similar to: [LLVMdev] Memory leak?

On Wed, Dec 10, 2014 at 05:27:45PM -0800, Duncan P. N. Exon Smith wrote: > +zalman at google.com > Hi Duncan, This patch plus another small change fixes the assertion failure for me. With the patch alone, the void* overload of addGarbageObject() was being used by MDNode::getTemporary(), so I had to cast the object as an MDNode*: diff --git a/lib/IR/Metadata.cpp b/lib/IR/Metadata.cpp

> On 2014 Dec 10, at 14:08, Tom Stellard <tom at stellard.net> wrote: > > On Wed, Dec 10, 2014 at 11:21:08AM -0800, Duncan P. N. Exon Smith wrote: >> >>> On 2014 Dec 10, at 08:40, Tom Stellard <tom at stellard.net> wrote: >>> >>> On Tue, Dec 09, 2014 at 09:22:16PM -0800, Duncan P. N. Exon Smith wrote: >>>> The `Metadata`/`Value`

I committed: r224058 = 966942da9e68b59c31ce770e7f94c55a63482c6b r224060 = da75f7277e3a129aed8ef8aa4e0d84de40b76fd4 r224061 = f88e4c8e9171045454b2c8e05054c2af8da3fe4f Let me know if somehow you're still hitting the problem. r224061 removes leak detection entirely from `MachineInstr`. There aren't any leaks to be had there, since they're allocated in a custom allocator. They're

Hi I just had some code which filtered via isa<UnaryInstruction>. Strangely enough not only the instructions which inherited UnaryInstructions but also: GetElemPtrInst, SwitchInst , xor, shl where true on isa<UnaryInstruction>. While the first two seem to be at least semantically ok the latter seem somehow wrong? I didn't find the place where the isa_impl templates where

On Dec 7, 2007, at 9:44 AM, ST wrote: > Hi > > I just had some code which filtered via isa<UnaryInstruction>. > Strangely > enough not only the instructions which inherited UnaryInstructions > but also: > GetElemPtrInst, SwitchInst , xor, shl > where true on isa<UnaryInstruction>. While the first two seem to be > at least > semantically ok the

> On 2014 Dec 10, at 08:40, Tom Stellard <tom at stellard.net> wrote: > > On Tue, Dec 09, 2014 at 09:22:16PM -0800, Duncan P. N. Exon Smith wrote: >> The `Metadata`/`Value` split (PR21532) landed in r223802 -- at least, the >> C++ side of it. This was a rocky day, but I suppose that's what I get >> for failing to stage the change in smaller pieces. >>

Hello, I have a problem of generating a load instruction. The LLVM bytecode is: ------------------------ entry: ... %2 = call i32 (...)* @atoi(i8*%1) nounwind /*<- Insertpos*/ ... -- bb1: .. %5 = icmp sgt i32 %2, %i.0 ... ----------------- Now I have pb: pointer to the Value object of *%2* of bb1. Here, I want to generate a load instruction and I did it as: new LoadInst(pb,

Can anyone help with these crazy linking errors?? /home/kgibbs/CS321/CVS/llvm/lib/Debug/vmcore.o(.gnu.linkonce.t._ZN4llvm12Lea kDetector16addGarbageObjectEPKNS_5ValueE+0xd): In function `llvm::LeakDetector::addGarbageObject(llvm::Value const*)': /usr/include/c++/3.2.1/bits/stl_pair.h:148: undefined reference to `llvm::LeakDetector::addGarbageObjectImpl(llvm::Value const*)'

Hi all, I'm having a bit of trouble with the online doxygen documentation. As far is I know, it should reflect current svn trunk. However, it seems out of date. For example, the ExtractValueInst [1] is not a UnaryInstruction yet and still takes Value* as indices. To add to the confusing, the mainpage [2] says it is documentation version "2.1svn" which seems even more weird. What

You'll need to include the "support" library to resolve these LeakDetector symbols. You can do this by either placing the path to "support.o" or -lsupport on your link line (in your Makefile). Note that if you use the -l option, you'll likely need to use -L<path> as well to tell the linker where to find "libsupport.a" (file included by the

Hi all, I am having some some trouble with LLVM 3.8... I built it with GCC 4.8.2 with cmake and no special CXX/C flags. That means in case of GCC that RTTI is switched on, I believe. However, building my MCJIT application against the just built LLVM and linking results in: libqdp.a(qdp_llvm.o):(.rodata._ZTIN4llvm8FCmpInstE[_ZTIN4llvm8FCmpInstE]+0x10): undefined reference to `typeinfo for

Zheng Wang <jason.wangz at gmail.com> writes: > I have a problem of generating a load instruction. The LLVM bytecode is: > > ------------------------ > entry: > ... > %2 = call i32 (...)* @atoi(i8*%1) nounwind > /*<- Insertpos*/ > ... > > -- > bb1: > .. > %5 = icmp sgt i32 %2, %i.0 > ... > ----------------- > > > Now I have >

Hi, I am trying to figure out the best way to add some extra metadata to the store and load llvm-ir instructions: The metadata content is a 'Value*' expression representing a side channel, containing dependency information that is used to help the Scoped Alias Analysis. Optimizations that don't know about this side channel can safely drop it, where the only effect would be on the

Ok that works but why on earth would I get a runtime error of: Type.cpp:132: const llvm::Type* llvm::Type::getForwardedTypeInternal() const: Assertion `ForwardType && "This type is not being forwarded to another type!"' failed. Aborted For the line of code: MyModule = new Module( std::string("IDontWork")); > -----Original Message----- > From: llvmdev-admin

> I'm having a bit of trouble with the online doxygen documentation. As far is I > know, it should reflect current svn trunk. However, it seems out of date. For > example, the ExtractValueInst [1] is not a UnaryInstruction yet and still > takes Value* as indices. > > To add to the confusing, the mainpage [2] says it is documentation version > "2.1svn" which seems

I'm sure there is a clever way to do the following, but I've been unable to find it on this forum or by writing my own functions. I have 8 years worth of weekly data but would like to restrict the labels on the x-axis to months only. I've included the first year's worth of data below. My line of thought has been along these lines

There is a phi-node "%phi = phi i64 [%cast1, %if], [%cast2, %else]" in the common successor. The actual control flow is a bit more complex, but there is a common successor block, and %cast1 and %cast2 are the two values that the phi node in the common successor takes. Does the existence of the phi node affect optimization? Thanks, Taewook From: <mats.o.petersson at

On Apr 16, 2008, at 2:50 AM, heisenbug wrote: > > And now here is my educated speculation: > There are 2 things that became slower > 1) Use::getUser() > 2) Use::get/set due to tagging. > > The former is seldom called: > > $ find lib -name "*.cpp" | xargs grep "getUser(" | wc -l > 41 The majority of those aren't actually Use::getUser, but

Hello, I found that GVN doesn't promote identical sequence of instructions in if and else branch to their common predecessors. For example, for the following code snippet pred: … br i1 %cmp, label %if, label %else if: %incdec.ptr.1 = getelementptr inbounds i8, i8* %ptr, i64 1 %cast1 = ptrtoint i8* %incdec.ptr.1 to i64 … else: %incdec.ptr.2 = getelementptr inbounds i8, i8* %ptr,

and by "right thing" i mean it can hoist if you want and it can prove it will not extend the live range. Note that VBE (very busy expressions) is a code size optimization only. It does not save time. On Tue, Feb 9, 2016 at 12:26 PM, Daniel Berlin <dberlin at dberlin.org> wrote: > This GVN does not do that, this is correct. It is a very simple GVN. All > phi nodes are