similar to: ReName

On Sep 27, 2011, at 12:11 AM, Leo Romanoff wrote: > Quite some of these register allocation proposals are also able to handle overlapping register classes. That's interesting. Do you have any references? /jakob -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110927/e4b1f13d/attachment.html>

Hi Jakob, Yes. I have references. For example, from the top of my head I would name the following papers: 1)  A Generalized Algorithm for Graph-Coloring Register Allocation by Michael D. Smith, Norman Ramsey and Glenn Holloway      http://www.cs.tufts.edu/~nr/pubs/gcra-abstract.html 2) Register allocation by puzzle solving by Fernando Magno Quintão Pereira, Jens Palsberg    

On Sep 27, 2011, at 12:11 AM, Leo Romanoff wrote: > > > It is true that names are not always reflecting the essense. But on the other hand, there is a lot of ongoing research on register allocation (and compilers in general) and it looks like more and more such efforts choose LLVM as a platform for experimentation. Quite some results and comparisons are published. So, it would be nice

Hi Jakob, Hi Andy, First of all, thanks a lot for very elaborative and interesting answers! > It may be more helpful to explain how LLVM's register allocator came > into existence before debating the high level algorithm. > > When I began working on LLVM last October, Jakob was developing an > infrastructure for global live range splitting. It was just becoming > clear

On Sep 28, 2011, at 3:08 PM, Chris Lattner wrote: > > On Sep 27, 2011, at 12:11 AM, Leo Romanoff wrote: > >> >> >> It is true that names are not always reflecting the essense. But on the other hand, there is a lot of ongoing research on register allocation (and compilers in general) and it looks like more and more such efforts choose LLVM as a platform for

After some attempt to implement reduce-add in LLVM, I found out a easier way to detect reduce-add without introducing new IR operations. The basic idea is annotating phi node instead of add (so that it is easier to handle other reduction operations). In PHINode class, we can add a flag indicating if the phi node is a reduction one (the flag can be set in loop vectorizer for vectorized phi nodes).

On Wed, Nov 25, 2015 at 2:32 PM, Hal Finkel <hfinkel at anl.gov> wrote: > Hi Cong, > > After reading the original RFC and this update, I'm still not entirely sure I understand the semantics of the flag you're proposing to add. Does it having something to do with the ordering of the reduction operations? The flag is only useful for vectorized reduction for now. I'll give

----- Original Message ----- > From: "Xinliang David Li" <davidxl at google.com> > To: "Cong Hou" <congh at google.com> > Cc: "Hal Finkel" <hfinkel at anl.gov>, "llvm-dev" <llvm-dev at lists.llvm.org> > Sent: Wednesday, November 25, 2015 5:17:58 PM > Subject: Re: [llvm-dev] [RFC] Introducing a vector reduction add

On Sep 26, 2011, at 4:22 AM, Leo Romanoff wrote: > Hi Jakob, > > Thanks for a very interesting description of the new register allocation algorithm in LLVM!!! > > Could you elaborate a bit on the following topics: > > 1) Do you have any plans to publish something more formal and detailed about this algorithm, e.g. a paper or something similar? It would be nice to better

Hello, I have got a linear model that looks like this: lm(criterion ~ variable.A*variable.a + variable.B*variable.b + variable.C *variable.c) The output computed with stdCoeff() seems to be all right, but it does not show the coefficients of the interaction of the first pair of variables. Instead, it shows "NA": (Intercept) NA

Does anyone know of R functions for doing composite quantile regression (Hou and Yuan Ann Stat 36:1108, 2008)? The paper's authors do not talk about software in their paper or on their web sites. Thanks Frank ----- Frank Harrell Department of Biostatistics, Vanderbilt University -- View this message in context: http://r.789695.n4.nabble.com/Composite-Quantile-Regression-tp4663463.html

Help ! I am very happy with my 20001026 wine, running my favourites windows applications. Unfortunately, I still have a minor problem: My keyboard in perfectly recognized as the Brazilian ABNT layout, I had to change something in the keyboard.c file, but now is ok. So when I start my application I can get every character I need, but I can tell windows that a '~' followed by a

I have spent the last 4-5 hours scrounging google articles on this and have found 2 solutions. The problem is one of them is something that we will not do (as MS will not support extending AD with Services For Unix(SFU)). The other is simply configuring kerberos and pam on the linux system. No problem there from what I can tell. Ticket cache: FILE:/tmp/krb5cc_0 Default principal: userid at

Hello, I followed a howto found on samba website. I use Samba v3.2.5 shares whom access are authentified against Active Directory, I have an error 233, saying that there is no process at the end of the chanel (excuse me if my english is not so good). The kerberos setup is good and kinit works fine. net ads join ... worked fine. Except a dns error message, I can see the server in Active

Hello list, using AD with rfc2307 provisioned and NIS extensions are available. In ADUC tool I choose the group "Domain Admins" and click on the [UNIX Attributes] tab. I activate it for my domain and choose the GID=500. When I execute on my member server "net cache flush && getent group 500" I get the result domain admins:x:500:johndoe,name1,name2 So far so good,

I came across a problem when one of our clients was not able to access any of the servers on our network. This person has never connected to us before and now for this first time was trying to do it from his home is Houston, TX using earthlink cable service provided by Time Warner. All this information, I think, is important because when I started examining my shorewall logs I found out

> >I suppose basically my question is: Is there anyway to get Windows to copy >the profile but exclude the 'My Documents' folder? Maybe you can find your answer at: http://www.windowsnetworking.com/kbase/WindowsTips/WindowsNT/RegistryTips/Disable/Disablecachingofroamingprofiles.html There hou can read: If you want roaming profiles to not include certain folders, you can

Hi Guys, I've again been surprised when something like > m <- matrix("", nrow=3,ncol=3) > index_i <- c(2,1,3) > index_j <- c(1,3,2) > vals <- c("a","b","c") > m[ index_i, index_j ] <- vals > m gives block-wise application of the assignment: [,1] [,2] [,3] [1,] "b" "b" "b" [2,]

According to the schedule (e.g. on the right on llvm.org), we should have tagged the release by now, but we haven't, so we're officially behind schedule. I'm still optimistic that we can wrap this up pretty soon, though. This is what's blocking us: - PR26509: Crash in InnerLoopVectorizer::vectorizeLoop() I'm waiting to hear what Cong comes up with, otherwise we can revert

I am trying to build R on a Centos server without admin privileges (or support), so I can call R functions from NetLogo models (via the NetLogo R extension or possibly Rserve), but I'm getting stuck near the end of make. Any suggestions? Here's the setup... $ R_SHELL=/bin/sh $ R_HOME=~/R-devel $ cd $R_HOME; ./configure; make returns ... begin installing recommended package MASS