- Brief history of
GPU computing:
- Before 2006:
graphics cards perform
specialized graphics rendering
functions
- 2006:
introduction of
the general purpose Graphics Processing Unit
(GPGPU)
- The GPU can be
programmed using a
high-level programming language
- Library functions were provided to support GPU programming
- The GPU can be
programmed using a
high-level programming language
Read more in this article: click here
- Before 2006:
graphics cards perform
specialized graphics rendering
functions
- CUDA:
- CUDA =
acronym for
Compute Unified Device Architecture
CUDA is a (vector) parallel computing platform and application programming interface (API) model created by Nvidia (a graphics card company).
- CUDA is similar to the Open Computing Language (OpenCL) standard
- CUDA =
acronym for
Compute Unified Device Architecture
- OpenCL
- OpenCL =
Open Computing Language --
it is an open standard
(CUDA is
a NVidia product)
OpenCL = a framework for writing programs that runs on platforms (= computers) consisting of
- Central Processing Units (CPUs)
- Graphics Processing Units (GPUs)
- Digital Signal Prcessors (DSPs)
- etc
- OpenCL =
Open Computing Language --
it is an open standard
(CUDA is
a NVidia product)
- The CUDA architecture:
Important facts about the CUDA architecture:
- The host system and the
GPU system
are 2 separate systems:
- Host memory (=RAM) that is only accessible by the CPU
- Device memory that is only accessible by the GPU
Consequently:
- Data stored in
Host memory(=RAM)/Device memory must be
transfered to
each other through a
bus by a DMA (Direct Memory Access) device
:
- A CUDA program consists of
2 parts:
- A CPU part:
This part (of the computer program) is stored in the host memory and runs (= executes) on the host computer
- A GPU part:
This part (of the computer program) is stored in the device memory and runs (= executes) on the GPU (parallel) computer
- A CPU part:
- The host system and the
GPU system
are 2 separate systems:
- A CUDA C program consists of
2 parts:
- A CPU part that consists of
CPU instructions
(i.e., "normal" computer instructions
stored in
the RAM memory
that is executed by
the CPU
- The CPU program code is
written in the C programming language
and can use
functions in the
C library
Example:
int main( ) { printf("Hello World\n"); // Standard C programming statement // uses printf( ) in C library }
Learn more about the C library: click here
- The CPU program code is
written in the C programming language
and can use
functions in the
C library
- A GPU part that
consists of GPU (graphics processor)
instructions stored in
the (GPU) device memory
that
is executed by
the vector processing
GPU (graphics processor)
- The GPU program code is
written in the "CUDA C"
programming language
and can use
functions in the
CUDA C library
Example:
__global__ void hello( ) { printf("Hello World\n"); // CUDA C code // uses printf( ) in CUDA C library }
Note:
- CUDA C is designed to be similar to the C language to make it easy to learn !!!
- The GPU program code is
written in the "CUDA C"
programming language
and can use
functions in the
CUDA C library
- A CPU part that consists of
CPU instructions
(i.e., "normal" computer instructions
stored in
the RAM memory
that is executed by
the CPU
- How the
CUDA architecure execute
a (parallel) program:
- A GPU operates
like a vector processor:
- The same instruction is
simultaneously executed using
multiple input data (operands)
(i.e.: SIMD)
- The instruction execution is performed by
multiple processors ("cores")
A program execution on a processor is called a "thread" (of execution)
Each thread is executed by a CUDA core under the control of the thread execution control unit:
- The same instruction is
simultaneously executed using
multiple input data (operands)
(i.e.: SIMD)
- A GPU operates
like a vector processor: