From 208108a8e0d2295493cfb3554724e379fe1ff7ae Mon Sep 17 00:00:00 2001 From: Andrew Kaylor Date: Mon, 22 Jul 2013 18:47:24 +0000 Subject: [PATCH] Adding example source to support MCJIT/Kaleidoscope blog posts. llvm-svn: 186854 --- examples/Kaleidoscope/MCJIT/README.txt | 12 + examples/Kaleidoscope/MCJIT/cached/Makefile | 11 + examples/Kaleidoscope/MCJIT/cached/README.txt | 28 + .../Kaleidoscope/MCJIT/cached/genk-timing.py | 219 +++ .../Kaleidoscope/MCJIT/cached/split-lib.py | 70 + .../Kaleidoscope/MCJIT/cached/toy-jit.cpp | 1207 ++++++++++++ examples/Kaleidoscope/MCJIT/cached/toy.cpp | 1551 +++++++++++++++ examples/Kaleidoscope/MCJIT/complete/Makefile | 4 + .../Kaleidoscope/MCJIT/complete/README.txt | 25 + .../MCJIT/complete/genk-timing.py | 224 +++ .../Kaleidoscope/MCJIT/complete/split-lib.py | 70 + examples/Kaleidoscope/MCJIT/complete/toy.cpp | 1710 +++++++++++++++++ examples/Kaleidoscope/MCJIT/initial/Makefile | 4 + .../Kaleidoscope/MCJIT/initial/README.txt | 18 + examples/Kaleidoscope/MCJIT/initial/toy.cpp | 1381 +++++++++++++ examples/Kaleidoscope/MCJIT/lazy/Makefile | 7 + examples/Kaleidoscope/MCJIT/lazy/README.txt | 25 + .../Kaleidoscope/MCJIT/lazy/genk-timing.py | 219 +++ examples/Kaleidoscope/MCJIT/lazy/toy-jit.cpp | 1167 +++++++++++ examples/Kaleidoscope/MCJIT/lazy/toy.cpp | 1422 ++++++++++++++ 20 files changed, 9374 insertions(+) create mode 100644 examples/Kaleidoscope/MCJIT/README.txt create mode 100644 examples/Kaleidoscope/MCJIT/cached/Makefile create mode 100644 examples/Kaleidoscope/MCJIT/cached/README.txt create mode 100644 examples/Kaleidoscope/MCJIT/cached/genk-timing.py create mode 100644 examples/Kaleidoscope/MCJIT/cached/split-lib.py create mode 100644 examples/Kaleidoscope/MCJIT/cached/toy-jit.cpp create mode 100644 examples/Kaleidoscope/MCJIT/cached/toy.cpp create mode 100644 examples/Kaleidoscope/MCJIT/complete/Makefile create mode 100644 examples/Kaleidoscope/MCJIT/complete/README.txt create mode 100644 examples/Kaleidoscope/MCJIT/complete/genk-timing.py create mode 100644 examples/Kaleidoscope/MCJIT/complete/split-lib.py create mode 100644 examples/Kaleidoscope/MCJIT/complete/toy.cpp create mode 100644 examples/Kaleidoscope/MCJIT/initial/Makefile create mode 100644 examples/Kaleidoscope/MCJIT/initial/README.txt create mode 100644 examples/Kaleidoscope/MCJIT/initial/toy.cpp create mode 100644 examples/Kaleidoscope/MCJIT/lazy/Makefile create mode 100644 examples/Kaleidoscope/MCJIT/lazy/README.txt create mode 100644 examples/Kaleidoscope/MCJIT/lazy/genk-timing.py create mode 100644 examples/Kaleidoscope/MCJIT/lazy/toy-jit.cpp create mode 100644 examples/Kaleidoscope/MCJIT/lazy/toy.cpp diff --git a/examples/Kaleidoscope/MCJIT/README.txt b/examples/Kaleidoscope/MCJIT/README.txt new file mode 100644 index 00000000000..ba742645acc --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/README.txt @@ -0,0 +1,12 @@ +//===----------------------------------------------------------------------===/ +// Kaleidoscope with MCJIT +//===----------------------------------------------------------------------===// + +The files in this directory are meant to accompany a series of blog posts +that describe the process of porting the Kaleidoscope tutorial to use the MCJIT +execution engine instead of the older JIT engine. + +When the blog posts are ready this file will be updated with links to the posts. + +These directories contain Makefiles that allow the code to be built in a +standalone manner, independent of the larger LLVM build infrastructure. \ No newline at end of file diff --git a/examples/Kaleidoscope/MCJIT/cached/Makefile b/examples/Kaleidoscope/MCJIT/cached/Makefile new file mode 100644 index 00000000000..dde39a75d76 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/cached/Makefile @@ -0,0 +1,11 @@ +all: toy-mcjit toy-jit toy-ir-gen + +toy-mcjit : toy.cpp + clang++ toy.cpp -g -O3 -rdynamic -fno-rtti `llvm-config --cppflags --ldflags --libs core mcjit native irreader` -o toy-mcjit + +toy-jit : toy-jit.cpp + clang++ toy-jit.cpp -g -O3 -rdynamic -fno-rtti `llvm-config --cppflags --ldflags --libs core jit native irreader` -o toy-jit + +# This is a special build for the purpose of converting Kaleidoscope input to an IR file +toy-ir-gen : toy-jit.cpp + clang++ toy-jit.cpp -g -O3 -rdynamic -fno-rtti -DDUMP_FINAL_MODULE `llvm-config --cppflags --ldflags --libs core jit native irreader` -o toy-ir-gen diff --git a/examples/Kaleidoscope/MCJIT/cached/README.txt b/examples/Kaleidoscope/MCJIT/cached/README.txt new file mode 100644 index 00000000000..6acaf057f28 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/cached/README.txt @@ -0,0 +1,28 @@ +//===----------------------------------------------------------------------===/ +// Kaleidoscope with MCJIT +//===----------------------------------------------------------------------===// + +The files in this directory are meant to accompany the first in a series of +three blog posts that describe the process of porting the Kaleidoscope tutorial +to use the MCJIT execution engine instead of the older JIT engine. + +When the blog post is ready this file will be updated with a link to the post. + +The source code in this directory demonstrates the third version of the +program, now modified to accept an input IR file on the command line and, +optionally, to use a basic caching mechanism to store generated object images. + +The toy-jit.cpp file contains a version of the original JIT-based source code +that has been modified to support the input IR file command line option. + +This directory contain a Makefile that allow the code to be built in a +standalone manner, independent of the larger LLVM build infrastructure. To build +the program you will need to have 'clang++' and 'llvm-config' in your path. If +you attempt to build using the LLVM 3.3 release, some minor modifications will +be required. + +This directory also contains a Python script that may be used to generate random +input for the program and test scripts to capture data for rough performance +comparisons. Another Python script will split generated input files into +definitions and function calls for the purpose of testing the IR input and +caching facilities. \ No newline at end of file diff --git a/examples/Kaleidoscope/MCJIT/cached/genk-timing.py b/examples/Kaleidoscope/MCJIT/cached/genk-timing.py new file mode 100644 index 00000000000..96dd6db5da3 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/cached/genk-timing.py @@ -0,0 +1,219 @@ +#!/usr/bin/env python + +import sys +import random + +class TimingScriptGenerator: + """Used to generate a bash script which will invoke the toy and time it""" + def __init__(self, scriptname, outputname): + self.timeFile = outputname + self.shfile = open(scriptname, 'w') + self.shfile.write("echo \"\" > %s\n" % self.timeFile) + + def writeTimingCall(self, filename, numFuncs, funcsCalled, totalCalls): + """Echo some comments and invoke both versions of toy""" + rootname = filename + if '.' in filename: + rootname = filename[:filename.rfind('.')] + self.shfile.write("echo \"%s: Calls %d of %d functions, %d total\" >> %s\n" % (filename, funcsCalled, numFuncs, totalCalls, self.timeFile)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-mcjit < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With JIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-jit < %s > %s-jit.out 2> %s-jit.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + +class KScriptGenerator: + """Used to generate random Kaleidoscope code""" + def __init__(self, filename): + self.kfile = open(filename, 'w') + self.nextFuncNum = 1 + self.lastFuncNum = None + self.callWeighting = 0.1 + # A mapping of calls within functions with no duplicates + self.calledFunctionTable = {} + # A list of function calls which will actually be executed + self.calledFunctions = [] + # A comprehensive mapping of calls within functions + # used for computing the total number of calls + self.comprehensiveCalledFunctionTable = {} + self.totalCallsExecuted = 0 + + def updateTotalCallCount(self, callee): + # Count this call + self.totalCallsExecuted += 1 + # Then count all the functions it calls + if callee in self.comprehensiveCalledFunctionTable: + for child in self.comprehensiveCalledFunctionTable[callee]: + self.updateTotalCallCount(child) + + def updateFunctionCallMap(self, caller, callee): + """Maintains a map of functions that are called from other functions""" + if not caller in self.calledFunctionTable: + self.calledFunctionTable[caller] = [] + if not callee in self.calledFunctionTable[caller]: + self.calledFunctionTable[caller].append(callee) + if not caller in self.comprehensiveCalledFunctionTable: + self.comprehensiveCalledFunctionTable[caller] = [] + self.comprehensiveCalledFunctionTable[caller].append(callee) + + def updateCalledFunctionList(self, callee): + """Maintains a list of functions that will actually be called""" + # Update the total call count + self.updateTotalCallCount(callee) + # If this function is already in the list, don't do anything else + if callee in self.calledFunctions: + return + # Add this function to the list of those that will be called. + self.calledFunctions.append(callee) + # If this function calls other functions, add them too + if callee in self.calledFunctionTable: + for subCallee in self.calledFunctionTable[callee]: + self.updateCalledFunctionList(subCallee) + + def setCallWeighting(self, weight): + """ Sets the probably of generating a function call""" + self.callWeighting = weight + + def writeln(self, line): + self.kfile.write(line + '\n') + + def writeComment(self, comment): + self.writeln('# ' + comment) + + def writeEmptyLine(self): + self.writeln("") + + def writePredefinedFunctions(self): + self.writeComment("Define ':' for sequencing: as a low-precedence operator that ignores operands") + self.writeComment("and just returns the RHS.") + self.writeln("def binary : 1 (x y) y;") + self.writeEmptyLine() + self.writeComment("Helper functions defined within toy") + self.writeln("extern putchard(x);") + self.writeln("extern printd(d);") + self.writeln("extern printlf();") + self.writeEmptyLine() + self.writeComment("Print the result of a function call") + self.writeln("def printresult(N Result)") + self.writeln(" # 'result('") + self.writeln(" putchard(114) : putchard(101) : putchard(115) : putchard(117) : putchard(108) : putchard(116) : putchard(40) :") + self.writeln(" printd(N) :"); + self.writeln(" # ') = '") + self.writeln(" putchard(41) : putchard(32) : putchard(61) : putchard(32) :") + self.writeln(" printd(Result) :"); + self.writeln(" printlf();") + self.writeEmptyLine() + + def writeRandomOperation(self, LValue, LHS, RHS): + shouldCallFunc = (self.lastFuncNum > 2 and random.random() < self.callWeighting) + if shouldCallFunc: + funcToCall = random.randrange(1, self.lastFuncNum - 1) + self.updateFunctionCallMap(self.lastFuncNum, funcToCall) + self.writeln(" %s = func%d(%s, %s) :" % (LValue, funcToCall, LHS, RHS)) + else: + possibleOperations = ["+", "-", "*", "/"] + operation = random.choice(possibleOperations) + if operation == "-": + # Don't let our intermediate value become zero + # This is complicated by the fact that '<' is our only comparison operator + self.writeln(" if %s < %s then" % (LHS, RHS)) + self.writeln(" %s = %s %s %s" % (LValue, LHS, operation, RHS)) + self.writeln(" else if %s < %s then" % (RHS, LHS)) + self.writeln(" %s = %s %s %s" % (LValue, LHS, operation, RHS)) + self.writeln(" else") + self.writeln(" %s = %s %s %f :" % (LValue, LHS, operation, random.uniform(1, 100))) + else: + self.writeln(" %s = %s %s %s :" % (LValue, LHS, operation, RHS)) + + def getNextFuncNum(self): + result = self.nextFuncNum + self.nextFuncNum += 1 + self.lastFuncNum = result + return result + + def writeFunction(self, elements): + funcNum = self.getNextFuncNum() + self.writeComment("Auto-generated function number %d" % funcNum) + self.writeln("def func%d(X Y)" % funcNum) + self.writeln(" var temp1 = X,") + self.writeln(" temp2 = Y,") + self.writeln(" temp3 in") + # Initialize the variable names to be rotated + first = "temp3" + second = "temp1" + third = "temp2" + # Write some random operations + for i in range(elements): + self.writeRandomOperation(first, second, third) + # Rotate the variables + temp = first + first = second + second = third + third = temp + self.writeln(" " + third + ";") + self.writeEmptyLine() + + def writeFunctionCall(self): + self.writeComment("Call the last function") + arg1 = random.uniform(1, 100) + arg2 = random.uniform(1, 100) + self.writeln("printresult(%d, func%d(%f, %f) )" % (self.lastFuncNum, self.lastFuncNum, arg1, arg2)) + self.writeEmptyLine() + self.updateCalledFunctionList(self.lastFuncNum) + + def writeFinalFunctionCounts(self): + self.writeComment("Called %d of %d functions" % (len(self.calledFunctions), self.lastFuncNum)) + +def generateKScript(filename, numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting, timingScript): + """ Generate a random Kaleidoscope script based on the given parameters """ + print "Generating " + filename + print(" %d functions, %d elements per function, %d functions between execution" % + (numFuncs, elementsPerFunc, funcsBetweenExec)) + print(" Call weighting = %f" % callWeighting) + script = KScriptGenerator(filename) + script.setCallWeighting(callWeighting) + script.writeComment("===========================================================================") + script.writeComment("Auto-generated script") + script.writeComment(" %d functions, %d elements per function, %d functions between execution" + % (numFuncs, elementsPerFunc, funcsBetweenExec)) + script.writeComment(" call weighting = %f" % callWeighting) + script.writeComment("===========================================================================") + script.writeEmptyLine() + script.writePredefinedFunctions() + funcsSinceLastExec = 0 + for i in range(numFuncs): + script.writeFunction(elementsPerFunc) + funcsSinceLastExec += 1 + if funcsSinceLastExec == funcsBetweenExec: + script.writeFunctionCall() + funcsSinceLastExec = 0 + # Always end with a function call + if funcsSinceLastExec > 0: + script.writeFunctionCall() + script.writeEmptyLine() + script.writeFinalFunctionCounts() + funcsCalled = len(script.calledFunctions) + print " Called %d of %d functions, %d total" % (funcsCalled, numFuncs, script.totalCallsExecuted) + timingScript.writeTimingCall(filename, numFuncs, funcsCalled, script.totalCallsExecuted) + +# Execution begins here +random.seed() + +timingScript = TimingScriptGenerator("time-toy.sh", "timing-data.txt") + +dataSets = [(5000, 3, 50, 0.50), (5000, 10, 100, 0.10), (5000, 10, 5, 0.10), (5000, 10, 1, 0.0), + (1000, 3, 10, 0.50), (1000, 10, 100, 0.10), (1000, 10, 5, 0.10), (1000, 10, 1, 0.0), + ( 200, 3, 2, 0.50), ( 200, 10, 40, 0.10), ( 200, 10, 2, 0.10), ( 200, 10, 1, 0.0)] + +# Generate the code +for (numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting) in dataSets: + filename = "test-%d-%d-%d-%d.k" % (numFuncs, elementsPerFunc, funcsBetweenExec, int(callWeighting * 100)) + generateKScript(filename, numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting, timingScript) +print "All done!" diff --git a/examples/Kaleidoscope/MCJIT/cached/split-lib.py b/examples/Kaleidoscope/MCJIT/cached/split-lib.py new file mode 100644 index 00000000000..5cdcc6d46c0 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/cached/split-lib.py @@ -0,0 +1,70 @@ +#!/usr/bin/env python + +class TimingScriptGenerator: + """Used to generate a bash script which will invoke the toy and time it""" + def __init__(self, scriptname, outputname): + self.shfile = open(scriptname, 'w') + self.timeFile = outputname + self.shfile.write("echo \"\" > %s\n" % self.timeFile) + + def writeTimingCall(self, irname, callname): + """Echo some comments and invoke both versions of toy""" + rootname = irname + if '.' in irname: + rootname = irname[:irname.rfind('.')] + self.shfile.write("echo \"%s: Calls %s\" >> %s\n" % (callname, irname, self.timeFile)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-mcjit -use-object-cache -input-IR=%s < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (irname, callname, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT again\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-mcjit -use-object-cache -input-IR=%s < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (irname, callname, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With JIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-jit -input-IR=%s < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (irname, callname, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + +class LibScriptGenerator: + """Used to generate a bash script which will convert Kaleidoscope files to IR""" + def __init__(self, filename): + self.shfile = open(filename, 'w') + + def writeLibGenCall(self, libname, irname): + self.shfile.write("./toy-ir-gen < %s 2> %s\n" % (libname, irname)) + +def splitScript(inputname, libGenScript, timingScript): + rootname = inputname[:-2] + libname = rootname + "-lib.k" + irname = rootname + "-lib.ir" + callname = rootname + "-call.k" + infile = open(inputname, "r") + libfile = open(libname, "w") + callfile = open(callname, "w") + print "Splitting %s into %s and %s" % (inputname, callname, libname) + for line in infile: + if not line.startswith("#"): + if line.startswith("print"): + callfile.write(line) + else: + libfile.write(line) + libGenScript.writeLibGenCall(libname, irname) + timingScript.writeTimingCall(irname, callname) + +# Execution begins here +libGenScript = LibScriptGenerator("make-libs.sh") +timingScript = TimingScriptGenerator("time-lib.sh", "lib-timing.txt") + +script_list = ["test-5000-3-50-50.k", "test-5000-10-100-10.k", "test-5000-10-5-10.k", "test-5000-10-1-0.k", + "test-1000-3-10-50.k", "test-1000-10-100-10.k", "test-1000-10-5-10.k", "test-1000-10-1-0.k", + "test-200-3-2-50.k", "test-200-10-40-10.k", "test-200-10-2-10.k", "test-200-10-1-0.k"] + +for script in script_list: + splitScript(script, libGenScript, timingScript) +print "All done!" diff --git a/examples/Kaleidoscope/MCJIT/cached/toy-jit.cpp b/examples/Kaleidoscope/MCJIT/cached/toy-jit.cpp new file mode 100644 index 00000000000..2539025aa2c --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/cached/toy-jit.cpp @@ -0,0 +1,1207 @@ +#define MINIMAL_STDERR_OUTPUT + +#include "llvm/Analysis/Passes.h" +#include "llvm/Analysis/Verifier.h" +#include "llvm/ExecutionEngine/ExecutionEngine.h" +#include "llvm/ExecutionEngine/JIT.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/IRReader/IRReader.h" +#include "llvm/PassManager.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/raw_ostream.h" +#include "llvm/Support/SourceMgr.h" +#include "llvm/Support/TargetSelect.h" +#include "llvm/Transforms/Scalar.h" +#include +#include +#include +#include + +using namespace llvm; + +//===----------------------------------------------------------------------===// +// Command-line options +//===----------------------------------------------------------------------===// + +namespace { + cl::opt + InputIR("input-IR", + cl::desc("Specify the name of an IR file to load for function definitions"), + cl::value_desc("input IR file name")); +} // namespace + +//===----------------------------------------------------------------------===// +// Lexer +//===----------------------------------------------------------------------===// + +// The lexer returns tokens [0-255] if it is an unknown character, otherwise one +// of these for known things. +enum Token { + tok_eof = -1, + + // commands + tok_def = -2, tok_extern = -3, + + // primary + tok_identifier = -4, tok_number = -5, + + // control + tok_if = -6, tok_then = -7, tok_else = -8, + tok_for = -9, tok_in = -10, + + // operators + tok_binary = -11, tok_unary = -12, + + // var definition + tok_var = -13 +}; + +static std::string IdentifierStr; // Filled in if tok_identifier +static double NumVal; // Filled in if tok_number + +/// gettok - Return the next token from standard input. +static int gettok() { + static int LastChar = ' '; + + // Skip any whitespace. + while (isspace(LastChar)) + LastChar = getchar(); + + if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* + IdentifierStr = LastChar; + while (isalnum((LastChar = getchar()))) + IdentifierStr += LastChar; + + if (IdentifierStr == "def") return tok_def; + if (IdentifierStr == "extern") return tok_extern; + if (IdentifierStr == "if") return tok_if; + if (IdentifierStr == "then") return tok_then; + if (IdentifierStr == "else") return tok_else; + if (IdentifierStr == "for") return tok_for; + if (IdentifierStr == "in") return tok_in; + if (IdentifierStr == "binary") return tok_binary; + if (IdentifierStr == "unary") return tok_unary; + if (IdentifierStr == "var") return tok_var; + return tok_identifier; + } + + if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ + std::string NumStr; + do { + NumStr += LastChar; + LastChar = getchar(); + } while (isdigit(LastChar) || LastChar == '.'); + + NumVal = strtod(NumStr.c_str(), 0); + return tok_number; + } + + if (LastChar == '#') { + // Comment until end of line. + do LastChar = getchar(); + while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); + + if (LastChar != EOF) + return gettok(); + } + + // Check for end of file. Don't eat the EOF. + if (LastChar == EOF) + return tok_eof; + + // Otherwise, just return the character as its ascii value. + int ThisChar = LastChar; + LastChar = getchar(); + return ThisChar; +} + +//===----------------------------------------------------------------------===// +// Abstract Syntax Tree (aka Parse Tree) +//===----------------------------------------------------------------------===// + +/// ExprAST - Base class for all expression nodes. +class ExprAST { +public: + virtual ~ExprAST() {} + virtual Value *Codegen() = 0; +}; + +/// NumberExprAST - Expression class for numeric literals like "1.0". +class NumberExprAST : public ExprAST { + double Val; +public: + NumberExprAST(double val) : Val(val) {} + virtual Value *Codegen(); +}; + +/// VariableExprAST - Expression class for referencing a variable, like "a". +class VariableExprAST : public ExprAST { + std::string Name; +public: + VariableExprAST(const std::string &name) : Name(name) {} + const std::string &getName() const { return Name; } + virtual Value *Codegen(); +}; + +/// UnaryExprAST - Expression class for a unary operator. +class UnaryExprAST : public ExprAST { + char Opcode; + ExprAST *Operand; +public: + UnaryExprAST(char opcode, ExprAST *operand) + : Opcode(opcode), Operand(operand) {} + virtual Value *Codegen(); +}; + +/// BinaryExprAST - Expression class for a binary operator. +class BinaryExprAST : public ExprAST { + char Op; + ExprAST *LHS, *RHS; +public: + BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) + : Op(op), LHS(lhs), RHS(rhs) {} + virtual Value *Codegen(); +}; + +/// CallExprAST - Expression class for function calls. +class CallExprAST : public ExprAST { + std::string Callee; + std::vector Args; +public: + CallExprAST(const std::string &callee, std::vector &args) + : Callee(callee), Args(args) {} + virtual Value *Codegen(); +}; + +/// IfExprAST - Expression class for if/then/else. +class IfExprAST : public ExprAST { + ExprAST *Cond, *Then, *Else; +public: + IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) + : Cond(cond), Then(then), Else(_else) {} + virtual Value *Codegen(); +}; + +/// ForExprAST - Expression class for for/in. +class ForExprAST : public ExprAST { + std::string VarName; + ExprAST *Start, *End, *Step, *Body; +public: + ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, + ExprAST *step, ExprAST *body) + : VarName(varname), Start(start), End(end), Step(step), Body(body) {} + virtual Value *Codegen(); +}; + +/// VarExprAST - Expression class for var/in +class VarExprAST : public ExprAST { + std::vector > VarNames; + ExprAST *Body; +public: + VarExprAST(const std::vector > &varnames, + ExprAST *body) + : VarNames(varnames), Body(body) {} + + virtual Value *Codegen(); +}; + +/// PrototypeAST - This class represents the "prototype" for a function, +/// which captures its argument names as well as if it is an operator. +class PrototypeAST { + std::string Name; + std::vector Args; + bool isOperator; + unsigned Precedence; // Precedence if a binary op. +public: + PrototypeAST(const std::string &name, const std::vector &args, + bool isoperator = false, unsigned prec = 0) + : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} + + bool isUnaryOp() const { return isOperator && Args.size() == 1; } + bool isBinaryOp() const { return isOperator && Args.size() == 2; } + + char getOperatorName() const { + assert(isUnaryOp() || isBinaryOp()); + return Name[Name.size()-1]; + } + + unsigned getBinaryPrecedence() const { return Precedence; } + + Function *Codegen(); + + void CreateArgumentAllocas(Function *F); +}; + +/// FunctionAST - This class represents a function definition itself. +class FunctionAST { + PrototypeAST *Proto; + ExprAST *Body; +public: + FunctionAST(PrototypeAST *proto, ExprAST *body) + : Proto(proto), Body(body) {} + + Function *Codegen(); +}; + +//===----------------------------------------------------------------------===// +// Parser +//===----------------------------------------------------------------------===// + +/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current +/// token the parser is looking at. getNextToken reads another token from the +/// lexer and updates CurTok with its results. +static int CurTok; +static int getNextToken() { + return CurTok = gettok(); +} + +/// BinopPrecedence - This holds the precedence for each binary operator that is +/// defined. +static std::map BinopPrecedence; + +/// GetTokPrecedence - Get the precedence of the pending binary operator token. +static int GetTokPrecedence() { + if (!isascii(CurTok)) + return -1; + + // Make sure it's a declared binop. + int TokPrec = BinopPrecedence[CurTok]; + if (TokPrec <= 0) return -1; + return TokPrec; +} + +/// Error* - These are little helper functions for error handling. +ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} +PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } +FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } + +static ExprAST *ParseExpression(); + +/// identifierexpr +/// ::= identifier +/// ::= identifier '(' expression* ')' +static ExprAST *ParseIdentifierExpr() { + std::string IdName = IdentifierStr; + + getNextToken(); // eat identifier. + + if (CurTok != '(') // Simple variable ref. + return new VariableExprAST(IdName); + + // Call. + getNextToken(); // eat ( + std::vector Args; + if (CurTok != ')') { + while (1) { + ExprAST *Arg = ParseExpression(); + if (!Arg) return 0; + Args.push_back(Arg); + + if (CurTok == ')') break; + + if (CurTok != ',') + return Error("Expected ')' or ',' in argument list"); + getNextToken(); + } + } + + // Eat the ')'. + getNextToken(); + + return new CallExprAST(IdName, Args); +} + +/// numberexpr ::= number +static ExprAST *ParseNumberExpr() { + ExprAST *Result = new NumberExprAST(NumVal); + getNextToken(); // consume the number + return Result; +} + +/// parenexpr ::= '(' expression ')' +static ExprAST *ParseParenExpr() { + getNextToken(); // eat (. + ExprAST *V = ParseExpression(); + if (!V) return 0; + + if (CurTok != ')') + return Error("expected ')'"); + getNextToken(); // eat ). + return V; +} + +/// ifexpr ::= 'if' expression 'then' expression 'else' expression +static ExprAST *ParseIfExpr() { + getNextToken(); // eat the if. + + // condition. + ExprAST *Cond = ParseExpression(); + if (!Cond) return 0; + + if (CurTok != tok_then) + return Error("expected then"); + getNextToken(); // eat the then + + ExprAST *Then = ParseExpression(); + if (Then == 0) return 0; + + if (CurTok != tok_else) + return Error("expected else"); + + getNextToken(); + + ExprAST *Else = ParseExpression(); + if (!Else) return 0; + + return new IfExprAST(Cond, Then, Else); +} + +/// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression +static ExprAST *ParseForExpr() { + getNextToken(); // eat the for. + + if (CurTok != tok_identifier) + return Error("expected identifier after for"); + + std::string IdName = IdentifierStr; + getNextToken(); // eat identifier. + + if (CurTok != '=') + return Error("expected '=' after for"); + getNextToken(); // eat '='. + + + ExprAST *Start = ParseExpression(); + if (Start == 0) return 0; + if (CurTok != ',') + return Error("expected ',' after for start value"); + getNextToken(); + + ExprAST *End = ParseExpression(); + if (End == 0) return 0; + + // The step value is optional. + ExprAST *Step = 0; + if (CurTok == ',') { + getNextToken(); + Step = ParseExpression(); + if (Step == 0) return 0; + } + + if (CurTok != tok_in) + return Error("expected 'in' after for"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new ForExprAST(IdName, Start, End, Step, Body); +} + +/// varexpr ::= 'var' identifier ('=' expression)? +// (',' identifier ('=' expression)?)