llvm dev - Mar 2005 - [LLVMdev] Regarding PRE in LLVM

hi,
        While surfing on net I found that LLVM is using sparse version of
E-path_PRE approach in file PRE.cpp but I could not get what actually code
is doing and I could not find the name of the developer in the file.
        Just want to know to because I could not able to find any
paper on sparse version of this algorithm.

        Thanks in Advance

-- 
Dheeraj Kumar
M. Tech. I yr (CSE)
H.No. 7 Room No. 249
IITB Bombay
Mumbai-400076
Mobile No. +91-9819939038

On Tue, 29 Mar 2005, Dheeraj Kumar Arora wrote:
>        While surfing on net I found that LLVM is using sparse version of
> E-path_PRE approach in file PRE.cpp but I could not get what actually code
> is doing and I could not find the name of the developer in the file.
>        Just want to know to because I could not able to find any
> paper on sparse version of this algorithm.
I wrote the code, using Dhamdhere's "E-path PRE---partial redundancy 
elimination made easy" paper as inspiration and starting from first 
principles.

The code in LLVM CVS is about 75% done.  In handles a bunch of cases, but 
misses others.  It has not been used in quite some time, so it has 
probably bitrotten some.

If you're interested in sparse PRE, I suggest chasing down references for 
GVN-PRE by Thomas VanDrunen.  His thesis is here: 
ftp://ftp.cs.purdue.edu/pub/hosking/papers/vandrunen.pdf

-Chris

-- 
http://nondot.org/sabre/
http://llvm.cs.uiuc.edu/