search for: pingali

...st of post-dominator trees, with one rooted at each exit node. The problem this can cause is that, by nature, these trees omit some nodes which are not post-dominated by any one exit. This invalidates analyses which build on a traversal of the post-dominator tree. (In my case, I'm implementing Pingali & Bilardi's optimal control dependence algorithms: http://doi.acm.org/10.1145/256167.256217.) Consider, for example, a simple function with a single branch and two exits: define i32 @f(i32 %X) { entry: %0 = icmp eq i32 %X, 0 br i1 %0, label %exit1, label %exit2 exit1...

I think you can also avoid the RDF computation using a more directed form of control dependence testing such as described in Keshav Pingali and Gianfranco Bilardi. 1997. Optimal control dependence computation and the Roman chariots problem. ACM Trans. Program. Lang. Syst. 19, 3 (May 1997), 462-491. DOI=http://dx.doi.org/10.1145/256167.256217 However one challenge seems to be fixing the SSA graph after deleting essentially arbitrary co...

...ith one > rooted at each exit node. The problem this can cause is that, by > nature, these trees omit some nodes which are not post-dominated by > any one exit. This invalidates analyses which build on a traversal of > the post-dominator tree. > > (In my case, I'm implementing Pingali & Bilardi's optimal control > dependence algorithms: http://doi.acm.org/10.1145/256167.256217.) > > Consider, for example, a simple function with a single branch and two > exits: > > define i32 @f(i32 %X) { > entry: > %0 = icmp eq i32 %X, 0 > br i1 %...

...not aware of any > analysis > which requires traditional dataflow, or that cannot be implemented more > efficiently in SSA form. If you are aware of one, please let me know. > :) Backward dataflow analyses, like live variable analysis, cannot be performed with SSA [Choi91, JohnsonPingali93]. Attached to this email is an example (color pdf) that illustrates why. Do you convert from SSA back to a non-SSA LLVM to do the live-variable analysis that you have implemented in the backends? At Argonne one of the application specific analyses we need to perform, activity analysis, has...

...n most cases. Cd-dce is not slower in GCC than dce On Fri, Jan 29, 2016, 8:31 PM David Callahan <dcallahan at fb.com<mailto:dcallahan at fb.com>> wrote: I think you can also avoid the RDF computation using a more directed form of control dependence testing such as described in Keshav Pingali and Gianfranco Bilardi. 1997. Optimal control dependence computation and the Roman chariots problem. ACM Trans. Program. Lang. Syst. 19, 3 (May 1997), 462-491. DOI=http://dx.doi.org/10.1145/256167.256217<https://urldefense.proofpoint.com/v2/url?u=http-3A__dx.doi.org_10.1145_256167.256217&d=C...

...d with undef in most cases. Cd-dce is not slower in GCC than dce On Fri, Jan 29, 2016, 8:31 PM David Callahan <dcallahan at fb.com> wrote: > I think you can also avoid the RDF computation using a more directed form > of control dependence testing such as described in > > Keshav Pingali and Gianfranco Bilardi. 1997. Optimal control dependence > computation and the Roman chariots problem. ACM Trans. Program. Lang. Syst. > 19, 3 (May 1997), 462-491. DOI=http://dx.doi.org/10.1145/256167.256217 > > However one challenge seems to be fixing the SSA graph after deleting >...

...in this mail: http://lists.cs.uiuc.edu/pipermail/llvmdev/2010-March/029996.html I just copied the relevant part for you: ----------------------------------------------------------------------- The basic ideas are taken from "The Program Structure Tree - Richard Johnson, David Pearson, Keshav Pingali - 1994", however enriched with ideas from "The Refined Process Structure Tree - Jussi Vanhatalo, Hagen Voelyer, Jana Koehler - 2009". The algorithm to calculate these data structures however is completely different, as it takes advantage of existing information already available i...

...n most cases. Cd-dce is not slower in GCC than dce On Fri, Jan 29, 2016, 8:31 PM David Callahan <dcallahan at fb.com<mailto:dcallahan at fb.com>> wrote: I think you can also avoid the RDF computation using a more directed form of control dependence testing such as described in Keshav Pingali and Gianfranco Bilardi. 1997. Optimal control dependence computation and the Roman chariots problem. ACM Trans. Program. Lang. Syst. 19, 3 (May 1997), 462-491. DOI=http://dx.doi.org/10.1145/256167.256217<https://urldefense.proofpoint.com/v2/url?u=http-3A__dx.doi.org_10.1145_256167.256217&d=C...

On Sat, Jan 7, 2012 at 12:14 AM, Cameron Zwarich <zwarich at apple.com> wrote: > On Jan 6, 2012, at 8:27 PM, Daniel Berlin wrote: > > Note: GCC takes exactly the same approach as LLVM here, for exactly > the reason chris specifies. > In fact, until we started local SSA updating (which is now many years > ago, but ...), dominance frontier calculation for ssa updating was in

...Regions can also be used to solve a problem in a divide and conquer style. Some information about the algorithm itself: ------------------------------------------------------------------------ The basic ideas are taken from "The Program Structure Tree - Richard Johnson, David Pearson, Keshav Pingali - 1994", however enriched with ideas from "The Refined Process Structure Tree - Jussi Vanhatalo, Hagen Voelyer, Jana Koehler - 2009". The algorithm to calculate these data structures however is completely different, as it takes advantage of existing information already available i...

On Thu, 6 Nov 2003, Michelle Strout wrote: > Those of us working on the OpenAnalysis project have been looking at > LLVM recently (excellent job on the website BTW). Thanks! I'm just one of the many people who have worked on it though, the praise belongs to them as much as it does to me. :) > This includes researchers at Rice, Argonne, and LLNL. Great! > John Mellor-Crummey

On Thu, Jan 28, 2016 at 10:09 PM, Hal Finkel <hfinkel at anl.gov> wrote: > > ------------------------------ > > *From: *"David Callahan via llvm-dev" <llvm-dev at lists.llvm.org> > *To: *"Daniel Berlin" <dberlin at dberlin.org>, "LLVM Dev Mailing list" < > llvm-dev at lists.llvm.org> > *Sent: *Thursday, January 28, 2016