data-level parallelism
nounUK /ˈdeɪtə ˈlɛvəl ˈpærəlɛlɪzəm/
US /ˈdeɪtə ˈlɛvəl ˈpærəlɛlɪzəm/
Data-level parallelism is when a computer does the same job on lots of pieces of information at the same time.
Graphics chips use data-level parallelism to color millions of pixels all at once.
RELATED WORDS
OTHER MEANINGS
2. A programming style where the same operation is applied to every element of a large data set simultaneously.
The image filter was rewritten with data-level parallelism so each pixel is processed in parallel.
3. A hardware feature where one instruction acts on many data elements at once.
The processor's data-level parallelism lets a single instruction add eight numbers together.
4. A way to speed up machine learning by processing many training examples at the same time.
Training used data-level parallelism across batches to cut the time in half.