copy device result back to host
cudaMemcpy (&c, dev_c, size, cudaMemcpyDeviceToHost);

cudaFree ( dev_a ); cudaFree ( dev_b ); cudaFree ( dev_c ); 
}

Are we missing anything ? That code does not look parallel! Solution: Let's look at what's inside the triple brackets in the kernel call and make some changes :

add <<< N, 1 >>> (dev_a, dev_b, dev_c);

Here we replaced 1 by N, so that N different CUDA blocks will be executed at the same time. However, in order to achieve parallelism we need to make some changes to the kernel as well:

__global__   void add (int *a, int *b, int *c){

	c[blockIdx.x] = a[blockIdx.x] + b[blockIdx.x];

where blockIdx.x is the unique number identifying a CUDA block. This way each CUDA block adds a value from a[ ] to b[ ].

CUDA blocks-based parallelism.

Can we again make some modifications in those triple brackets ?

add <<< 1, '''N''' >>> (dev_a, dev_b, dev_c);

Now instead of blocks, the job is distributed across parallel threads. What is the advantage of having parallel threads ? Unlike blocks, threads can communicate between each other: in other words, we parallelize across multiple threads in the block when heavy communication is involved. The chunks of code that can run independently, i.e. with little or no communication, are distributed across parallel blocks.

Advantages of shared memory

So far all the memory transfers in the kernel have been done via the regular GPU (global) memory which is relatively slow. Often we have so many communications between the threads that the performance decreases significantly. In order to address this issue there exists another type of memory called shared memory which can be used to speed-up the memory operations between the threads. However the trick is that only the threads within a block can communicate. In order to demonstrate the usage of such shared memory we consider the dot product example where two vectors are multiplied together element by element and then summed. Below is the kernel:

__global__   void dot(int *a, int *b, int *c){
        int temp = a[threadIdx.x]*b[threadIdx.x]; 
}

After each thread computes its portion, we need to add everything together: each thread has to share its data. However, the problem is that each copy of thread's temp variable is private. This can be resolved by the use of shared memory. Below is the kernel with the modifications to use shared memory:

#define N 512
__global__   void dot(int *a, int *b, int *c){
   __shared__ int temp[N];
   temp[threadIdx.x] = a[threadIdx.x]*b[threadIdx.x];
   __syncthreads(); 
   if(threadIdx.x==0){
	int sum; for(int i=0;i<N;i++) sum+= temp[i];
	*c=sum; } 
}

Basic performance considerations

Memory transfer

• PCI-e is extremely slow (4-6 GB/s) compared to both host and device memories
• Minimize host-to-device and device-to-host memory copies
• Keep data on the device as long as possible
• Sometimes it is not effificient to make the host (CPU) do non-optimal jobs; executing it on the GPU may still be faster than copying to CPU, executing, and copying back
• Use memcpy times to analyse the execution times

Bandwidth

• Always keep CUDA bandwidth limitations in mind when changing your code
• Know the theoretical peak bandwidth of the various data links
• Count bytes read/written and compare to the theoretical peak
• Utilize the various memory spaces depending on the situation: global, shared, constant

Common GPU programming strategies

• Constant memory also resides in DRAM - much slower access than shared memory
  • BUT, it’s cached !!!
  • highly efficient access for read-only, broadcast
• Carefully divide data acording to access patterns:
  • read-only: constant memory (very fast if in cache)
  • read/write within block: shared memory (very fast)
  • read/write within thread: registers (very fast)
  • read/write input/results: global memory (very slow)