* 'in' expression +static ExprAST *ParseVarExpr() { + getNextToken(); // eat the var. + + std::vector > VarNames; + + // At least one variable name is required. + if (CurTok != tok_identifier) + return Error("expected identifier after var"); + + while (1) { + std::string Name = IdentifierStr; + getNextToken(); // eat identifier. + + // Read the optional initializer. + ExprAST *Init = 0; + if (CurTok == '=') { + getNextToken(); // eat the '='. + + Init = ParseExpression(); + if (Init == 0) return 0; + } + + VarNames.push_back(std::make_pair(Name, Init)); + + // End of var list, exit loop. + if (CurTok != ',') break; + getNextToken(); // eat the ','. + + if (CurTok != tok_identifier) + return Error("expected identifier list after var"); + } + + // At this point, we have to have 'in'. + if (CurTok != tok_in) + return Error("expected 'in' keyword after 'var'"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new VarExprAST(VarNames, Body); +} + +/// primary +/// ::= identifierexpr +/// ::= numberexpr +/// ::= parenexpr +/// ::= ifexpr +/// ::= forexpr +/// ::= varexpr +static ExprAST *ParsePrimary() { + switch (CurTok) { + default: return Error("unknown token when expecting an expression"); + case tok_identifier: return ParseIdentifierExpr(); + case tok_number: return ParseNumberExpr(); + case '(': return ParseParenExpr(); + case tok_if: return ParseIfExpr(); + case tok_for: return ParseForExpr(); + case tok_var: return ParseVarExpr(); + } +} + +/// unary +/// ::= primary +/// ::= '!' unary +static ExprAST *ParseUnary() { + // If the current token is not an operator, it must be a primary expr. + if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') + return ParsePrimary(); + + // If this is a unary operator, read it. + int Opc = CurTok; + getNextToken(); + if (ExprAST *Operand = ParseUnary()) + return new UnaryExprAST(Opc, Operand); + return 0; +} + +/// binoprhs +/// ::= ('+' unary)* +static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { + // If this is a binop, find its precedence. + while (1) { + int TokPrec = GetTokPrecedence(); + + // If this is a binop that binds at least as tightly as the current binop, + // consume it, otherwise we are done. + if (TokPrec < ExprPrec) + return LHS; + + // Okay, we know this is a binop. + int BinOp = CurTok; + getNextToken(); // eat binop + + // Parse the unary expression after the binary operator. + ExprAST *RHS = ParseUnary(); + if (!RHS) return 0; + + // If BinOp binds less tightly with RHS than the operator after RHS, let + // the pending operator take RHS as its LHS. + int NextPrec = GetTokPrecedence(); + if (TokPrec < NextPrec) { + RHS = ParseBinOpRHS(TokPrec+1, RHS); + if (RHS == 0) return 0; + } + + // Merge LHS/RHS. + LHS = new BinaryExprAST(BinOp, LHS, RHS); + } +} + +/// expression +/// ::= unary binoprhs +/// +static ExprAST *ParseExpression() { + ExprAST *LHS = ParseUnary(); + if (!LHS) return 0; + + return ParseBinOpRHS(0, LHS); +} + +/// prototype +/// ::= id '(' id* ')' +/// ::= binary LETTER number? (id, id) +/// ::= unary LETTER (id) +static PrototypeAST *ParsePrototype() { + std::string FnName; + + unsigned Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. + unsigned BinaryPrecedence = 30; + + switch (CurTok) { + default: + return ErrorP("Expected function name in prototype"); + case tok_identifier: + FnName = IdentifierStr; + Kind = 0; + getNextToken(); + break; + case tok_unary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected unary operator"); + FnName = "unary"; + FnName += (char)CurTok; + Kind = 1; + getNextToken(); + break; + case tok_binary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected binary operator"); + FnName = "binary"; + FnName += (char)CurTok; + Kind = 2; + getNextToken(); + + // Read the precedence if present. + if (CurTok == tok_number) { + if (NumVal < 1 || NumVal > 100) + return ErrorP("Invalid precedecnce: must be 1..100"); + BinaryPrecedence = (unsigned)NumVal; + getNextToken(); + } + break; + } + + if (CurTok != '(') + return ErrorP("Expected '(' in prototype"); + + std::vector ArgNames; + while (getNextToken() == tok_identifier) + ArgNames.push_back(IdentifierStr); + if (CurTok != ')') + return ErrorP("Expected ')' in prototype"); + + // success. + getNextToken(); // eat ')'. + + // Verify right number of names for operator. + if (Kind && ArgNames.size() != Kind) + return ErrorP("Invalid number of operands for operator"); + + return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); +} + +/// definition ::= 'def' prototype expression +static FunctionAST *ParseDefinition() { + getNextToken(); // eat def. + PrototypeAST *Proto = ParsePrototype(); + if (Proto == 0) return 0; + + if (ExprAST *E = ParseExpression()) + return new FunctionAST(Proto, E); + return 0; +} + +/// toplevelexpr ::= expression +static FunctionAST *ParseTopLevelExpr() { + if (ExprAST *E = ParseExpression()) { + // Make an anonymous proto. + PrototypeAST *Proto = new PrototypeAST("", std::vector()); + return new FunctionAST(Proto, E); + } + return 0; +} + +/// external ::= 'extern' prototype +static PrototypeAST *ParseExtern() { + getNextToken(); // eat extern. + return ParsePrototype(); +} + +//===----------------------------------------------------------------------===// +// Code Generation +//===----------------------------------------------------------------------===// + +static Module *TheModule; +static FunctionPassManager *TheFPM; +static IRBuilder<> Builder(getGlobalContext()); +static std::map NamedValues; + +Value *ErrorV(const char *Str) { Error(Str); return 0; } + +/// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of +/// the function. This is used for mutable variables etc. +static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction, + const std::string &VarName) { + IRBuilder<> TmpB(&TheFunction->getEntryBlock(), + TheFunction->getEntryBlock().begin()); + return TmpB.CreateAlloca(Type::getDoubleTy(getGlobalContext()), 0, + VarName.c_str()); +} + +Value *NumberExprAST::Codegen() { + return ConstantFP::get(getGlobalContext(), APFloat(Val)); +} + +Value *VariableExprAST::Codegen() { + // Look this variable up in the function. + Value *V = NamedValues[Name]; + if (V == 0) return ErrorV("Unknown variable name"); + + // Load the value. + return Builder.CreateLoad(V, Name.c_str()); +} + +Value *UnaryExprAST::Codegen() { + Value *OperandV = Operand->Codegen(); + if (OperandV == 0) return 0; +#ifdef USE_MCJIT + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("unary")+Opcode)); +#else + Function *F = TheModule->getFunction(std::string("unary")+Opcode); +#endif + if (F == 0) + return ErrorV("Unknown unary operator"); + + return Builder.CreateCall(F, OperandV, "unop"); +} + +Value *BinaryExprAST::Codegen() { + // Special case '=' because we don't want to emit the LHS as an expression. + if (Op == '=') { + // Assignment requires the LHS to be an identifier. + // For now, I'm building without RTTI because LLVM builds that way by + // default and so we need to build that way to use the command line supprt. + // If you build LLVM with RTTI this can be changed back to a dynamic_cast. + VariableExprAST *LHSE = reinterpret_cast(LHS); + if (!LHSE) + return ErrorV("destination of '=' must be a variable"); + // Codegen the RHS. + Value *Val = RHS->Codegen(); + if (Val == 0) return 0; + + // Look up the name. + Value *Variable = NamedValues[LHSE->getName()]; + if (Variable == 0) return ErrorV("Unknown variable name"); + + Builder.CreateStore(Val, Variable); + return Val; + } + + Value *L = LHS->Codegen(); + Value *R = RHS->Codegen(); + if (L == 0 || R == 0) return 0; + + switch (Op) { + case '+': return Builder.CreateFAdd(L, R, "addtmp"); + case '-': return Builder.CreateFSub(L, R, "subtmp"); + case '*': return Builder.CreateFMul(L, R, "multmp"); + case '/': return Builder.CreateFDiv(L, R, "divtmp"); + case '<': + L = Builder.CreateFCmpULT(L, R, "cmptmp"); + // Convert bool 0/1 to double 0.0 or 1.0 + return Builder.CreateUIToFP(L, Type::getDoubleTy(getGlobalContext()), + "booltmp"); + default: break; + } + + // If it wasn't a builtin binary operator, it must be a user defined one. Emit + // a call to it. + Function *F = TheModule->getFunction(std::string("binary")+Op); + assert(F && "binary operator not found!"); + + Value *Ops[] = { L, R }; + return Builder.CreateCall(F, Ops, "binop"); +} + +Value *CallExprAST::Codegen() { + // Look up the name in the global module table. + Function *CalleeF = TheModule->getFunction(Callee); + if (CalleeF == 0) { + char error_str[64]; + sprintf(error_str, "Unknown function referenced %s", Callee.c_str()); + return ErrorV(error_str); + } + + // If argument mismatch error. + if (CalleeF->arg_size() != Args.size()) + return ErrorV("Incorrect # arguments passed"); + + std::vector ArgsV; + for (unsigned i = 0, e = Args.size(); i != e; ++i) { + ArgsV.push_back(Args[i]->Codegen()); + if (ArgsV.back() == 0) return 0; + } + + return Builder.CreateCall(CalleeF, ArgsV, "calltmp"); +} + +Value *IfExprAST::Codegen() { + Value *CondV = Cond->Codegen(); + if (CondV == 0) return 0; + + // Convert condition to a bool by comparing equal to 0.0. + CondV = Builder.CreateFCmpONE(CondV, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "ifcond"); + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create blocks for the then and else cases. Insert the 'then' block at the + // end of the function. + BasicBlock *ThenBB = BasicBlock::Create(getGlobalContext(), "then", TheFunction); + BasicBlock *ElseBB = BasicBlock::Create(getGlobalContext(), "else"); + BasicBlock *MergeBB = BasicBlock::Create(getGlobalContext(), "ifcont"); + + Builder.CreateCondBr(CondV, ThenBB, ElseBB); + + // Emit then value. + Builder.SetInsertPoint(ThenBB); + + Value *ThenV = Then->Codegen(); + if (ThenV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Then' can change the current block, update ThenBB for the PHI. + ThenBB = Builder.GetInsertBlock(); + + // Emit else block. + TheFunction->getBasicBlockList().push_back(ElseBB); + Builder.SetInsertPoint(ElseBB); + + Value *ElseV = Else->Codegen(); + if (ElseV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Else' can change the current block, update ElseBB for the PHI. + ElseBB = Builder.GetInsertBlock(); + + // Emit merge block. + TheFunction->getBasicBlockList().push_back(MergeBB); + Builder.SetInsertPoint(MergeBB); + PHINode *PN = Builder.CreatePHI(Type::getDoubleTy(getGlobalContext()), 2, + "iftmp"); + + PN->addIncoming(ThenV, ThenBB); + PN->addIncoming(ElseV, ElseBB); + return PN; +} + +Value *ForExprAST::Codegen() { + // Output this as: + // var = alloca double + // ... + // start = startexpr + // store start -> var + // goto loop + // loop: + // ... + // bodyexpr + // ... + // loopend: + // step = stepexpr + // endcond = endexpr + // + // curvar = load var + // nextvar = curvar + step + // store nextvar -> var + // br endcond, loop, endloop + // outloop: + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create an alloca for the variable in the entry block. + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + + // Emit the start code first, without 'variable' in scope. + Value *StartVal = Start->Codegen(); + if (StartVal == 0) return 0; + + // Store the value into the alloca. + Builder.CreateStore(StartVal, Alloca); + + // Make the new basic block for the loop header, inserting after current + // block. + BasicBlock *LoopBB = BasicBlock::Create(getGlobalContext(), "loop", TheFunction); + + // Insert an explicit fall through from the current block to the LoopBB. + Builder.CreateBr(LoopBB); + + // Start insertion in LoopBB. + Builder.SetInsertPoint(LoopBB); + + // Within the loop, the variable is defined equal to the PHI node. If it + // shadows an existing variable, we have to restore it, so save it now. + AllocaInst *OldVal = NamedValues[VarName]; + NamedValues[VarName] = Alloca; + + // Emit the body of the loop. This, like any other expr, can change the + // current BB. Note that we ignore the value computed by the body, but don't + // allow an error. + if (Body->Codegen() == 0) + return 0; + + // Emit the step value. + Value *StepVal; + if (Step) { + StepVal = Step->Codegen(); + if (StepVal == 0) return 0; + } else { + // If not specified, use 1.0. + StepVal = ConstantFP::get(getGlobalContext(), APFloat(1.0)); + } + + // Compute the end condition. + Value *EndCond = End->Codegen(); + if (EndCond == 0) return EndCond; + + // Reload, increment, and restore the alloca. This handles the case where + // the body of the loop mutates the variable. + Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str()); + Value *NextVar = Builder.CreateFAdd(CurVar, StepVal, "nextvar"); + Builder.CreateStore(NextVar, Alloca); + + // Convert condition to a bool by comparing equal to 0.0. + EndCond = Builder.CreateFCmpONE(EndCond, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "loopcond"); + + // Create the "after loop" block and insert it. + BasicBlock *AfterBB = BasicBlock::Create(getGlobalContext(), "afterloop", TheFunction); + + // Insert the conditional branch into the end of LoopEndBB. + Builder.CreateCondBr(EndCond, LoopBB, AfterBB); + + // Any new code will be inserted in AfterBB. + Builder.SetInsertPoint(AfterBB); + + // Restore the unshadowed variable. + if (OldVal) + NamedValues[VarName] = OldVal; + else + NamedValues.erase(VarName); + + + // for expr always returns 0.0. + return Constant::getNullValue(Type::getDoubleTy(getGlobalContext())); +} + +Value *VarExprAST::Codegen() { + std::vector OldBindings; + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Register all variables and emit their initializer. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) { + const std::string &VarName = VarNames[i].first; + ExprAST *Init = VarNames[i].second; + + // Emit the initializer before adding the variable to scope, this prevents + // the initializer from referencing the variable itself, and permits stuff + // like this: + // var a = 1 in + // var a = a in ... # refers to outer 'a'. + Value *InitVal; + if (Init) { + InitVal = Init->Codegen(); + if (InitVal == 0) return 0; + } else { // If not specified, use 0.0. + InitVal = ConstantFP::get(getGlobalContext(), APFloat(0.0)); + } + + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + Builder.CreateStore(InitVal, Alloca); + + // Remember the old variable binding so that we can restore the binding when + // we unrecurse. + OldBindings.push_back(NamedValues[VarName]); + + // Remember this binding. + NamedValues[VarName] = Alloca; + } + + // Codegen the body, now that all vars are in scope. + Value *BodyVal = Body->Codegen(); + if (BodyVal == 0) return 0; + + // Pop all our variables from scope. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) + NamedValues[VarNames[i].first] = OldBindings[i]; + + // Return the body computation. + return BodyVal; +} + +Function *PrototypeAST::Codegen() { + // Make the function type: double(double,double) etc. + std::vector Doubles(Args.size(), + Type::getDoubleTy(getGlobalContext())); + FunctionType *FT = FunctionType::get(Type::getDoubleTy(getGlobalContext()), + Doubles, false); + + Function *F = Function::Create(FT, Function::ExternalLinkage, Name, TheModule); + // If F conflicted, there was already something named 'Name'. If it has a + // body, don't allow redefinition or reextern. + if (F->getName() != Name) { + // Delete the one we just made and get the existing one. + F->eraseFromParent(); + F = TheModule->getFunction(Name); + // If F already has a body, reject this. + if (!F->empty()) { + ErrorF("redefinition of function"); + return 0; + } + // If F took a different number of args, reject. + if (F->arg_size() != Args.size()) { + ErrorF("redefinition of function with different # args"); + return 0; + } + } + + // Set names for all arguments. + unsigned Idx = 0; + for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); + ++AI, ++Idx) + AI->setName(Args[Idx]); + + return F; +} + +/// CreateArgumentAllocas - Create an alloca for each argument and register the +/// argument in the symbol table so that references to it will succeed. +void PrototypeAST::CreateArgumentAllocas(Function *F) { + Function::arg_iterator AI = F->arg_begin(); + for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) { + // Create an alloca for this variable. + AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]); + + // Store the initial value into the alloca. + Builder.CreateStore(AI, Alloca); + + // Add arguments to variable symbol table. + NamedValues[Args[Idx]] = Alloca; + } +} + +Function *FunctionAST::Codegen() { + NamedValues.clear(); + + Function *TheFunction = Proto->Codegen(); + if (TheFunction == 0) + return 0; + + // If this is an operator, install it. + if (Proto->isBinaryOp()) + BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); + + // Create a new basic block to start insertion into. + BasicBlock *BB = BasicBlock::Create(getGlobalContext(), "entry", TheFunction); + Builder.SetInsertPoint(BB); + + // Add all arguments to the symbol table and create their allocas. + Proto->CreateArgumentAllocas(TheFunction); + + if (Value *RetVal = Body->Codegen()) { + // Finish off the function. + Builder.CreateRet(RetVal); + + // Validate the generated code, checking for consistency. + verifyFunction(*TheFunction); + + // Optimize the function. + TheFPM->run(*TheFunction); + + return TheFunction; + } + + // Error reading body, remove function. + TheFunction->eraseFromParent(); + + if (Proto->isBinaryOp()) + BinopPrecedence.erase(Proto->getOperatorName()); + return 0; +} + +//===----------------------------------------------------------------------===// +// Top-Level parsing and JIT Driver +//===----------------------------------------------------------------------===// + +static ExecutionEngine *TheExecutionEngine; + +static void HandleDefinition() { + if (FunctionAST *F = ParseDefinition()) { + if (Function *LF = F->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read function definition:"); + LF->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleExtern() { + if (PrototypeAST *P = ParseExtern()) { + if (Function *F = P->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read extern: "); + F->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleTopLevelExpression() { + // Evaluate a top-level expression into an anonymous function. + if (FunctionAST *F = ParseTopLevelExpr()) { + if (Function *LF = F->Codegen()) { + // JIT the function, returning a function pointer. + void *FPtr = TheExecutionEngine->getPointerToFunction(LF); + // Cast it to the right type (takes no arguments, returns a double) so we + // can call it as a native function. + double (*FP)() = (double (*)())(intptr_t)FPtr; +#ifdef MINIMAL_STDERR_OUTPUT + FP(); +#else + fprintf(stderr, "Evaluated to %f\n", FP()); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +/// top ::= definition | external | expression | ';' +static void MainLoop() { + while (1) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + switch (CurTok) { + case tok_eof: return; + case ';': getNextToken(); break; // ignore top-level semicolons. + case tok_def: HandleDefinition(); break; + case tok_extern: HandleExtern(); break; + default: HandleTopLevelExpression(); break; + } + } +} + +//===----------------------------------------------------------------------===// +// "Library" functions that can be "extern'd" from user code. +//===----------------------------------------------------------------------===// + +/// putchard - putchar that takes a double and returns 0. +extern "C" +double putchard(double X) { + putchar((char)X); + return 0; +} + +/// printd - printf that takes a double prints it as "%f\n", returning 0. +extern "C" +double printd(double X) { + printf("%f", X); + return 0; +} + +extern "C" +double printlf() { + printf("\n"); + return 0; +} + +//===----------------------------------------------------------------------===// +// Command line input file handlers +//===----------------------------------------------------------------------===// + +Module* parseInputIR(std::string InputFile) { + SMDiagnostic Err; + Module *M = ParseIRFile(InputFile, Err, getGlobalContext()); + if (!M) { + Err.print("IR parsing failed: ", errs()); + return NULL; + } + + return M; +} + +//===----------------------------------------------------------------------===// +// Main driver code. +//===----------------------------------------------------------------------===// + +int main(int argc, char **argv) { + InitializeNativeTarget(); + LLVMContext &Context = getGlobalContext(); + + cl::ParseCommandLineOptions(argc, argv, + "Kaleidoscope example program\n"); + + // Install standard binary operators. + // 1 is lowest precedence. + BinopPrecedence['='] = 2; + BinopPrecedence['<'] = 10; + BinopPrecedence['+'] = 20; + BinopPrecedence['-'] = 20; + BinopPrecedence['/'] = 40; + BinopPrecedence['*'] = 40; // highest. + + // Make the module, which holds all the code. + if (!InputIR.empty()) { + TheModule = parseInputIR(InputIR); + } else { + TheModule = new Module("my cool jit", Context); + } + + // Create the JIT. This takes ownership of the module. + std::string ErrStr; + TheExecutionEngine = EngineBuilder(TheModule).setErrorStr(&ErrStr).create(); + if (!TheExecutionEngine) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + FunctionPassManager OurFPM(TheModule); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + OurFPM.add(new DataLayout(*TheExecutionEngine->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + OurFPM.add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + OurFPM.add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + OurFPM.add(createInstructionCombiningPass()); + // Reassociate expressions. + OurFPM.add(createReassociatePass()); + // Eliminate Common SubExpressions. + OurFPM.add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + OurFPM.add(createCFGSimplificationPass()); + + OurFPM.doInitialization(); + + // Set the global so the code gen can use this. + TheFPM = &OurFPM; + + // Prime the first token. +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + getNextToken(); + + // Run the main "interpreter loop" now. + MainLoop(); + + // Print out all of the generated code. + TheFPM = 0; +#if !defined(MINIMAL_STDERR_OUTPUT) || defined(DUMP_FINAL_MODULE) + TheModule->dump(); +#endif + return 0; +} diff --git a/examples/Kaleidoscope/MCJIT/cached/toy.cpp b/examples/Kaleidoscope/MCJIT/cached/toy.cpp new file mode 100644 index 00000000000..6c8e38b13f5 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/cached/toy.cpp @@ -0,0 +1,1551 @@ +#define MINIMAL_STDERR_OUTPUT + +#include "llvm/Analysis/Passes.h" +#include "llvm/Analysis/Verifier.h" +#include "llvm/ExecutionEngine/ExecutionEngine.h" +#include "llvm/ExecutionEngine/MCJIT.h" +#include "llvm/ExecutionEngine/ObjectCache.h" +#include "llvm/ExecutionEngine/SectionMemoryManager.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/IRReader/IRReader.h" +#include "llvm/PassManager.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/FileSystem.h" +#include "llvm/Support/Path.h" +#include "llvm/Support/raw_ostream.h" +#include "llvm/Support/SourceMgr.h" +#include "llvm/Support/TargetSelect.h" +#include "llvm/Transforms/Scalar.h" +#include +#include +#include +#include +using namespace llvm; + +//===----------------------------------------------------------------------===// +// Command-line options +//===----------------------------------------------------------------------===// + +cl::opt +InputIR("input-IR", + cl::desc("Specify the name of an IR file to load for function definitions"), + cl::value_desc("input IR file name")); + +cl::opt +UseObjectCache("use-object-cache", + cl::desc("Enable use of the MCJIT object caching"), + cl::init(false)); + +//===----------------------------------------------------------------------===// +// Lexer +//===----------------------------------------------------------------------===// + +// The lexer returns tokens [0-255] if it is an unknown character, otherwise one +// of these for known things. +enum Token { + tok_eof = -1, + + // commands + tok_def = -2, tok_extern = -3, + + // primary + tok_identifier = -4, tok_number = -5, + + // control + tok_if = -6, tok_then = -7, tok_else = -8, + tok_for = -9, tok_in = -10, + + // operators + tok_binary = -11, tok_unary = -12, + + // var definition + tok_var = -13 +}; + +static std::string IdentifierStr; // Filled in if tok_identifier +static double NumVal; // Filled in if tok_number + +/// gettok - Return the next token from standard input. +static int gettok() { + static int LastChar = ' '; + + // Skip any whitespace. + while (isspace(LastChar)) + LastChar = getchar(); + + if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* + IdentifierStr = LastChar; + while (isalnum((LastChar = getchar()))) + IdentifierStr += LastChar; + + if (IdentifierStr == "def") return tok_def; + if (IdentifierStr == "extern") return tok_extern; + if (IdentifierStr == "if") return tok_if; + if (IdentifierStr == "then") return tok_then; + if (IdentifierStr == "else") return tok_else; + if (IdentifierStr == "for") return tok_for; + if (IdentifierStr == "in") return tok_in; + if (IdentifierStr == "binary") return tok_binary; + if (IdentifierStr == "unary") return tok_unary; + if (IdentifierStr == "var") return tok_var; + return tok_identifier; + } + + if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ + std::string NumStr; + do { + NumStr += LastChar; + LastChar = getchar(); + } while (isdigit(LastChar) || LastChar == '.'); + + NumVal = strtod(NumStr.c_str(), 0); + return tok_number; + } + + if (LastChar == '#') { + // Comment until end of line. + do LastChar = getchar(); + while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); + + if (LastChar != EOF) + return gettok(); + } + + // Check for end of file. Don't eat the EOF. + if (LastChar == EOF) + return tok_eof; + + // Otherwise, just return the character as its ascii value. + int ThisChar = LastChar; + LastChar = getchar(); + return ThisChar; +} + +//===----------------------------------------------------------------------===// +// Abstract Syntax Tree (aka Parse Tree) +//===----------------------------------------------------------------------===// + +/// ExprAST - Base class for all expression nodes. +class ExprAST { +public: + virtual ~ExprAST() {} + virtual Value *Codegen() = 0; +}; + +/// NumberExprAST - Expression class for numeric literals like "1.0". +class NumberExprAST : public ExprAST { + double Val; +public: + NumberExprAST(double val) : Val(val) {} + virtual Value *Codegen(); +}; + +/// VariableExprAST - Expression class for referencing a variable, like "a". +class VariableExprAST : public ExprAST { + std::string Name; +public: + VariableExprAST(const std::string &name) : Name(name) {} + const std::string &getName() const { return Name; } + virtual Value *Codegen(); +}; + +/// UnaryExprAST - Expression class for a unary operator. +class UnaryExprAST : public ExprAST { + char Opcode; + ExprAST *Operand; +public: + UnaryExprAST(char opcode, ExprAST *operand) + : Opcode(opcode), Operand(operand) {} + virtual Value *Codegen(); +}; + +/// BinaryExprAST - Expression class for a binary operator. +class BinaryExprAST : public ExprAST { + char Op; + ExprAST *LHS, *RHS; +public: + BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) + : Op(op), LHS(lhs), RHS(rhs) {} + virtual Value *Codegen(); +}; + +/// CallExprAST - Expression class for function calls. +class CallExprAST : public ExprAST { + std::string Callee; + std::vector Args; +public: + CallExprAST(const std::string &callee, std::vector &args) + : Callee(callee), Args(args) {} + virtual Value *Codegen(); +}; + +/// IfExprAST - Expression class for if/then/else. +class IfExprAST : public ExprAST { + ExprAST *Cond, *Then, *Else; +public: + IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) + : Cond(cond), Then(then), Else(_else) {} + virtual Value *Codegen(); +}; + +/// ForExprAST - Expression class for for/in. +class ForExprAST : public ExprAST { + std::string VarName; + ExprAST *Start, *End, *Step, *Body; +public: + ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, + ExprAST *step, ExprAST *body) + : VarName(varname), Start(start), End(end), Step(step), Body(body) {} + virtual Value *Codegen(); +}; + +/// VarExprAST - Expression class for var/in +class VarExprAST : public ExprAST { + std::vector > VarNames; + ExprAST *Body; +public: + VarExprAST(const std::vector > &varnames, + ExprAST *body) + : VarNames(varnames), Body(body) {} + + virtual Value *Codegen(); +}; + +/// PrototypeAST - This class represents the "prototype" for a function, +/// which captures its argument names as well as if it is an operator. +class PrototypeAST { + std::string Name; + std::vector Args; + bool isOperator; + unsigned Precedence; // Precedence if a binary op. +public: + PrototypeAST(const std::string &name, const std::vector &args, + bool isoperator = false, unsigned prec = 0) + : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} + + bool isUnaryOp() const { return isOperator && Args.size() == 1; } + bool isBinaryOp() const { return isOperator && Args.size() == 2; } + + char getOperatorName() const { + assert(isUnaryOp() || isBinaryOp()); + return Name[Name.size()-1]; + } + + unsigned getBinaryPrecedence() const { return Precedence; } + + Function *Codegen(); + + void CreateArgumentAllocas(Function *F); +}; + +/// FunctionAST - This class represents a function definition itself. +class FunctionAST { + PrototypeAST *Proto; + ExprAST *Body; +public: + FunctionAST(PrototypeAST *proto, ExprAST *body) + : Proto(proto), Body(body) {} + + Function *Codegen(); +}; + +//===----------------------------------------------------------------------===// +// Parser +//===----------------------------------------------------------------------===// + +/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current +/// token the parser is looking at. getNextToken reads another token from the +/// lexer and updates CurTok with its results. +static int CurTok; +static int getNextToken() { + return CurTok = gettok(); +} + +/// BinopPrecedence - This holds the precedence for each binary operator that is +/// defined. +static std::map BinopPrecedence; + +/// GetTokPrecedence - Get the precedence of the pending binary operator token. +static int GetTokPrecedence() { + if (!isascii(CurTok)) + return -1; + + // Make sure it's a declared binop. + int TokPrec = BinopPrecedence[CurTok]; + if (TokPrec <= 0) return -1; + return TokPrec; +} + +/// Error* - These are little helper functions for error handling. +ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} +PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } +FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } + +static ExprAST *ParseExpression(); + +/// identifierexpr +/// ::= identifier +/// ::= identifier '(' expression* ')' +static ExprAST *ParseIdentifierExpr() { + std::string IdName = IdentifierStr; + + getNextToken(); // eat identifier. + + if (CurTok != '(') // Simple variable ref. + return new VariableExprAST(IdName); + + // Call. + getNextToken(); // eat ( + std::vector Args; + if (CurTok != ')') { + while (1) { + ExprAST *Arg = ParseExpression(); + if (!Arg) return 0; + Args.push_back(Arg); + + if (CurTok == ')') break; + + if (CurTok != ',') + return Error("Expected ')' or ',' in argument list"); + getNextToken(); + } + } + + // Eat the ')'. + getNextToken(); + + return new CallExprAST(IdName, Args); +} + +/// numberexpr ::= number +static ExprAST *ParseNumberExpr() { + ExprAST *Result = new NumberExprAST(NumVal); + getNextToken(); // consume the number + return Result; +} + +/// parenexpr ::= '(' expression ')' +static ExprAST *ParseParenExpr() { + getNextToken(); // eat (. + ExprAST *V = ParseExpression(); + if (!V) return 0; + + if (CurTok != ')') + return Error("expected ')'"); + getNextToken(); // eat ). + return V; +} + +/// ifexpr ::= 'if' expression 'then' expression 'else' expression +static ExprAST *ParseIfExpr() { + getNextToken(); // eat the if. + + // condition. + ExprAST *Cond = ParseExpression(); + if (!Cond) return 0; + + if (CurTok != tok_then) + return Error("expected then"); + getNextToken(); // eat the then + + ExprAST *Then = ParseExpression(); + if (Then == 0) return 0; + + if (CurTok != tok_else) + return Error("expected else"); + + getNextToken(); + + ExprAST *Else = ParseExpression(); + if (!Else) return 0; + + return new IfExprAST(Cond, Then, Else); +} + +/// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression +static ExprAST *ParseForExpr() { + getNextToken(); // eat the for. + + if (CurTok != tok_identifier) + return Error("expected identifier after for"); + + std::string IdName = IdentifierStr; + getNextToken(); // eat identifier. + + if (CurTok != '=') + return Error("expected '=' after for"); + getNextToken(); // eat '='. + + + ExprAST *Start = ParseExpression(); + if (Start == 0) return 0; + if (CurTok != ',') + return Error("expected ',' after for start value"); + getNextToken(); + + ExprAST *End = ParseExpression(); + if (End == 0) return 0; + + // The step value is optional. + ExprAST *Step = 0; + if (CurTok == ',') { + getNextToken(); + Step = ParseExpression(); + if (Step == 0) return 0; + } + + if (CurTok != tok_in) + return Error("expected 'in' after for"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new ForExprAST(IdName, Start, End, Step, Body); +} + +/// varexpr ::= 'var' identifier ('=' expression)? +// (',' identifier ('=' expression)?)