Linear analysis and optimization of stream programs

13 years 9 months ago

Download groups.csail.mit.edu

As more complex DSP algorithms are realized in practice, an increasing need for high-level stream abstractions that can be compiled without sacriﬁcing eﬃciency. Toward this end, we present a set of aggressive optimizations that target linear sections of a stream program. Our input language is StreamIt, which represents programs as a hierarchical graph of autonomous ﬁlters. A ﬁlter is linear if each of its outputs can be represented as an aﬃne combination of its inputs. Linearity is common in DSP components; examples include FIR ﬁlters, expanders, compressors, FFTs and DCTs. We demonstrate that several algorithmic transformations, traditionally hand-tuned by DSP experts, can be completely automated by the compiler. First, we present a linear extraction analysis that automatically detects linear ﬁlters from the C-like code in their work function. Then, we give a procedure for combining adjacent linear ﬁlters into a single ﬁlter, as well as for translating a linear ﬁl...

Andrew A. Lamb, William Thies, Saman P. Amarasingh

Real-time Traffic