similar to: [LLVMdev] How to get the Instruction where one function use the global variable.

How can I go about evaluating constant expressions and providing them as the initial value to globals? I allow my globals to be initialized by an expression. In C syntax this looks like: int const global_val = const_expr( 1, 2, 3 ); I use the "const_expr" to indicate construction-time constant folding will not be possible. The IR optimizers however will likely reduce this to a single

Hi Everyone, I'm quite new to LLVM and I want to update value of global variable in LLVM IR. I created new global variable in ModulePass: bool runOnModule(llvm::Module &M) { IRBuilder<> Builder(M.getContext()); Instruction *I = &*inst_begin(M.getFunction("main")); Builder.SetInsertPoint(I); M.getOrInsertGlobal("globalKey",

I don't think it's the same problem as you described. By printing I meant calling printf function and passing my global variable as one of the arguments. My code: Instruction* InstructionVisitor::incrementGlobalKey(Instruction* I) { IRBuilder<> Builder(I->getContext()); Builder.SetInsertPoint(I->getNextNode()); GlobalVariable* key =

Hi Dmitry, > I'm attaching a patch that should fix the issue mentioned above. It > simply makes the same check seen in the same file for global > variables: > > emitPTXAddressSpace(PTy->getAddressSpace(), O); > if (GVar->getAlignment() == 0) > O << " .align " << (int) TD->getPrefTypeAlignment(ETy); > else > O

Hi Duncan, You're right, global variables use preferred alignment. And - yes, preferred alignment in this case is bigger: 8 instead of 4. NVIDIA's prop. compiler gives 4. However, since CUDA 5.0 ptx modules are linkable with each other, I think alignments for externally visible functions and data should all follow ABI rules. Is there a guide on making tests? I have ~5 pending patches

Hi everyone, I am doing some work with LLVM IR, I need to use LLVM IR to do operation on C variables. Code emission is done by LLVM JIT. That variable is C thread local , for example __thread int* gvar; I think some methods to convert that variable to LLVM IR, (1) use external function I know LLVM IR is able to call an external function, so I can write codes that look like: int* load()

Hi Dmitry, > You're right, global variables use preferred alignment. And - yes, > preferred alignment in this case is bigger: 8 instead of 4. NVIDIA's > prop. compiler gives 4. However, since CUDA 5.0 ptx modules are > linkable with each other, I think alignments for externally visible > functions and data should all follow ABI rules. giving it an alignment of 8 does

Test cases exist under test/CodeGen/NVPTX (name changed in May). Now that I'm back at NVIDIA, I'm going to be running through the bugzilla issues (thanks Dmitry for the reports!). I have practically the exact same patch here in my queue. :) In this case, I would prefer ABI alignment for compatibility with the vendor compiler. It should work either way, but I do need to audit the

Dear all, I'm attaching a patch that should fix the issue mentioned above. It simply makes the same check seen in the same file for global variables: emitPTXAddressSpace(PTy->getAddressSpace(), O); if (GVar->getAlignment() == 0) O << " .align " << (int) TD->getPrefTypeAlignment(ETy); else O << " .align " <<

Hi, I am brand new to R and not familiar with the language, though I have been reading the manuals and making some slow going progress. I am working with some source code from a Global Vector Auto -Regressive program written by Ranier Puhr from the R-forge group. I need help interpreting the processes of the following code. I am going to post in parts since it's pretty long: GVAR

Not to distract, but the word, `procedurization' is not an English word. It's just leaping out at me when it is either procedure(s) (noun) or proceduralize (verb). Even processes would make sense. I couldn't help myself because the word was distracting. - Marc P.S. Not that my vote counts, but I'm more in the camp of Hal whose approach to tackling the parallelization

Hi All, We'd like to make a proposal for OpenMP representation in LLVM IR. Our goal is to reach an agreement in the community on a simple, complete and extensible representation of OpenMP language constructs in LLVM IR. Hopefully, this would serve as a common ground and would enable further development of OpenMP support both in Clang and LLVM compiler toolchain. We seek feedback on the

Hi, I'm trying to define a mutable variable, outside functions. The code below is trying to evaluate an expression much like "x = 1" at the module level in Python. It appears that when it tries to JIT the function there is an error because there is no storage allocated for 'x'. How do I do that? thanks, Rob ----- dumped IR ----- ; ModuleID = 'repl-module' %0 =

Hello, FYI, this is a bug http://llvm.org/bugs/show_bug.cgi?id=13324 When compiling the following code for sm_20, func params are by some reason given with .align 0, which is invalid. Problem does not occur if compiled for sm_10. > cat test.ll ; ModuleID = '__kernelgen_main_module' target datalayout = "e-p:64:64-i64:64:64-f64:64:64-n1:8:16:32:64" target triple =

So it looks like the verifier doesn't catch this condition - I think it should. The attached program reproduces the problem - verification succeeds, but the linker fails with a type assertion. BTW, if no one has the bandwidth to work on this I'm willing to attempt a fix, assuming that you agree that the verifier should discover this condition. Michael Muller wrote: > >

I have written two functions which do useful things with panel data a.k.a. longitudinal data, where one unit of observation (a firm or a person or an animal) is observed on a uniform time grid: - The first function makes lagged values of variables of your choice. - The second function makes growth rates w.r.t. q observations ago, for variables of your choice. These strike me as

Andrey, While I think that it will be relatively easy to have the intrinsics serve as code-motion barriers for other code that might be threads sensitive (like other external function calls), we would need to think through exactly how this would work. The easiest thing would be to make the intrinsics have having unmodeled side effects, although we might want to do something more intelligent.

=?ISO-8859-1?Q?Rafael_=C1vila_de_Esp=EDndola?= wrote: > > I can reproduce this consistently, but only under an extremely large pile of > > code :-) I haven't tried to strip it down to minimal a test case yet, but I > > will. One salient difference with the code you've provided is that in my > > code, @a is a struct type. However, changing the example to use a

Hello everyone ! I followed http://stackoverflow.com/questions/4976298/modern-equivalent-of-llvm-annotationmanagerin order to get annotations from my target bytecode. All the examples that I give in the following is related to the code from that link. I have `__attribute__((annotate("DS"))) int f=0;` into the target C++ program and the related IR code: @.str = private unnamed_addr

> Method 2 > > use LLVM InlineAsm to load that gvar > so It may look like > %x = call InlineAsm(... ); > > but I am not really sure how to write LLVM InlineAsm. Are you looking for this? http://llvm.org/docs/LangRef.html#inline-assembler-expressions HTH, chenwj -- Wei-Ren Chen (陳韋任) Computer Systems Lab, Institute of Information Science, Academia Sinica, Taiwan