* 'in' expression +static ExprAST *ParseVarExpr() { + getNextToken(); // eat the var. + + std::vector > VarNames; + + // At least one variable name is required. + if (CurTok != tok_identifier) + return Error("expected identifier after var"); + + while (1) { + std::string Name = IdentifierStr; + getNextToken(); // eat identifier. + + // Read the optional initializer. + ExprAST *Init = 0; + if (CurTok == '=') { + getNextToken(); // eat the '='. + + Init = ParseExpression(); + if (Init == 0) return 0; + } + + VarNames.push_back(std::make_pair(Name, Init)); + + // End of var list, exit loop. + if (CurTok != ',') break; + getNextToken(); // eat the ','. + + if (CurTok != tok_identifier) + return Error("expected identifier list after var"); + } + + // At this point, we have to have 'in'. + if (CurTok != tok_in) + return Error("expected 'in' keyword after 'var'"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new VarExprAST(VarNames, Body); +} + +/// primary +/// ::= identifierexpr +/// ::= numberexpr +/// ::= parenexpr +/// ::= ifexpr +/// ::= forexpr +/// ::= varexpr +static ExprAST *ParsePrimary() { + switch (CurTok) { + default: return Error("unknown token when expecting an expression"); + case tok_identifier: return ParseIdentifierExpr(); + case tok_number: return ParseNumberExpr(); + case '(': return ParseParenExpr(); + case tok_if: return ParseIfExpr(); + case tok_for: return ParseForExpr(); + case tok_var: return ParseVarExpr(); + } +} + +/// unary +/// ::= primary +/// ::= '!' unary +static ExprAST *ParseUnary() { + // If the current token is not an operator, it must be a primary expr. + if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') + return ParsePrimary(); + + // If this is a unary operator, read it. + int Opc = CurTok; + getNextToken(); + if (ExprAST *Operand = ParseUnary()) + return new UnaryExprAST(Opc, Operand); + return 0; +} + +/// binoprhs +/// ::= ('+' unary)* +static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { + // If this is a binop, find its precedence. + while (1) { + int TokPrec = GetTokPrecedence(); + + // If this is a binop that binds at least as tightly as the current binop, + // consume it, otherwise we are done. + if (TokPrec < ExprPrec) + return LHS; + + // Okay, we know this is a binop. + int BinOp = CurTok; + getNextToken(); // eat binop + + // Parse the unary expression after the binary operator. + ExprAST *RHS = ParseUnary(); + if (!RHS) return 0; + + // If BinOp binds less tightly with RHS than the operator after RHS, let + // the pending operator take RHS as its LHS. + int NextPrec = GetTokPrecedence(); + if (TokPrec < NextPrec) { + RHS = ParseBinOpRHS(TokPrec+1, RHS); + if (RHS == 0) return 0; + } + + // Merge LHS/RHS. + LHS = new BinaryExprAST(BinOp, LHS, RHS); + } +} + +/// expression +/// ::= unary binoprhs +/// +static ExprAST *ParseExpression() { + ExprAST *LHS = ParseUnary(); + if (!LHS) return 0; + + return ParseBinOpRHS(0, LHS); +} + +/// prototype +/// ::= id '(' id* ')' +/// ::= binary LETTER number? (id, id) +/// ::= unary LETTER (id) +static PrototypeAST *ParsePrototype() { + std::string FnName; + + unsigned Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. + unsigned BinaryPrecedence = 30; + + switch (CurTok) { + default: + return ErrorP("Expected function name in prototype"); + case tok_identifier: + FnName = IdentifierStr; + Kind = 0; + getNextToken(); + break; + case tok_unary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected unary operator"); + FnName = "unary"; + FnName += (char)CurTok; + Kind = 1; + getNextToken(); + break; + case tok_binary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected binary operator"); + FnName = "binary"; + FnName += (char)CurTok; + Kind = 2; + getNextToken(); + + // Read the precedence if present. + if (CurTok == tok_number) { + if (NumVal < 1 || NumVal > 100) + return ErrorP("Invalid precedecnce: must be 1..100"); + BinaryPrecedence = (unsigned)NumVal; + getNextToken(); + } + break; + } + + if (CurTok != '(') + return ErrorP("Expected '(' in prototype"); + + std::vector ArgNames; + while (getNextToken() == tok_identifier) + ArgNames.push_back(IdentifierStr); + if (CurTok != ')') + return ErrorP("Expected ')' in prototype"); + + // success. + getNextToken(); // eat ')'. + + // Verify right number of names for operator. + if (Kind && ArgNames.size() != Kind) + return ErrorP("Invalid number of operands for operator"); + + return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); +} + +/// definition ::= 'def' prototype expression +static FunctionAST *ParseDefinition() { + getNextToken(); // eat def. + PrototypeAST *Proto = ParsePrototype(); + if (Proto == 0) return 0; + + if (ExprAST *E = ParseExpression()) + return new FunctionAST(Proto, E); + return 0; +} + +/// toplevelexpr ::= expression +static FunctionAST *ParseTopLevelExpr() { + if (ExprAST *E = ParseExpression()) { + // Make an anonymous proto. + PrototypeAST *Proto = new PrototypeAST("", std::vector()); + return new FunctionAST(Proto, E); + } + return 0; +} + +/// external ::= 'extern' prototype +static PrototypeAST *ParseExtern() { + getNextToken(); // eat extern. + return ParsePrototype(); +} + +//===----------------------------------------------------------------------===// +// Quick and dirty hack +//===----------------------------------------------------------------------===// + +// FIXME: Obviously we can do better than this +std::string GenerateUniqueName(const char *root) +{ + static int i = 0; + char s[16]; + sprintf(s, "%s%d", root, i++); + std::string S = s; + return S; +} + +std::string MakeLegalFunctionName(std::string Name) +{ + std::string NewName; + if (!Name.length()) + return GenerateUniqueName("anon_func_"); + + // Start with what we have + NewName = Name; + + // Look for a numberic first character + if (NewName.find_first_of("0123456789") == 0) { + NewName.insert(0, 1, 'n'); + } + + // Replace illegal characters with their ASCII equivalent + std::string legal_elements = "_abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"; + size_t pos; + while ((pos = NewName.find_first_not_of(legal_elements)) != std::string::npos) { + char old_c = NewName.at(pos); + char new_str[16]; + sprintf(new_str, "%d", (int)old_c); + NewName = NewName.replace(pos, 1, new_str); + } + + return NewName; +} + +//===----------------------------------------------------------------------===// +// MCJIT object cache class +//===----------------------------------------------------------------------===// + +class MCJITObjectCache : public ObjectCache { +public: + MCJITObjectCache() { + // Set IR cache directory + sys::fs::current_path(CacheDir); + sys::path::append(CacheDir, "toy_object_cache"); + } + + virtual ~MCJITObjectCache() { + } + + virtual void notifyObjectCompiled(const Module *M, const MemoryBuffer *Obj) { + // Get the ModuleID + const std::string ModuleID = M->getModuleIdentifier(); + + // If we've flagged this as an IR file, cache it + if (0 == ModuleID.compare(0, 3, "IR:")) { + std::string IRFileName = ModuleID.substr(3); + SmallString<128>IRCacheFile = CacheDir; + sys::path::append(IRCacheFile, IRFileName); + if (!sys::fs::exists(CacheDir.str()) && sys::fs::create_directory(CacheDir.str())) { + fprintf(stderr, "Unable to create cache directory\n"); + return; + } + std::string ErrStr; + raw_fd_ostream IRObjectFile(IRCacheFile.c_str(), ErrStr, raw_fd_ostream::F_Binary); + IRObjectFile << Obj->getBuffer(); + } + } + + // MCJIT will call this function before compiling any module + // MCJIT takes ownership of both the MemoryBuffer object and the memory + // to which it refers. + virtual MemoryBuffer* getObject(const Module* M) { + // Get the ModuleID + const std::string ModuleID = M->getModuleIdentifier(); + + // If we've flagged this as an IR file, cache it + if (0 == ModuleID.compare(0, 3, "IR:")) { + std::string IRFileName = ModuleID.substr(3); + SmallString<128> IRCacheFile = CacheDir; + sys::path::append(IRCacheFile, IRFileName); + if (!sys::fs::exists(IRCacheFile.str())) { + // This file isn't in our cache + return NULL; + } + OwningPtr IRObjectBuffer; + MemoryBuffer::getFile(IRCacheFile.c_str(), IRObjectBuffer, -1, false); + // MCJIT will want to write into this buffer, and we don't want that + // because the file has probably just been mmapped. Instead we make + // a copy. The filed-based buffer will be released when it goes + // out of scope. + return MemoryBuffer::getMemBufferCopy(IRObjectBuffer->getBuffer()); + } + + return NULL; + } + +private: + SmallString<128> CacheDir; +}; + +//===----------------------------------------------------------------------===// +// MCJIT helper class +//===----------------------------------------------------------------------===// + +class MCJITHelper +{ +public: + MCJITHelper(LLVMContext& C) : Context(C), OpenModule(NULL) {} + ~MCJITHelper(); + + Function *getFunction(const std::string FnName); + Module *getModuleForNewFunction(); + void *getPointerToFunction(Function* F); + void *getPointerToNamedFunction(const std::string &Name); + ExecutionEngine *compileModule(Module *M); + void closeCurrentModule(); + void addModule(Module *M); + void dump(); + +private: + typedef std::vector ModuleVector; + + LLVMContext &Context; + Module *OpenModule; + ModuleVector Modules; + std::map EngineMap; + MCJITObjectCache OurObjectCache; +}; + +class HelpingMemoryManager : public SectionMemoryManager +{ + HelpingMemoryManager(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + void operator=(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + +public: + HelpingMemoryManager(MCJITHelper *Helper) : MasterHelper(Helper) {} + virtual ~HelpingMemoryManager() {} + + /// This method returns the address of the specified function. + /// Our implementation will attempt to find functions in other + /// modules associated with the MCJITHelper to cross link functions + /// from one generated module to another. + /// + /// If \p AbortOnFailure is false and no function with the given name is + /// found, this function returns a null pointer. Otherwise, it prints a + /// message to stderr and aborts. + virtual void *getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure = true); +private: + MCJITHelper *MasterHelper; +}; + +void *HelpingMemoryManager::getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure) +{ + // Try the standard symbol resolution first, but ask it not to abort. + void *pfn = SectionMemoryManager::getPointerToNamedFunction(Name, false); + if (pfn) + return pfn; + + pfn = MasterHelper->getPointerToNamedFunction(Name); + if (!pfn && AbortOnFailure) + report_fatal_error("Program used external function '" + Name + + "' which could not be resolved!"); + return pfn; +} + +MCJITHelper::~MCJITHelper() +{ + // Walk the vector of modules. + ModuleVector::iterator it, end; + for (it = Modules.begin(), end = Modules.end(); + it != end; ++it) { + // See if we have an execution engine for this module. + std::map::iterator mapIt = EngineMap.find(*it); + // If we have an EE, the EE owns the module so just delete the EE. + if (mapIt != EngineMap.end()) { + delete mapIt->second; + } else { + // Otherwise, we still own the module. Delete it now. + delete *it; + } + } +} + +Function *MCJITHelper::getFunction(const std::string FnName) { + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(FnName); + if (F) { + if (*it == OpenModule) + return F; + + assert(OpenModule != NULL); + + // This function is in a module that has already been JITed. + // We need to generate a new prototype for external linkage. + Function *PF = OpenModule->getFunction(FnName); + if (PF && !PF->empty()) { + ErrorF("redefinition of function across modules"); + return 0; + } + + // If we don't have a prototype yet, create one. + if (!PF) + PF = Function::Create(F->getFunctionType(), + Function::ExternalLinkage, + FnName, + OpenModule); + return PF; + } + } + return NULL; +} + +Module *MCJITHelper::getModuleForNewFunction() { + // If we have a Module that hasn't been JITed, use that. + if (OpenModule) + return OpenModule; + + // Otherwise create a new Module. + std::string ModName = GenerateUniqueName("mcjit_module_"); + Module *M = new Module(ModName, Context); + Modules.push_back(M); + OpenModule = M; + return M; +} + +void *MCJITHelper::getPointerToFunction(Function* F) { + // Look for this function in an existing module + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + std::string FnName = F->getName(); + for (it = begin; it != end; ++it) { + Function *MF = (*it)->getFunction(FnName); + if (MF == F) { + std::map::iterator eeIt = EngineMap.find(*it); + if (eeIt != EngineMap.end()) { + void *P = eeIt->second->getPointerToFunction(F); + if (P) + return P; + } else { + ExecutionEngine *EE = compileModule(*it); + void *P = EE->getPointerToFunction(F); + if (P) + return P; + } + } + } + return NULL; +} + +void MCJITHelper::closeCurrentModule() { + OpenModule = NULL; +} + +ExecutionEngine *MCJITHelper::compileModule(Module *M) { + if (M == OpenModule) + closeCurrentModule(); + + std::string ErrStr; + ExecutionEngine *NewEngine = EngineBuilder(M) + .setErrorStr(&ErrStr) + .setUseMCJIT(true) + .setMCJITMemoryManager(new HelpingMemoryManager(this)) + .create(); + if (!NewEngine) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + if (UseObjectCache) + NewEngine->setObjectCache(&OurObjectCache); + + // Get the ModuleID so we can identify IR input files + const std::string ModuleID = M->getModuleIdentifier(); + + // If we've flagged this as an IR file, it doesn't need function passes run. + if (0 != ModuleID.compare(0, 3, "IR:")) { + // Create a function pass manager for this engine + FunctionPassManager *FPM = new FunctionPassManager(M); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + FPM->add(new DataLayout(*NewEngine->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + FPM->add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + FPM->add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + FPM->add(createInstructionCombiningPass()); + // Reassociate expressions. + FPM->add(createReassociatePass()); + // Eliminate Common SubExpressions. + FPM->add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + FPM->add(createCFGSimplificationPass()); + FPM->doInitialization(); + + // For each function in the module + Module::iterator it; + Module::iterator end = M->end(); + for (it = M->begin(); it != end; ++it) { + // Run the FPM on this function + FPM->run(*it); + } + + // We don't need this anymore + delete FPM; + } + + // Store this engine + EngineMap[M] = NewEngine; + NewEngine->finalizeObject(); + + return NewEngine; +} + +void *MCJITHelper::getPointerToNamedFunction(const std::string &Name) +{ + // Look for the functions in our modules, compiling only as necessary + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(Name); + if (F && !F->empty()) { + std::map::iterator eeIt = EngineMap.find(*it); + if (eeIt != EngineMap.end()) { + void *P = eeIt->second->getPointerToFunction(F); + if (P) + return P; + } else { + ExecutionEngine *EE = compileModule(*it); + void *P = EE->getPointerToFunction(F); + if (P) + return P; + } + } + } + return NULL; +} + +void MCJITHelper::addModule(Module* M) { + Modules.push_back(M); +} + +void MCJITHelper::dump() +{ + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) + (*it)->dump(); +} + +//===----------------------------------------------------------------------===// +// Code Generation +//===----------------------------------------------------------------------===// + +static MCJITHelper *TheHelper; +static IRBuilder<> Builder(getGlobalContext()); +static std::map NamedValues; + +Value *ErrorV(const char *Str) { Error(Str); return 0; } + +/// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of +/// the function. This is used for mutable variables etc. +static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction, + const std::string &VarName) { + IRBuilder<> TmpB(&TheFunction->getEntryBlock(), + TheFunction->getEntryBlock().begin()); + return TmpB.CreateAlloca(Type::getDoubleTy(getGlobalContext()), 0, + VarName.c_str()); +} + +Value *NumberExprAST::Codegen() { + return ConstantFP::get(getGlobalContext(), APFloat(Val)); +} + +Value *VariableExprAST::Codegen() { + // Look this variable up in the function. + Value *V = NamedValues[Name]; + char ErrStr[256]; + sprintf(ErrStr, "Unknown variable name %s", Name.c_str()); + if (V == 0) return ErrorV(ErrStr); + + // Load the value. + return Builder.CreateLoad(V, Name.c_str()); +} + +Value *UnaryExprAST::Codegen() { + Value *OperandV = Operand->Codegen(); + if (OperandV == 0) return 0; + + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("unary")+Opcode)); + if (F == 0) + return ErrorV("Unknown unary operator"); + + return Builder.CreateCall(F, OperandV, "unop"); +} + +Value *BinaryExprAST::Codegen() { + // Special case '=' because we don't want to emit the LHS as an expression. + if (Op == '=') { + // Assignment requires the LHS to be an identifier. + VariableExprAST *LHSE = reinterpret_cast(LHS); + if (!LHSE) + return ErrorV("destination of '=' must be a variable"); + // Codegen the RHS. + Value *Val = RHS->Codegen(); + if (Val == 0) return 0; + + // Look up the name. + Value *Variable = NamedValues[LHSE->getName()]; + if (Variable == 0) return ErrorV("Unknown variable name"); + + Builder.CreateStore(Val, Variable); + return Val; + } + + Value *L = LHS->Codegen(); + Value *R = RHS->Codegen(); + if (L == 0 || R == 0) return 0; + + switch (Op) { + case '+': return Builder.CreateFAdd(L, R, "addtmp"); + case '-': return Builder.CreateFSub(L, R, "subtmp"); + case '*': return Builder.CreateFMul(L, R, "multmp"); + case '/': return Builder.CreateFDiv(L, R, "divtmp"); + case '<': + L = Builder.CreateFCmpULT(L, R, "cmptmp"); + // Convert bool 0/1 to double 0.0 or 1.0 + return Builder.CreateUIToFP(L, Type::getDoubleTy(getGlobalContext()), + "booltmp"); + default: break; + } + + // If it wasn't a builtin binary operator, it must be a user defined one. Emit + // a call to it. + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("binary")+Op)); + assert(F && "binary operator not found!"); + + Value *Ops[] = { L, R }; + return Builder.CreateCall(F, Ops, "binop"); +} + +Value *CallExprAST::Codegen() { + // Look up the name in the global module table. + Function *CalleeF = TheHelper->getFunction(Callee); + if (CalleeF == 0) + return ErrorV("Unknown function referenced"); + + // If argument mismatch error. + if (CalleeF->arg_size() != Args.size()) + return ErrorV("Incorrect # arguments passed"); + + std::vector ArgsV; + for (unsigned i = 0, e = Args.size(); i != e; ++i) { + ArgsV.push_back(Args[i]->Codegen()); + if (ArgsV.back() == 0) return 0; + } + + return Builder.CreateCall(CalleeF, ArgsV, "calltmp"); +} + +Value *IfExprAST::Codegen() { + Value *CondV = Cond->Codegen(); + if (CondV == 0) return 0; + + // Convert condition to a bool by comparing equal to 0.0. + CondV = Builder.CreateFCmpONE(CondV, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "ifcond"); + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create blocks for the then and else cases. Insert the 'then' block at the + // end of the function. + BasicBlock *ThenBB = BasicBlock::Create(getGlobalContext(), "then", TheFunction); + BasicBlock *ElseBB = BasicBlock::Create(getGlobalContext(), "else"); + BasicBlock *MergeBB = BasicBlock::Create(getGlobalContext(), "ifcont"); + + Builder.CreateCondBr(CondV, ThenBB, ElseBB); + + // Emit then value. + Builder.SetInsertPoint(ThenBB); + + Value *ThenV = Then->Codegen(); + if (ThenV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Then' can change the current block, update ThenBB for the PHI. + ThenBB = Builder.GetInsertBlock(); + + // Emit else block. + TheFunction->getBasicBlockList().push_back(ElseBB); + Builder.SetInsertPoint(ElseBB); + + Value *ElseV = Else->Codegen(); + if (ElseV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Else' can change the current block, update ElseBB for the PHI. + ElseBB = Builder.GetInsertBlock(); + + // Emit merge block. + TheFunction->getBasicBlockList().push_back(MergeBB); + Builder.SetInsertPoint(MergeBB); + PHINode *PN = Builder.CreatePHI(Type::getDoubleTy(getGlobalContext()), 2, + "iftmp"); + + PN->addIncoming(ThenV, ThenBB); + PN->addIncoming(ElseV, ElseBB); + return PN; +} + +Value *ForExprAST::Codegen() { + // Output this as: + // var = alloca double + // ... + // start = startexpr + // store start -> var + // goto loop + // loop: + // ... + // bodyexpr + // ... + // loopend: + // step = stepexpr + // endcond = endexpr + // + // curvar = load var + // nextvar = curvar + step + // store nextvar -> var + // br endcond, loop, endloop + // outloop: + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create an alloca for the variable in the entry block. + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + + // Emit the start code first, without 'variable' in scope. + Value *StartVal = Start->Codegen(); + if (StartVal == 0) return 0; + + // Store the value into the alloca. + Builder.CreateStore(StartVal, Alloca); + + // Make the new basic block for the loop header, inserting after current + // block. + BasicBlock *LoopBB = BasicBlock::Create(getGlobalContext(), "loop", TheFunction); + + // Insert an explicit fall through from the current block to the LoopBB. + Builder.CreateBr(LoopBB); + + // Start insertion in LoopBB. + Builder.SetInsertPoint(LoopBB); + + // Within the loop, the variable is defined equal to the PHI node. If it + // shadows an existing variable, we have to restore it, so save it now. + AllocaInst *OldVal = NamedValues[VarName]; + NamedValues[VarName] = Alloca; + + // Emit the body of the loop. This, like any other expr, can change the + // current BB. Note that we ignore the value computed by the body, but don't + // allow an error. + if (Body->Codegen() == 0) + return 0; + + // Emit the step value. + Value *StepVal; + if (Step) { + StepVal = Step->Codegen(); + if (StepVal == 0) return 0; + } else { + // If not specified, use 1.0. + StepVal = ConstantFP::get(getGlobalContext(), APFloat(1.0)); + } + + // Compute the end condition. + Value *EndCond = End->Codegen(); + if (EndCond == 0) return EndCond; + + // Reload, increment, and restore the alloca. This handles the case where + // the body of the loop mutates the variable. + Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str()); + Value *NextVar = Builder.CreateFAdd(CurVar, StepVal, "nextvar"); + Builder.CreateStore(NextVar, Alloca); + + // Convert condition to a bool by comparing equal to 0.0. + EndCond = Builder.CreateFCmpONE(EndCond, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "loopcond"); + + // Create the "after loop" block and insert it. + BasicBlock *AfterBB = BasicBlock::Create(getGlobalContext(), "afterloop", TheFunction); + + // Insert the conditional branch into the end of LoopEndBB. + Builder.CreateCondBr(EndCond, LoopBB, AfterBB); + + // Any new code will be inserted in AfterBB. + Builder.SetInsertPoint(AfterBB); + + // Restore the unshadowed variable. + if (OldVal) + NamedValues[VarName] = OldVal; + else + NamedValues.erase(VarName); + + + // for expr always returns 0.0. + return Constant::getNullValue(Type::getDoubleTy(getGlobalContext())); +} + +Value *VarExprAST::Codegen() { + std::vector OldBindings; + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Register all variables and emit their initializer. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) { + const std::string &VarName = VarNames[i].first; + ExprAST *Init = VarNames[i].second; + + // Emit the initializer before adding the variable to scope, this prevents + // the initializer from referencing the variable itself, and permits stuff + // like this: + // var a = 1 in + // var a = a in ... # refers to outer 'a'. + Value *InitVal; + if (Init) { + InitVal = Init->Codegen(); + if (InitVal == 0) return 0; + } else { // If not specified, use 0.0. + InitVal = ConstantFP::get(getGlobalContext(), APFloat(0.0)); + } + + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + Builder.CreateStore(InitVal, Alloca); + + // Remember the old variable binding so that we can restore the binding when + // we unrecurse. + OldBindings.push_back(NamedValues[VarName]); + + // Remember this binding. + NamedValues[VarName] = Alloca; + } + + // Codegen the body, now that all vars are in scope. + Value *BodyVal = Body->Codegen(); + if (BodyVal == 0) return 0; + + // Pop all our variables from scope. