Editing LLVM (section)

===Intermediate representation===
[[File:Mesa layers of crap 2016.svg|thumb|LLVM IR is used e.g., by radeonsi and by llvmpipe. Both are part of [[Mesa 3D]].]]

The core of LLVM is the [[intermediate representation]] (IR), a [[low-level programming language]] similar to assembly. IR is a strongly typed [[reduced instruction set computer]] (RISC) instruction set which abstracts away most details of the target. For example, the [[calling convention]] is abstracted through <code>call</code> and <code>ret</code> instructions with explicit arguments. Also, instead of a fixed set of registers, IR uses an infinite set of temporaries of the form %0, %1, etc. LLVM supports three equivalent forms of IR: a human-readable assembly format,<ref name=IRref>{{cite web |url=https://llvm.org/docs/LangRef.html |title=LLVM Language Reference Manual |author=<!-- Unstated staff writers --> |date=10 January 2023 |website=LLVM.org}}</ref> an in-memory format suitable for frontends, and a dense bitcode format for serializing. A simple [["Hello, world!" program]] in the human-readable IR format:

<syntaxhighlight lang="llvm">
@.str = internal constant [14 x i8] c"Hello, world\0A\00"

declare i32 @printf(ptr, ...)

define i32 @main(i32 %argc, ptr %argv) nounwind {
entry:
    %tmp1 = getelementptr [14 x i8], ptr @.str, i32 0, i32 0
    %tmp2 = call i32 (ptr, ...) @printf( ptr %tmp1 ) nounwind
    ret i32 0
}
</syntaxhighlight>

The many different conventions used and features provided by different targets mean that LLVM cannot truly produce a target-independent IR and retarget it without breaking some established rules. Examples of target dependence beyond what is explicitly mentioned in the documentation can be found in a 2011 proposal for "wordcode", a fully target-independent variant of LLVM IR intended for online distribution.<ref>{{cite web |last1=Kang |first1=Jin-Gu |title=Wordcode: more target independent LLVM bitcode |url=https://llvm.org/devmtg/2011-09-16/EuroLLVM2011-MoreTargetIndependentLLVMBitcode.pdf |access-date=1 December 2019}}</ref> A more practical example is [[PNaCl]].<ref>{{cite web|url=http://nativeclient.googlecode.com/svn/data/site/pnacl.pdf |title=PNaCl: Portable Native Client Executables |access-date=25 April 2012 |url-status=dead |archive-url=https://web.archive.org/web/20120502135033/http://nativeclient.googlecode.com/svn/data/site/pnacl.pdf |archive-date=2 May 2012 |df=dmy-all}}</ref>

The LLVM project also introduces another type of intermediate representation named [[MLIR (software)|MLIR]]<ref>{{Cite web |title=MLIR |url=https://mlir.llvm.org/ |access-date=2022-06-07 |website=mlir.llvm.org}}</ref> which helps build reusable and extensible compiler infrastructure by employing a plugin architecture named Dialect.<ref>{{Cite web |title=Dialects - MLIR |url=https://mlir.llvm.org/docs/Dialects/ |access-date=2022-06-07 |website=mlir.llvm.org}}</ref> It enables the use of higher-level information on the program structure in the process of optimization including [[Polytope model|polyhedral compilation]].