1====================================================== 2Kaleidoscope: Conclusion and other useful LLVM tidbits 3====================================================== 4 5.. contents:: 6 :local: 7 8Tutorial Conclusion 9=================== 10 11Welcome to the final chapter of the "`Implementing a language with 12LLVM <index.html>`_" tutorial. In the course of this tutorial, we have 13grown our little Kaleidoscope language from being a useless toy, to 14being a semi-interesting (but probably still useless) toy. :) 15 16It is interesting to see how far we've come, and how little code it has 17taken. We built the entire lexer, parser, AST, code generator, an 18interactive run-loop (with a JIT!), and emitted debug information in 19standalone executables - all in under 1000 lines of (non-comment/non-blank) 20code. 21 22Our little language supports a couple of interesting features: it 23supports user defined binary and unary operators, it uses JIT 24compilation for immediate evaluation, and it supports a few control flow 25constructs with SSA construction. 26 27Part of the idea of this tutorial was to show you how easy and fun it 28can be to define, build, and play with languages. Building a compiler 29need not be a scary or mystical process! Now that you've seen some of 30the basics, I strongly encourage you to take the code and hack on it. 31For example, try adding: 32 33- **global variables** - While global variables have questional value 34 in modern software engineering, they are often useful when putting 35 together quick little hacks like the Kaleidoscope compiler itself. 36 Fortunately, our current setup makes it very easy to add global 37 variables: just have value lookup check to see if an unresolved 38 variable is in the global variable symbol table before rejecting it. 39 To create a new global variable, make an instance of the LLVM 40 ``GlobalVariable`` class. 41- **typed variables** - Kaleidoscope currently only supports variables 42 of type double. This gives the language a very nice elegance, because 43 only supporting one type means that you never have to specify types. 44 Different languages have different ways of handling this. The easiest 45 way is to require the user to specify types for every variable 46 definition, and record the type of the variable in the symbol table 47 along with its Value\*. 48- **arrays, structs, vectors, etc** - Once you add types, you can start 49 extending the type system in all sorts of interesting ways. Simple 50 arrays are very easy and are quite useful for many different 51 applications. Adding them is mostly an exercise in learning how the 52 LLVM `getelementptr <../LangRef.html#getelementptr-instruction>`_ instruction 53 works: it is so nifty/unconventional, it `has its own 54 FAQ <../GetElementPtr.html>`_! 55- **standard runtime** - Our current language allows the user to access 56 arbitrary external functions, and we use it for things like "printd" 57 and "putchard". As you extend the language to add higher-level 58 constructs, often these constructs make the most sense if they are 59 lowered to calls into a language-supplied runtime. For example, if 60 you add hash tables to the language, it would probably make sense to 61 add the routines to a runtime, instead of inlining them all the way. 62- **memory management** - Currently we can only access the stack in 63 Kaleidoscope. It would also be useful to be able to allocate heap 64 memory, either with calls to the standard libc malloc/free interface 65 or with a garbage collector. If you would like to use garbage 66 collection, note that LLVM fully supports `Accurate Garbage 67 Collection <../GarbageCollection.html>`_ including algorithms that 68 move objects and need to scan/update the stack. 69- **exception handling support** - LLVM supports generation of `zero 70 cost exceptions <../ExceptionHandling.html>`_ which interoperate with 71 code compiled in other languages. You could also generate code by 72 implicitly making every function return an error value and checking 73 it. You could also make explicit use of setjmp/longjmp. There are 74 many different ways to go here. 75- **object orientation, generics, database access, complex numbers, 76 geometric programming, ...** - Really, there is no end of crazy 77 features that you can add to the language. 78- **unusual domains** - We've been talking about applying LLVM to a 79 domain that many people are interested in: building a compiler for a 80 specific language. However, there are many other domains that can use 81 compiler technology that are not typically considered. For example, 82 LLVM has been used to implement OpenGL graphics acceleration, 83 translate C++ code to ActionScript, and many other cute and clever 84 things. Maybe you will be the first to JIT compile a regular 85 expression interpreter into native code with LLVM? 86 87Have fun - try doing something crazy and unusual. Building a language 88like everyone else always has, is much less fun than trying something a 89little crazy or off the wall and seeing how it turns out. If you get 90stuck or want to talk about it, feel free to email the `llvm-dev mailing 91list <http://lists.llvm.org/mailman/listinfo/llvm-dev>`_: it has lots 92of people who are interested in languages and are often willing to help 93out. 94 95Before we end this tutorial, I want to talk about some "tips and tricks" 96for generating LLVM IR. These are some of the more subtle things that 97may not be obvious, but are very useful if you want to take advantage of 98LLVM's capabilities. 99 100Properties of the LLVM IR 101========================= 102 103We have a couple of common questions about code in the LLVM IR form - 104let's just get these out of the way right now, shall we? 105 106Target Independence 107------------------- 108 109Kaleidoscope is an example of a "portable language": any program written 110in Kaleidoscope will work the same way on any target that it runs on. 111Many other languages have this property, e.g. lisp, java, haskell, 112javascript, python, etc (note that while these languages are portable, 113not all their libraries are). 114 115One nice aspect of LLVM is that it is often capable of preserving target 116independence in the IR: you can take the LLVM IR for a 117Kaleidoscope-compiled program and run it on any target that LLVM 118supports, even emitting C code and compiling that on targets that LLVM 119doesn't support natively. You can trivially tell that the Kaleidoscope 120compiler generates target-independent code because it never queries for 121any target-specific information when generating code. 122 123The fact that LLVM provides a compact, target-independent, 124representation for code gets a lot of people excited. Unfortunately, 125these people are usually thinking about C or a language from the C 126family when they are asking questions about language portability. I say 127"unfortunately", because there is really no way to make (fully general) 128C code portable, other than shipping the source code around (and of 129course, C source code is not actually portable in general either - ever 130port a really old application from 32- to 64-bits?). 