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) + NamedValues[VarNames[i].first] = OldBindings[i]; + + // Return the body computation. + return BodyVal; +} + +Function *PrototypeAST::Codegen() { + // Make the function type: double(double,double) etc. + std::vector Doubles(Args.size(), + Type::getDoubleTy(getGlobalContext())); + FunctionType *FT = FunctionType::get(Type::getDoubleTy(getGlobalContext()), + Doubles, false); + + std::string FnName = MakeLegalFunctionName(Name); + + Module* M = TheHelper->getModuleForNewFunction(); + + Function *F = Function::Create(FT, Function::ExternalLinkage, FnName, M); + + // If F conflicted, there was already something named 'FnName'. If it has a + // body, don't allow redefinition or reextern. + if (F->getName() != FnName) { + // Delete the one we just made and get the existing one. + F->eraseFromParent(); + F = M->getFunction(Name); + + // If F already has a body, reject this. + if (!F->empty()) { + ErrorF("redefinition of function"); + return 0; + } + + // If F took a different number of args, reject. + if (F->arg_size() != Args.size()) { + ErrorF("redefinition of function with different # args"); + return 0; + } + } + + // Set names for all arguments. + unsigned Idx = 0; + for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); + ++AI, ++Idx) + AI->setName(Args[Idx]); + + return F; +} + +/// CreateArgumentAllocas - Create an alloca for each argument and register the +/// argument in the symbol table so that references to it will succeed. +void PrototypeAST::CreateArgumentAllocas(Function *F) { + Function::arg_iterator AI = F->arg_begin(); + for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) { + // Create an alloca for this variable. + AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]); + + // Store the initial value into the alloca. + Builder.CreateStore(AI, Alloca); + + // Add arguments to variable symbol table. + NamedValues[Args[Idx]] = Alloca; + } +} + +Function *FunctionAST::Codegen() { + NamedValues.clear(); + + Function *TheFunction = Proto->Codegen(); + if (TheFunction == 0) + return 0; + + // If this is an operator, install it. + if (Proto->isBinaryOp()) + BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); + + // Create a new basic block to start insertion into. + BasicBlock *BB = BasicBlock::Create(getGlobalContext(), "entry", TheFunction); + Builder.SetInsertPoint(BB); + + // Add all arguments to the symbol table and create their allocas. + Proto->CreateArgumentAllocas(TheFunction); + + if (Value *RetVal = Body->Codegen()) { + // Finish off the function. + Builder.CreateRet(RetVal); + + // Validate the generated code, checking for consistency. + verifyFunction(*TheFunction); + + return TheFunction; + } + + // Error reading body, remove function. + TheFunction->eraseFromParent(); + + if (Proto->isBinaryOp()) + BinopPrecedence.erase(Proto->getOperatorName()); + return 0; +} + +//===----------------------------------------------------------------------===// +// Top-Level parsing and JIT Driver +//===----------------------------------------------------------------------===// + +static void HandleDefinition() { + if (FunctionAST *F = ParseDefinition()) { + TheHelper->closeCurrentModule(); + if (Function *LF = F->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read function definition:"); + LF->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleExtern() { + if (PrototypeAST *P = ParseExtern()) { + if (Function *F = P->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read extern: "); + F->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleTopLevelExpression() { + // Evaluate a top-level expression into an anonymous function. + if (FunctionAST *F = ParseTopLevelExpr()) { + if (Function *LF = F->Codegen()) { + // JIT the function, returning a function pointer. + void *FPtr = TheHelper->getPointerToFunction(LF); + + // Cast it to the right type (takes no arguments, returns a double) so we + // can call it as a native function. + double (*FP)() = (double (*)())(intptr_t)FPtr; +#ifdef MINIMAL_STDERR_OUTPUT + FP(); +#else + fprintf(stderr, "Evaluated to %f\n", FP()); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +/// top ::= definition | external | expression | ';' +static void MainLoop() { + while (1) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + switch (CurTok) { + case tok_eof: return; + case ';': getNextToken(); break; // ignore top-level semicolons. + case tok_def: HandleDefinition(); break; + case tok_extern: HandleExtern(); break; + default: HandleTopLevelExpression(); break; + } + } +} + +//===----------------------------------------------------------------------===// +// "Library" functions that can be "extern'd" from user code. +//===----------------------------------------------------------------------===// + +/// putchard - putchar that takes a double and returns 0. +extern "C" +double putchard(double X) { + putchar((char)X); + return 0; +} + +/// printd - printf that takes a double prints it as "%f\n", returning 0. +extern "C" +double printd(double X) { + printf("%f", X); + return 0; +} + +extern "C" +double printlf() { + printf("\n"); + return 0; +} + +//===----------------------------------------------------------------------===// +// Command line input file handler +//===----------------------------------------------------------------------===// + +Module* parseInputIR(std::string InputFile) { + SMDiagnostic Err; + Module *M = ParseIRFile(InputFile, Err, getGlobalContext()); + if (!M) { + Err.print("IR parsing failed: ", errs()); + return NULL; + } + + char ModID[256]; + sprintf(ModID, "IR:%s", InputFile.c_str()); + M->setModuleIdentifier(ModID); + + TheHelper->addModule(M); + return M; +} + +//===----------------------------------------------------------------------===// +// Main driver code. +//===----------------------------------------------------------------------===// + +int main(int argc, char **argv) { + InitializeNativeTarget(); + InitializeNativeTargetAsmPrinter(); + InitializeNativeTargetAsmParser(); + LLVMContext &Context = getGlobalContext(); + + cl::ParseCommandLineOptions(argc, argv, + "Kaleidoscope example program\n"); + + // Install standard binary operators. + // 1 is lowest precedence. + BinopPrecedence['='] = 2; + BinopPrecedence['<'] = 10; + BinopPrecedence['+'] = 20; + BinopPrecedence['-'] = 20; + BinopPrecedence['/'] = 40; + BinopPrecedence['*'] = 40; // highest. + + // Prime the first token. +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + getNextToken(); + + // Make the helper, which holds all the code. + TheHelper = new MCJITHelper(Context); + + if (!InputIR.empty()) { + parseInputIR(InputIR); + } + + // Run the main "interpreter loop" now. + MainLoop(); + +#ifndef MINIMAL_STDERR_OUTPUT + // Print out all of the generated code. + TheHelper->dump(); +#endif + + return 0; +} diff --git a/examples/Kaleidoscope/MCJIT/complete/Makefile b/examples/Kaleidoscope/MCJIT/complete/Makefile new file mode 100644 index 00000000000..9e45d17a69f --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/complete/Makefile @@ -0,0 +1,4 @@ +all: toy + +toy : toy.cpp + clang++ toy.cpp -g -O3 -rdynamic -fno-rtti `llvm-config --cppflags --ldflags --libs core jit mcjit native irreader` -o toy diff --git a/examples/Kaleidoscope/MCJIT/complete/README.txt b/examples/Kaleidoscope/MCJIT/complete/README.txt new file mode 100644 index 00000000000..82bc397a4eb --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/complete/README.txt @@ -0,0 +1,25 @@ +//===----------------------------------------------------------------------===/ +// Kaleidoscope with MCJIT +//===----------------------------------------------------------------------===// + +The files in this directory are meant to accompany the first in a series of +three blog posts that describe the process of porting the Kaleidoscope tutorial +to use the MCJIT execution engine instead of the older JIT engine. + +When the blog post is ready this file will be updated with a link to the post. + +The source code in this directory combines all previous versions, including the +old JIT-based implementation, into a single file for easy comparison with +command line options to select between the various possibilities. + +This directory contain a Makefile that allow the code to be built in a +standalone manner, independent of the larger LLVM build infrastructure. To build +the program you will need to have 'clang++' and 'llvm-config' in your path. If +you attempt to build using the LLVM 3.3 release, some minor modifications will +be required. + +This directory also contains a Python script that may be used to generate random +input for the program and test scripts to capture data for rough performance +comparisons. Another Python script will split generated input files into +definitions and function calls for the purpose of testing the IR input and +caching facilities. \ No newline at end of file diff --git a/examples/Kaleidoscope/MCJIT/complete/genk-timing.py b/examples/Kaleidoscope/MCJIT/complete/genk-timing.py new file mode 100644 index 00000000000..72591fe8bd4 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/complete/genk-timing.py @@ -0,0 +1,224 @@ +#!/usr/bin/env python + +import sys +import random + +class TimingScriptGenerator: + """Used to generate a bash script which will invoke the toy and time it""" + def __init__(self, scriptname, outputname): + self.timeFile = outputname + self.shfile = open(scriptname, 'w') + self.shfile.write("echo \"\" > %s\n" % self.timeFile) + + def writeTimingCall(self, filename, numFuncs, funcsCalled, totalCalls): + """Echo some comments and invoke both versions of toy""" + rootname = filename + if '.' in filename: + rootname = filename[:filename.rfind('.')] + self.shfile.write("echo \"%s: Calls %d of %d functions, %d total\" >> %s\n" % (filename, funcsCalled, numFuncs, totalCalls, self.timeFile)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT (original)\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy -suppress-prompts -use-mcjit=true -enable-lazy-compilation=false < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT (lazy)\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy -suppress-prompts -use-mcjit=true -enable-lazy-compilation=true < %s > %s-mcjit-lazy.out 2> %s-mcjit-lazy.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With JIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy -suppress-prompts -use-mcjit=false < %s > %s-jit.out 2> %s-jit.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + +class KScriptGenerator: + """Used to generate random Kaleidoscope code""" + def __init__(self, filename): + self.kfile = open(filename, 'w') + self.nextFuncNum = 1 + self.lastFuncNum = None + self.callWeighting = 0.1 + # A mapping of calls within functions with no duplicates + self.calledFunctionTable = {} + # A list of function calls which will actually be executed + self.calledFunctions = [] + # A comprehensive mapping of calls within functions + # used for computing the total number of calls + self.comprehensiveCalledFunctionTable = {} + self.totalCallsExecuted = 0 + + def updateTotalCallCount(self, callee): + # Count this call + self.totalCallsExecuted += 1 + # Then count all the functions it calls + if callee in self.comprehensiveCalledFunctionTable: + for child in self.comprehensiveCalledFunctionTable[callee]: + self.updateTotalCallCount(child) + + def updateFunctionCallMap(self, caller, callee): + """Maintains a map of functions that are called from other functions""" + if not caller in self.calledFunctionTable: + self.calledFunctionTable[caller] = [] + if not callee in self.calledFunctionTable[caller]: + self.calledFunctionTable[caller].append(callee) + if not caller in self.comprehensiveCalledFunctionTable: + self.comprehensiveCalledFunctionTable[caller] = [] + self.comprehensiveCalledFunctionTable[caller].append(callee) + + def updateCalledFunctionList(self, callee): + """Maintains a list of functions that will actually be called""" + # Update the total call count + self.updateTotalCallCount(callee) + # If this function is already in the list, don't do anything else + if callee in self.calledFunctions: + return + # Add this function to the list of those that will be called. + self.calledFunctions.append(callee) + # If this function calls other functions, add them too + if callee in self.calledFunctionTable: + for subCallee in self.calledFunctionTable[callee]: + self.updateCalledFunctionList(subCallee) + + def setCallWeighting(self, weight): + """ Sets the probably of generating a function call""" + self.callWeighting = weight + + def writeln(self, line): + self.kfile.write(line + '\n') + + def writeComment(self, comment): + self.writeln('# ' + comment) + + def writeEmptyLine(self): + self.writeln("") + + def writePredefinedFunctions(self): + self.writeComment("Define ':' for sequencing: as a low-precedence operator that ignores operands") + self.writeComment("and just returns the RHS.") + self.writeln("def binary : 1 (x y) y;") + self.writeEmptyLine() + self.writeComment("Helper functions defined within toy") + self.writeln("extern putchard(x);") + self.writeln("extern printd(d);") + self.writeln("extern printlf();") + self.writeEmptyLine() + self.writeComment("Print the result of a function call") + self.writeln("def printresult(N Result)") + self.writeln(" # 'result('") + self.writeln(" putchard(114) : putchard(101) : putchard(115) : putchard(117) : putchard(108) : putchard(116) : putchard(40) :") + self.writeln(" printd(N) :"); + self.writeln(" # ') = '") + self.writeln(" putchard(41) : putchard(32) : putchard(61) : putchard(32) :") + self.writeln(" printd(Result) :"); + self.writeln(" printlf();") + self.writeEmptyLine() + + def writeRandomOperation(self, LValue, LHS, RHS): + shouldCallFunc = (self.lastFuncNum > 2 and random.random() < self.callWeighting) + if shouldCallFunc: + funcToCall = random.randrange(1, self.lastFuncNum - 1) + self.updateFunctionCallMap(self.lastFuncNum, funcToCall) + self.writeln(" %s = func%d(%s, %s) :" % (LValue, funcToCall, LHS, RHS)) + else: + possibleOperations = ["+", "-", "*", "/"] + operation = random.choice(possibleOperations) + if operation == "-": + # Don't let our intermediate value become zero + # This is complicated by the fact that '<' is our only comparison operator + self.writeln(" if %s < %s then" % (LHS, RHS)) + self.writeln(" %s = %s %s %s" % (LValue, LHS, operation, RHS)) + self.writeln(" else if %s < %s then" % (RHS, LHS)) + self.writeln(" %s = %s %s %s" % (LValue, LHS, operation, RHS)) + self.writeln(" else") + self.writeln(" %s = %s %s %f :" % (LValue, LHS, operation, random.uniform(1, 100))) + else: + self.writeln(" %s = %s %s %s :" % (LValue, LHS, operation, RHS)) + + def getNextFuncNum(self): + result = self.nextFuncNum + self.nextFuncNum += 1 + self.lastFuncNum = result + return result + + def writeFunction(self, elements): + funcNum = self.getNextFuncNum() + self.writeComment("Auto-generated function number %d" % funcNum) + self.writeln("def func%d(X Y)" % funcNum) + self.writeln(" var temp1 = X,") + self.writeln(" temp2 = Y,") + self.writeln(" temp3 in") + # Initialize the variable names to be rotated + first = "temp3" + second = "temp1" + third = "temp2" + # Write some random operations + for i in range(elements): + self.writeRandomOperation(first, second, third) + # Rotate the variables + temp = first + first = second + second = third + third = temp + self.writeln(" " + third + ";") + self.writeEmptyLine() + + def writeFunctionCall(self): + self.writeComment("Call the last function") + arg1 = random.uniform(1, 100) + arg2 = random.uniform(1, 100) + self.writeln("printresult(%d, func%d(%f, %f) )" % (self.lastFuncNum, self.lastFuncNum, arg1, arg2)) + self.writeEmptyLine() + self.updateCalledFunctionList(self.lastFuncNum) + + def writeFinalFunctionCounts(self): + self.writeComment("Called %d of %d functions" % (len(self.calledFunctions), self.lastFuncNum)) + +def generateKScript(filename, numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting, timingScript): + """ Generate a random Kaleidoscope script based on the given parameters """ + print "Generating " + filename + print(" %d functions, %d elements per function, %d functions between execution" % + (numFuncs, elementsPerFunc, funcsBetweenExec)) + print(" Call weighting = %f" % callWeighting) + script = KScriptGenerator(filename) + script.setCallWeighting(callWeighting) + script.writeComment("===========================================================================") + script.writeComment("Auto-generated script") + script.writeComment(" %d functions, %d elements per function, %d functions between execution" + % (numFuncs, elementsPerFunc, funcsBetweenExec)) + script.writeComment(" call weighting = %f" % callWeighting) + script.writeComment("===========================================================================") + script.writeEmptyLine() + script.writePredefinedFunctions() + funcsSinceLastExec = 0 + for i in range(numFuncs): + script.writeFunction(elementsPerFunc) + funcsSinceLastExec += 1 + if funcsSinceLastExec == funcsBetweenExec: + script.writeFunctionCall() + funcsSinceLastExec = 0 + # Always end with a function call + if funcsSinceLastExec > 0: + script.writeFunctionCall() + script.writeEmptyLine() + script.writeFinalFunctionCounts() + funcsCalled = len(script.calledFunctions) + print " Called %d of %d functions, %d total" % (funcsCalled, numFuncs, script.totalCallsExecuted) + timingScript.writeTimingCall(filename, numFuncs, funcsCalled, script.totalCallsExecuted) + +# Execution begins here +random.seed() + +timingScript = TimingScriptGenerator("time-toy.sh", "timing-data.txt") + +dataSets = [(5000, 3, 50, 0.50), (5000, 10, 100, 0.10), (5000, 10, 5, 0.10), (5000, 10, 1, 0.0), + (1000, 3, 10, 0.50), (1000, 10, 100, 0.10), (1000, 10, 5, 0.10), (1000, 10, 1, 0.0), + ( 200, 3, 2, 0.50), ( 200, 10, 40, 0.10), ( 200, 10, 2, 0.10), ( 200, 10, 1, 0.0)] + +# Generate the code +for (numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting) in dataSets: + filename = "test-%d-%d-%d-%d.k" % (numFuncs, elementsPerFunc, funcsBetweenExec, int(callWeighting * 100)) + generateKScript(filename, numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting, timingScript) +print "All done!" diff --git a/examples/Kaleidoscope/MCJIT/complete/split-lib.py b/examples/Kaleidoscope/MCJIT/complete/split-lib.py new file mode 100644 index 00000000000..f6bec0276d8 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/complete/split-lib.py @@ -0,0 +1,70 @@ +#!/usr/bin/env python + +class TimingScriptGenerator: + """Used to generate a bash script which will invoke the toy and time it""" + def __init__(self, scriptname, outputname): + self.shfile = open(scriptname, 'w') + self.timeFile = outputname + self.shfile.write("echo \"\" > %s\n" % self.timeFile) + + def writeTimingCall(self, irname, callname): + """Echo some comments and invoke both versions of toy""" + rootname = irname + if '.' in irname: + rootname = irname[:irname.rfind('.')] + self.shfile.write("echo \"%s: Calls %s\" >> %s\n" % (callname, irname, self.timeFile)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy -suppress-prompts -use-mcjit=true -enable-lazy-compilation=true -use-object-cache -input-IR=%s < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (irname, callname, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT again\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy -suppress-prompts -use-mcjit=true -enable-lazy-compilation=true -use-object-cache -input-IR=%s < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (irname, callname, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With JIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy -suppress-prompts -use-mcjit=false -input-IR=%s < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (irname, callname, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + +class LibScriptGenerator: + """Used to generate a bash script which will invoke the toy and time it""" + def __init__(self, filename): + self.shfile = open(filename, 'w') + + def writeLibGenCall(self, libname, irname): + self.shfile.write("./toy -suppress-prompts -use-mcjit=false -dump-modules < %s 2> %s\n" % (libname, irname)) + +def splitScript(inputname, libGenScript, timingScript): + rootname = inputname[:-2] + libname = rootname + "-lib.k" + irname = rootname + "-lib.ir" + callname = rootname + "-call.k" + infile = open(inputname, "r") + libfile = open(libname, "w") + callfile = open(callname, "w") + print "Splitting %s into %s and %s" % (inputname, callname, libname) + for line in infile: + if not line.startswith("#"): + if line.startswith("print"): + callfile.write(line) + else: + libfile.write(line) + libGenScript.writeLibGenCall(libname, irname) + timingScript.writeTimingCall(irname, callname) + +# Execution begins here +libGenScript = LibScriptGenerator("make-libs.sh") +timingScript = TimingScriptGenerator("time-lib.sh", "lib-timing.txt") + +script_list = ["test-5000-3-50-50.k", "test-5000-10-100-10.k", "test-5000-10-5-10.k", "test-5000-10-1-0.k", + "test-1000-3-10-50.k", "test-1000-10-100-10.k", "test-1000-10-5-10.k", "test-1000-10-1-0.k", + "test-200-3-2-50.k", "test-200-10-40-10.k", "test-200-10-2-10.k", "test-200-10-1-0.k"] + +for script in script_list: + splitScript(script, libGenScript, timingScript) +print "All done!" diff --git a/examples/Kaleidoscope/MCJIT/complete/toy.cpp b/examples/Kaleidoscope/MCJIT/complete/toy.cpp new file mode 100644 index 00000000000..da3f3b17693 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/complete/toy.cpp @@ -0,0 +1,1710 @@ +#include "llvm/Analysis/Passes.h" +#include "llvm/Analysis/Verifier.h" +#include "llvm/ExecutionEngine/ExecutionEngine.h" +#include "llvm/ExecutionEngine/JIT.h" +#include "llvm/ExecutionEngine/MCJIT.h" +#include "llvm/ExecutionEngine/ObjectCache.h" +#include "llvm/ExecutionEngine/SectionMemoryManager.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/IRReader/IRReader.h" +#include "llvm/PassManager.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/FileSystem.h" +#include "llvm/Support/Path.h" +#include "llvm/Support/raw_ostream.h" +#include "llvm/Support/SourceMgr.h" +#include "llvm/Support/TargetSelect.h" +#include "llvm/Transforms/Scalar.h" +#include +#include +#include +#include + +using namespace llvm; + +//===----------------------------------------------------------------------===// +// Command-line options +//===----------------------------------------------------------------------===// + +namespace { + cl::opt + InputIR("input-IR", + cl::desc("Specify the name of an IR file to load for function definitions"), + cl::value_desc("input IR file name")); + + cl::opt + VerboseOutput("verbose", + cl::desc("Enable verbose output (results, IR, etc.) to stderr"), + cl::init(false)); + + cl::opt + SuppressPrompts("suppress-prompts", + cl::desc("Disable printing the 'ready' prompt"), + cl::init(false)); + + cl::opt + DumpModulesOnExit("dump-modules", + cl::desc("Dump IR from modules to stderr on shutdown"), + cl::init(false)); + + cl::opt UseMCJIT( + "use-mcjit", cl::desc("Use the MCJIT execution engine"), + cl::init(true)); + + cl::opt EnableLazyCompilation( + "enable-lazy-compilation", cl::desc("Enable lazy compilation when using the MCJIT engine"), + cl::init(true)); + + cl::opt UseObjectCache( + "use-object-cache", cl::desc("Enable use of the MCJIT object caching"), + cl::init(false)); +} // namespace + +//===----------------------------------------------------------------------===// +// Lexer +//===----------------------------------------------------------------------===// + +// The lexer returns tokens [0-255] if it is an unknown character, otherwise one +// of these for known things. +enum Token { + tok_eof = -1, + + // commands + tok_def = -2, tok_extern = -3, + + // primary + tok_identifier = -4, tok_number = -5, + + // control + tok_if = -6, tok_then = -7, tok_else = -8, + tok_for = -9, tok_in = -10, + + // operators + tok_binary = -11, tok_unary = -12, + + // var definition + tok_var = -13 +}; + +static std::string IdentifierStr; // Filled in if tok_identifier +static double NumVal; // Filled in if tok_number + +/// gettok - Return the next token from standard input. +static int gettok() { + static int LastChar = ' '; + + // Skip any whitespace. + while (isspace(LastChar)) + LastChar = getchar(); + + if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* + IdentifierStr = LastChar; + while (isalnum((LastChar = getchar()))) + IdentifierStr += LastChar; + + if (IdentifierStr == "def") return tok_def; + if (IdentifierStr == "extern") return tok_extern; + if (IdentifierStr == "if") return tok_if; + if (IdentifierStr == "then") return tok_then; + if (IdentifierStr == "else") return tok_else; + if (IdentifierStr == "for") return tok_for; + if (IdentifierStr == "in") return tok_in; + if (IdentifierStr == "binary") return tok_binary; + if (IdentifierStr == "unary") return tok_unary; + if (IdentifierStr == "var") return tok_var; + return tok_identifier; + } + + if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ + std::string NumStr; + do { + NumStr += LastChar; + LastChar = getchar(); + } while (isdigit(LastChar) || LastChar == '.'); + + NumVal = strtod(NumStr.c_str(), 0); + return tok_number; + } + + if (LastChar == '#') { + // Comment until end of line. + do LastChar = getchar(); + while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); + + if (LastChar != EOF) + return gettok(); + } + + // Check for end of file. Don't eat the EOF. + if (LastChar == EOF) + return tok_eof; + + // Otherwise, just return the character as its ascii value. + int ThisChar = LastChar; + LastChar = getchar(); + return ThisChar; +} + +//===----------------------------------------------------------------------===// +// Abstract Syntax Tree (aka Parse Tree) +//===----------------------------------------------------------------------===// + +/// ExprAST - Base class for all expression nodes. +class ExprAST { +public: + virtual ~ExprAST() {} + virtual Value *Codegen() = 0; +}; + +/// NumberExprAST - Expression class for numeric literals like "1.0". +class NumberExprAST : public ExprAST { + double Val; +public: + NumberExprAST(double val) : Val(val) {} + virtual Value *Codegen(); +}; + +/// VariableExprAST - Expression class for referencing a variable, like "a". +class VariableExprAST : public ExprAST { + std::string Name; +public: + VariableExprAST(const std::string &name) : Name(name) {} + const std::string &getName() const { return Name; } + virtual Value *Codegen(); +}; + +/// UnaryExprAST - Expression class for a unary operator. +class UnaryExprAST : public ExprAST { + char Opcode; + ExprAST *Operand; +public: + UnaryExprAST(char opcode, ExprAST *operand) + : Opcode(opcode), Operand(operand) {} + virtual Value *Codegen(); +}; + +/// BinaryExprAST - Expression class for a binary operator. +class BinaryExprAST : public ExprAST { + char Op; + ExprAST *LHS, *RHS; +public: + BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) + : Op(op), LHS(lhs), RHS(rhs) {} + virtual Value *Codegen(); +}; + +/// CallExprAST - Expression class for function calls. +class CallExprAST : public ExprAST { + std::string Callee; + std::vector Args; +public: + CallExprAST(const std::string &callee, std::vector &args) + : Callee(callee), Args(args) {} + virtual Value *Codegen(); +}; + +/// IfExprAST - Expression class for if/then/else. +class IfExprAST : public ExprAST { + ExprAST *Cond, *Then, *Else; +public: + IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) + : Cond(cond), Then(then), Else(_else) {} + virtual Value *Codegen(); +}; + +/// ForExprAST - Expression class for for/in. +class ForExprAST : public ExprAST { + std::string VarName; + ExprAST *Start, *End, *Step, *Body; +public: + ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, + ExprAST *step, ExprAST *body) + : VarName(varname), Start(start), End(end), Step(step), Body(body) {} + virtual Value *Codegen(); +}; + +/// VarExprAST - Expression class for var/in +class VarExprAST : public ExprAST { + std::vector > VarNames; + ExprAST *Body; +public: + VarExprAST(const std::vector > &varnames, + ExprAST *body) + : VarNames(varnames), Body(body) {} + + virtual Value *Codegen(); +}; + +/// PrototypeAST - This class represents the "prototype" for a function, +/// which captures its argument names as well as if it is an operator. +class PrototypeAST { + std::string Name; + std::vector Args; + bool isOperator; + unsigned Precedence; // Precedence if a binary op. +public: + PrototypeAST(const std::string &name, const std::vector &args, + bool isoperator = false, unsigned prec = 0) + : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} + + bool isUnaryOp() const { return isOperator && Args.size() == 1; } + bool isBinaryOp() const { return isOperator && Args.size() == 2; } + + char getOperatorName() const { + assert(isUnaryOp() || isBinaryOp()); + return Name[Name.size()-1]; + } + + unsigned getBinaryPrecedence() const { return Precedence; } + + Function *Codegen(); + + void CreateArgumentAllocas(Function *F); +}; + +/// FunctionAST - This class represents a function definition itself. +class FunctionAST { + PrototypeAST *Proto; + ExprAST *Body; +public: + FunctionAST(PrototypeAST *proto, ExprAST *body) + : Proto(proto), Body(body) {} + + Function *Codegen(); +}; + +//===----------------------------------------------------------------------===// +// Parser +//===----------------------------------------------------------------------===// + +/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current +/// token the parser is looking at. getNextToken reads another token from the +/// lexer and updates CurTok with its results. +static int CurTok; +static int getNextToken() { + return CurTok = gettok(); +} + +/// BinopPrecedence - This holds the precedence for each binary operator that is +/// defined. +static std::map BinopPrecedence; + +/// GetTokPrecedence - Get the precedence of the pending binary operator token. +static int GetTokPrecedence() { + if (!isascii(CurTok)) + return -1; + + // Make sure it's a declared binop. + int TokPrec = BinopPrecedence[CurTok]; + if (TokPrec <= 0) return -1; + return TokPrec; +} + +/// Error* - These are little helper functions for error handling. +ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} +PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } +FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } + +static ExprAST *ParseExpression(); + +/// identifierexpr +/// ::= identifier +/// ::= identifier '(' expression* ')' +static ExprAST *ParseIdentifierExpr() { + std::string IdName = IdentifierStr; + + getNextToken(); // eat identifier. + + if (CurTok != '(') // Simple variable ref. + return new VariableExprAST(IdName); + + // Call. + getNextToken(); // eat ( + std::vector Args; + if (CurTok != ')') { + while (1) { + ExprAST *Arg = ParseExpression(); + if (!Arg) return 0; + Args.push_back(Arg); + + if (CurTok == ')') break; + + if (CurTok != ',') + return Error("Expected ')' or ',' in argument list"); + getNextToken(); + } + } + + // Eat the ')'. + getNextToken(); + + return new CallExprAST(IdName, Args); +} + +/// numberexpr ::= number +static ExprAST *ParseNumberExpr() { + ExprAST *Result = new NumberExprAST(NumVal); + getNextToken(); // consume the number + return Result; +} + +/// parenexpr ::= '(' expression ')' +static ExprAST *ParseParenExpr() { + getNextToken(); // eat (. + ExprAST *V = ParseExpression(); + if (!V) return 0; + + if (CurTok != ')') + return Error("expected ')'"); + getNextToken(); // eat ). + return V; +} + +/// ifexpr ::= 'if' expression 'then' expression 'else' expression +static ExprAST *ParseIfExpr() { + getNextToken(); // eat the if. + + // condition. + ExprAST *Cond = ParseExpression(); + if (!Cond) return 0; + + if (CurTok != tok_then) + return Error("expected then"); + getNextToken(); // eat the then + + ExprAST *Then = ParseExpression(); + if (Then == 0) return 0; + + if (CurTok != tok_else) + return Error("expected else"); + + getNextToken(); + + ExprAST *Else = ParseExpression(); + if (!Else) return 0; + + return new IfExprAST(Cond, Then, Else); +} + +/// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression +static ExprAST *ParseForExpr() { + getNextToken(); // eat the for. + + if (CurTok != tok_identifier) + return Error("expected identifier after for"); + + std::string IdName = IdentifierStr; + getNextToken(); // eat identifier. + + if (CurTok != '=') + return Error("expected '=' after for"); + getNextToken(); // eat '='. + + + ExprAST *Start = ParseExpression(); + if (Start == 0) return 0; + if (CurTok != ',') + return Error("expected ',' after for start value"); + getNextToken(); + + ExprAST *End = ParseExpression(); + if (End == 0) return 0; + + // The step value is optional. + ExprAST *Step = 0; + if (CurTok == ',') { + getNextToken(); + Step = ParseExpression(); + if (Step == 0) return 0; + } + + if (CurTok != tok_in) + return Error("expected 'in' after for"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new ForExprAST(IdName, Start, End, Step, Body); +} + +/// varexpr ::= 'var' identifier ('=' expression)? +// (',' identifier ('=' expression)?)