Imagine: Media Processing with Streams

By Brucek Khailany, William (Bill) J. Dally, U. Kapasi, Peter R. Mattson, Jinyung Namkoong, John Douglas Owens, Brian Towles, Andrew Chang, Scott Rixner. IEEE Micro, 2001

384 citations recorded by the source corpus.

The power-efficient Imagine stream processor achieves performance densities comparable to those of special-purpose embedded processors. Executing programs mapped to streams and kernels, a single Imagine processor is expected to have a peak performance of 20 gflops and sustain 18.3 gops on mpeg-2 encoding.

Subfields: Computer Vision and Pattern Recognition, Electrical and Electronic Engineering, Signal Processing

View Paper: https://www.semanticscholar.org/paper/7564661f026abd1d472707c15357494fd79e63c0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Jens Harald Krüger
2. Naga K. Govindaraju
3. Kayvon Fatahalian
4. David P. Luebke
5. Mattan Erez
6. Mark J. Harris
7. P. Hanrahan
8. I. Wald

In Electrical and Electronic Engineering

1. Samuel Williams
2. John M. Shalf
3. Nuwan Jayasena
4. William (Bill) J. Dally

In Signal Processing

1. David P. Luebke
2. P. Hanrahan
3. Samuel Williams
4. Christian Bienia
5. Naga K. Govindaraju
6. Kai Li
7. William (Bill) J. Dally
8. Jeffrey Heer

Citations in the corpus

• Ray Tracing on the Cell Processor (2006, background): “1A streaming compute model is one in which small kernels operate on entire streams of data items, such as shading a stream of fragments [13].”
• Merrimac: Supercomputing with Streams (2003, methodology): “This stream instruction set closely follows that of the Imagine streaming media processor [5, 6].”
• The potential of the cell processor for scientific computing (2005, background): “Over the last five years, a plethora of alternatives to conventional cache-based architectures have been suggested including scratchpad memories [9,16,30], paged on-chip memories [12, 17], and explicit three-level memory architectures [18, 19].”
• GPGPU: general purpose computation on graphics hardware (2004, background): “Stream processors are well suited for a variety of media applications since they are characterized by high levels of parallelism, high computation per memory access ra­tios and high tolerance of memory latency [Khailany et al. 2001].”
• Benchmarking modern multiprocessors (2011, background): “Stream programming is a parallelization approach that decomposes a program into a parallel network of specialized kernels which are then mapped to processing elements [18, 29, 46].”
• Brook for GPUs: Stream Computing on Graphics Hardware (2004, background): “This is a major di(cid:11)erence between the vector and stream processor abstraction [Khailany et al. 2001].”