17 Jan 1998 00:05:06 -0500

Related articles |
---|

Dataflow Analysis of Scalar Vars goren@cse.ucsc.edu (1998-01-17) |

From: | goren@cse.ucsc.edu (Sezer Goren) |

Newsgroups: | comp.compilers |

Date: | 17 Jan 1998 00:05:06 -0500 |

Organization: | UC Santa Cruz CIS/CE |

Keywords: | analysis, question |

Dear Compiler People:

My specialty is behavioral synthesis which is nothing but hardware

compilation from an HDL specification such as Verilog. I have

developed a state-of-the-art tool for that.

In doing that, I believe I devised a very useful algorithm for

dataflow analysis of scalar variables. The algorithm is of quadratic

complexity even for non-reducible flow graphs. According to the DRAGON

book, the worst case complexity of known algorithms for non-reducible

flow graphs is exponential. I am not sure if I made a breakthrough, or

am not aware of the new research on the subject, or am making a

mistake in my algorithm. If I am making a mistake, it should not be a

trivial one as my algorithm is used in hundreds of Verilog designs and

always compiles correct hardware, and also it is very solid in terms

of its graph theoretical basis.

I would like to publish my research. Since I am not a compiler

reasearcher, I would like to team up with a compiler expert and

publish a joint paper.

I have contacted a well-known prof at Stanford but I don't think I

will get enough attention if any. I am looking for interested

researchers or maybe you could point me to the right person.

Thanx.

PS: I am posting this from my fiancee's account since we have a

firewall at work. Please reply me at fatih@aluxs.micro.lucent.com.

A short bio:

I got my PhD from CWRU in Jan 95. I joined General Motors Research in

March 93 before I finished my PhD. I worked on behavioral synthesis

there until Aug 97 when I joined Lucent Microelectronics in Silicon

Valley as an ASIC design consultant.

--

Fatih Ugurdag

Lucent Technologies, Santa Clara, CA

http://www.lookup.com/Homepages/66915/fatih.html

--

Post a followup to this message

Return to the
comp.compilers page.

Search the
comp.compilers archives again.