* 'in' expression +static ExprAST *ParseVarExpr() { + getNextToken(); // eat the var. + + std::vector > VarNames; + + // At least one variable name is required. + if (CurTok != tok_identifier) + return Error("expected identifier after var"); + + while (1) { + std::string Name = IdentifierStr; + getNextToken(); // eat identifier. + + // Read the optional initializer. + ExprAST *Init = 0; + if (CurTok == '=') { + getNextToken(); // eat the '='. + + Init = ParseExpression(); + if (Init == 0) return 0; + } + + VarNames.push_back(std::make_pair(Name, Init)); + + // End of var list, exit loop. + if (CurTok != ',') break; + getNextToken(); // eat the ','. + + if (CurTok != tok_identifier) + return Error("expected identifier list after var"); + } + + // At this point, we have to have 'in'. + if (CurTok != tok_in) + return Error("expected 'in' keyword after 'var'"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new VarExprAST(VarNames, Body); +} + +/// primary +/// ::= identifierexpr +/// ::= numberexpr +/// ::= parenexpr +/// ::= ifexpr +/// ::= forexpr +/// ::= varexpr +static ExprAST *ParsePrimary() { + switch (CurTok) { + default: return Error("unknown token when expecting an expression"); + case tok_identifier: return ParseIdentifierExpr(); + case tok_number: return ParseNumberExpr(); + case '(': return ParseParenExpr(); + case tok_if: return ParseIfExpr(); + case tok_for: return ParseForExpr(); + case tok_var: return ParseVarExpr(); + } +} + +/// unary +/// ::= primary +/// ::= '!' unary +static ExprAST *ParseUnary() { + // If the current token is not an operator, it must be a primary expr. + if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') + return ParsePrimary(); + + // If this is a unary operator, read it. + int Opc = CurTok; + getNextToken(); + if (ExprAST *Operand = ParseUnary()) + return new UnaryExprAST(Opc, Operand); + return 0; +} + +/// binoprhs +/// ::= ('+' unary)* +static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { + // If this is a binop, find its precedence. + while (1) { + int TokPrec = GetTokPrecedence(); + + // If this is a binop that binds at least as tightly as the current binop, + // consume it, otherwise we are done. + if (TokPrec < ExprPrec) + return LHS; + + // Okay, we know this is a binop. + int BinOp = CurTok; + getNextToken(); // eat binop + + // Parse the unary expression after the binary operator. + ExprAST *RHS = ParseUnary(); + if (!RHS) return 0; + + // If BinOp binds less tightly with RHS than the operator after RHS, let + // the pending operator take RHS as its LHS. + int NextPrec = GetTokPrecedence(); + if (TokPrec < NextPrec) { + RHS = ParseBinOpRHS(TokPrec+1, RHS); + if (RHS == 0) return 0; + } + + // Merge LHS/RHS. + LHS = new BinaryExprAST(BinOp, LHS, RHS); + } +} + +/// expression +/// ::= unary binoprhs +/// +static ExprAST *ParseExpression() { + ExprAST *LHS = ParseUnary(); + if (!LHS) return 0; + + return ParseBinOpRHS(0, LHS); +} + +/// prototype +/// ::= id '(' id* ')' +/// ::= binary LETTER number? (id, id) +/// ::= unary LETTER (id) +static PrototypeAST *ParsePrototype() { + std::string FnName; + + unsigned Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. + unsigned BinaryPrecedence = 30; + + switch (CurTok) { + default: + return ErrorP("Expected function name in prototype"); + case tok_identifier: + FnName = IdentifierStr; + Kind = 0; + getNextToken(); + break; + case tok_unary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected unary operator"); + FnName = "unary"; + FnName += (char)CurTok; + Kind = 1; + getNextToken(); + break; + case tok_binary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected binary operator"); + FnName = "binary"; + FnName += (char)CurTok; + Kind = 2; + getNextToken(); + + // Read the precedence if present. + if (CurTok == tok_number) { + if (NumVal < 1 || NumVal > 100) + return ErrorP("Invalid precedecnce: must be 1..100"); + BinaryPrecedence = (unsigned)NumVal; + getNextToken(); + } + break; + } + + if (CurTok != '(') + return ErrorP("Expected '(' in prototype"); + + std::vector ArgNames; + while (getNextToken() == tok_identifier) + ArgNames.push_back(IdentifierStr); + if (CurTok != ')') + return ErrorP("Expected ')' in prototype"); + + // success. + getNextToken(); // eat ')'. + + // Verify right number of names for operator. + if (Kind && ArgNames.size() != Kind) + return ErrorP("Invalid number of operands for operator"); + + return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); +} + +/// definition ::= 'def' prototype expression +static FunctionAST *ParseDefinition() { + getNextToken(); // eat def. + PrototypeAST *Proto = ParsePrototype(); + if (Proto == 0) return 0; + + if (ExprAST *E = ParseExpression()) + return new FunctionAST(Proto, E); + return 0; +} + +/// toplevelexpr ::= expression +static FunctionAST *ParseTopLevelExpr() { + if (ExprAST *E = ParseExpression()) { + // Make an anonymous proto. + PrototypeAST *Proto = new PrototypeAST("", std::vector()); + return new FunctionAST(Proto, E); + } + return 0; +} + +/// external ::= 'extern' prototype +static PrototypeAST *ParseExtern() { + getNextToken(); // eat extern. + return ParsePrototype(); +} + +//===----------------------------------------------------------------------===// +// Quick and dirty hack +//===----------------------------------------------------------------------===// + +// FIXME: Obviously we can do better than this +std::string GenerateUniqueName(const char *root) +{ + static int i = 0; + char s[16]; + sprintf(s, "%s%d", root, i++); + std::string S = s; + return S; +} + +std::string MakeLegalFunctionName(std::string Name) +{ + std::string NewName; + if (!Name.length()) + return GenerateUniqueName("anon_func_"); + + // Start with what we have + NewName = Name; + + // Look for a numberic first character + if (NewName.find_first_of("0123456789") == 0) { + NewName.insert(0, 1, 'n'); + } + + // Replace illegal characters with their ASCII equivalent + std::string legal_elements = "_abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"; + size_t pos; + while ((pos = NewName.find_first_not_of(legal_elements)) != std::string::npos) { + char old_c = NewName.at(pos); + char new_str[16]; + sprintf(new_str, "%d", (int)old_c); + NewName = NewName.replace(pos, 1, new_str); + } + + return NewName; +} + +//===----------------------------------------------------------------------===// +// MCJIT object cache class +//===----------------------------------------------------------------------===// + +class MCJITObjectCache : public ObjectCache { +public: + MCJITObjectCache() { + // Set IR cache directory + sys::fs::current_path(CacheDir); + sys::path::append(CacheDir, "toy_object_cache"); + } + + virtual ~MCJITObjectCache() { + } + + virtual void notifyObjectCompiled(const Module *M, const MemoryBuffer *Obj) { + // Get the ModuleID + const std::string ModuleID = M->getModuleIdentifier(); + + // If we've flagged this as an IR file, cache it + if (0 == ModuleID.compare(0, 3, "IR:")) { + std::string IRFileName = ModuleID.substr(3); + SmallString<128>IRCacheFile = CacheDir; + sys::path::append(IRCacheFile, IRFileName); + if (!sys::fs::exists(CacheDir.str()) && sys::fs::create_directory(CacheDir.str())) { + fprintf(stderr, "Unable to create cache directory\n"); + return; + } + std::string ErrStr; + raw_fd_ostream IRObjectFile(IRCacheFile.c_str(), ErrStr, raw_fd_ostream::F_Binary); + IRObjectFile << Obj->getBuffer(); + } + } + + // MCJIT will call this function before compiling any module + // MCJIT takes ownership of both the MemoryBuffer object and the memory + // to which it refers. + virtual MemoryBuffer* getObject(const Module* M) { + // Get the ModuleID + const std::string ModuleID = M->getModuleIdentifier(); + + // If we've flagged this as an IR file, cache it + if (0 == ModuleID.compare(0, 3, "IR:")) { + std::string IRFileName = ModuleID.substr(3); + SmallString<128> IRCacheFile = CacheDir; + sys::path::append(IRCacheFile, IRFileName); + if (!sys::fs::exists(IRCacheFile.str())) { + // This file isn't in our cache + return NULL; + } + OwningPtr IRObjectBuffer; + MemoryBuffer::getFile(IRCacheFile.c_str(), IRObjectBuffer, -1, false); + // MCJIT will want to write into this buffer, and we don't want that + // because the file has probably just been mmapped. Instead we make + // a copy. The filed-based buffer will be released when it goes + // out of scope. + return MemoryBuffer::getMemBufferCopy(IRObjectBuffer->getBuffer()); + } + + return NULL; + } + +private: + SmallString<128> CacheDir; +}; + +//===----------------------------------------------------------------------===// +// IR input file handler +//===----------------------------------------------------------------------===// + +Module* parseInputIR(std::string InputFile, LLVMContext &Context) { + SMDiagnostic Err; + Module *M = ParseIRFile(InputFile, Err, Context); + if (!M) { + Err.print("IR parsing failed: ", errs()); + return NULL; + } + + char ModID[256]; + sprintf(ModID, "IR:%s", InputFile.c_str()); + M->setModuleIdentifier(ModID); + return M; +} + +//===----------------------------------------------------------------------===// +// Helper class for execution engine abstraction +//===----------------------------------------------------------------------===// + +class BaseHelper +{ +public: + BaseHelper() {} + virtual ~BaseHelper() {} + + virtual Function *getFunction(const std::string FnName) = 0; + virtual Module *getModuleForNewFunction() = 0; + virtual void *getPointerToFunction(Function* F) = 0; + virtual void *getPointerToNamedFunction(const std::string &Name) = 0; + virtual void closeCurrentModule() = 0; + virtual void runFPM(Function &F) = 0; + virtual void dump(); +}; + +//===----------------------------------------------------------------------===// +// Helper class for JIT execution engine +//===----------------------------------------------------------------------===// + +class JITHelper : public BaseHelper { +public: + JITHelper(LLVMContext &Context) { + // Make the module, which holds all the code. + if (!InputIR.empty()) { + TheModule = parseInputIR(InputIR, Context); + } else { + TheModule = new Module("my cool jit", Context); + } + + // Create the JIT. This takes ownership of the module. + std::string ErrStr; + TheExecutionEngine = EngineBuilder(TheModule).setErrorStr(&ErrStr).create(); + if (!TheExecutionEngine) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + TheFPM = new FunctionPassManager(TheModule); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + TheFPM->add(new DataLayout(*TheExecutionEngine->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + TheFPM->add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + TheFPM->add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + TheFPM->add(createInstructionCombiningPass()); + // Reassociate expressions. + TheFPM->add(createReassociatePass()); + // Eliminate Common SubExpressions. + TheFPM->add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + TheFPM->add(createCFGSimplificationPass()); + + TheFPM->doInitialization(); + } + + virtual ~JITHelper() { + if (TheFPM) + delete TheFPM; + if (TheExecutionEngine) + delete TheExecutionEngine; + } + + virtual Function *getFunction(const std::string FnName) { + assert(TheModule); + return TheModule->getFunction(FnName); + } + + virtual Module *getModuleForNewFunction() { + assert(TheModule); + return TheModule; + } + + virtual void *getPointerToFunction(Function* F) { + assert(TheExecutionEngine); + return TheExecutionEngine->getPointerToFunction(F); + } + + virtual void *getPointerToNamedFunction(const std::string &Name) { + return TheExecutionEngine->getPointerToNamedFunction(Name); + } + + virtual void runFPM(Function &F) { + assert(TheFPM); + TheFPM->run(F); + } + + virtual void closeCurrentModule() { + // This should never be called for JIT + assert(false); + } + + virtual void dump() { + assert(TheModule); + TheModule->dump(); + } + +private: + Module *TheModule; + ExecutionEngine *TheExecutionEngine; + FunctionPassManager *TheFPM; +}; + +//===----------------------------------------------------------------------===// +// MCJIT helper class +//===----------------------------------------------------------------------===// + +class MCJITHelper : public BaseHelper +{ +public: + MCJITHelper(LLVMContext& C) : Context(C), CurrentModule(NULL) { + if (!InputIR.empty()) { + Module *M = parseInputIR(InputIR, Context); + Modules.push_back(M); + if (!EnableLazyCompilation) + compileModule(M); + } + } + ~MCJITHelper(); + + Function *getFunction(const std::string FnName); + Module *getModuleForNewFunction(); + void *getPointerToFunction(Function* F); + void *getPointerToNamedFunction(const std::string &Name); + void closeCurrentModule(); + virtual void runFPM(Function &F) {} // Not needed, see compileModule + void dump(); + +protected: + ExecutionEngine *compileModule(Module *M); + +private: + typedef std::vector ModuleVector; + + MCJITObjectCache OurObjectCache; + + LLVMContext &Context; + ModuleVector Modules; + + std::map EngineMap; + + Module *CurrentModule; +}; + +class HelpingMemoryManager : public SectionMemoryManager +{ + HelpingMemoryManager(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + void operator=(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + +public: + HelpingMemoryManager(MCJITHelper *Helper) : MasterHelper(Helper) {} + virtual ~HelpingMemoryManager() {} + + /// This method returns the address of the specified function. + /// Our implementation will attempt to find functions in other + /// modules associated with the MCJITHelper to cross link functions + /// from one generated module to another. + /// + /// If \p AbortOnFailure is false and no function with the given name is + /// found, this function returns a null pointer. Otherwise, it prints a + /// message to stderr and aborts. + virtual void *getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure = true); +private: + MCJITHelper *MasterHelper; +}; + +void *HelpingMemoryManager::getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure) +{ + // Try the standard symbol resolution first, but ask it not to abort. + void *pfn = RTDyldMemoryManager::getPointerToNamedFunction(Name, false); + if (pfn) + return pfn; + + pfn = MasterHelper->getPointerToNamedFunction(Name); + if (!pfn && AbortOnFailure) + report_fatal_error("Program used external function '" + Name + + "' which could not be resolved!"); + return pfn; +} + +MCJITHelper::~MCJITHelper() +{ + // Walk the vector of modules. + ModuleVector::iterator it, end; + for (it = Modules.begin(), end = Modules.end(); + it != end; ++it) { + // See if we have an execution engine for this module. + std::map::iterator mapIt = EngineMap.find(*it); + // If we have an EE, the EE owns the module so just delete the EE. + if (mapIt != EngineMap.end()) { + delete mapIt->second; + } else { + // Otherwise, we still own the module. Delete it now. + delete *it; + } + } +} + +Function *MCJITHelper::getFunction(const std::string FnName) { + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(FnName); + if (F) { + if (*it == CurrentModule) + return F; + + assert(CurrentModule != NULL); + + // This function is in a module that has already been JITed. + // We just need a prototype for external linkage. + Function *PF = CurrentModule->getFunction(FnName); + if (PF && !PF->empty()) { + ErrorF("redefinition of function across modules"); + return 0; + } + + // If we don't have a prototype yet, create one. + if (!PF) + PF = Function::Create(F->getFunctionType(), + Function::ExternalLinkage, + FnName, + CurrentModule); + return PF; + } + } + return NULL; +} + +Module *MCJITHelper::getModuleForNewFunction() { + // If we have a Module that hasn't been JITed, use that. + if (CurrentModule) + return CurrentModule; + + // Otherwise create a new Module. + std::string ModName = GenerateUniqueName("mcjit_module_"); + Module *M = new Module(ModName, Context); + Modules.push_back(M); + CurrentModule = M; + + return M; +} + +ExecutionEngine *MCJITHelper::compileModule(Module *M) { + assert(EngineMap.find(M) == EngineMap.end()); + + if (M == CurrentModule) + closeCurrentModule(); + + std::string ErrStr; + ExecutionEngine *EE = EngineBuilder(M) + .setErrorStr(&ErrStr) + .setUseMCJIT(true) + .setMCJITMemoryManager(new HelpingMemoryManager(this)) + .create(); + if (!EE) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + if (UseObjectCache) + EE->setObjectCache(&OurObjectCache); + // Get the ModuleID so we can identify IR input files + const std::string ModuleID = M->getModuleIdentifier(); + + // If we've flagged this as an IR file, it doesn't need function passes run. + if (0 != ModuleID.compare(0, 3, "IR:")) { + FunctionPassManager *FPM = 0; + + // Create a FPM for this module + FPM = new FunctionPassManager(M); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + FPM->add(new DataLayout(*EE->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + FPM->add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + FPM->add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + FPM->add(createInstructionCombiningPass()); + // Reassociate expressions. + FPM->add(createReassociatePass()); + // Eliminate Common SubExpressions. + FPM->add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + FPM->add(createCFGSimplificationPass()); + + FPM->doInitialization(); + + // For each function in the module + Module::iterator it; + Module::iterator end = M->end(); + for (it = M->begin(); it != end; ++it) { + // Run the FPM on this function + FPM->run(*it); + } + + delete FPM; + } + + EE->finalizeObject(); + + // Store this engine + EngineMap[M] = EE; + + return EE; +} + +void *MCJITHelper::getPointerToFunction(Function* F) { + // Look for this function in an existing module + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + std::string FnName = F->getName(); + for (it = begin; it != end; ++it) { + Function *MF = (*it)->getFunction(FnName); + if (MF == F) { + std::map::iterator eeIt = EngineMap.find(*it); + if (eeIt != EngineMap.end()) { + void *P = eeIt->second->getPointerToFunction(F); + if (P) + return P; + } else { + ExecutionEngine *EE = compileModule(*it); + void *P = EE->getPointerToFunction(F); + if (P) + return P; + } + } + } + return NULL; +} + +void MCJITHelper::closeCurrentModule() { + // If we have an open module (and we should), pack it up + if (CurrentModule) { + CurrentModule = NULL; + } +} + +void *MCJITHelper::getPointerToNamedFunction(const std::string &Name) +{ + // Look for the functions in our modules, compiling only as necessary + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(Name); + if (F && !F->empty()) { + std::map::iterator eeIt = EngineMap.find(*it); + if (eeIt != EngineMap.end()) { + void *P = eeIt->second->getPointerToFunction(F); + if (P) + return P; + } else { + ExecutionEngine *EE = compileModule(*it); + void *P = EE->getPointerToFunction(F); + if (P) + return P; + } + } + } + return NULL; +} + +void MCJITHelper::dump() +{ + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) + (*it)->dump(); +} + +//===----------------------------------------------------------------------===// +// Code Generation +//===----------------------------------------------------------------------===// + +static BaseHelper *TheHelper; +static IRBuilder<> Builder(getGlobalContext()); +static std::map NamedValues; + +Value *ErrorV(const char *Str) { Error(Str); return 0; } + +/// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of +/// the function. This is used for mutable variables etc. +static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction, + const std::string &VarName) { + IRBuilder<> TmpB(&TheFunction->getEntryBlock(), + TheFunction->getEntryBlock().begin()); + return TmpB.CreateAlloca(Type::getDoubleTy(getGlobalContext()), 0, + VarName.c_str()); +} + +Value *NumberExprAST::Codegen() { + return ConstantFP::get(getGlobalContext(), APFloat(Val)); +} + +Value *VariableExprAST::Codegen() { + // Look this variable up in the function. + Value *V = NamedValues[Name]; + if (V == 0) return ErrorV("Unknown variable name"); + + // Load the value. + return Builder.CreateLoad(V, Name.c_str()); +} + +Value *UnaryExprAST::Codegen() { + Value *OperandV = Operand->Codegen(); + if (OperandV == 0) return 0; + Function *F; + if (UseMCJIT) + F = TheHelper->getFunction(MakeLegalFunctionName(std::string("unary")+Opcode)); + else + F = TheHelper->getFunction(std::string("unary")+Opcode); + if (F == 0) + return ErrorV("Unknown unary operator"); + + return Builder.CreateCall(F, OperandV, "unop"); +} + +Value *BinaryExprAST::Codegen() { + // Special case '=' because we don't want to emit the LHS as an expression. + if (Op == '=') { + // Assignment requires the LHS to be an identifier. + // This assume we're building without RTTI because LLVM builds that way by + // default. If you build LLVM with RTTI this can be changed to a + // dynamic_cast for automatic error checking. + VariableExprAST *LHSE = reinterpret_cast(LHS); + if (!LHSE) + return ErrorV("destination of '=' must be a variable"); + // Codegen the RHS. + Value *Val = RHS->Codegen(); + if (Val == 0) return 0; + + // Look up the name. + Value *Variable = NamedValues[LHSE->getName()]; + if (Variable == 0) return ErrorV("Unknown variable name"); + + Builder.CreateStore(Val, Variable); + return Val; + } + + Value *L = LHS->Codegen(); + Value *R = RHS->Codegen(); + if (L == 0 || R == 0) return 0; + + switch (Op) { + case '+': return Builder.CreateFAdd(L, R, "addtmp"); + case '-': return Builder.CreateFSub(L, R, "subtmp"); + case '*': return Builder.CreateFMul(L, R, "multmp"); + case '/': return Builder.CreateFDiv(L, R, "divtmp"); + case '<': + L = Builder.CreateFCmpULT(L, R, "cmptmp"); + // Convert bool 0/1 to double 0.0 or 1.0 + return Builder.CreateUIToFP(L, Type::getDoubleTy(getGlobalContext()), + "booltmp"); + default: break; + } + + // If it wasn't a builtin binary operator, it must be a user defined one. Emit + // a call to it. + Function *F; + if (UseMCJIT) + F = TheHelper->getFunction(MakeLegalFunctionName(std::string("binary")+Op)); + else + F = TheHelper->getFunction(std::string("binary")+Op); + assert(F && "binary operator not found!"); + + Value *Ops[] = { L, R }; + return Builder.CreateCall(F, Ops, "binop"); +} + +Value *CallExprAST::Codegen() { + // Look up the name in the global module table. + Function *CalleeF = TheHelper->getFunction(Callee); + if (CalleeF == 0) { + char error_str[64]; + sprintf(error_str, "Unknown function referenced %s", Callee.c_str()); + return ErrorV(error_str); + } + + // If argument mismatch error. + if (CalleeF->arg_size() != Args.size()) + return ErrorV("Incorrect # arguments passed"); + + std::vector ArgsV; + for (unsigned i = 0, e = Args.size(); i != e; ++i) { + ArgsV.push_back(Args[i]->Codegen()); + if (ArgsV.back() == 0) return 0; + } + + return Builder.CreateCall(CalleeF, ArgsV, "calltmp"); +} + +Value *IfExprAST::Codegen() { + Value *CondV = Cond->Codegen(); + if (CondV == 0) return 0; + + // Convert condition to a bool by comparing equal to 0.0. + CondV = Builder.CreateFCmpONE(CondV, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "ifcond"); + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create blocks for the then and else cases. Insert the 'then' block at the + // end of the function. + BasicBlock *ThenBB = BasicBlock::Create(getGlobalContext(), "then", TheFunction); + BasicBlock *ElseBB = BasicBlock::Create(getGlobalContext(), "else"); + BasicBlock *MergeBB = BasicBlock::Create(getGlobalContext(), "ifcont"); + + Builder.CreateCondBr(CondV, ThenBB, ElseBB); + + // Emit then value. + Builder.SetInsertPoint(ThenBB); + + Value *ThenV = Then->Codegen(); + if (ThenV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Then' can change the current block, update ThenBB for the PHI. + ThenBB = Builder.GetInsertBlock(); + + // Emit else block. + TheFunction->getBasicBlockList().push_back(ElseBB); + Builder.SetInsertPoint(ElseBB); + + Value *ElseV = Else->Codegen(); + if (ElseV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Else' can change the current block, update ElseBB for the PHI. + ElseBB = Builder.GetInsertBlock(); + + // Emit merge block. + TheFunction->getBasicBlockList().push_back(MergeBB); + Builder.SetInsertPoint(MergeBB); + PHINode *PN = Builder.CreatePHI(Type::getDoubleTy(getGlobalContext()), 2, + "iftmp"); + + PN->addIncoming(ThenV, ThenBB); + PN->addIncoming(ElseV, ElseBB); + return PN; +} + +Value *ForExprAST::Codegen() { + // Output this as: + // var = alloca double + // ... + // start = startexpr + // store start -> var + // goto loop + // loop: + // ... + // bodyexpr + // ... + // loopend: + // step = stepexpr + // endcond = endexpr + // + // curvar = load var + // nextvar = curvar + step + // store nextvar -> var + // br endcond, loop, endloop + // outloop: + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create an alloca for the variable in the entry block. + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + + // Emit the start code first, without 'variable' in scope. + Value *StartVal = Start->Codegen(); + if (StartVal == 0) return 0; + + // Store the value into the alloca. + Builder.CreateStore(StartVal, Alloca); + + // Make the new basic block for the loop header, inserting after current + // block. + BasicBlock *LoopBB = BasicBlock::Create(getGlobalContext(), "loop", TheFunction); + + // Insert an explicit fall through from the current block to the LoopBB. + Builder.CreateBr(LoopBB); + + // Start insertion in LoopBB. + Builder.SetInsertPoint(LoopBB); + + // Within the loop, the variable is defined equal to the PHI node. If it + // shadows an existing variable, we have to restore it, so save it now. + AllocaInst *OldVal = NamedValues[VarName]; + NamedValues[VarName] = Alloca; + + // Emit the body of the loop. This, like any other expr, can change the + // current BB. Note that we ignore the value computed by the body, but don't + // allow an error. + if (Body->Codegen() == 0) + return 0; + + // Emit the step value. + Value *StepVal; + if (Step) { + StepVal = Step->Codegen(); + if (StepVal == 0) return 0; + } else { + // If not specified, use 1.0. + StepVal = ConstantFP::get(getGlobalContext(), APFloat(1.0)); + } + + // Compute the end condition. + Value *EndCond = End->Codegen(); + if (EndCond == 0) return EndCond; + + // Reload, increment, and restore the alloca. This handles the case where + // the body of the loop mutates the variable. + Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str()); + Value *NextVar = Builder.CreateFAdd(CurVar, StepVal, "nextvar"); + Builder.CreateStore(NextVar, Alloca); + + // Convert condition to a bool by comparing equal to 0.0. + EndCond = Builder.CreateFCmpONE(EndCond, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "loopcond"); + + // Create the "after loop" block and insert it. + BasicBlock *AfterBB = BasicBlock::Create(getGlobalContext(), "afterloop", TheFunction); + + // Insert the conditional branch into the end of LoopEndBB. + Builder.CreateCondBr(EndCond, LoopBB, AfterBB); + + // Any new code will be inserted in AfterBB. + Builder.SetInsertPoint(AfterBB); + + // Restore the unshadowed variable. + if (OldVal) + NamedValues[VarName] = OldVal; + else + NamedValues.erase(VarName); + + + // for expr always returns 0.0. + return Constant::getNullValue(Type::getDoubleTy(getGlobalContext())); +} + +Value *VarExprAST::Codegen() { + std::vector OldBindings; + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Register all variables and emit their initializer. