29.01.2013 Views

Tutorial CUDA

Tutorial CUDA

Tutorial CUDA

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

More on Memory Spaces<br />

Each thread can:<br />

Read/write per-thread registers<br />

Read/write per-block shared memory<br />

Read/write per-grid global memory<br />

Most important, commonly used<br />

Each thread can also:<br />

Read/write per-thread local memory<br />

Read only per-grid constant memory<br />

Read only per-grid texture memory<br />

Used for convenience/performance<br />

More details later<br />

The host can read/write global,<br />

constant, and texture memory<br />

(stored in DRAM)<br />

© NVIDIA Corporation 2008<br />

Host<br />

Grid<br />

Block (0, 0)<br />

Shared Memory<br />

Registers<br />

Thread (0, 0)<br />

Local<br />

Memory<br />

Global<br />

Memory<br />

Constant<br />

Memory<br />

Texture<br />

Memory<br />

Registers<br />

Thread (1, 0)<br />

Local<br />

Memory<br />

Block (1, 0)<br />

Local<br />

Memory<br />

Shared Memory<br />

Registers<br />

Thread (0, 0)<br />

Registers<br />

Thread (1, 0)<br />

Local<br />

Memory

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!