29.01.2013 Views

Tutorial CUDA

Tutorial CUDA

Tutorial CUDA

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

What You Need To Know<br />

FP64 instructions will be slower than FP32<br />

It takes more than just wider data paths to implement<br />

double precision<br />

For best performance, use FP64 judiciously<br />

Analyze your computations<br />

Use FP64 only for precision/range-sensitive computations<br />

Use FP32 for computations that are accurate and robust<br />

with 32-bit floating point<br />

<strong>CUDA</strong> compiler supports mixed usage of float and<br />

double<br />

Supported since <strong>CUDA</strong> 1.0<br />

© NVIDIA Corporation 2008

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

Saved successfully!

Ooh no, something went wrong!