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) { + const std::string &VarName = VarNames[i].first; + ExprAST *Init = VarNames[i].second; + + // Emit the initializer before adding the variable to scope, this prevents + // the initializer from referencing the variable itself, and permits stuff + // like this: + // var a = 1 in + // var a = a in ... # refers to outer 'a'. + Value *InitVal; + if (Init) { + InitVal = Init->Codegen(); + if (InitVal == 0) return 0; + } else { // If not specified, use 0.0. + InitVal = ConstantFP::get(getGlobalContext(), APFloat(0.0)); + } + + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + Builder.CreateStore(InitVal, Alloca); + + // Remember the old variable binding so that we can restore the binding when + // we unrecurse. + OldBindings.push_back(NamedValues[VarName]); + + // Remember this binding. + NamedValues[VarName] = Alloca; + } + + // Codegen the body, now that all vars are in scope. + Value *BodyVal = Body->Codegen(); + if (BodyVal == 0) return 0; + + // Pop all our variables from scope. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) + NamedValues[VarNames[i].first] = OldBindings[i]; + + // Return the body computation. + return BodyVal; +} + +Function *PrototypeAST::Codegen() { + // Make the function type: double(double,double) etc. + std::vector Doubles(Args.size(), + Type::getDoubleTy(getGlobalContext())); + FunctionType *FT = FunctionType::get(Type::getDoubleTy(getGlobalContext()), + Doubles, false); + + std::string FnName; + if (UseMCJIT) + FnName = MakeLegalFunctionName(Name); + else + FnName = Name; + + Module* M = TheHelper->getModuleForNewFunction(); + Function *F = Function::Create(FT, Function::ExternalLinkage, FnName, M); + + // FIXME: Implement duplicate function detection. + // The check below will only work if the duplicate is in the open module. + // If F conflicted, there was already something named 'Name'. If it has a + // body, don't allow redefinition or reextern. + if (F->getName() != FnName) { + // Delete the one we just made and get the existing one. + F->eraseFromParent(); + F = M->getFunction(FnName); + // If F already has a body, reject this. + if (!F->empty()) { + ErrorF("redefinition of function"); + return 0; + } + // If F took a different number of args, reject. + if (F->arg_size() != Args.size()) { + ErrorF("redefinition of function with different # args"); + return 0; + } + } + + // Set names for all arguments. + unsigned Idx = 0; + for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); + ++AI, ++Idx) + AI->setName(Args[Idx]); + + return F; +} + +/// CreateArgumentAllocas - Create an alloca for each argument and register the +/// argument in the symbol table so that references to it will succeed. +void PrototypeAST::CreateArgumentAllocas(Function *F) { + Function::arg_iterator AI = F->arg_begin(); + for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) { + // Create an alloca for this variable. + AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]); + + // Store the initial value into the alloca. + Builder.CreateStore(AI, Alloca); + + // Add arguments to variable symbol table. + NamedValues[Args[Idx]] = Alloca; + } +} + +Function *FunctionAST::Codegen() { + NamedValues.clear(); + + Function *TheFunction = Proto->Codegen(); + if (TheFunction == 0) + return 0; + + // If this is an operator, install it. + if (Proto->isBinaryOp()) + BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); + + // Create a new basic block to start insertion into. + BasicBlock *BB = BasicBlock::Create(getGlobalContext(), "entry", TheFunction); + Builder.SetInsertPoint(BB); + + // Add all arguments to the symbol table and create their allocas. + Proto->CreateArgumentAllocas(TheFunction); + + if (Value *RetVal = Body->Codegen()) { + // Finish off the function. + Builder.CreateRet(RetVal); + + // Validate the generated code, checking for consistency. + verifyFunction(*TheFunction); + + // Optimize the function. + if (!UseMCJIT) + TheHelper->runFPM(*TheFunction); + + return TheFunction; + } + + // Error reading body, remove function. + TheFunction->eraseFromParent(); + + if (Proto->isBinaryOp()) + BinopPrecedence.erase(Proto->getOperatorName()); + return 0; +} + +//===----------------------------------------------------------------------===// +// Top-Level parsing and JIT Driver +//===----------------------------------------------------------------------===// + +static void HandleDefinition() { + if (FunctionAST *F = ParseDefinition()) { + if (UseMCJIT && EnableLazyCompilation) + TheHelper->closeCurrentModule(); + Function *LF = F->Codegen(); + if (LF && VerboseOutput) { + fprintf(stderr, "Read function definition:"); + LF->dump(); + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleExtern() { + if (PrototypeAST *P = ParseExtern()) { + Function *F = P->Codegen(); + if (F && VerboseOutput) { + fprintf(stderr, "Read extern: "); + F->dump(); + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleTopLevelExpression() { + // Evaluate a top-level expression into an anonymous function. + if (FunctionAST *F = ParseTopLevelExpr()) { + if (Function *LF = F->Codegen()) { + // JIT the function, returning a function pointer. + void *FPtr = TheHelper->getPointerToFunction(LF); + // Cast it to the right type (takes no arguments, returns a double) so we + // can call it as a native function. + double (*FP)() = (double (*)())(intptr_t)FPtr; + double Result = FP(); + if (VerboseOutput) + fprintf(stderr, "Evaluated to %f\n", Result); + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +/// top ::= definition | external | expression | ';' +static void MainLoop() { + while (1) { + if (!SuppressPrompts) + fprintf(stderr, "ready> "); + switch (CurTok) { + case tok_eof: return; + case ';': getNextToken(); break; // ignore top-level semicolons. + case tok_def: HandleDefinition(); break; + case tok_extern: HandleExtern(); break; + default: HandleTopLevelExpression(); break; + } + } +} + +//===----------------------------------------------------------------------===// +// "Library" functions that can be "extern'd" from user code. +//===----------------------------------------------------------------------===// + +/// putchard - putchar that takes a double and returns 0. +extern "C" +double putchard(double X) { + putchar((char)X); + return 0; +} + +/// printd - printf that takes a double prints it as "%f\n", returning 0. +extern "C" +double printd(double X) { + printf("%f", X); + return 0; +} + +extern "C" +double printlf() { + printf("\n"); + return 0; +} + +//===----------------------------------------------------------------------===// +// Main driver code. +//===----------------------------------------------------------------------===// + +int main(int argc, char **argv) { + InitializeNativeTarget(); + if (UseMCJIT) { + InitializeNativeTargetAsmPrinter(); + InitializeNativeTargetAsmParser(); + } + LLVMContext &Context = getGlobalContext(); + + cl::ParseCommandLineOptions(argc, argv, + "Kaleidoscope example program\n"); + + // Install standard binary operators. + // 1 is lowest precedence. + BinopPrecedence['='] = 2; + BinopPrecedence['<'] = 10; + BinopPrecedence['+'] = 20; + BinopPrecedence['-'] = 20; + BinopPrecedence['/'] = 40; + BinopPrecedence['*'] = 40; // highest. + + // Make the Helper, which holds all the code. + if (UseMCJIT) + TheHelper = new MCJITHelper(Context); + else + TheHelper = new JITHelper(Context); + + // Prime the first token. + if (!SuppressPrompts) + fprintf(stderr, "ready> "); + getNextToken(); + + // Run the main "interpreter loop" now. + MainLoop(); + + // Print out all of the generated code. + if (DumpModulesOnExit) + TheHelper->dump(); + + return 0; +} diff --git a/examples/Kaleidoscope/MCJIT/initial/Makefile b/examples/Kaleidoscope/MCJIT/initial/Makefile new file mode 100644 index 00000000000..2989832d3c7 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/initial/Makefile @@ -0,0 +1,4 @@ +all: toy-mcjit + +toy-mcjit : toy.cpp + clang++ toy.cpp -g -O3 -rdynamic -fno-rtti `llvm-config --cppflags --ldflags --libs core mcjit native` -o toy-mcjit diff --git a/examples/Kaleidoscope/MCJIT/initial/README.txt b/examples/Kaleidoscope/MCJIT/initial/README.txt new file mode 100644 index 00000000000..b352a784522 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/initial/README.txt @@ -0,0 +1,18 @@ +//===----------------------------------------------------------------------===/ +// Kaleidoscope with MCJIT +//===----------------------------------------------------------------------===// + +The files in this directory are meant to accompany the first in a series of +three blog posts that describe the process of porting the Kaleidoscope tutorial +to use the MCJIT execution engine instead of the older JIT engine. + +When the blog post is ready this file will be updated with a link to the post. + +The source code in this directory demonstrates the initial working version of +the program before subsequent performance improvements are applied. + +This directory contain a Makefile that allow the code to be built in a +standalone manner, independent of the larger LLVM build infrastructure. To build +the program you will need to have 'clang++' and 'llvm-config' in your path. If +you attempt to build using the LLVM 3.3 release, some minor modifications will +be required, as mentioned in the blog posts. \ No newline at end of file diff --git a/examples/Kaleidoscope/MCJIT/initial/toy.cpp b/examples/Kaleidoscope/MCJIT/initial/toy.cpp new file mode 100644 index 00000000000..c6829076fcb --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/initial/toy.cpp @@ -0,0 +1,1381 @@ +#include "llvm/Analysis/Passes.h" +#include "llvm/Analysis/Verifier.h" +#include "llvm/ExecutionEngine/ExecutionEngine.h" +#include "llvm/ExecutionEngine/MCJIT.h" +#include "llvm/ExecutionEngine/SectionMemoryManager.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/PassManager.h" +#include "llvm/Support/TargetSelect.h" +#include "llvm/Transforms/Scalar.h" +#include +#include +#include +#include +using namespace llvm; + +//===----------------------------------------------------------------------===// +// Lexer +//===----------------------------------------------------------------------===// + +// The lexer returns tokens [0-255] if it is an unknown character, otherwise one +// of these for known things. +enum Token { + tok_eof = -1, + + // commands + tok_def = -2, tok_extern = -3, + + // primary + tok_identifier = -4, tok_number = -5, + + // control + tok_if = -6, tok_then = -7, tok_else = -8, + tok_for = -9, tok_in = -10, + + // operators + tok_binary = -11, tok_unary = -12, + + // var definition + tok_var = -13 +}; + +static std::string IdentifierStr; // Filled in if tok_identifier +static double NumVal; // Filled in if tok_number + +/// gettok - Return the next token from standard input. +static int gettok() { + static int LastChar = ' '; + + // Skip any whitespace. + while (isspace(LastChar)) + LastChar = getchar(); + + if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* + IdentifierStr = LastChar; + while (isalnum((LastChar = getchar()))) + IdentifierStr += LastChar; + + if (IdentifierStr == "def") return tok_def; + if (IdentifierStr == "extern") return tok_extern; + if (IdentifierStr == "if") return tok_if; + if (IdentifierStr == "then") return tok_then; + if (IdentifierStr == "else") return tok_else; + if (IdentifierStr == "for") return tok_for; + if (IdentifierStr == "in") return tok_in; + if (IdentifierStr == "binary") return tok_binary; + if (IdentifierStr == "unary") return tok_unary; + if (IdentifierStr == "var") return tok_var; + return tok_identifier; + } + + if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ + std::string NumStr; + do { + NumStr += LastChar; + LastChar = getchar(); + } while (isdigit(LastChar) || LastChar == '.'); + + NumVal = strtod(NumStr.c_str(), 0); + return tok_number; + } + + if (LastChar == '#') { + // Comment until end of line. + do LastChar = getchar(); + while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); + + if (LastChar != EOF) + return gettok(); + } + + // Check for end of file. Don't eat the EOF. + if (LastChar == EOF) + return tok_eof; + + // Otherwise, just return the character as its ascii value. + int ThisChar = LastChar; + LastChar = getchar(); + return ThisChar; +} + +//===----------------------------------------------------------------------===// +// Abstract Syntax Tree (aka Parse Tree) +//===----------------------------------------------------------------------===// + +/// ExprAST - Base class for all expression nodes. +class ExprAST { +public: + virtual ~ExprAST() {} + virtual Value *Codegen() = 0; +}; + +/// NumberExprAST - Expression class for numeric literals like "1.0". +class NumberExprAST : public ExprAST { + double Val; +public: + NumberExprAST(double val) : Val(val) {} + virtual Value *Codegen(); +}; + +/// VariableExprAST - Expression class for referencing a variable, like "a". +class VariableExprAST : public ExprAST { + std::string Name; +public: + VariableExprAST(const std::string &name) : Name(name) {} + const std::string &getName() const { return Name; } + virtual Value *Codegen(); +}; + +/// UnaryExprAST - Expression class for a unary operator. +class UnaryExprAST : public ExprAST { + char Opcode; + ExprAST *Operand; +public: + UnaryExprAST(char opcode, ExprAST *operand) + : Opcode(opcode), Operand(operand) {} + virtual Value *Codegen(); +}; + +/// BinaryExprAST - Expression class for a binary operator. +class BinaryExprAST : public ExprAST { + char Op; + ExprAST *LHS, *RHS; +public: + BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) + : Op(op), LHS(lhs), RHS(rhs) {} + virtual Value *Codegen(); +}; + +/// CallExprAST - Expression class for function calls. +class CallExprAST : public ExprAST { + std::string Callee; + std::vector Args; +public: + CallExprAST(const std::string &callee, std::vector &args) + : Callee(callee), Args(args) {} + virtual Value *Codegen(); +}; + +/// IfExprAST - Expression class for if/then/else. +class IfExprAST : public ExprAST { + ExprAST *Cond, *Then, *Else; +public: + IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) + : Cond(cond), Then(then), Else(_else) {} + virtual Value *Codegen(); +}; + +/// ForExprAST - Expression class for for/in. +class ForExprAST : public ExprAST { + std::string VarName; + ExprAST *Start, *End, *Step, *Body; +public: + ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, + ExprAST *step, ExprAST *body) + : VarName(varname), Start(start), End(end), Step(step), Body(body) {} + virtual Value *Codegen(); +}; + +/// VarExprAST - Expression class for var/in +class VarExprAST : public ExprAST { + std::vector > VarNames; + ExprAST *Body; +public: + VarExprAST(const std::vector > &varnames, + ExprAST *body) + : VarNames(varnames), Body(body) {} + + virtual Value *Codegen(); +}; + +/// PrototypeAST - This class represents the "prototype" for a function, +/// which captures its argument names as well as if it is an operator. +class PrototypeAST { + std::string Name; + std::vector Args; + bool isOperator; + unsigned Precedence; // Precedence if a binary op. +public: + PrototypeAST(const std::string &name, const std::vector &args, + bool isoperator = false, unsigned prec = 0) + : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} + + bool isUnaryOp() const { return isOperator && Args.size() == 1; } + bool isBinaryOp() const { return isOperator && Args.size() == 2; } + + char getOperatorName() const { + assert(isUnaryOp() || isBinaryOp()); + return Name[Name.size()-1]; + } + + unsigned getBinaryPrecedence() const { return Precedence; } + + Function *Codegen(); + + void CreateArgumentAllocas(Function *F); +}; + +/// FunctionAST - This class represents a function definition itself. +class FunctionAST { + PrototypeAST *Proto; + ExprAST *Body; +public: + FunctionAST(PrototypeAST *proto, ExprAST *body) + : Proto(proto), Body(body) {} + + Function *Codegen(); +}; + +//===----------------------------------------------------------------------===// +// Parser +//===----------------------------------------------------------------------===// + +/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current +/// token the parser is looking at. getNextToken reads another token from the +/// lexer and updates CurTok with its results. +static int CurTok; +static int getNextToken() { + return CurTok = gettok(); +} + +/// BinopPrecedence - This holds the precedence for each binary operator that is +/// defined. +static std::map BinopPrecedence; + +/// GetTokPrecedence - Get the precedence of the pending binary operator token. +static int GetTokPrecedence() { + if (!isascii(CurTok)) + return -1; + + // Make sure it's a declared binop. + int TokPrec = BinopPrecedence[CurTok]; + if (TokPrec <= 0) return -1; + return TokPrec; +} + +/// Error* - These are little helper functions for error handling. +ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} +PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } +FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } + +static ExprAST *ParseExpression(); + +/// identifierexpr +/// ::= identifier +/// ::= identifier '(' expression* ')' +static ExprAST *ParseIdentifierExpr() { + std::string IdName = IdentifierStr; + + getNextToken(); // eat identifier. + + if (CurTok != '(') // Simple variable ref. + return new VariableExprAST(IdName); + + // Call. + getNextToken(); // eat ( + std::vector Args; + if (CurTok != ')') { + while (1) { + ExprAST *Arg = ParseExpression(); + if (!Arg) return 0; + Args.push_back(Arg); + + if (CurTok == ')') break; + + if (CurTok != ',') + return Error("Expected ')' or ',' in argument list"); + getNextToken(); + } + } + + // Eat the ')'. + getNextToken(); + + return new CallExprAST(IdName, Args); +} + +/// numberexpr ::= number +static ExprAST *ParseNumberExpr() { + ExprAST *Result = new NumberExprAST(NumVal); + getNextToken(); // consume the number + return Result; +} + +/// parenexpr ::= '(' expression ')' +static ExprAST *ParseParenExpr() { + getNextToken(); // eat (. + ExprAST *V = ParseExpression(); + if (!V) return 0; + + if (CurTok != ')') + return Error("expected ')'"); + getNextToken(); // eat ). + return V; +} + +/// ifexpr ::= 'if' expression 'then' expression 'else' expression +static ExprAST *ParseIfExpr() { + getNextToken(); // eat the if. + + // condition. + ExprAST *Cond = ParseExpression(); + if (!Cond) return 0; + + if (CurTok != tok_then) + return Error("expected then"); + getNextToken(); // eat the then + + ExprAST *Then = ParseExpression(); + if (Then == 0) return 0; + + if (CurTok != tok_else) + return Error("expected else"); + + getNextToken(); + + ExprAST *Else = ParseExpression(); + if (!Else) return 0; + + return new IfExprAST(Cond, Then, Else); +} + +/// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression +static ExprAST *ParseForExpr() { + getNextToken(); // eat the for. + + if (CurTok != tok_identifier) + return Error("expected identifier after for"); + + std::string IdName = IdentifierStr; + getNextToken(); // eat identifier. + + if (CurTok != '=') + return Error("expected '=' after for"); + getNextToken(); // eat '='. + + + ExprAST *Start = ParseExpression(); + if (Start == 0) return 0; + if (CurTok != ',') + return Error("expected ',' after for start value"); + getNextToken(); + + ExprAST *End = ParseExpression(); + if (End == 0) return 0; + + // The step value is optional. + ExprAST *Step = 0; + if (CurTok == ',') { + getNextToken(); + Step = ParseExpression(); + if (Step == 0) return 0; + } + + if (CurTok != tok_in) + return Error("expected 'in' after for"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new ForExprAST(IdName, Start, End, Step, Body); +} + +/// varexpr ::= 'var' identifier ('=' expression)? +// (',' identifier ('=' expression)?)* 'in' expression +static ExprAST *ParseVarExpr() { + getNextToken(); // eat the var. + + std::vector > VarNames; + + // At least one variable name is required. + if (CurTok != tok_identifier) + return Error("expected identifier after var"); + + while (1) { + std::string Name = IdentifierStr; + getNextToken(); // eat identifier. + + // Read the optional initializer. + ExprAST *Init = 0; + if (CurTok == '=') { + getNextToken(); // eat the '='. + + Init = ParseExpression(); + if (Init == 0) return 0; + } + + VarNames.push_back(std::make_pair(Name, Init)); + + // End of var list, exit loop. + if (CurTok != ',') break; + getNextToken(); // eat the ','. + + if (CurTok != tok_identifier) + return Error("expected identifier list after var"); + } + + // At this point, we have to have 'in'. + if (CurTok != tok_in) + return Error("expected 'in' keyword after 'var'"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new VarExprAST(VarNames, Body); +} + +/// primary +/// ::= identifierexpr +/// ::= numberexpr +/// ::= parenexpr +/// ::= ifexpr +/// ::= forexpr +/// ::= varexpr +static ExprAST *ParsePrimary() { + switch (CurTok) { + default: return Error("unknown token when expecting an expression"); + case tok_identifier: return ParseIdentifierExpr(); + case tok_number: return ParseNumberExpr(); + case '(': return ParseParenExpr(); + case tok_if: return ParseIfExpr(); + case tok_for: return ParseForExpr(); + case tok_var: return ParseVarExpr(); + } +} + +/// unary +/// ::= primary +/// ::= '!' unary +static ExprAST *ParseUnary() { + // If the current token is not an operator, it must be a primary expr. + if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') + return ParsePrimary(); + + // If this is a unary operator, read it. + int Opc = CurTok; + getNextToken(); + if (ExprAST *Operand = ParseUnary()) + return new UnaryExprAST(Opc, Operand); + return 0; +} + +/// binoprhs +/// ::= ('+' unary)* +static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { + // If this is a binop, find its precedence. + while (1) { + int TokPrec = GetTokPrecedence(); + + // If this is a binop that binds at least as tightly as the current binop, + // consume it, otherwise we are done. + if (TokPrec < ExprPrec) + return LHS; + + // Okay, we know this is a binop. + int BinOp = CurTok; + getNextToken(); // eat binop + + // Parse the unary expression after the binary operator. + ExprAST *RHS = ParseUnary(); + if (!RHS) return 0; + + // If BinOp binds less tightly with RHS than the operator after RHS, let + // the pending operator take RHS as its LHS. + int NextPrec = GetTokPrecedence(); + if (TokPrec < NextPrec) { + RHS = ParseBinOpRHS(TokPrec+1, RHS); + if (RHS == 0) return 0; + } + + // Merge LHS/RHS. + LHS = new BinaryExprAST(BinOp, LHS, RHS); + } +} + +/// expression +/// ::= unary binoprhs +/// +static ExprAST *ParseExpression() { + ExprAST *LHS = ParseUnary(); + if (!LHS) return 0; + + return ParseBinOpRHS(0, LHS); +} + +/// prototype +/// ::= id '(' id* ')' +/// ::= binary LETTER number? (id, id) +/// ::= unary LETTER (id) +static PrototypeAST *ParsePrototype() { + std::string FnName; + + unsigned Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. + unsigned BinaryPrecedence = 30; + + switch (CurTok) { + default: + return ErrorP("Expected function name in prototype"); + case tok_identifier: + FnName = IdentifierStr; + Kind = 0; + getNextToken(); + break; + case tok_unary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected unary operator"); + FnName = "unary"; + FnName += (char)CurTok; + Kind = 1; + getNextToken(); + break; + case tok_binary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected binary operator"); + FnName = "binary"; + FnName += (char)CurTok; + Kind = 2; + getNextToken(); + + // Read the precedence if present. + if (CurTok == tok_number) { + if (NumVal < 1 || NumVal > 100) + return ErrorP("Invalid precedecnce: must be 1..100"); + BinaryPrecedence = (unsigned)NumVal; + getNextToken(); + } + break; + } + + if (CurTok != '(') + return ErrorP("Expected '(' in prototype"); + + std::vector ArgNames; + while (getNextToken() == tok_identifier) + ArgNames.push_back(IdentifierStr); + if (CurTok != ')') + return ErrorP("Expected ')' in prototype"); + + // success. + getNextToken(); // eat ')'. + + // Verify right number of names for operator. + if (Kind && ArgNames.size() != Kind) + return ErrorP("Invalid number of operands for operator"); + + return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); +} + +/// definition ::= 'def' prototype expression +static FunctionAST *ParseDefinition() { + getNextToken(); // eat def. + PrototypeAST *Proto = ParsePrototype(); + if (Proto == 0) return 0; + + if (ExprAST *E = ParseExpression()) + return new FunctionAST(Proto, E); + return 0; +} + +/// toplevelexpr ::= expression +static FunctionAST *ParseTopLevelExpr() { + if (ExprAST *E = ParseExpression()) { + // Make an anonymous proto. + PrototypeAST *Proto = new PrototypeAST("", std::vector()); + return new FunctionAST(Proto, E); + } + return 0; +} + +/// external ::= 'extern' prototype +static PrototypeAST *ParseExtern() { + getNextToken(); // eat extern. + return ParsePrototype(); +} + +//===----------------------------------------------------------------------===// +// Quick and dirty hack +//===----------------------------------------------------------------------===// + +// FIXME: Obviously we can do better than this +std::string GenerateUniqueName(const char *root) +{ + static int i = 0; + char s[16]; + sprintf(s, "%s%d", root, i++); + std::string S = s; + return S; +} + +std::string MakeLegalFunctionName(std::string Name) +{ + std::string NewName; + if (!Name.length()) + return GenerateUniqueName("anon_func_"); + + // Start with what we have + NewName = Name; + + // Look for a numberic first character + if (NewName.find_first_of("0123456789") == 0) { + NewName.insert(0, 1, 'n'); + } + + // Replace illegal characters with their ASCII equivalent + std::string legal_elements = "_abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"; + size_t pos; + while ((pos = NewName.find_first_not_of(legal_elements)) != std::string::npos) { + char old_c = NewName.at(pos); + char new_str[16]; + sprintf(new_str, "%d", (int)old_c); + NewName = NewName.replace(pos, 1, new_str); + } + + return NewName; +} + +//===----------------------------------------------------------------------===// +// MCJIT helper class +//===----------------------------------------------------------------------===// + +class MCJITHelper +{ +public: + MCJITHelper(LLVMContext& C) : Context(C), OpenModule(NULL) {} + ~MCJITHelper(); + + Function *getFunction(const std::string FnName); + Module *getModuleForNewFunction(); + void *getPointerToFunction(Function* F); + void *getPointerToNamedFunction(const std::string &Name); + void dump(); + +private: + typedef std::vector ModuleVector; + typedef std::vector EngineVector; + + LLVMContext &Context; + Module *OpenModule; + ModuleVector Modules; + EngineVector Engines; +}; + +class HelpingMemoryManager : public SectionMemoryManager +{ + HelpingMemoryManager(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + void operator=(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + +public: + HelpingMemoryManager(MCJITHelper *Helper) : MasterHelper(Helper) {} + virtual ~HelpingMemoryManager() {} + + /// This method returns the address of the specified function. + /// Our implementation will attempt to find functions in other + /// modules associated with the MCJITHelper to cross link functions + /// from one generated module to another. + /// + /// If \p AbortOnFailure is false and no function with the given name is + /// found, this function returns a null pointer. Otherwise, it prints a + /// message to stderr and aborts. + virtual void *getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure = true); +private: + MCJITHelper *MasterHelper; +}; + +void *HelpingMemoryManager::getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure) +{ + // Try the standard symbol resolution first, but ask it not to abort. + void *pfn = SectionMemoryManager::getPointerToNamedFunction(Name, false); + if (pfn) + return pfn; + + pfn = MasterHelper->getPointerToNamedFunction(Name); + if (!pfn && AbortOnFailure) + report_fatal_error("Program used external function '" + Name + + "' which could not be resolved!"); + return pfn; +} + +MCJITHelper::~MCJITHelper() +{ + if (OpenModule) + delete OpenModule; + EngineVector::iterator begin = Engines.begin(); + EngineVector::iterator end = Engines.end(); + EngineVector::iterator it; + for (it = begin; it != end; ++it) + delete *it; +} + +Function *MCJITHelper::getFunction(const std::string FnName) { + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(FnName); + if (F) { + if (*it == OpenModule) + return F; + + assert(OpenModule != NULL); + + // This function is in a module that has already been JITed. + // We need to generate a new prototype for external linkage. + Function *PF = OpenModule->getFunction(FnName); + if (PF && !PF->empty()) { + ErrorF("redefinition of function across modules"); + return 0; + } + + // If we don't have a prototype yet, create one. + if (!PF) + PF = Function::Create(F->getFunctionType(), + Function::ExternalLinkage, + FnName, + OpenModule); + return PF; + } + } + return NULL; +} + +Module *MCJITHelper::getModuleForNewFunction() { + // If we have a Module that hasn't been JITed, use that. + if (OpenModule) + return OpenModule; + + // Otherwise create a new Module. + std::string ModName = GenerateUniqueName("mcjit_module_"); + Module *M = new Module(ModName, Context); + Modules.push_back(M); + OpenModule = M; + return M; +} + +void *MCJITHelper::getPointerToFunction(Function* F) { + // See if an existing instance of MCJIT has this function. + EngineVector::iterator begin = Engines.begin(); + EngineVector::iterator end = Engines.end(); + EngineVector::iterator it; + for (it = begin; it != end; ++it) { + void *P = (*it)->getPointerToFunction(F); + if (P) + return P; + } + + // If we didn't find the function, see if we can generate it. + if (OpenModule) { + std::string ErrStr; + ExecutionEngine *NewEngine = EngineBuilder(OpenModule) + .setErrorStr(&ErrStr) + .setUseMCJIT(true) + .setMCJITMemoryManager(new