+

CUDA

Antonyus Pyetro do Amaral Ferreira

+The problem

The advent of multicore CPUs and manycore GPUs means that mainstream processor chips are now parallel systems.

The challenge is to develop application software that transparently scales its parallelism.

+A solution

CUDA is a parallel programming model and software environmenr.

A compiled CUDA program can therefore execute on any number of processor cores, and only the runtime system needs to know the physical processor count.

+CPU vs. GPU

+CPU vs. GPU

The GPU is especially well-suited to address problems that can be expressed as data-parallel computations.

Because the same program is executed for each data element, there is a lower requirement for sophisticated flow control.

+Applications?

General signal processing or physics simulation to computational finance or computational biology.

The latest generation of NVIDIA GPUs, based on the Tesla architecture, supports the CUDA programming model

+CUDA - Hello world

+What is CUDA?

CUDA extends C by allowing the programmer to define C functions, called kernels,that are executed N times in parallel by N different CUDA threads.

Each of the threads that execute a kernel is given a unique thread ID that is accessible within the kernel through the built-in threadIdx variable.

+CUDA Sum of vectors

+Concurrency

Threads within a block can cooperate among themselves by sharing data through some shared memory.

__syncthreads() acts as a barrier at which all threads in the block must wait before any are allowed to proceed.

+Process Hierarchy

+Memory Hierarchy

Per-thread local memory

Per-block shared memory

+Memory Hierarchy

+Host and Device

CUDA assumes that the CUDA threads may execute on a physically separate device that operates as a coprocessor to the host.

CUDA also assumes that both the host and the device maintain their own DRAM, referred to as host memory and device memory

+Software Stack

CUDA LibrariesCUDA Libraries

CUDA RuntimeCUDA Runtime

CUDA DriverCUDA Driver

ApplicatiApplicationon

HostHost

DevicDevicee

+Language Extensions

Function type qualifiers to specify whether a function executes on the host or on the device and whether it is callable from the host or from the device.

Variable type qualifiers to specify the memory location on the device of a variable .

+Language Extensions

A new directive to specify how a kernel is executed on the device from the host.

vecAdd<<<1, N>>>(A, B, C);

Four built-in variables that specify the grid and block dimensions and the block and thread indices

+Function Type Qualifiers

__device__ Executed on the device Callable from the device

only.

__global__ Executed on the

device Callable from the host

only.__host__

Executed on the host Callable from the host

only. Default type

+Variable Type Qualifiers

__device__

global memory space

Is accessible from all the threads within the grid

__constant__

constant memory space

Is accessible from all the threads within the grid

__shared__

space of a thread block

Is only accessible from all the threads within the block

+Execution Configuration

Any call to a __global__ function must specify the execution configuration for that call.

The execution configuration defines the dimension of the grid and blocks that will be used to execute the function on the device.

+Execution Configuration

<<< Dg, Db, Ns, S >>>

Dg is of type dim3 and specifies the dimension and size of the grid, such that Dg.x * Dg.y equals the number of blocks being launched; Dg.z is unused;

Db is of type dim3 and specifies the dimension and size of each block, such that Db.x * Db.y * Db.z equals the number of threads per block;

Ns is of type size_t and specifies the number of bytes in shared memory that is dynamically allocated per block for this call in addition to the statically allocated memory; this dynamically allocated memory is used by any of the variables declared as an external array; Ns is an optional argument which defaults to 0;

S is of type cudaStream_t and specifies the associated stream; S is an optional argument which defaults to 0.

+Built-in Variables

gridDim

This variable is of type dim3 and contains the dimensions of the grid.

blockIdx

This variable is of type uint3 and contains the block index within the grid.

blockDim

This variable is of type dim3 and contains the dimensions of the block. 

threadIdx

This variable is of type uint3 and contains the thread indexwithin the block.

+Example – Matrix multiplication

Task: C = A(hA,wA) X B(wA, wB)

Each thread block is responsible for computing one square sub-matrix Csub of C;

Each thread within the block is responsible for computing one element of Csub.

+Example – Matrix multiplication

Csub is equal to the product of two rectangular matrices:

the sub-matrix of A of dimension (wA, block_size) and the sub-matrix of B of dimension (block_size, wA)

these two rectangular matrices are divided into as many square matrices of dimension block_size as necessary.

Csub is computed as the sum of the products of these square matrices.

+Example – Matrix multiplication

+Compilation with NVCC

MS – visual studio 2005:

tools>options>Project&solutions >VC++ directories>include files\libraries files

Point to:

C:\Program Files\NVIDIA Corporation\NVIDIA CUDA SDK\common\inc

+CUDA Interoperability

OpenGL

Direct3D