similar to: [LLVMdev] Native MCAsmStreamer?

Hi Craig and Rail, At Movidius, we have had to make a few changes to ‘MCAsmStreamer’ to support our assembler which is not ‘gas’ compliant. Earlier versions of LLVM (3.1 and 3.2) did have a separate header for ‘MCAsmStreamer’, and we had previously sub-classed this. The following are modifications that we have had to make because although ‘MCAsmStreamer’ does most of what we need,

This is just a cosmetic change in MCAsmStreamer.cpp/EmitSymbolAttribute: all attributes have now a \t before and after, as done for '.type'. This makes the output look consistent, as well as help some third party assemblers expecting the attributes to be in the second column. The patch applies cleanly on the svn head. Best regards, -- Arnaud de Grandmaison -------------- next part

I'm writing a backend using a mid-October svn snapshot of LLVM. I'm having a hard time figuring out the relationship between my XXXAsmPrinter and MCAsmStreamer. Can someone explain what each is responsible for? Looking at the existing targets, the XXXAsmPrinter implementations seem to implement both the legacy "create a .s file" behavior, using printInstruction/printOperand/etc

Does anybody know if there is a particular reason why MCAsmStreamer doesn't have its own .h file? https://github.com/llvm-mirror/llvm/blob/0e66a5f53c74056f95d178c86531d7d9cfb23da9/lib/MC/MCAsmStreamer.cpp#L41 It seems like it is a good idea to have this class declared as its own module ( its own .cpp and .h files). That would make it easier to inherit from it if there is a need (like in my

On Thu, Jan 21, 2016 at 4:04 PM, Craig Topper <craig.topper at gmail.com> wrote: > Isn't it also marked 'final' so it can't be inherited from anyway? What's > your need to inherit from it? > -- > ~Craig > Oops, missed the final part. I need to change the alignment. For my target it is sort of independent of the data layout. I was going to overwrite void

On Thu, Jan 21, 2016 at 5:03 PM, Craig Topper <craig.topper at gmail.com> wrote: > I dont' know why its final. That routine is just a method to force the > current write pointer to a specific alignment. Shouldn't you be changing > the callers to give you the alignment you want or don't want. > > > > -- > ~Craig > It looks like the callers get their

I posted to llvmdev a few months ago to ask for advice on the best way to avoid the SelectionDAGBuilder from imposing a constraint whereby a volatile load would be serialised relative to all pending loads. The LLVM LRM says that a volatile load only needs to be serialised relative to other volatile loads and, while it may not matter to most targets, the current behaviour of the SelectionDAGBuilder

Really sorry I missed this. Just found it looking for something else in my inbox. I think we should support this but * We should still err on other identifiers ".cfi_startport bar" is invalid. * If I read the gas documentation correctly, the effect of "simple" is to skip the initial cfi instructions. We should test if that is the case and implement it too. Accepting and

Hello everyone, In the file *lib/CodeGen/AsmPrinter/AsmPrinter.cpp*, I would like to obtain an MCInst corresponding to its MachineInstr. Can anyone tell me a way to do that? If that is not possible, then, I would like to know if a given MachineInstr is an *lea *instruction and I would like to know if the symbol involved with this lea instruction is a jump-table. For instance, given a

Hi, I'm trying to implement a LLVM backend but have actually a problem and don't know how to handle. Actually I'm at the very beginning. I tried to (carefully) copy from existing backends (Sparc and Hexogon), but I messed up with something. My "LLVMInitializeMyTargetAsmPrinter" function is not called. All the other "LLVMInitializeMyTargetSomeThing" functions to

Hi all, I think that I came somewhat closer to a solution for splitting a MachineBasicBlock for a PSEUDO_CALL_R instruction and having a label to the new MBB: For following piece of code: --- typedef int callme_t(int a, int b); callme_t* c01; int foo(int a, int b) { return c01(a,b); // MachineBasicBlock will be split at call instruction } --- I have initially following correspondence: BB1

Hi, What is the best way to debug an llc crash using bugpoint? I am getting the following crash that I would like to reduce llc: /home/vadve/aggarwa4/llvm29/llvm-2.9/lib/MC/MCAsmStreamer.cpp:273: virtual void<unnamed>::MCAsmStreamer::EmitLabel(llvm::MCSymbol*): Assertion `Symbol->isUndefined() && "Cannot define a symbol twice!"' failed. 0 llc

Do something like this: bugpoint -run-llc <bitcode file> --tool-args -- <llc arguments> Cameron On May 23, 2011, at 1:33 PM, Arushi Aggarwal wrote: > Hi, > > What is the best way to debug an llc crash using bugpoint? > > I am getting the following crash that I would like to reduce > > llc: /home/vadve/aggarwa4/llvm29/llvm-2.9/lib/MC/MCAsmStreamer.cpp:273:

Hi, I'm trying to implement a 10-bit relocation that does not start at the beginning of a byte-boundary and I'm not entirely sure I understand the use by some targets of MCFixup.Offset and MCFixupKindInfo.TargetOffset. LLVM's documentation states that: MCFixup.Offset -"/// The byte index of start of the relocation inside the encoded instruction."

I have a question regarding implementation of subclasses of MCInstPrinter. I am implementing Machine IR layer to MC layer lowering for Mips. What's the best way to print the value of "Register" in the following code in MCAsmStreamer::EmitRegisterName? Do I have to convert the LLVM register number back to its corresponding dwarf register number in function

Hi, After your change, EmitCFIStartProcImpl in MCAsmStreamer does not match the signature of the EmitCFIStartProcImpl in MCStreamer and you end up not overriding the original function. One of the places where "virtual void EmitCFIStartProcImpl(MCDwarfFrameInfo &Frame) = 0" would have helped. :) Specifically, adding virtual void EmitCFIStartProcImpl(MCDwarfFrameInfo &Frame)

The current implementation of llvm::Mangler does not check if a symbol name exists before mangling. This can cause problems if the mangled name aliases another symbol in the IR. For example, consider the following: @"a(b" = addrspace(1) global float 1.0 @"a_28_b" = addrspace(1) global float 1.0 If I run this through the x86 backend, I get: llc:

Hi, I noticed that the following regression tests fail when building a config generated with CMake. When building with a config generated with configure, all is well. I get the same results when using gcc-4.2.1 and clang-trunk. It seems to me that there's a glitch in de CMake config somewhere. I doing a verbose build right now to investigate, but maybe someone already knows what's

Hi, I am wondering if someone can guide me in adding metadata to IR which already contains some metadata. I am trying to add dbg.declare inst for a local variable I added to a function. I used the DIBuilder to build a DIVariable. When I try to compile llc fails with following message. llc: MCAsmStreamer.cpp:273: virtual void<unnamed>::MCAsmStreamer::EmitLabel(llvm::MCSymbol*): Assertion

Hello everyone! Could somebody, advice me what's wrong with my AsmPrinter? I've created ColdFireAsmPrinter.cpp file, and added initialization: extern "C" void LLVMInitializeColdFireAsmPrinter() { RegisterAsmPrinter<ColdFireAsmPrinter> X(TheColdFireTarget); } But, unfortunately, after build AsmPrinters.def doesn't contain LLVM_ASM_PRINTER(ColdFire). Maybe I've