HelpingMemoryManager(this)) + .create(); + if (!NewEngine) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + // Create a function pass manager for this engine + FunctionPassManager *FPM = new FunctionPassManager(OpenModule); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + FPM->add(new DataLayout(*NewEngine->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + FPM->add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + FPM->add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + FPM->add(createInstructionCombiningPass()); + // Reassociate expressions. + FPM->add(createReassociatePass()); + // Eliminate Common SubExpressions. + FPM->add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + FPM->add(createCFGSimplificationPass()); + FPM->doInitialization(); + + // For each function in the module + Module::iterator it; + Module::iterator end = OpenModule->end(); + for (it = OpenModule->begin(); it != end; ++it) { + // Run the FPM on this function + FPM->run(*it); + } + + // We don't need this anymore + delete FPM; + + OpenModule = NULL; + Engines.push_back(NewEngine); + NewEngine->finalizeObject(); + return NewEngine->getPointerToFunction(F); + } + return NULL; +} + +void *MCJITHelper::getPointerToNamedFunction(const std::string &Name) +{ + // Look for the function in each of our execution engines. + EngineVector::iterator begin = Engines.begin(); + EngineVector::iterator end = Engines.end(); + EngineVector::iterator it; + for (it = begin; it != end; ++it) { + if (Function *F = (*it)->FindFunctionNamed(Name.c_str())) + return (*it)->getPointerToFunction(F); + } + + return NULL; +} + +void MCJITHelper::dump() +{ + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) + (*it)->dump(); +} + +//===----------------------------------------------------------------------===// +// Code Generation +//===----------------------------------------------------------------------===// + +static MCJITHelper *TheHelper; +static IRBuilder<> Builder(getGlobalContext()); +static std::map NamedValues; + +Value *ErrorV(const char *Str) { Error(Str); return 0; } + +/// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of +/// the function. This is used for mutable variables etc. +static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction, + const std::string &VarName) { + IRBuilder<> TmpB(&TheFunction->getEntryBlock(), + TheFunction->getEntryBlock().begin()); + return TmpB.CreateAlloca(Type::getDoubleTy(getGlobalContext()), 0, + VarName.c_str()); +} + +Value *NumberExprAST::Codegen() { + return ConstantFP::get(getGlobalContext(), APFloat(Val)); +} + +Value *VariableExprAST::Codegen() { + // Look this variable up in the function. + Value *V = NamedValues[Name]; + char ErrStr[256]; + sprintf(ErrStr, "Unknown variable name %s", Name.c_str()); + if (V == 0) return ErrorV(ErrStr); + + // Load the value. + return Builder.CreateLoad(V, Name.c_str()); +} + +Value *UnaryExprAST::Codegen() { + Value *OperandV = Operand->Codegen(); + if (OperandV == 0) return 0; + + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("unary")+Opcode)); + if (F == 0) + return ErrorV("Unknown unary operator"); + + return Builder.CreateCall(F, OperandV, "unop"); +} + +Value *BinaryExprAST::Codegen() { + // Special case '=' because we don't want to emit the LHS as an expression. + if (Op == '=') { + // Assignment requires the LHS to be an identifier. + VariableExprAST *LHSE = reinterpret_cast(LHS); + if (!LHSE) + return ErrorV("destination of '=' must be a variable"); + // Codegen the RHS. + Value *Val = RHS->Codegen(); + if (Val == 0) return 0; + + // Look up the name. + Value *Variable = NamedValues[LHSE->getName()]; + if (Variable == 0) return ErrorV("Unknown variable name"); + + Builder.CreateStore(Val, Variable); + return Val; + } + + Value *L = LHS->Codegen(); + Value *R = RHS->Codegen(); + if (L == 0 || R == 0) return 0; + + switch (Op) { + case '+': return Builder.CreateFAdd(L, R, "addtmp"); + case '-': return Builder.CreateFSub(L, R, "subtmp"); + case '*': return Builder.CreateFMul(L, R, "multmp"); + case '/': return Builder.CreateFDiv(L, R, "divtmp"); + case '<': + L = Builder.CreateFCmpULT(L, R, "cmptmp"); + // Convert bool 0/1 to double 0.0 or 1.0 + return Builder.CreateUIToFP(L, Type::getDoubleTy(getGlobalContext()), + "booltmp"); + default: break; + } + + // If it wasn't a builtin binary operator, it must be a user defined one. Emit + // a call to it. + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("binary")+Op)); + assert(F && "binary operator not found!"); + + Value *Ops[] = { L, R }; + return Builder.CreateCall(F, Ops, "binop"); +} + +Value *CallExprAST::Codegen() { + // Look up the name in the global module table. + Function *CalleeF = TheHelper->getFunction(Callee); + if (CalleeF == 0) + return ErrorV("Unknown function referenced"); + + // If argument mismatch error. + if (CalleeF->arg_size() != Args.size()) + return ErrorV("Incorrect # arguments passed"); + + std::vector ArgsV; + for (unsigned i = 0, e = Args.size(); i != e; ++i) { + ArgsV.push_back(Args[i]->Codegen()); + if (ArgsV.back() == 0) return 0; + } + + return Builder.CreateCall(CalleeF, ArgsV, "calltmp"); +} + +Value *IfExprAST::Codegen() { + Value *CondV = Cond->Codegen(); + if (CondV == 0) return 0; + + // Convert condition to a bool by comparing equal to 0.0. + CondV = Builder.CreateFCmpONE(CondV, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "ifcond"); + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create blocks for the then and else cases. Insert the 'then' block at the + // end of the function. + BasicBlock *ThenBB = BasicBlock::Create(getGlobalContext(), "then", TheFunction); + BasicBlock *ElseBB = BasicBlock::Create(getGlobalContext(), "else"); + BasicBlock *MergeBB = BasicBlock::Create(getGlobalContext(), "ifcont"); + + Builder.CreateCondBr(CondV, ThenBB, ElseBB); + + // Emit then value. + Builder.SetInsertPoint(ThenBB); + + Value *ThenV = Then->Codegen(); + if (ThenV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Then' can change the current block, update ThenBB for the PHI. + ThenBB = Builder.GetInsertBlock(); + + // Emit else block. + TheFunction->getBasicBlockList().push_back(ElseBB); + Builder.SetInsertPoint(ElseBB); + + Value *ElseV = Else->Codegen(); + if (ElseV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Else' can change the current block, update ElseBB for the PHI. + ElseBB = Builder.GetInsertBlock(); + + // Emit merge block. + TheFunction->getBasicBlockList().push_back(MergeBB); + Builder.SetInsertPoint(MergeBB); + PHINode *PN = Builder.CreatePHI(Type::getDoubleTy(getGlobalContext()), 2, + "iftmp"); + + PN->addIncoming(ThenV, ThenBB); + PN->addIncoming(ElseV, ElseBB); + return PN; +} + +Value *ForExprAST::Codegen() { + // Output this as: + // var = alloca double + // ... + // start = startexpr + // store start -> var + // goto loop + // loop: + // ... + // bodyexpr + // ... + // loopend: + // step = stepexpr + // endcond = endexpr + // + // curvar = load var + // nextvar = curvar + step + // store nextvar -> var + // br endcond, loop, endloop + // outloop: + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create an alloca for the variable in the entry block. + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + + // Emit the start code first, without 'variable' in scope. + Value *StartVal = Start->Codegen(); + if (StartVal == 0) return 0; + + // Store the value into the alloca. + Builder.CreateStore(StartVal, Alloca); + + // Make the new basic block for the loop header, inserting after current + // block. + BasicBlock *LoopBB = BasicBlock::Create(getGlobalContext(), "loop", TheFunction); + + // Insert an explicit fall through from the current block to the LoopBB. + Builder.CreateBr(LoopBB); + + // Start insertion in LoopBB. + Builder.SetInsertPoint(LoopBB); + + // Within the loop, the variable is defined equal to the PHI node. If it + // shadows an existing variable, we have to restore it, so save it now. + AllocaInst *OldVal = NamedValues[VarName]; + NamedValues[VarName] = Alloca; + + // Emit the body of the loop. This, like any other expr, can change the + // current BB. Note that we ignore the value computed by the body, but don't + // allow an error. + if (Body->Codegen() == 0) + return 0; + + // Emit the step value. + Value *StepVal; + if (Step) { + StepVal = Step->Codegen(); + if (StepVal == 0) return 0; + } else { + // If not specified, use 1.0. + StepVal = ConstantFP::get(getGlobalContext(), APFloat(1.0)); + } + + // Compute the end condition. + Value *EndCond = End->Codegen(); + if (EndCond == 0) return EndCond; + + // Reload, increment, and restore the alloca. This handles the case where + // the body of the loop mutates the variable. + Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str()); + Value *NextVar = Builder.CreateFAdd(CurVar, StepVal, "nextvar"); + Builder.CreateStore(NextVar, Alloca); + + // Convert condition to a bool by comparing equal to 0.0. + EndCond = Builder.CreateFCmpONE(EndCond, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "loopcond"); + + // Create the "after loop" block and insert it. + BasicBlock *AfterBB = BasicBlock::Create(getGlobalContext(), "afterloop", TheFunction); + + // Insert the conditional branch into the end of LoopEndBB. + Builder.CreateCondBr(EndCond, LoopBB, AfterBB); + + // Any new code will be inserted in AfterBB. + Builder.SetInsertPoint(AfterBB); + + // Restore the unshadowed variable. + if (OldVal) + NamedValues[VarName] = OldVal; + else + NamedValues.erase(VarName); + + + // for expr always returns 0.0. + return Constant::getNullValue(Type::getDoubleTy(getGlobalContext())); +} + +Value *VarExprAST::Codegen() { + std::vector OldBindings; + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Register all variables and emit their initializer. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) { + const std::string &VarName = VarNames[i].first; + ExprAST *Init = VarNames[i].second; + + // Emit the initializer before adding the variable to scope, this prevents + // the initializer from referencing the variable itself, and permits stuff + // like this: + // var a = 1 in + // var a = a in ... # refers to outer 'a'. + Value *InitVal; + if (Init) { + InitVal = Init->Codegen(); + if (InitVal == 0) return 0; + } else { // If not specified, use 0.0. + InitVal = ConstantFP::get(getGlobalContext(), APFloat(0.0)); + } + + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + Builder.CreateStore(InitVal, Alloca); + + // Remember the old variable binding so that we can restore the binding when + // we unrecurse. + OldBindings.push_back(NamedValues[VarName]); + + // Remember this binding. + NamedValues[VarName] = Alloca; + } + + // Codegen the body, now that all vars are in scope. + Value *BodyVal = Body->Codegen(); + if (BodyVal == 0) return 0; + + // Pop all our variables from scope. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) + NamedValues[VarNames[i].first] = OldBindings[i]; + + // Return the body computation. + return BodyVal; +} + +Function *PrototypeAST::Codegen() { + // Make the function type: double(double,double) etc. + std::vector Doubles(Args.size(), + Type::getDoubleTy(getGlobalContext())); + FunctionType *FT = FunctionType::get(Type::getDoubleTy(getGlobalContext()), + Doubles, false); + + std::string FnName = MakeLegalFunctionName(Name); + + Module* M = TheHelper->getModuleForNewFunction(); + + Function *F = Function::Create(FT, Function::ExternalLinkage, FnName, M); + + // If F conflicted, there was already something named 'FnName'. If it has a + // body, don't allow redefinition or reextern. + if (F->getName() != FnName) { + // Delete the one we just made and get the existing one. + F->eraseFromParent(); + F = M->getFunction(Name); + + // If F already has a body, reject this. + if (!F->empty()) { + ErrorF("redefinition of function"); + return 0; + } + + // If F took a different number of args, reject. + if (F->arg_size() != Args.size()) { + ErrorF("redefinition of function with different # args"); + return 0; + } + } + + // Set names for all arguments. + unsigned Idx = 0; + for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); + ++AI, ++Idx) + AI->setName(Args[Idx]); + + return F; +} + +/// CreateArgumentAllocas - Create an alloca for each argument and register the +/// argument in the symbol table so that references to it will succeed. +void PrototypeAST::CreateArgumentAllocas(Function *F) { + Function::arg_iterator AI = F->arg_begin(); + for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) { + // Create an alloca for this variable. + AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]); + + // Store the initial value into the alloca. + Builder.CreateStore(AI, Alloca); + + // Add arguments to variable symbol table. + NamedValues[Args[Idx]] = Alloca; + } +} + +Function *FunctionAST::Codegen() { + NamedValues.clear(); + + Function *TheFunction = Proto->Codegen(); + if (TheFunction == 0) + return 0; + + // If this is an operator, install it. + if (Proto->isBinaryOp()) + BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); + + // Create a new basic block to start insertion into. + BasicBlock *BB = BasicBlock::Create(getGlobalContext(), "entry", TheFunction); + Builder.SetInsertPoint(BB); + + // Add all arguments to the symbol table and create their allocas. + Proto->CreateArgumentAllocas(TheFunction); + + if (Value *RetVal = Body->Codegen()) { + // Finish off the function. + Builder.CreateRet(RetVal); + + // Validate the generated code, checking for consistency. + verifyFunction(*TheFunction); + + return TheFunction; + } + + // Error reading body, remove function. + TheFunction->eraseFromParent(); + + if (Proto->isBinaryOp()) + BinopPrecedence.erase(Proto->getOperatorName()); + return 0; +} + +//===----------------------------------------------------------------------===// +// Top-Level parsing and JIT Driver +//===----------------------------------------------------------------------===// + +static void HandleDefinition() { + if (FunctionAST *F = ParseDefinition()) { + if (Function *LF = F->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read function definition:"); + LF->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleExtern() { + if (PrototypeAST *P = ParseExtern()) { + if (Function *F = P->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read extern: "); + F->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleTopLevelExpression() { + // Evaluate a top-level expression into an anonymous function. + if (FunctionAST *F = ParseTopLevelExpr()) { + if (Function *LF = F->Codegen()) { + // JIT the function, returning a function pointer. + void *FPtr = TheHelper->getPointerToFunction(LF); + + // Cast it to the right type (takes no arguments, returns a double) so we + // can call it as a native function. + double (*FP)() = (double (*)())(intptr_t)FPtr; +#ifdef MINIMAL_STDERR_OUTPUT + FP(); +#else + fprintf(stderr, "Evaluated to %f\n", FP()); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +/// top ::= definition | external | expression | ';' +static void MainLoop() { + while (1) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + switch (CurTok) { + case tok_eof: return; + case ';': getNextToken(); break; // ignore top-level semicolons. + case tok_def: HandleDefinition(); break; + case tok_extern: HandleExtern(); break; + default: HandleTopLevelExpression(); break; + } + } +} + +//===----------------------------------------------------------------------===// +// "Library" functions that can be "extern'd" from user code. +//===----------------------------------------------------------------------===// + +/// putchard - putchar that takes a double and returns 0. +extern "C" +double putchard(double X) { + putchar((char)X); + return 0; +} + +/// printd - printf that takes a double prints it as "%f\n", returning 0. +extern "C" +double printd(double X) { + printf("%f", X); + return 0; +} + +extern "C" +double printlf() { + printf("\n"); + return 0; +} + +//===----------------------------------------------------------------------===// +// Main driver code. +//===----------------------------------------------------------------------===// + +int main() { + InitializeNativeTarget(); + InitializeNativeTargetAsmPrinter(); + InitializeNativeTargetAsmParser(); + LLVMContext &Context = getGlobalContext(); + + // Install standard binary operators. + // 1 is lowest precedence. + BinopPrecedence['='] = 2; + BinopPrecedence['<'] = 10; + BinopPrecedence['+'] = 20; + BinopPrecedence['-'] = 20; + BinopPrecedence['/'] = 40; + BinopPrecedence['*'] = 40; // highest. + + // Prime the first token. +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + getNextToken(); + + // Make the helper, which holds all the code. + TheHelper = new MCJITHelper(Context); + + // Run the main "interpreter loop" now. + MainLoop(); + +#ifndef MINIMAL_STDERR_OUTPUT + // Print out all of the generated code. + TheHelper->dump(); +#endif + + return 0; +} diff --git a/examples/Kaleidoscope/MCJIT/lazy/Makefile b/examples/Kaleidoscope/MCJIT/lazy/Makefile new file mode 100644 index 00000000000..21cbc18d6e2 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/lazy/Makefile @@ -0,0 +1,7 @@ +all: toy-mcjit toy-jit + +toy-mcjit : toy.cpp + clang++ toy.cpp -g -O3 -rdynamic -fno-rtti `llvm-config --cppflags --ldflags --libs core mcjit native` -o toy-mcjit + +toy-jit : toy-jit.cpp + clang++ toy-jit.cpp -g -O3 -rdynamic `llvm-config --cppflags --ldflags --libs core jit native` -o toy-jit diff --git a/examples/Kaleidoscope/MCJIT/lazy/README.txt b/examples/Kaleidoscope/MCJIT/lazy/README.txt new file mode 100644 index 00000000000..c5c271e02f6 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/lazy/README.txt @@ -0,0 +1,25 @@ +//===----------------------------------------------------------------------===/ +// Kaleidoscope with MCJIT +//===----------------------------------------------------------------------===// + +The files in this directory are meant to accompany the first in a series of +three blog posts that describe the process of porting the Kaleidoscope tutorial +to use the MCJIT execution engine instead of the older JIT engine. + +When the blog post is ready this file will be updated with a link to the post. + +The source code in this directory demonstrates the second version of the +program, now modified to implement a sort of 'lazy' compilation. + +The toy-jit.cpp file contains a version of the original JIT-based source code +that has been modified to disable most stderr output for timing purposes. + +This directory contain a Makefile that allow the code to be built in a +standalone manner, independent of the larger LLVM build infrastructure. To build +the program you will need to have 'clang++' and 'llvm-config' in your path. If +you attempt to build using the LLVM 3.3 release, some minor modifications will +be required. + +This directory also contains a Python script that may be used to generate random +input for the program and test scripts to capture data for rough performance +comparisons. \ No newline at end of file diff --git a/examples/Kaleidoscope/MCJIT/lazy/genk-timing.py b/examples/Kaleidoscope/MCJIT/lazy/genk-timing.py new file mode 100644 index 00000000000..96dd6db5da3 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/lazy/genk-timing.py @@ -0,0 +1,219 @@ +#!/usr/bin/env python + +import sys +import random + +class TimingScriptGenerator: + """Used to generate a bash script which will invoke the toy and time it""" + def __init__(self, scriptname, outputname): + self.timeFile = outputname + self.shfile = open(scriptname, 'w') + self.shfile.write("echo \"\" > %s\n" % self.timeFile) + + def writeTimingCall(self, filename, numFuncs, funcsCalled, totalCalls): + """Echo some comments and invoke both versions of toy""" + rootname = filename + if '.' in filename: + rootname = filename[:filename.rfind('.')] + self.shfile.write("echo \"%s: Calls %d of %d functions, %d total\" >> %s\n" % (filename, funcsCalled, numFuncs, totalCalls, self.timeFile)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With MCJIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-mcjit < %s > %s-mcjit.out 2> %s-mcjit.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"With JIT\" >> %s\n" % self.timeFile) + self.shfile.write("/usr/bin/time -f \"Command %C\\n\\tuser time: %U s\\n\\tsytem time: %S s\\n\\tmax set: %M kb\"") + self.shfile.write(" -o %s -a " % self.timeFile) + self.shfile.write("./toy-jit < %s > %s-jit.out 2> %s-jit.err\n" % (filename, rootname, rootname)) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + self.shfile.write("echo \"\" >> %s\n" % self.timeFile) + +class KScriptGenerator: + """Used to generate random Kaleidoscope code""" + def __init__(self, filename): + self.kfile = open(filename, 'w') + self.nextFuncNum = 1 + self.lastFuncNum = None + self.callWeighting = 0.1 + # A mapping of calls within functions with no duplicates + self.calledFunctionTable = {} + # A list of function calls which will actually be executed + self.calledFunctions = [] + # A comprehensive mapping of calls within functions + # used for computing the total number of calls + self.comprehensiveCalledFunctionTable = {} + self.totalCallsExecuted = 0 + + def updateTotalCallCount(self, callee): + # Count this call + self.totalCallsExecuted += 1 + # Then count all the functions it calls + if callee in self.comprehensiveCalledFunctionTable: + for child in self.comprehensiveCalledFunctionTable[callee]: + self.updateTotalCallCount(child) + + def updateFunctionCallMap(self, caller, callee): + """Maintains a map of functions that are called from other functions""" + if not caller in self.calledFunctionTable: + self.calledFunctionTable[caller] = [] + if not callee in self.calledFunctionTable[caller]: + self.calledFunctionTable[caller].append(callee) + if not caller in self.comprehensiveCalledFunctionTable: + self.comprehensiveCalledFunctionTable[caller] = [] + self.comprehensiveCalledFunctionTable[caller].append(callee) + + def updateCalledFunctionList(self, callee): + """Maintains a list of functions that will actually be called""" + # Update the total call count + self.updateTotalCallCount(callee) + # If this function is already in the list, don't do anything else + if callee in self.calledFunctions: + return + # Add this function to the list of those that will be called. + self.calledFunctions.append(callee) + # If this function calls other functions, add them too + if callee in self.calledFunctionTable: + for subCallee in self.calledFunctionTable[callee]: + self.updateCalledFunctionList(subCallee) + + def setCallWeighting(self, weight): + """ Sets the probably of generating a function call""" + self.callWeighting = weight + + def writeln(self, line): + self.kfile.write(line + '\n') + + def writeComment(self, comment): + self.writeln('# ' + comment) + + def writeEmptyLine(self): + self.writeln("") + + def writePredefinedFunctions(self): + self.writeComment("Define ':' for sequencing: as a low-precedence operator that ignores operands") + self.writeComment("and just returns the RHS.") + self.writeln("def binary : 1 (x y) y;") + self.writeEmptyLine() + self.writeComment("Helper functions defined within toy") + self.writeln("extern putchard(x);") + self.writeln("extern printd(d);") + self.writeln("extern printlf();") + self.writeEmptyLine() + self.writeComment("Print the result of a function call") + self.writeln("def printresult(N Result)") + self.writeln(" # 'result('") + self.writeln(" putchard(114) : putchard(101) : putchard(115) : putchard(117) : putchard(108) : putchard(116) : putchard(40) :") + self.writeln(" printd(N) :"); + self.writeln(" # ') = '") + self.writeln(" putchard(41) : putchard(32) : putchard(61) : putchard(32) :") + self.writeln(" printd(Result) :"); + self.writeln(" printlf();") + self.writeEmptyLine() + + def writeRandomOperation(self, LValue, LHS, RHS): + shouldCallFunc = (self.lastFuncNum > 2 and random.random() < self.callWeighting) + if shouldCallFunc: + funcToCall = random.randrange(1, self.lastFuncNum - 1) + self.updateFunctionCallMap(self.lastFuncNum, funcToCall) + self.writeln(" %s = func%d(%s, %s) :" % (LValue, funcToCall, LHS, RHS)) + else: + possibleOperations = ["+", "-", "*", "/"] + operation = random.choice(possibleOperations) + if operation == "-": + # Don't let our intermediate value become zero + # This is complicated by the fact that '<' is our only comparison operator + self.writeln(" if %s < %s then" % (LHS, RHS)) + self.writeln(" %s = %s %s %s" % (LValue, LHS, operation, RHS)) + self.writeln(" else if %s < %s then" % (RHS, LHS)) + self.writeln(" %s = %s %s %s" % (LValue, LHS, operation, RHS)) + self.writeln(" else") + self.writeln(" %s = %s %s %f :" % (LValue, LHS, operation, random.uniform(1, 100))) + else: + self.writeln(" %s = %s %s %s :" % (LValue, LHS, operation, RHS)) + + def getNextFuncNum(self): + result = self.nextFuncNum + self.nextFuncNum += 1 + self.lastFuncNum = result + return result + + def writeFunction(self, elements): + funcNum = self.getNextFuncNum() + self.writeComment("Auto-generated function number %d" % funcNum) + self.writeln("def func%d(X Y)" % funcNum) + self.writeln(" var temp1 = X,") + self.writeln(" temp2 = Y,") + self.writeln(" temp3 in") + # Initialize the variable names to be rotated + first = "temp3" + second = "temp1" + third = "temp2" + # Write some random operations + for i in range(elements): + self.writeRandomOperation(first, second, third) + # Rotate the variables + temp = first + first = second + second = third + third = temp + self.writeln(" " + third + ";") + self.writeEmptyLine() + + def writeFunctionCall(self): + self.writeComment("Call the last function") + arg1 = random.uniform(1, 100) + arg2 = random.uniform(1, 100) + self.writeln("printresult(%d, func%d(%f, %f) )" % (self.lastFuncNum, self.lastFuncNum, arg1, arg2)) + self.writeEmptyLine() + self.updateCalledFunctionList(self.lastFuncNum) + + def writeFinalFunctionCounts(self): + self.writeComment("Called %d of %d functions" % (len(self.calledFunctions), self.lastFuncNum)) + +def generateKScript(filename, numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting, timingScript): + """ Generate a random Kaleidoscope script based on the given parameters """ + print "Generating " + filename + print(" %d functions, %d elements per function, %d functions between execution" % + (numFuncs, elementsPerFunc, funcsBetweenExec)) + print(" Call weighting = %f" % callWeighting) + script = KScriptGenerator(filename) + script.setCallWeighting(callWeighting) + script.writeComment("===========================================================================") + script.writeComment("Auto-generated script") + script.writeComment(" %d functions, %d elements per function, %d functions between execution" + % (numFuncs, elementsPerFunc, funcsBetweenExec)) + script.writeComment(" call weighting = %f" % callWeighting) + script.writeComment("===========================================================================") + script.writeEmptyLine() + script.writePredefinedFunctions() + funcsSinceLastExec = 0 + for i in range(numFuncs): + script.writeFunction(elementsPerFunc) + funcsSinceLastExec += 1 + if funcsSinceLastExec == funcsBetweenExec: + script.writeFunctionCall() + funcsSinceLastExec = 0 + # Always end with a function call + if funcsSinceLastExec > 0: + script.writeFunctionCall() + script.writeEmptyLine() + script.writeFinalFunctionCounts() + funcsCalled = len(script.calledFunctions) + print " Called %d of %d functions, %d total" % (funcsCalled, numFuncs, script.totalCallsExecuted) + timingScript.writeTimingCall(filename, numFuncs, funcsCalled, script.totalCallsExecuted) + +# Execution begins here +random.seed() + +timingScript = TimingScriptGenerator("time-toy.sh", "timing-data.txt") + +dataSets = [(5000, 3, 50, 0.50), (5000, 10, 100, 0.10), (5000, 10, 5, 0.10), (5000, 10, 1, 0.0), + (1000, 3, 10, 0.50), (1000, 10, 100, 0.10), (1000, 10, 5, 0.10), (1000, 10, 1, 0.0), + ( 200, 3, 2, 0.50), ( 200, 10, 40, 0.10), ( 200, 10, 2, 0.10), ( 200, 10, 1, 0.0)] + +# Generate the code +for (numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting) in dataSets: + filename = "test-%d-%d-%d-%d.k" % (numFuncs, elementsPerFunc, funcsBetweenExec, int(callWeighting * 100)) + generateKScript(filename, numFuncs, elementsPerFunc, funcsBetweenExec, callWeighting, timingScript) +print "All done!" diff --git a/examples/Kaleidoscope/MCJIT/lazy/toy-jit.cpp b/examples/Kaleidoscope/MCJIT/lazy/toy-jit.cpp new file mode 100644 index 00000000000..8650019fe3a --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/lazy/toy-jit.cpp @@ -0,0 +1,1167 @@ +#define MINIMAL_STDERR_OUTPUT + +#include "llvm/Analysis/Passes.h" +#include "llvm/Analysis/Verifier.h" +#include "llvm/ExecutionEngine/ExecutionEngine.h" +#include "llvm/ExecutionEngine/JIT.