similar to: [LLVMdev] How are OutputOperandList and InputOperandList used?

Hi all, I started to write an LLVM backend for custom CPU. I created XXXInstrInfo but there are some problems. I searched for it but I couldn't find anything. Can anyone help me? include "XXXInstrFormats.td" def simm16 : Operand<i32> { let DecoderMethod = "DecodeSimm16"; } def mem : Operand<i32> { let PrintMethod = "printMemOperand"; let

Could generated assembly with option "-x86-asm-symtab=intel" be assembled by nasm directly? On Thu, Jun 21, 2012 at 2:30 PM, Sean Silva <silvas at purdue.edu> wrote: > If by "NASM format" you mean Intel syntax, then yes. In my experience most > LLVM tools refer to it with the option "-x86-asm-syntax=intel". For > example, tools/llvm-objdump has this

I just found that some samples used link.exe from Visual Studio to generate the final image, does LLVM has a replacement for link.exe to generate the final binary? -- Thanks Thomson -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20111015/1b33f012/attachment.html>

The blow snippet in target.td shows the regList in RegisterClass is typed as dag. Why not make it a simple list, such as list<Register>? class RegisterClass<string namespace, list<ValueType> regTypes, int alignment, dag regList> Thanks, -Thomson -------------- next part -------------- An HTML attachment was scrubbed... URL:

I saw there is reference to dead def on registers in LLVM source code. I am not aware of this concept from the traditional course material. What are the properties of dead def? Cheers Thomson -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150101/ea39d1e4/attachment.html>

I saw some LLVM generated assembly in NASM format, but did find this support in the official release. Is this supported? -- Thanks Thomson Tan -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120621/c571b488/attachment.html>

So I thought I'd try to use the documentation on llvm backends to try to create a DLX backend. I think I've got most of the stuff for the .td files done but I've got some problems. * Do I need to represent the PC in my XXXRegisterInfo.td file; the branch instruction effects it but you can directly access it ... I'm thinking not. * In my Instruction subclasses (in

To generate code for a target which doesn't have immediate constant as instruction operand, do I (the target specific back-end, XXXTarget) need to provide code to break up the SDNode with constant (like ISD::ADD $reg1, #1) to 2 SDNodes (ISD::LOAD $reg2, #1; ISD::ADD $reg1, $reg2) in my XXXTargetLowering::LowerOperation, or LLVM target-independent framework can do such transformation

I found that LLVM optimized the IR by replacing printf with puts. I wondered which pass did this optimization? And is it common that puts is faster (and some other metric) than printf? -- Thanks Thomson -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120617/08aa6c45/attachment.html>

On Sat, Oct 15, 2011 at 7:38 AM, Thomson <lilotom at gmail.com> wrote: > I just found that some samples used link.exe from Visual Studio to generate > the final image, does LLVM has a replacement for link.exe to generate the > final binary? > > -- > Thanks > Thomson Not currently, but a linker is being worked on. Is there any reason why link.exe is undesirable? You can

If by "NASM format" you mean Intel syntax, then yes. In my experience most LLVM tools refer to it with the option "-x86-asm-syntax=intel". For example, tools/llvm-objdump has this flag. --Sean Silva On Wed, Jun 20, 2012 at 10:12 PM, Thomson <lilotom at gmail.com> wrote: > I saw some LLVM generated assembly in NASM format, but did find this > support in the

I tried to build an LLVM pass in visual studio with cmake, but it seems add_LLVM_loadable_module macro doesn't support non-unix and non-cygwin platform. Is loadable module necessary for LLVM pass? Could I just build a static library and pass it to opt? If static library is not supported, is there any way to build an LLVM pass through visual studio? Or through clang with cmake? -- Thanks

On 19 January 2014 10:19, Thomson <lilotom at gmail.com> wrote: > The blow snippet in target.td shows the regList in RegisterClass is typed as > dag. Why not make it a simple list, such as list<Register>? I don't know about the original reason, but these days we have a few operators in use to make constructing those sets easier which would be much harder to do for lists.

On Sat, Oct 15, 2011 at 12:26 PM, Michael Spencer <bigcheesegs at gmail.com> wrote: > On Sat, Oct 15, 2011 at 7:38 AM, Thomson <lilotom at gmail.com> wrote: >> I just found that some samples used link.exe from Visual Studio to generate >> the final image, does LLVM has a replacement for link.exe to generate the >> final binary? I'm not related to the topic

Ok. So it seems CI->eraseFromParent() removed the old instruction and the new one is inserted right after this one in the inner function in the case of printf->puts. There is another line CI->repalceAllUsesWith(Result). I think this line could also do the replacement besides inserting the new one in the inner function. What's the difference of these 2 replacement methods? Also

Could you give an example of the old instruction cannot be removed safely? Since the new instruction would produce the same context/result of the old one, I suppose it is safe to remove the old one when the new instruction is ready (inserted after the old one). Anything I missed here? Thanks, -Thomson On Mon, Jun 18, 2012 at 7:31 PM, Christoph Erhardt <christoph at sicherha.de>wrote: >

This makes sense now. So if the instruction doesn't make any assignment, it would be unnecessary to replace any operand reference and the same instruction(CI) would be returned from the inner function to avoid this replacement, is this right? Thanks, -Thomson On Mon, Jun 18, 2012 at 8:09 PM, Christoph Erhardt <christoph at sicherha.de>wrote: > Hi Thomson, > > > Could you

I experimented to collect edge profile in llvm, and the report output by llvm-prof included call in the middle of the block. So it seems llvm-prof doesn't treat call as a basic block terminator, is this expected? Thanks, -Thomson -------------- next part -------------- An HTML attachment was scrubbed... URL:

I found that ProfileInfo is contained in an analysis group, but ProfileInfo class itself is not derived from any base classes of passes, so it appears not to be a pass. Anything I missed here? Thanks, -Thomson -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120624/7f26ca93/attachment.html>

Thanks, John. This sounds reasonable that the analysis group just provides an interface to its implementors (pass). Then I have another question, is it allowed to specify multiple pass in the same analysis group in the command line for opt? If yes, which pass will be used for the call of getAnaysis<AnalysisGroup>()? -Thomson On Mon, Jun 25, 2012 at 10:34 PM, John Criswell <criswell at