Paper on compressed video processing appears at Multimedia 2009.

First Ph.D. thesis defense as part of the StreamIt project (thesis, defense slides, related video)

StreamIt is featured in Technology Review.

MIT class on multicore programming features StreamIt language and compiler.

Version 2.1.1 of the StreamIt compiler is available for download. This version fixes a simple performance bug in the multicore backend, leading to a 3-4x speedup on 4-core machines.

Version 2.1 of the StreamIt compiler is now available for download. You may download either the source distribution and natively build the compiler, or a prebuilt distribution of the full compiler. The distribution also includes a set of streaming benchmarks.

Our paper on exploiting coarse-grained parallelism in stream programs will appear at ASPLOS in October 2006. This paper maps StreamIt to Raw, a 16-core architecture, demonstrating an 11.2X mean speedup over a single-core baseline.

Our paper on MPEG-2 decoding in StreamIt appears at IPDPS. The paper shows that MPEG decoding is a good match for StreamIt, leading to clean, reusable, and portable code. The source code and related resources are available for download.

Our paper on linear state space analysis appears in CASES 2005. This paper extends our previous work on optimizing linear filters to encompass linear filters with state, including new optimizations for reducing the number of states and non-zero system parameters.

A paper on flexible graphics pipelines, in collaboration with Jiawen Chen, Matthias Zwicker, Kari Pulli, and Frédo Durand, appears in Graphics Hardware 2005. As part of this project, we extended StreamIt's Raw backend to support dynamic communication rates between filters.

Our paper on cache optimizations for embedded and general-purpose processors appears in LCTES 2005. Also, our paper on teleport messaging, a new language construct for out-of-band communication, appears in PPoPP 2005.

A paper on programming by sketching for bit-streaming programs, a collaborative effort with Armando Solar-Lezama and Ras Bodik of UC Berkeley and Kemal Ebcioglu of IBM Research, wins the Best Paper Award at PLDI 2005.

Our paper on the StreamIt Development Tool for high-productivity stream programming, including a user study that demonstrates improved debugging of stream programs, appears in P-PHEC 2005. The development tool is an Eclipse plugin and is available for download.

Mani Narayanan and Kathy Yelick of UC Berkeley publish a paper in the 4th Workshop on Media and Streaming Processors, describing automatic generation of permutation instructions for VIRAM using StreamIt as a high-level input language.

Version 2.0 of the StreamIt compiler is now available for download. You may download either the source distribution and natively build the compiler, or a prebuilt distribution of the full compiler. The distribution also includes a set of streaming benchmarks.