similar to: always allow canonicalizing to 8- and 16-bit ops?

Thanks for the perf testing. I assume that DAG legalization is equipped to handle these cases fairly well, or someone would've complained by now... FWIW (and at least some of this can be blamed on me), instcombine already does the narrowing transforms without checking shouldChangeType() for binops like and/or/xor/udiv. The justification was that narrower ops are always better for

Hello Thanks for looking into this. I can't be very confident what the knock on result of a change like that would be, especially on architectures that are not Arm. What I can do though, is run some benchmarks and look at that results. Using this patch: --- a/lib/Transforms/InstCombine/InstructionCombining.cpp +++ b/lib/Transforms/InstCombine/InstructionCombining.cpp @@ -150,6 +150,9 @@

Thanks Krzysztof. That's very helpful - I was missing the distinction between a register containing an undefined value and a register marked as containing an undefined value. Cheers! On 1/23/18, via llvm-dev <llvm-dev at lists.llvm.org> wrote: > Send llvm-dev mailing list submissions to > llvm-dev at lists.llvm.org > > To subscribe or unsubscribe via the World Wide Web,

Hi Sanjay, 2017-01-30 22:22 GMT+01:00 Sanjay Patel <spatel at rotateright.com>: > My minimal patch idea is to ease the restriction in ShouldChangeType > because i1 is special. I tried the patch below, and it works on the > example...and nothing in 'make check' failed. :) > Yeah, that would work for me as well, I just wasn't sure about the implications that has.

Hi Sanjay, unfortunately that patch does not help in my case. Here's the IR that fails to get fully optimized: target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" target triple = "x86_64-unknown-linux-gnu" define fastcc zeroext i1 @testfunc(i8** noalias nocapture readonly dereferenceable(8)) unnamed_addr { entry-block: %1 = load i8*, i8**

On Tue, Jan 31, 2017 at 3:09 AM, Björn Steinbrink <bsteinbr at gmail.com> wrote: > Hi Sanjay, > > 2017-01-30 22:22 GMT+01:00 Sanjay Patel <spatel at rotateright.com>: > >> My minimal patch idea is to ease the restriction in ShouldChangeType >> because i1 is special. I tried the patch below, and it works on the >> example...and nothing in 'make

2017-01-31 16:59 GMT+01:00 Sanjay Patel <spatel at rotateright.com>: > > > On Tue, Jan 31, 2017 at 3:09 AM, Björn Steinbrink <bsteinbr at gmail.com> > wrote: > >> Hi Sanjay, >> >> 2017-01-30 22:22 GMT+01:00 Sanjay Patel <spatel at rotateright.com>: >> >>> My minimal patch idea is to ease the restriction in ShouldChangeType

Hi Björn and Daniel, FoldPHIArgOpIntoPHI() is for unary ops (casts), but it calls FoldPHIArgBinOpIntoPHI() for binops which takes care of almost everything else? My minimal patch idea is to ease the restriction in ShouldChangeType because i1 is special. I tried the patch below, and it works on the example...and nothing in 'make check' failed. :) Index:

Greetings, A couple general questions regarding the use of splines to interpolate depth profile data. Here is an example of a set of depths, with associated attributes for a given soil profile, along with a function for calculating midpoints from a set of soil horizon boundaries: #calculate midpoints: mid <- function(x) { for( i in 1:length(x)) { if( i > 1) { a[i] = (x[i] -

Hi, I'm Luca from Savona, Italy. I'm using Compiz and I'm writing an application with Xlib to draw two windows, let's say window_A and window_B. I would like to keep fixed the relative position between two established corners of them that they have when mapped. So that if I move or resize one of the window, the other if necessary automatically moves to observe the condition.

Hi Luca, In recent versions of compiz (0.9.10) if you have lazy positioning on (org.compiz.move 'lazy-positioning') other clients won't get a configure notification until the end of the move. This is a performance optimization to avoid a slow path on the nvidia drivers with vsync. If you want such notifications you need to turn this option off. Either that or implement this as a

Hi, I wonder why my attempt to extend an existing loess fit to a new data set is producing error. I was trying the following: dat = read.csv(choose.files()) x = dat[,2]; y = dat[,1] x.sort = sort(x) y.loess = loess(y~x, span=0.75) # For testing the above fit with a new dataset: test = read.csv(choose.files()) # test data new_x = test [,1]; new_y = test[,2] new_x.sort = sort(new_x) predicted

Buenas Tengo un Data table de la siguiente manera: datos<-data.table(grupo=rep(c("a","b"),5),x=c(1:10),y=rnorm(10,2,1)) Lo que quiero es añadir una fila por cada grupo y en esa nueva fila, al valor de la x ponerle el valor anterior de la y Lo que hago es añadir una nueva fila por grupo, con: datos[,.SD[1:(.N+1)],by=grupo] Y para añadir el valor anterior uso la función

Hi everyone, this should be pretty basic but I need asking for help as I got stuck. I am running simple linear regression models on R with k regressors where k > 1. In order to automate my code I packed all the regressors in a matrix X so that lm(y~X) will always produce the results I want regardless of the variables in X. I am new to R but I found this advice somewhere so I guess it is

Hi all, I have a data.frame with the following colnames pattern: x1 y11 x2 y21 y22 y23 x3 y31 y32 ... I.e. I have an x followed by a few y's. What I would like to do is turn this wide format into a tall format with two columns: "x", "y". The structure is that xi needs to be associated with yij (e.g. x1 should next to y11 and y12, x2 should be next to y21, y22, and

Changes from v3: - mini-os configuration files moved into stubdom/ - mini-os extra console support now a config option - Fewer #ifdefs - grant table setup uses hypercall bounce - Xenstore stub domain syslog support re-enabled Changes from v2: - configuration support added to mini-os build system - add mini-os support for conditionally compiling frontends, xenbus -

Hello, I have been trying for weeks to compile a 64-bit Rblas. I started with ATLAS where I have had success in the past, but 64 bit was not behaving, and as each compilation takes between 9 and 12 hours, "test, check, and revise" was not going to be really viable. I therefore switched to OpenBLas (OPBL). I was successful in compiling R-3.0.1 and an OPBL-based BLAS for Windows 64bit

I am trying to allow for AND type ''find'' but to allow for simply a single set of find criteria. controller code... @vw_string = @vw_string1 = @vw_string2 = @vw_string3 = [] if params[:beg_intake_date] != "" then @vw_string1 = ["intake_date between ? and ?", params[:beg_intake_date], params[:end_intake_date] ] end if

Hi, I'm trying to do some constrained non-linear optimisation, but my function does not have second order derivatives everywhere. To be a little more specific (the actual function is huge and horrible, so it would probably be better to just describe it) my model has four variables and I'm using optim to minimise an error term. My data is split into discreet days and I have two types of

After searching in R- Introduction, FAQ, help... I don't understand this: I write a function in a file (.R): tt <- function(mc) { date() mc<-read.csv2("machines.txt",na.strings="") date() } I source it in R and I type tt(). The answer is > tt() [1] "Mon Aug 14 11:18:25 2000" > The instructions following the first "date()" are ignored. Why?