similar to: [LLVMdev] Performance problems with FORTRAN allocatable arrays

Hi Wonsun, can you please provide a testcase. Best wishes, Duncan. > I've noticed that LLVM does a bad job of optimizing array indexing > code for FORTRAN arrays declared using the ALLOCATABLE keyword. > > For example if you have something like the following: > > DOUBLE PRECISION,ALLOCATABLE,DIMENSION(:,:,:,:) :: QAV > ... > ALLOCATE( QAV(

Hi Peter, > Why not do this too? I've applied this - thanks for the patch! Ciao, Duncan.

Duncan, Do a quick check here on x86_64-apple-darwin10 with svn llvm and svn dragon-egg against release gcc 4.5.0, the results from the himenoBMTxpa benchmark compiled at -O3 look pretty good. With stock gcc-4.5.0, we get... Grid-size = M mimax = 128 mjmax = 128 mkmax = 256 imax = 127 jmax = 127 kmax =255 Start rehearsal measurement process. Measure the performance in 3 times. MFLOPS:

Hi, I'm not a programmer, so I have a question about R functions, I run the Random Forest regression models, but I would like to run the random forest model 1000 times with different random sample set. to check the uncertainty of the regression model estimates. exemple of data: ################################# table= all Y: all$AGB X variables: Variables=as.matrix(all[, c( "min",

I'm doing convolutions in the frequency domain and need to know the layout of the Fourier modes returned by fft. (This is leading up to a more involved question about moment generating functions, but I need to know if I've got this part correct first.) I think in 1D the pattern is: 0 1 2 3 -2 1 (even) 0 1 2 3 -3 2 1 (odd) In 2D is it simply (for a square matrix): 0 1 2 -1 (horizontal)

There was some talk a year ago about a patch that merges PARSEC into the test suite. Here is a link to the thread: http://lists.cs.uiuc.edu/pipermail/llvmdev/2010-August/033877.html Has this just fallen through the cracks? Any help on this subject is welcome. Thanks! -- Wonsun -------------- next part -------------- An HTML attachment was scrubbed... URL:

Hello, I used a mac G5, R.2.1.1, and G77 3.4.4 and I would like to use and call a fortran subroutine. The trouble is that it seems I am not able to correctly load the compiled code. Here is what I have done: In the terminal this how I compiled my fortran code: R CMD SHLIB ~/Desktop/Fortan_kmeans/kmeans3.f There is the wrapper I have paste inside de kmeans3.f file: c

Hi Petr, Many thanks for your response. Basically I want to create a probability matrix to be used in a trinomial tree going forward. This is the reason why I thought to build the matrix around 0 would be much more efficient. I need to loop through because the probabilities will depend on my node and is not always the same per row (e.g. if N> jmax, jmax being defined in another function) I

Hi Well, I am not an expert in this field so I cannot comment your approach. I wanted only to point out that building matrix your way is like scratching your left ear with right hand, especially in R. What if you want increase size of your matrix? E.g. you use function ProbUP once for row "0" and than for rows different from jmax (if I correctly understand your code). Use of any

I want to populate the matrix prb through the function HWMProb <- function (a,j,dt) that encapsulates different functions (please see code below), using j= 0:2 for each j. It only populates prb if I specify each function independently in the global environment and then run the loop with the iF statement, as per below. for (j in 0:2) { if (j==0) { prb["0","1"] <-

Hi Eric I did not see any answer and frankly speaking I cannot provide you with canned help. AFAIK if a function is defined within another function (which is your case) it cannot be called directly so it is necessary to define it in global environment. > fff <- function(x) { + myf <- function(a) a+2 + myf(x)^2} > > fff(5) [1] 49 > myf(5) Error in myf(5) : could not find

Hallo, i am evaluating the possibility of using llvm for an application, and i wondering if this is the right place where to pose questions. I haven't found a llvm users mailing list ... Maurizio De Cecco

On Jul 22, 2010, at 4:18 PMPDT, Reza Yazdani wrote: > Hi, > > I ran Spec2006 with -O4. All integer benchmarks passed, but only 8 > out 17 of floating point benchmarks passed. Is this normal or I > made a mistake in my build? Hi Reza. Somebody on Linux should answer, but I don't think it's normal. You may have checked out the source at a moment when it had a bug

Hi, I ran Spec2006 with -O4. All integer benchmarks passed, but only 8 out 17 of floating point benchmarks passed. Is this normal or I made a mistake in my build? Reza -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20100722/4c4a81a9/attachment.html>

Hey there! I got a pseudo code and don't know how to apply it to R, maybe someone can help me: Input: A dataset X, kmax: maximum number of clusters, num_subsamples: number of subsamples. Output: S(i; k) - a distribution of similarities between partitions into k clusters of a reference clustering and clustering of subsamples; i = 1 to num_subsamples Requires: T = cluster(X): A hierarchical

This is my code i don't understand the error message: library(rgenoud) rm(list=ls()) set.seed(666) ######################################################### # As a first step, it is assumed that all input parameters are independent of ageing : ######################################################### InputDim <-20 # Max number of ageings in the inputs CPIRate <- rep(0.02 , InputDim ) #

I have been looking for a while to LLVM as a possible technology to build a DSP execution engine, providing the runtime flexibility needed by real-time interactive DSP applications (like patcher languages), or by plug-in based processors, and in the same moment the powerful link time optimizations that such a system can provide. Such a task is daunting for a single developer project like

On Wednesday 30 July 2008 16:03, David Greene wrote: > Ok, I've been tracking down a remat bug for over a week. I think I finally > understand what is going on. This happens in specfp2006 leslie3d > compiled with our frontend and optimizer. Unfortunately, getting a > testcase is going to be impossible due to restrictions on SPEC > redistribution and the complexity of the

Hi, We are currently working on revising a journal article that describes our work on pre-allocation scheduling using LLVM and have some questions about LLVM's pre-allocation scheduler. The answers to these question will help us better document and analyze the results of our benchmark tests that compare our algorithm with LLVM's pre-allocation scheduling algorithm. First, here is a

dear all, I apologize for my delay in replying you. Here my contribution, maybe just for completeness: Similar to "earth", "segmented" also fits piecewise linear relationships with the number of breakpoints being selected by the AIC or BIC (recommended). #code (example and code from Martin Maechler previous email) library(segmented) o<-selgmented(y, ~x, Kmax=20,