131 132The problem with C (again, in its full generality) is that it is heavily 133laden with target specific assumptions. As one simple example, the 134preprocessor often destructively removes target-independence from the 135code when it processes the input text: 136 137.. code-block:: c 138 139 #ifdef __i386__ 140 int X = 1; 141 #else 142 int X = 42; 143 #endif 144 145While it is possible to engineer more and more complex solutions to 146problems like this, it cannot be solved in full generality in a way that 147is better than shipping the actual source code. 148 149That said, there are interesting subsets of C that can be made portable. 150If you are willing to fix primitive types to a fixed size (say int = 15132-bits, and long = 64-bits), don't care about ABI compatibility with 152existing binaries, and are willing to give up some other minor features, 153you can have portable code. This can make sense for specialized domains 154such as an in-kernel language. 155 156Safety Guarantees 157----------------- 158 159Many of the languages above are also "safe" languages: it is impossible 160for a program written in Java to corrupt its address space and crash the 161process (assuming the JVM has no bugs). Safety is an interesting 162property that requires a combination of language design, runtime 163support, and often operating system support. 164 165It is certainly possible to implement a safe language in LLVM, but LLVM 166IR does not itself guarantee safety. The LLVM IR allows unsafe pointer 167casts, use after free bugs, buffer over-runs, and a variety of other 168problems. Safety needs to be implemented as a layer on top of LLVM and, 169conveniently, several groups have investigated this. Ask on the `llvm-dev 170mailing list <http://lists.llvm.org/mailman/listinfo/llvm-dev>`_ if 171you are interested in more details. 172 173Language-Specific Optimizations 174------------------------------- 175 176One thing about LLVM that turns off many people is that it does not 177solve all the world's problems in one system (sorry 'world hunger', 178someone else will have to solve you some other day). One specific 179complaint is that people perceive LLVM as being incapable of performing 180high-level language-specific optimization: LLVM "loses too much 181information". 182 183Unfortunately, this is really not the place to give you a full and 184unified version of "Chris Lattner's theory of compiler design". Instead, 185I'll make a few observations: 186 187First, you're right that LLVM does lose information. For example, as of 188this writing, there is no way to distinguish in the LLVM IR whether an 189SSA-value came from a C "int" or a C "long" on an ILP32 machine (other 190than debug info). Both get compiled down to an 'i32' value and the 191information about what it came from is lost. The more general issue 192here, is that the LLVM type system uses "structural equivalence" instead 193of "name equivalence". Another place this surprises people is if you 194have two types in a high-level language that have the same structure 195(e.g. two different structs that have a single int field): these types 196will compile down into a single LLVM type and it will be impossible to 197tell what it came from. 198 199Second, while LLVM does lose information, LLVM is not a fixed target: we 200continue to enhance and improve it in many different ways. In addition 201to adding new features (LLVM did not always support exceptions or debug 202info), we also extend the IR to capture important information for 203optimization (e.g. whether an argument is sign or zero extended, 204information about pointers aliasing, etc). Many of the enhancements are 205user-driven: people want LLVM to include some specific feature, so they 206go ahead and extend it. 207 208Third, it is *possible and easy* to add language-specific optimizations, 209and you have a number of choices in how to do it. As one trivial 210example, it is easy to add language-specific optimization passes that 211"know" things about code compiled for a language. In the case of the C 212family, there is an optimization pass that "knows" about the standard C 213library functions. If you call "exit(0)" in main(), it knows that it is 214safe to optimize that into "return 0;" because C specifies what the 215'exit' function does. 216 217In addition to simple library knowledge, it is possible to embed a 218variety of other language-specific information into the LLVM IR. If you 219have a specific need and run into a wall, please bring the topic up on 220the llvm-dev list. At the very worst, you can always treat LLVM as if it 221were a "dumb code generator" and implement the high-level optimizations 222you desire in your front-end, on the language-specific AST. 223 224Tips and Tricks 225=============== 226 227There is a variety of useful tips and tricks that you come to know after 228working on/with LLVM that aren't obvious at first glance. Instead of 229letting everyone rediscover them, this section talks about some of these 230issues. 231 232Implementing portable offsetof/sizeof 233------------------------------------- 234 235One interesting thing that comes up, if you are trying to keep the code 236generated by your compiler "target independent", is that you often need 237to know the size of some LLVM type or the offset of some field in an 238llvm structure. For example, you might need to pass the size of a type 239into a function that allocates memory. 240 241Unfortunately, this can vary widely across targets: for example the 242width of a pointer is trivially target-specific. However, there is a 243`clever way to use the getelementptr 244instruction <http://nondot.org/sabre/LLVMNotes/SizeOf-OffsetOf-VariableSizedStructs.txt>`_ 245that allows you to compute this in a portable way. 246 247Garbage Collected Stack Frames 248------------------------------ 249 250Some languages want to explicitly manage their stack frames, often so 251that they are garbage collected or to allow easy implementation of 252closures. There are often better ways to implement these features than 253explicit stack frames, but `LLVM does support 254them, <http://nondot.org/sabre/LLVMNotes/ExplicitlyManagedStackFrames.txt>`_ 255if you want. It requires your front-end to convert the code into 256`Continuation Passing 257Style <http://en.wikipedia.org/wiki/Continuation-passing_style>`_ and 258the use of tail calls (which LLVM also supports). 259 260