similar to: [LLVMdev] LoopInfo fails

Hi! if I have this little function: float loopTest(float a, float b) { float x = 1.0f; int i = 0; if (a < b) goto l; for (i = 0; i < 10; ++i) { x *= a; l: x += b; } return x; } then llvm::LoopInfo does not detect that there is a loop. Is this a bug or is such a case not allowed? -Jochen

On Thursday 11 March 2010 12:36:55 Jochen Wilhelmy wrote: > Hi! > > I'd like to analyze a function with the llvm::LoopInfo pass and then > write some info to disk. I create a FunctionPassManager and insert it and > a DominatorTree which is required by LoopInfo. > As it turns out the analysis info is only available to other > passes and gets cleared when the run function of

Hi! I'd like to analyze a function with the llvm::LoopInfo pass and then write some info to disk. I create a FunctionPassManager and insert it and a DominatorTree which is required by LoopInfo. As it turns out the analysis info is only available to other passes and gets cleared when the run function of the pass manager returns. I therefore guess that I need to create an own pass. Is this

I think Chris is saying that the and is necessary because with your i1 trunc you're ignoring all of the high bits. The and implements that. If you don't want this behavior, don't generate the trunc in the first place and just compare the full width to zero. Reid On Wed, Dec 28, 2011 at 6:45 AM, Jochen Wilhelmy <j.wilhelmy at arcor.de>wrote: > > >> Hi! >

Hi! > I'm not sure what you mean by this new pass being of no use to other passes. > It's not generating any new information about the program (or if it is, it's > writing it to disk), so other passes wouldn't need it anyway. > So I know that I'm on the right way. I just think it's a bit of overhead implementing an own pass and register it using

On 25.11.2010 23:56, Michael Spencer wrote: > On Nov 25, 2010, at 5:01 PM, Jochen Wilhelmy <j.wilhelmy at arcor.de > <mailto:j.wilhelmy at arcor.de>> wrote: > >> Hi! >> >> Of course nobody wants to implement unicode support for windows >> because windows should support an utf8-locale and windows is obsolete >> anyway ;-) >> >> But

A natural loop is one whose header dominates all of the nodes in the loop. There is probably some block outside of the loop jumping to a block in the loop body. Cameron On Apr 29, 2011, at 7:43 PM, Bo Wu <bwu at cs.wm.edu> wrote: > Hi, > > I found that some loops can not be identified by LoopInfo pass. For example, the loop at line 3094 of rdopt.c of benchmark 464.h264ref from

On Nov 25, 2010, at 5:01 PM, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > Hi! > > Of course nobody wants to implement unicode support for windows > because windows should support an utf8-locale and windows is obsolete > anyway ;-) > > But there is a simple solution: use boost::filesystem::path everywhere you > use file names and paths, for example in

Thanks for the reply. The LoopInfo pass can actually identify the loop I mentioned. My former question is due to a bug in my pass. What I worried is that LoopInfo can not identify all the natural loops for the benchmarks I use, but now I haven't found any such cases. Bo On Sun, May 1, 2011 at 4:24 PM, John Criswell <criswell at illinois.edu> wrote: > On 4/30/11 8:52 AM, Bo Wu

On Fri, 02 Mar 2012 13:55:18 +0100 Jochen Wilhelmy <jochen.wilhelmy at googlemail.com> wrote: > Hi! > > in the llvm code there are several places with hardcoded function > names for e.g. sin, sinf, sqrt, sqrtf etc., namely > ConstantFolding.cpp > InlineCost.cpp > SelectionDAGBuilder.cpp > IntrinsicLowering.cpp > TargetLowering.cpp > > my question is:

On Fri, 26 Nov 2010 09:28:17 -0500 Michael Spencer <bigcheesegs at gmail.com> wrote: > On Fri, Nov 26, 2010 at 4:00 AM, Jochen Wilhelmy > <j.wilhelmy at arcor.de> wrote: > > No, this post was prompted since I switched to boost::filesystem > > version 3 in my own code and llvm/clang 2.8 > > was the only lib with no unicode support on windows. > > Will your

On Fri, Nov 26, 2010 at 4:00 AM, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > No, this post was prompted since I switched to boost::filesystem version 3 > in my own code and llvm/clang 2.8 > was the only lib with no unicode support on windows. > Will your code be api compatible to boost::filesystem? No. boost::filesystem makes extensive use of exceptions, which LLVM is

On 4/30/11 8:52 AM, Bo Wu wrote: > In C code, if a loop is not a natural loop, that means its loop body > should at least have one label, right? In that case, some BB out of > the loop can jump to the loop body, so the loop has more than on entry. Off the top of my head, I would think you are right: a C label would have to exist within the loop in order to have a jump from outside the

On 3 March 2010 11:56, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > > the sin and cos calls are folded, but not the mul by zero. > Is x*0 => 0 true if isnan(x)? And cos(x)*sin(x) makes me desperately want to fold it to sin(2*x)/2, but I suppose that's not allowed either.

Hello Jochen- Your expectation is incorrect, I'm afraid. Typically this would be expanded to something like this: … %cmp = … br i1 %cmp, label %ternary.true, label %ternary.false ternary.true: %truevalue = … br label %ternary.end ternary.false: %falsevalue = … br label %ternary.end ternary.end: %value = phi %type [%truevalue, %ternary.true], [%falsevalue,

On Tue, Feb 2, 2010 at 11:40 AM, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > Hi! > > I have a llvm::Module that references an external function > and a second mdoule that implements the function. > The implementation consists of only one instruction. > How can I explicitly inline this function? > > -Jochen > >

On Thu, Apr 8, 2010 at 2:10 AM, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > Hi! > > while trying to use llvm::DominatorTreeBase on a custom graph that > has nothing to do with llvm::BasicBlock I ran into some difficulties, > because llvm::DominatorTreeBase calls e.g. getParent()->front() > directly on the nodes and uses llvm::Inverse which forced me to >

Instcombine does of this, late codegen also does some of it. -Chris On May 14, 2010, at 5:58 AM, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > Hi! > > Is there a pass that optimizes vector operations? > If I have for examle a sequence of shufflevector instructions > that optimizes them? > (in opencl notation e.g. a.xyzw.wzyx.xxxx -> a.wwww) > > -Jochen

Can you be more precise about where the error occurs? Depending on where it is, LLVM may or may not be able to propagate the error safely. Reid On Tue, Oct 26, 2010 at 4:05 PM, Jochen Wilhelmy <j.wilhelmy at arcor.de> wrote: > Hi! > > I have this scenario: > executionEngine->getPointerToFunction fails because a symbol can not be > resolved. > > What can I do to

I think the way to do this is to generate a shared library which you reload. Shared libraries already handle the problems of PIC code and relocatable globals, so I think it would be best to leverage that. However, there's obviously some issues here. Presumably you don't have all the code up front, or you wouldn't be using a JIT in the first place. I don't think you can update