similar to: [LLVMdev] How to get debug dump of candidate pairs selected in BBVectorizer?

My question is: How do I make clang to generate assembly with vector instruction for my target? The back story is: I've added a few vector instructions to my target and confirmed that they are used by running my code on the test below and using a following command: opt i.esencia.ll -S -march=esencia -mcpu=esencia -loop-vectorize | llc -mcpu=esencia -o i.esencia.s target datalayout =

Hi, LLVM’s SLP-vectorizer is a new pass that combines similar independent instructions in a straight-line code. It is currently not enabled by default, and people who want to experiment with it can use the clang command line flag “-fslp-vectorize”. I ran LLVM’s test suite with and without the SLP vectorizer on a Sandybridge mac (using SSE4, w/o AVX). Based on my performance measurements

Hi All, I was going through http://en.wikipedia.org/wiki/Normalized_loop and some other documents on Normalized Loops. LLVM currently doesn't seem to normalize the loop as mentioned in the link (i.e. to start with 0 and have a unit step). I wrote a pass to convert a loop into normalized form as in the link (i.e. to have start value of 0 and step of 1 whenever possible) but I'm not sure

Hi All, I was debugging a clang binary when i found this problem. The following code is complied with clang. typedef struct s { short s; } SVAL; void recurse (SVAL a, int depth) { a.s = --depth; if (depth == 0) return; else recurse(a,depth); } int main () { SVAL s; s.s = 5; recurse (s, 5); return 0; } When i try to access value of a.s in function recurse through gdb(i.e

On Jul 14, 2013, at 9:52 PM, Chris Lattner <clattner at apple.com> wrote: > > On Jul 13, 2013, at 11:30 PM, Nadav Rotem <nrotem at apple.com> wrote: > >> Hi, >> >> LLVM’s SLP-vectorizer is a new pass that combines similar independent instructions in a straight-line code. It is currently not enabled by default, and people who want to experiment with it

Hi David, I think it might not be exactly PR13303 which might be causing the corruption of struct when accessed through GDB. This seems to be an ABI problem in clang. The problem seems to be that when we have pass by value of struct (having indirect arguments) stack is not aligned properly. I tried realigning the stack for indirect arguments in(TargetInfo.cpp) - ABIArgInfo

On Jul 13, 2013, at 11:30 PM, Nadav Rotem <nrotem at apple.com> wrote: > Hi, > > LLVM’s SLP-vectorizer is a new pass that combines similar independent instructions in a straight-line code. It is currently not enabled by default, and people who want to experiment with it can use the clang command line flag “-fslp-vectorize”. I ran LLVM’s test suite with and without the SLP

If you don't use BBVectorize at all, you can ignore this. Hal suggested this in a thread in 2014: http://lists.llvm.org/pipermail/llvm-dev/2014-November/079091.html None objected then, and I don't think any new uses have arisen so I plan to just remove it. It is causing maintenance burden, complexity, and is a set of features I'd rather not port to the new PM. Just an FYI email to

Hi Relph, I'm trying to print the value of 'a' while executing a.s = --depth; I have used break line number instead of break function so that the initial prologue part gets executed. The problem seems to be happening when parameters are pushed into stack and we call a function recursively. For example in the code when we have a int s; inside the struct instead of short s; gdb is able

Hi Everyone, I propose that we remove BBVectorize from trunk. Here's why: - It never made it from "interesting experiment" to "production quality" (it is not part of any in-tree optimization pipeline). - We now have an SLP vectorizer that we do use in production, had have for some time. - BBVectorize otherwise needs refactoring, and the implementation has lots of

On Thu, Dec 6, 2012 at 12:33 AM, Karthik Bhat <karthikthecool at gmail.com> wrote: > Hi David, > > I think it might not be exactly PR13303 which might be causing the > corruption of struct when accessed through GDB. > This seems to be an ABI problem in clang. > The problem seems to be that when we have pass by value of struct > (having indirect arguments) stack is not

Hi All, I compiled the following code on my linux PC using clang with PIE option- struct struct4 {char a; char b; char c; char d; }; struct struct4 foo4 = {'a','2','c','4'}; struct struct4 fun4() { return foo4; } int main() { fun4(); return 0; } > clang -g -fPIE structs.c In GDB session when i call p/c fun4() i get something like below- (gdb) p/c

Hi, Sorry for the delay in response. I measured the code size change and noticed small changes in both directions for individual programs. I found a 30k binary size growth for the entire testsuite + SPEC. I attached an updated performance report that includes both compile time and performance measurements. Thanks, Nadav On Jul 14, 2013, at 10:55 PM, Nadav Rotem <nrotem at apple.com>

> > I assume with target you mean the backend? Consider the x86 backend. It > supports 32bit and 64bit mode, with the GNU x32 ABI in between. There > are three different executable formats support (ELF, PE, MachO) with > different constraints. Some platforms require 32bit alignment of the > stack, others require 128bit alignment. The list goes on. The triple > specifies >

On Fri, 2012-02-03 at 10:28 +0100, Tobias Grosser wrote: > Hi Hal, > > this is one of the first test cases, I would love to have improved > vectorizer support. I sent it out earlier, but I think it is a good time > to look into it again, after the vectorizer was committed. > > The basic examples is a set of scalar loads that load for consecutive > elements and store

Karthik, At what point within recurse() are you asking gdb to display the value of argument a? What I'm wondering about is if the debug information gdb is using to get at a might not be correct at the particular point you are checking a, particularly if that is before the prolog has completed execution. The way debug information for arguments pushed on the stack is represented in

This seems to be another case of PR13303 - since GDB can't figure out where to break for this function based on the debug info (you'll notice when you "break recurse" that it's not breaking on a line or source file, just an address) it's breaking at the very start, before the prologue I'm about to commit a fix to this. On Tue, Dec 4, 2012 at 5:34 AM, Karthik Bhat

Hello, Thanks for your work on the bb-vectorizer. It looks like a promising pass to be used for multi-work-item-vectorization in pocl. On 02/04/2012 06:21 AM, Hal Finkel wrote: > Try it now (after r149761). If this "solution" causes other problems, > then we may need to think of something more sophisticated. I wonder if the case where a store is the last user of the value could

Hi guys im having a problem getting R to numerically integrate for some function, say f(bhat)*optimize(G(bhat)), over bhat. Where id like to integrate this over some finite range, so that here as we integrate over bhat optimize would return a different optimum. For instance consider this simple example for which I cannot get R to return the desired result: f <- function(bhat) exp(bhat) g

I'm trying to vectorize a simple C code. My problem is that I don't quite understand the relationship between clang --target option and the cores mentioned in X86.td as well as other X86 related options (such as -mtune). Below are the command and the code that I'm trying to vectorize. The code compiles but I don't see any vectors. What am I doing wrong? Any help is appreciated.