similar to: [LLVMdev] CodeExtractor is not aware of debugger debugger intrinsics

Hi Pablo, Your tool sounds really cool. It goes well beyond what I’m trying to do, which is really just extracting blocks of code, serializing and sending the inputs over to another core, running the code over there, and then sending the outputs back to the original caller (like an automatic RPC). So it sounds like most of the things your tool can do would be overkill for my use case. What I’ve

Hi, I've been experimenting with the CodeExtractor class and have been running into a problem where the verifier pass fails with the error "function-local metadata used in wrong function" when I compile with debugging symbols enabled (-g flag). The issue seems to be that after extraction we have metadata that references variables that aren't allocated in the extracted function.

On Fri, Jan 08, 2010 at 05:04:17PM -0800, Chris Lattner wrote: > On Jan 8, 2010, at 5:01 PM, Julien Lerouge wrote: > >Hello, > > > >The CodeExtractor contains a std::set<BasicBlock*> to keep track > >of the > >blocks to extract. Iterators on this set are not deterministic, and so > >the functions that are generated are not (the order of the >

Hello, The CodeExtractor contains a std::set<BasicBlock*> to keep track of the blocks to extract. Iterators on this set are not deterministic, and so the functions that are generated are not (the order of the inputs/outputs can change). The attached patch uses a SetVector instead. Ok to apply ? Thanks, Julien -- Julien Lerouge PGP Key Id: 0xB1964A62 PGP Fingerprint: 392D 4BAD DB8B CE7F

Hello guys, Do we have a pass to do unconditional branch folding except the one in Codegen? I checked in the lib/Transforms/Scalar, did not see any promising one. thank you, yuanfang

I am working on a pass to extract small regions of code to run somewhere else (different node in a cluster). Basically what I need is the ability to isolate a region of code, get its inputs and outputs, create a new function with the extracted code and code aggregating the in and out parameters as structs that can be cast for a “void*”-based interface. It looks like the CodeExtractor

On Jan 8, 2010, at 5:01 PM, Julien Lerouge wrote: > Hello, > > The CodeExtractor contains a std::set<BasicBlock*> to keep track of > the > blocks to extract. Iterators on this set are not deterministic, and so > the functions that are generated are not (the order of the > inputs/outputs can change). > > The attached patch uses a SetVector instead. Ok to apply ?

On 18/11/13 21:43, Brandon Holt wrote: > I am working on a pass to extract small regions of code to run > somewhere else (different node in a cluster). Basically what I need is > the ability to isolate a region of code, get its inputs and outputs, > create a new function with the extracted code and code aggregating the > in and out parameters as structs that can be cast for a

Hello guys, For single scalar pointer parameter, we have 'byval' to specify the pointee is not changed. But for base+range parameters like: declare void @_gfortran_transfer_integer_write(%struct.__st_parameter_dt*, i8*, i32) (This is the fortran runtime api to 'printf' one integer) The 2nd & 3rd args are used to specify the memory of that integer. I didn't find a

On Nov 19, 2008, at 7:41 AM, Jack Tzu-Han Hung wrote: > Hi, > > I found a problem in CodeExtractor::severSplitPHINodes() > <CodeExtractor.cpp>. > > The algorithm first separates the header block into two, one > containing only PHI nodes and the other containing the remaining non- > PHI nodes. The variable NewBB holds the pointer to the latter half > block.

Hi Devang, Thanks for your reply. But if you look at the comment of BasicBlock::splitBasicBlock(), it says that "...an unconditional branch is added to the new BB, and the rest of the instructions in the BB are moved to the newBB, including the old terminator." So, the terminator of the newBB is exactly the same as the terminator of the oldBB. IF the oldBB has multiple successors,

Hi, I found a problem in CodeExtractor::severSplitPHINodes() <CodeExtractor.cpp>. The algorithm first separates the header block into two, one containing only PHI nodes and the other containing the remaining non-PHI nodes. The variable NewBB holds the pointer to the latter half block. Later, it tries to update DT information. if (DT) DT->splitBlock(NewBB); In splitBlock, it checks

On Dec 31, 2010, at 11:20 AM, Vu Le wrote: > There might be a misuse of DominatorTree::splitBasicBlock in CodeExtractor.cpp, line 145. > Header is splited into two (OldPred->NewBB). > > Line 145 updates the dominator tree by calling DT->splitBasicBlock(NewBB). > I think it should be DT->splitBasicBlock(OldPred). > > When I tried to extract a code region whose

I'd just note that not every pass you can run with "opt" is actually part of the optimization pipeline. There are a few important IR-level passes that only run in the codegen pipeline, but are still nameable with opt to run individually for testing purposes. Switching over doesn't need to block on these passes being migrated. So I'm not sure this method of determining

> From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] > On Behalf Of Wei Dang > Subject: [LLVMdev] Question about method CodeExtractor::severSplitPHINodes > In the implementation of serverSplitPHINodes(), why it only checks the > first PHI node for possible multiple inputs from outside the region to > extract. There could be more than one PHI nodes in

One thing I want to mention. I believe in the current legacy pass manager implementation only one MachineFunction ever exists at a time. It is deleted before the next MachineFunction is created. This is very important for memory usage. I think the MachineOutliner being in the pipeline may create an exception to this. I think the initial version of retpoline used a ModulePass and that had to be

I forgot to cc llvmdev. Here is my original message. I'm a bit confused on DominatorTreeBase::Split() ( http://llvm.org/docs/doxygen/html/Dominators_8h_source.html#l00229) If a basic block A splits into A->B, when I call Split(), which is NewBB? A or B? The semantics shows that NewBB is the newly split basic block B. But the assertion at line 229 doesn't seem right. 00229

Dear Chandler, > Please let me know if you have out-of-tree users of this logic At KernelGen we have an out-of-tree variation of CodeExractor called BranchedCodeExractor [1], which instead of taking a code region into a new function, does it conditionally: ORIGINAL_CODE; ->> if (extracted_code_function(args) != -1) { ORIGINAL_CODE; } I think many hybrid and parallelizing tools

> From: Wei Dang [mailto:jacdang at gmail.com] > Subject: Re: [LLVMdev] Question about method CodeExtractor::severSplitPHINodes > Please excuse me if I'm not supposed to reply to all. You should do reply-all, to make sure the list sees all of the thread. > Are you saying all PHI nodes are required to include all its predecessor blocks > no matter they have input or not?

On Wed, Jul 15, 2020 at 6:39 PM Chen, Yuanfang via llvm-dev <llvm-dev at lists.llvm.org> wrote: > Indeed, but there is a distinction about their position in the pipeline. We run opt & codegen pipeline separately, Why, though? Is there a reason why this inherently makes sense, or is it just a historical accident? At least to me it seems that it would make more sense to run all passes