similar to: [LLVMdev] GlobalsModRef (and thus LTO) is completely broken

----- Original Message ----- > From: "Chris Lattner" <clattner at apple.com> > To: "Chandler Carruth" <chandlerc at gmail.com> > Cc: "LLVM Developers Mailing List" <llvmdev at cs.uiuc.edu>, "Hal Finkel" <hfinkel at anl.gov>, "Justin Bogner" > <mail at justinbogner.com>, "Duncan Exon Smith"

On Mon, Jul 13, 2015 at 10:21 PM Chris Lattner <clattner at apple.com> wrote: > > > On Jul 13, 2015, at 8:19 PM, Chandler Carruth <chandlerc at gmail.com> > wrote: > > > > Ok folks, > > > > I wrote up the general high-level thoughts I have about stateful AA in a > separate thread. But we need to sort out the completely and horribly broken >

Greetings folks, As I'm working on refactoring the AA interfaces in LLVM to prepare for the new pass manager, I keep hitting issues. Some of the complexity that is hitting this stems from the update API, which consists of three virtual functions: deleteValue(Value *V) copyValue(Value *From, Value *To) addEscapingUse(Use &U) These interfaces are *very* rarely called. Here are the only

----- Original Message ----- > From: "Hal Finkel" <hfinkel at anl.gov> > To: "Chandler Carruth" <chandlerc at gmail.com> > Cc: "LLVM Developers Mailing List" <llvmdev at cs.uiuc.edu>, "Daniel Berlin" <dannyb at google.com> > Sent: Thursday, July 2, 2015 3:14:38 PM > Subject: Re: AliasAnalysis update interface - a tale

Hello I am trying to enable globalsmodref-aa by default. globalmodref-aa is a ModulePass and therefore can be invalidated and need to be rerun. I see globalsmodref-aa is enabled in LTO passmanager by adding the analysis pass explicitly. I wonder whether globalsmodref-aa can be enabled/run based on pass dependencies, i.e. those indicated in getAnalysisUsage(). Thanks, Xin -------------- next

Can you say what Benchmark or give a test case so we understand the nature of the regression? As Gerolf said, that will be important to understand what is best to do. On Fri, Jul 17, 2015, 06:43 Evgeny Astigeevich <Evgeny.Astigeevich at arm.com> wrote: > Yes, the regression is stable. I double checked this. A full benchmark > run consists of at least 10 sub-runs to validate the

Based on function names and structures, this is some version of GCC :) Any way you can post the entire .ll file? Because it's globalsmodref, it's hard to debug without the other functions, since it goes over all the functions to determine address takenness, etc :) On Tue, Jul 21, 2015 at 3:23 PM, Michael Zolotukhin <mzolotukhin at apple.com> wrote: > Hi Chandler, > > We

Before the fix, the compiler may simply return 'noalias' for cases it can not really prove to be noalias, but actually correct by luck (or even wrong noalias, but does not result in miscompile). It would be useful to find out the set of missed noalias queries from GlobalModRef with your benchmark and examine if there is some improvement can be done. David On Fri, Jul 17, 2015 at 6:32

Hello, I'm playing with alias analysis, using the following program: %i = external global int ; <int*> [#uses=2] implementation ; Functions: int %_Z3bari(int %p) { entry: %tmp.0 = load int* %i ; <int> [#uses=1] %tmp.1 = setgt int %tmp.0, 10 ; <bool> [#uses=1] br bool %tmp.1, label %then, label %UnifiedReturnBlock then:

Hi Chandler, I would like to run some benchmarks on ARM hardware and to look at impact of your patches on LTO. Kind regards, Evgeny Astigeevich From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Chandler Carruth Sent: 15 July 2015 10:45 To: Chandler Carruth; Gerolf Hoflehner Cc: LLVM Developers Mailing List Subject: Re: [LLVMdev] GlobalsModRef

Hey, thanks for benchmarking. How stable is the 2% regression? Michael ran some benchmarks with GlobalsModRef completely disabled and the only differences were in the noise. This was a complete spec2k6 run along with some others. Based on the number of benchmarks run there, I'm going to go ahead and submit these patches, but if you can clarify the impact here, we can look at potentially some

On Fri, Jul 17, 2015 at 9:13 AM Evgeny Astigeevich < evgeny.astigeevich at arm.com> wrote: > It’s Dhrystone. > Dhrystone has historically not been a good indicator of real-world performance fluctuations, especially at this small of a shift. I'd like to see if we see any fluctuation on larger and more realistic application benchmarks. One advantage of the flag being set is that we

Greetings folks! I would like to enable globalsmodref-aa in the non-LTO pass pipeline so that it gets tested more and there are fewer differences between the two. For all of my benchmarks, this is performance neutral, but I'd appreciate others benchmarking this combination to see if they see any benefits or regressions. You can demo this mode easily: -mllvm -enable-non-lto-gmr Please let me

Replying here, but several of the questions raised boil down to "couldn't you make the usage of GetUnderlyingObject conservatively correct?". I'll try and address that. I think this *is* the right approach, but I think it is very hard to do without effectively disabling this part of GlobalsModRef. That is, the easy ways are likely to fire very frequently IMO. The core idea is

On Monday 25 April 2005 14:43, Vladimir Prus wrote: > The 'i' variable is never modified in the program, however, all analyses > except for -globalsmodref-aa report that the > > %tmp.3 = call int %_Z3bari( int %p ) ; <int> [#uses=1] > > instruction can modify 'i'. I'm somewhat surprised, because it looks like > -globalsmodref-aa is the simplest

# Front matter # First, I want to emphasize that this is not a short-term plan. This is a long-term plan. When I was looking at refactoring the existing Alias Analysis infrastructure in LLVM in ways that would break stateful AA, Hal Finkel asked me to figure out how things would work with the new pass manager and make sure that was sane. This plan assumes the new pass manager. There is nothing

Hi there, I am doing: opt -print-memdeps ./test.bc -analyze -globalsmodref-aa by adding globalsmodref-aa, I am hoping that globalsmodref alias analysis will be used. However, it does not turn out to be so. I found this out by adding some "errs() << " into the source code for that alias analysis. So my question is what should I do to let memory dependency analysis use

----- Original Message ----- > From: "James Molloy via llvm-dev" <llvm-dev at lists.llvm.org> > To: "Vaivaswatha Nagaraj" <vn at compilertree.com> > Cc: "LLVM Dev" <llvm-dev at lists.llvm.org> > Sent: Thursday, December 3, 2015 4:41:46 AM > Subject: Re: [llvm-dev] Function attributes for LibFunc and its impact on GlobalsAA > >

On Mon, 25 Apr 2005, Vladimir Prus wrote: > The GlobalsModRef::getModRefInfo has this logic: > > // If we are asking for mod/ref info of a direct call with a pointer to a > // global we are tracking, return information if we have it. > if (GlobalValue *GV = const_cast<GlobalValue*>(getUnderlyingObject(P))) > if (GV->hasInternalLinkage()) > > So, no

Jack Howarth wrote: > I've had better luck compiling all of pymol 1.1r2 with > -O4 on darwin9. Everythink links and there appears to be > no regressions in the resulting code. I take it that LTO > in llvm 2.5 is still limited to dead code elimination, > correct? No. libLTO does the equivalent to opt -internalize -ipsccp -globalopt -constmerge -deadargelim -instcombine