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/PassManager.h" +#include "llvm/Support/TargetSelect.h" +#include "llvm/Transforms/Scalar.h" +#include +#include +#include +#include + +using namespace llvm; + +//===----------------------------------------------------------------------===// +// Lexer +//===----------------------------------------------------------------------===// + +// The lexer returns tokens [0-255] if it is an unknown character, otherwise one +// of these for known things. +enum Token { + tok_eof = -1, + + // commands + tok_def = -2, tok_extern = -3, + + // primary + tok_identifier = -4, tok_number = -5, + + // control + tok_if = -6, tok_then = -7, tok_else = -8, + tok_for = -9, tok_in = -10, + + // operators + tok_binary = -11, tok_unary = -12, + + // var definition + tok_var = -13 +}; + +static std::string IdentifierStr; // Filled in if tok_identifier +static double NumVal; // Filled in if tok_number + +/// gettok - Return the next token from standard input. +static int gettok() { + static int LastChar = ' '; + + // Skip any whitespace. + while (isspace(LastChar)) + LastChar = getchar(); + + if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* + IdentifierStr = LastChar; + while (isalnum((LastChar = getchar()))) + IdentifierStr += LastChar; + + if (IdentifierStr == "def") return tok_def; + if (IdentifierStr == "extern") return tok_extern; + if (IdentifierStr == "if") return tok_if; + if (IdentifierStr == "then") return tok_then; + if (IdentifierStr == "else") return tok_else; + if (IdentifierStr == "for") return tok_for; + if (IdentifierStr == "in") return tok_in; + if (IdentifierStr == "binary") return tok_binary; + if (IdentifierStr == "unary") return tok_unary; + if (IdentifierStr == "var") return tok_var; + return tok_identifier; + } + + if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ + std::string NumStr; + do { + NumStr += LastChar; + LastChar = getchar(); + } while (isdigit(LastChar) || LastChar == '.'); + + NumVal = strtod(NumStr.c_str(), 0); + return tok_number; + } + + if (LastChar == '#') { + // Comment until end of line. + do LastChar = getchar(); + while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); + + if (LastChar != EOF) + return gettok(); + } + + // Check for end of file. Don't eat the EOF. + if (LastChar == EOF) + return tok_eof; + + // Otherwise, just return the character as its ascii value. + int ThisChar = LastChar; + LastChar = getchar(); + return ThisChar; +} + +//===----------------------------------------------------------------------===// +// Abstract Syntax Tree (aka Parse Tree) +//===----------------------------------------------------------------------===// + +/// ExprAST - Base class for all expression nodes. +class ExprAST { +public: + virtual ~ExprAST() {} + virtual Value *Codegen() = 0; +}; + +/// NumberExprAST - Expression class for numeric literals like "1.0". +class NumberExprAST : public ExprAST { + double Val; +public: + NumberExprAST(double val) : Val(val) {} + virtual Value *Codegen(); +}; + +/// VariableExprAST - Expression class for referencing a variable, like "a". +class VariableExprAST : public ExprAST { + std::string Name; +public: + VariableExprAST(const std::string &name) : Name(name) {} + const std::string &getName() const { return Name; } + virtual Value *Codegen(); +}; + +/// UnaryExprAST - Expression class for a unary operator. +class UnaryExprAST : public ExprAST { + char Opcode; + ExprAST *Operand; +public: + UnaryExprAST(char opcode, ExprAST *operand) + : Opcode(opcode), Operand(operand) {} + virtual Value *Codegen(); +}; + +/// BinaryExprAST - Expression class for a binary operator. +class BinaryExprAST : public ExprAST { + char Op; + ExprAST *LHS, *RHS; +public: + BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) + : Op(op), LHS(lhs), RHS(rhs) {} + virtual Value *Codegen(); +}; + +/// CallExprAST - Expression class for function calls. +class CallExprAST : public ExprAST { + std::string Callee; + std::vector Args; +public: + CallExprAST(const std::string &callee, std::vector &args) + : Callee(callee), Args(args) {} + virtual Value *Codegen(); +}; + +/// IfExprAST - Expression class for if/then/else. +class IfExprAST : public ExprAST { + ExprAST *Cond, *Then, *Else; +public: + IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) + : Cond(cond), Then(then), Else(_else) {} + virtual Value *Codegen(); +}; + +/// ForExprAST - Expression class for for/in. +class ForExprAST : public ExprAST { + std::string VarName; + ExprAST *Start, *End, *Step, *Body; +public: + ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, + ExprAST *step, ExprAST *body) + : VarName(varname), Start(start), End(end), Step(step), Body(body) {} + virtual Value *Codegen(); +}; + +/// VarExprAST - Expression class for var/in +class VarExprAST : public ExprAST { + std::vector > VarNames; + ExprAST *Body; +public: + VarExprAST(const std::vector > &varnames, + ExprAST *body) + : VarNames(varnames), Body(body) {} + + virtual Value *Codegen(); +}; + +/// PrototypeAST - This class represents the "prototype" for a function, +/// which captures its argument names as well as if it is an operator. +class PrototypeAST { + std::string Name; + std::vector Args; + bool isOperator; + unsigned Precedence; // Precedence if a binary op. +public: + PrototypeAST(const std::string &name, const std::vector &args, + bool isoperator = false, unsigned prec = 0) + : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} + + bool isUnaryOp() const { return isOperator && Args.size() == 1; } + bool isBinaryOp() const { return isOperator && Args.size() == 2; } + + char getOperatorName() const { + assert(isUnaryOp() || isBinaryOp()); + return Name[Name.size()-1]; + } + + unsigned getBinaryPrecedence() const { return Precedence; } + + Function *Codegen(); + + void CreateArgumentAllocas(Function *F); +}; + +/// FunctionAST - This class represents a function definition itself. +class FunctionAST { + PrototypeAST *Proto; + ExprAST *Body; +public: + FunctionAST(PrototypeAST *proto, ExprAST *body) + : Proto(proto), Body(body) {} + + Function *Codegen(); +}; + +//===----------------------------------------------------------------------===// +// Parser +//===----------------------------------------------------------------------===// + +/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current +/// token the parser is looking at. getNextToken reads another token from the +/// lexer and updates CurTok with its results. +static int CurTok; +static int getNextToken() { + return CurTok = gettok(); +} + +/// BinopPrecedence - This holds the precedence for each binary operator that is +/// defined. +static std::map BinopPrecedence; + +/// GetTokPrecedence - Get the precedence of the pending binary operator token. +static int GetTokPrecedence() { + if (!isascii(CurTok)) + return -1; + + // Make sure it's a declared binop. + int TokPrec = BinopPrecedence[CurTok]; + if (TokPrec <= 0) return -1; + return TokPrec; +} + +/// Error* - These are little helper functions for error handling. +ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} +PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } +FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } + +static ExprAST *ParseExpression(); + +/// identifierexpr +/// ::= identifier +/// ::= identifier '(' expression* ')' +static ExprAST *ParseIdentifierExpr() { + std::string IdName = IdentifierStr; + + getNextToken(); // eat identifier. + + if (CurTok != '(') // Simple variable ref. + return new VariableExprAST(IdName); + + // Call. + getNextToken(); // eat ( + std::vector Args; + if (CurTok != ')') { + while (1) { + ExprAST *Arg = ParseExpression(); + if (!Arg) return 0; + Args.push_back(Arg); + + if (CurTok == ')') break; + + if (CurTok != ',') + return Error("Expected ')' or ',' in argument list"); + getNextToken(); + } + } + + // Eat the ')'. + getNextToken(); + + return new CallExprAST(IdName, Args); +} + +/// numberexpr ::= number +static ExprAST *ParseNumberExpr() { + ExprAST *Result = new NumberExprAST(NumVal); + getNextToken(); // consume the number + return Result; +} + +/// parenexpr ::= '(' expression ')' +static ExprAST *ParseParenExpr() { + getNextToken(); // eat (. + ExprAST *V = ParseExpression(); + if (!V) return 0; + + if (CurTok != ')') + return Error("expected ')'"); + getNextToken(); // eat ). + return V; +} + +/// ifexpr ::= 'if' expression 'then' expression 'else' expression +static ExprAST *ParseIfExpr() { + getNextToken(); // eat the if. + + // condition. + ExprAST *Cond = ParseExpression(); + if (!Cond) return 0; + + if (CurTok != tok_then) + return Error("expected then"); + getNextToken(); // eat the then + + ExprAST *Then = ParseExpression(); + if (Then == 0) return 0; + + if (CurTok != tok_else) + return Error("expected else"); + + getNextToken(); + + ExprAST *Else = ParseExpression(); + if (!Else) return 0; + + return new IfExprAST(Cond, Then, Else); +} + +/// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression +static ExprAST *ParseForExpr() { + getNextToken(); // eat the for. + + if (CurTok != tok_identifier) + return Error("expected identifier after for"); + + std::string IdName = IdentifierStr; + getNextToken(); // eat identifier. + + if (CurTok != '=') + return Error("expected '=' after for"); + getNextToken(); // eat '='. + + + ExprAST *Start = ParseExpression(); + if (Start == 0) return 0; + if (CurTok != ',') + return Error("expected ',' after for start value"); + getNextToken(); + + ExprAST *End = ParseExpression(); + if (End == 0) return 0; + + // The step value is optional. + ExprAST *Step = 0; + if (CurTok == ',') { + getNextToken(); + Step = ParseExpression(); + if (Step == 0) return 0; + } + + if (CurTok != tok_in) + return Error("expected 'in' after for"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new ForExprAST(IdName, Start, End, Step, Body); +} + +/// varexpr ::= 'var' identifier ('=' expression)? +// (',' identifier ('=' expression)?)* 'in' expression +static ExprAST *ParseVarExpr() { + getNextToken(); // eat the var. + + std::vector > VarNames; + + // At least one variable name is required. + if (CurTok != tok_identifier) + return Error("expected identifier after var"); + + while (1) { + std::string Name = IdentifierStr; + getNextToken(); // eat identifier. + + // Read the optional initializer. + ExprAST *Init = 0; + if (CurTok == '=') { + getNextToken(); // eat the '='. + + Init = ParseExpression(); + if (Init == 0) return 0; + } + + VarNames.push_back(std::make_pair(Name, Init)); + + // End of var list, exit loop. + if (CurTok != ',') break; + getNextToken(); // eat the ','. + + if (CurTok != tok_identifier) + return Error("expected identifier list after var"); + } + + // At this point, we have to have 'in'. + if (CurTok != tok_in) + return Error("expected 'in' keyword after 'var'"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new VarExprAST(VarNames, Body); +} + +/// primary +/// ::= identifierexpr +/// ::= numberexpr +/// ::= parenexpr +/// ::= ifexpr +/// ::= forexpr +/// ::= varexpr +static ExprAST *ParsePrimary() { + switch (CurTok) { + default: return Error("unknown token when expecting an expression"); + case tok_identifier: return ParseIdentifierExpr(); + case tok_number: return ParseNumberExpr(); + case '(': return ParseParenExpr(); + case tok_if: return ParseIfExpr(); + case tok_for: return ParseForExpr(); + case tok_var: return ParseVarExpr(); + } +} + +/// unary +/// ::= primary +/// ::= '!' unary +static ExprAST *ParseUnary() { + // If the current token is not an operator, it must be a primary expr. + if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') + return ParsePrimary(); + + // If this is a unary operator, read it. + int Opc = CurTok; + getNextToken(); + if (ExprAST *Operand = ParseUnary()) + return new UnaryExprAST(Opc, Operand); + return 0; +} + +/// binoprhs +/// ::= ('+' unary)* +static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { + // If this is a binop, find its precedence. + while (1) { + int TokPrec = GetTokPrecedence(); + + // If this is a binop that binds at least as tightly as the current binop, + // consume it, otherwise we are done. + if (TokPrec < ExprPrec) + return LHS; + + // Okay, we know this is a binop. + int BinOp = CurTok; + getNextToken(); // eat binop + + // Parse the unary expression after the binary operator. + ExprAST *RHS = ParseUnary(); + if (!RHS) return 0; + + // If BinOp binds less tightly with RHS than the operator after RHS, let + // the pending operator take RHS as its LHS. + int NextPrec = GetTokPrecedence(); + if (TokPrec < NextPrec) { + RHS = ParseBinOpRHS(TokPrec+1, RHS); + if (RHS == 0) return 0; + } + + // Merge LHS/RHS. + LHS = new BinaryExprAST(BinOp, LHS, RHS); + } +} + +/// expression +/// ::= unary binoprhs +/// +static ExprAST *ParseExpression() { + ExprAST *LHS = ParseUnary(); + if (!LHS) return 0; + + return ParseBinOpRHS(0, LHS); +} + +/// prototype +/// ::= id '(' id* ')' +/// ::= binary LETTER number? (id, id) +/// ::= unary LETTER (id) +static PrototypeAST *ParsePrototype() { + std::string FnName; + + unsigned Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. + unsigned BinaryPrecedence = 30; + + switch (CurTok) { + default: + return ErrorP("Expected function name in prototype"); + case tok_identifier: + FnName = IdentifierStr; + Kind = 0; + getNextToken(); + break; + case tok_unary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected unary operator"); + FnName = "unary"; + FnName += (char)CurTok; + Kind = 1; + getNextToken(); + break; + case tok_binary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected binary operator"); + FnName = "binary"; + FnName += (char)CurTok; + Kind = 2; + getNextToken(); + + // Read the precedence if present. + if (CurTok == tok_number) { + if (NumVal < 1 || NumVal > 100) + return ErrorP("Invalid precedecnce: must be 1..100"); + BinaryPrecedence = (unsigned)NumVal; + getNextToken(); + } + break; + } + + if (CurTok != '(') + return ErrorP("Expected '(' in prototype"); + + std::vector ArgNames; + while (getNextToken() == tok_identifier) + ArgNames.push_back(IdentifierStr); + if (CurTok != ')') + return ErrorP("Expected ')' in prototype"); + + // success. + getNextToken(); // eat ')'. + + // Verify right number of names for operator. + if (Kind && ArgNames.size() != Kind) + return ErrorP("Invalid number of operands for operator"); + + return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); +} + +/// definition ::= 'def' prototype expression +static FunctionAST *ParseDefinition() { + getNextToken(); // eat def. + PrototypeAST *Proto = ParsePrototype(); + if (Proto == 0) return 0; + + if (ExprAST *E = ParseExpression()) + return new FunctionAST(Proto, E); + return 0; +} + +/// toplevelexpr ::= expression +static FunctionAST *ParseTopLevelExpr() { + if (ExprAST *E = ParseExpression()) { + // Make an anonymous proto. + PrototypeAST *Proto = new PrototypeAST("", std::vector()); + return new FunctionAST(Proto, E); + } + return 0; +} + +/// external ::= 'extern' prototype +static PrototypeAST *ParseExtern() { + getNextToken(); // eat extern. + return ParsePrototype(); +} + +//===----------------------------------------------------------------------===// +// Code Generation +//===----------------------------------------------------------------------===// + +static Module *TheModule; +static FunctionPassManager *TheFPM; +static IRBuilder<> Builder(getGlobalContext()); +static std::map NamedValues; + +Value *ErrorV(const char *Str) { Error(Str); return 0; } + +/// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of +/// the function. This is used for mutable variables etc. +static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction, + const std::string &VarName) { + IRBuilder<> TmpB(&TheFunction->getEntryBlock(), + TheFunction->getEntryBlock().begin()); + return TmpB.CreateAlloca(Type::getDoubleTy(getGlobalContext()), 0, + VarName.c_str()); +} + +Value *NumberExprAST::Codegen() { + return ConstantFP::get(getGlobalContext(), APFloat(Val)); +} + +Value *VariableExprAST::Codegen() { + // Look this variable up in the function. + Value *V = NamedValues[Name]; + if (V == 0) return ErrorV("Unknown variable name"); + + // Load the value. + return Builder.CreateLoad(V, Name.c_str()); +} + +Value *UnaryExprAST::Codegen() { + Value *OperandV = Operand->Codegen(); + if (OperandV == 0) return 0; +#ifdef USE_MCJIT + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("unary")+Opcode)); +#else + Function *F = TheModule->getFunction(std::string("unary")+Opcode); +#endif + if (F == 0) + return ErrorV("Unknown unary operator"); + + return Builder.CreateCall(F, OperandV, "unop"); +} + +Value *BinaryExprAST::Codegen() { + // Special case '=' because we don't want to emit the LHS as an expression. + if (Op == '=') { + // Assignment requires the LHS to be an identifier. + VariableExprAST *LHSE = dynamic_cast(LHS); + if (!LHSE) + return ErrorV("destination of '=' must be a variable"); + // Codegen the RHS. + Value *Val = RHS->Codegen(); + if (Val == 0) return 0; + + // Look up the name. + Value *Variable = NamedValues[LHSE->getName()]; + if (Variable == 0) return ErrorV("Unknown variable name"); + + Builder.CreateStore(Val, Variable); + return Val; + } + + Value *L = LHS->Codegen(); + Value *R = RHS->Codegen(); + if (L == 0 || R == 0) return 0; + + switch (Op) { + case '+': return Builder.CreateFAdd(L, R, "addtmp"); + case '-': return Builder.CreateFSub(L, R, "subtmp"); + case '*': return Builder.CreateFMul(L, R, "multmp"); + case '/': return Builder.CreateFDiv(L, R, "divtmp"); + case '<': + L = Builder.CreateFCmpULT(L, R, "cmptmp"); + // Convert bool 0/1 to double 0.0 or 1.0 + return Builder.CreateUIToFP(L, Type::getDoubleTy(getGlobalContext()), + "booltmp"); + default: break; + } + + // If it wasn't a builtin binary operator, it must be a user defined one. Emit + // a call to it. + Function *F = TheModule->getFunction(std::string("binary")+Op); + assert(F && "binary operator not found!"); + + Value *Ops[] = { L, R }; + return Builder.CreateCall(F, Ops, "binop"); +} + +Value *CallExprAST::Codegen() { + // Look up the name in the global module table. + Function *CalleeF = TheModule->getFunction(Callee); + if (CalleeF == 0) { + char error_str[64]; + sprintf(error_str, "Unknown function referenced %s", Callee.c_str()); + return ErrorV(error_str); + } + + // If argument mismatch error. + if (CalleeF->arg_size() != Args.size()) + return ErrorV("Incorrect # arguments passed"); + + std::vector ArgsV; + for (unsigned i = 0, e = Args.size(); i != e; ++i) { + ArgsV.push_back(Args[i]->Codegen()); + if (ArgsV.back() == 0) return 0; + } + + return Builder.CreateCall(CalleeF, ArgsV, "calltmp"); +} + +Value *IfExprAST::Codegen() { + Value *CondV = Cond->Codegen(); + if (CondV == 0) return 0; + + // Convert condition to a bool by comparing equal to 0.0. + CondV = Builder.CreateFCmpONE(CondV, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "ifcond"); + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create blocks for the then and else cases. Insert the 'then' block at the + // end of the function. + BasicBlock *ThenBB = BasicBlock::Create(getGlobalContext(), "then", TheFunction); + BasicBlock *ElseBB = BasicBlock::Create(getGlobalContext(), "else"); + BasicBlock *MergeBB = BasicBlock::Create(getGlobalContext(), "ifcont"); + + Builder.CreateCondBr(CondV, ThenBB, ElseBB); + + // Emit then value. + Builder.SetInsertPoint(ThenBB); + + Value *ThenV = Then->Codegen(); + if (ThenV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Then' can change the current block, update ThenBB for the PHI. + ThenBB = Builder.GetInsertBlock(); + + // Emit else block. + TheFunction->getBasicBlockList().push_back(ElseBB); + Builder.SetInsertPoint(ElseBB); + + Value *ElseV = Else->Codegen(); + if (ElseV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Else' can change the current block, update ElseBB for the PHI. + ElseBB = Builder.GetInsertBlock(); + + // Emit merge block. + TheFunction->getBasicBlockList().push_back(MergeBB); + Builder.SetInsertPoint(MergeBB); + PHINode *PN = Builder.CreatePHI(Type::getDoubleTy(getGlobalContext()), 2, + "iftmp"); + + PN->addIncoming(ThenV, ThenBB); + PN->addIncoming(ElseV, ElseBB); + return PN; +} + +Value *ForExprAST::Codegen() { + // Output this as: + // var = alloca double + // ... + // start = startexpr + // store start -> var + // goto loop + // loop: + // ... + // bodyexpr + // ... + // loopend: + // step = stepexpr + // endcond = endexpr + // + // curvar = load var + // nextvar = curvar + step + // store nextvar -> var + // br endcond, loop, endloop + // outloop: + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create an alloca for the variable in the entry block. + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + + // Emit the start code first, without 'variable' in scope. + Value *StartVal = Start->Codegen(); + if (StartVal == 0) return 0; + + // Store the value into the alloca. + Builder.CreateStore(StartVal, Alloca); + + // Make the new basic block for the loop header, inserting after current + // block. + BasicBlock *LoopBB = BasicBlock::Create(getGlobalContext(), "loop", TheFunction); + + // Insert an explicit fall through from the current block to the LoopBB. + Builder.CreateBr(LoopBB); + + // Start insertion in LoopBB. + Builder.SetInsertPoint(LoopBB); + + // Within the loop, the variable is defined equal to the PHI node. If it + // shadows an existing variable, we have to restore it, so save it now. + AllocaInst *OldVal = NamedValues[VarName]; + NamedValues[VarName] = Alloca; + + // Emit the body of the loop. This, like any other expr, can change the + // current BB. Note that we ignore the value computed by the body, but don't + // allow an error. + if (Body->Codegen() == 0) + return 0; + + // Emit the step value. + Value *StepVal; + if (Step) { + StepVal = Step->Codegen(); + if (StepVal == 0) return 0; + } else { + // If not specified, use 1.0. + StepVal = ConstantFP::get(getGlobalContext(), APFloat(1.0)); + } + + // Compute the end condition. + Value *EndCond = End->Codegen(); + if (EndCond == 0) return EndCond; + + // Reload, increment, and restore the alloca. This handles the case where + // the body of the loop mutates the variable. + Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str()); + Value *NextVar = Builder.CreateFAdd(CurVar, StepVal, "nextvar"); + Builder.CreateStore(NextVar, Alloca); + + // Convert condition to a bool by comparing equal to 0.0. + EndCond = Builder.CreateFCmpONE(EndCond, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "loopcond"); + + // Create the "after loop" block and insert it. + BasicBlock *AfterBB = BasicBlock::Create(getGlobalContext(), "afterloop", TheFunction); + + // Insert the conditional branch into the end of LoopEndBB. + Builder.CreateCondBr(EndCond, LoopBB, AfterBB); + + // Any new code will be inserted in AfterBB. + Builder.SetInsertPoint(AfterBB); + + // Restore the unshadowed variable. + if (OldVal) + NamedValues[VarName] = OldVal; + else + NamedValues.erase(VarName); + + + // for expr always returns 0.0. + return Constant::getNullValue(Type::getDoubleTy(getGlobalContext())); +} + +Value *VarExprAST::Codegen() { + std::vector OldBindings; + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Register all variables and emit their initializer. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) { + const std::string &VarName = VarNames[i].first; + ExprAST *Init = VarNames[i].second; + + // Emit the initializer before adding the variable to scope, this prevents + // the initializer from referencing the variable itself, and permits stuff + // like this: + // var a = 1 in + // var a = a in ... # refers to outer 'a'. + Value *InitVal; + if (Init) { + InitVal = Init->Codegen(); + if (InitVal == 0) return 0; + } else { // If not specified, use 0.0. + InitVal = ConstantFP::get(getGlobalContext(), APFloat(0.0)); + } + + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + Builder.CreateStore(InitVal, Alloca); + + // Remember the old variable binding so that we can restore the binding when + // we unrecurse. + OldBindings.push_back(NamedValues[VarName]); + + // Remember this binding. + NamedValues[VarName] = Alloca; + } + + // Codegen the body, now that all vars are in scope. + Value *BodyVal = Body->Codegen(); + if (BodyVal == 0) return 0; + + // Pop all our variables from scope. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) + NamedValues[VarNames[i].first] = OldBindings[i]; + + // Return the body computation. + return BodyVal; +} + +Function *PrototypeAST::Codegen() { + // Make the function type: double(double,double) etc. + std::vector Doubles(Args.size(), + Type::getDoubleTy(getGlobalContext())); + FunctionType *FT = FunctionType::get(Type::getDoubleTy(getGlobalContext()), + Doubles, false); + + Function *F = Function::Create(FT, Function::ExternalLinkage, Name, TheModule); + // If F conflicted, there was already something named 'Name'. If it has a + // body, don't allow redefinition or reextern. + if (F->getName() != Name) { + // Delete the one we just made and get the existing one. + F->eraseFromParent(); + F = TheModule->getFunction(Name); + // If F already has a body, reject this. + if (!F->empty()) { + ErrorF("redefinition of function"); + return 0; + } + // If F took a different number of args, reject. + if (F->arg_size() != Args.size()) { + ErrorF("redefinition of function with different # args"); + return 0; + } + } + + // Set names for all arguments. + unsigned Idx = 0; + for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); + ++AI, ++Idx) + AI->setName(Args[Idx]); + + return F; +} + +/// CreateArgumentAllocas - Create an alloca for each argument and register the +/// argument in the symbol table so that references to it will succeed. +void PrototypeAST::CreateArgumentAllocas(Function *F) { + Function::arg_iterator AI = F->arg_begin(); + for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) { + // Create an alloca for this variable. + AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]); + + // Store the initial value into the alloca. + Builder.CreateStore(AI, Alloca); + + // Add arguments to variable symbol table. + NamedValues[Args[Idx]] = Alloca; + } +} + +Function *FunctionAST::Codegen() { + NamedValues.clear(); + + Function *TheFunction = Proto->Codegen(); + if (TheFunction == 0) + return 0; + + // If this is an operator, install it. + if (Proto->isBinaryOp()) + BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); + + // Create a new basic block to start insertion into. + BasicBlock *BB = BasicBlock::Create(getGlobalContext(), "entry", TheFunction); + Builder.SetInsertPoint(BB); + + // Add all arguments to the symbol table and create their allocas. + Proto->CreateArgumentAllocas(TheFunction); + + if (Value *RetVal = Body->Codegen()) { + // Finish off the function. + Builder.CreateRet(RetVal); + + // Validate the generated code, checking for consistency. + verifyFunction(*TheFunction); + + // Optimize the function. + TheFPM->run(*TheFunction); + + return TheFunction; + } + + // Error reading body, remove function. + TheFunction->eraseFromParent(); + + if (Proto->isBinaryOp()) + BinopPrecedence.erase(Proto->getOperatorName()); + return 0; +} + +//===----------------------------------------------------------------------===// +// Top-Level parsing and JIT Driver +//===----------------------------------------------------------------------===// + +static ExecutionEngine *TheExecutionEngine; + +static void HandleDefinition() { + if (FunctionAST *F = ParseDefinition()) { + if (Function *LF = F->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read function definition:"); + LF->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleExtern() { + if (PrototypeAST *P = ParseExtern()) { + if (Function *F = P->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read extern: "); + F->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleTopLevelExpression() { + // Evaluate a top-level expression into an anonymous function. + if (FunctionAST *F = ParseTopLevelExpr()) { + if (Function *LF = F->Codegen()) { + // JIT the function, returning a function pointer. + void *FPtr = TheExecutionEngine->getPointerToFunction(LF); + // Cast it to the right type (takes no arguments, returns a double) so we + // can call it as a native function. + double (*FP)() = (double (*)())(intptr_t)FPtr; +#ifdef MINIMAL_STDERR_OUTPUT + FP(); +#else + fprintf(stderr, "Evaluated to %f\n", FP()); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +/// top ::= definition | external | expression | ';' +static void MainLoop() { + while (1) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + switch (CurTok) { + case tok_eof: return; + case ';': getNextToken(); break; // ignore top-level semicolons. + case tok_def: HandleDefinition(); break; + case tok_extern: HandleExtern(); break; + default: HandleTopLevelExpression(); break; + } + } +} + +//===----------------------------------------------------------------------===// +// "Library" functions that can be "extern'd" from user code. +//===----------------------------------------------------------------------===// + +/// putchard - putchar that takes a double and returns 0. +extern "C" +double putchard(double X) { + putchar((char)X); + return 0; +} + +/// printd - printf that takes a double prints it as "%f\n", returning 0. +extern "C" +double printd(double X) { + printf("%f", X); + return 0; +} + +extern "C" +double printlf() { + printf("\n"); + return 0; +} + +//===----------------------------------------------------------------------===// +// Main driver code. +//===----------------------------------------------------------------------===// + +int main(int argc, char **argv) { + InitializeNativeTarget(); + LLVMContext &Context = getGlobalContext(); + + // Install standard binary operators. + // 1 is lowest precedence. + BinopPrecedence['='] = 2; + BinopPrecedence['<'] = 10; + BinopPrecedence['+'] = 20; + BinopPrecedence['-'] = 20; + BinopPrecedence['/'] = 40; + BinopPrecedence['*'] = 40; // highest. + + // Make the module, which holds all the code. + TheModule = new Module("my cool jit", Context); + + // Create the JIT. This takes ownership of the module. + std::string ErrStr; + TheExecutionEngine = EngineBuilder(TheModule).setErrorStr(&ErrStr).create(); + if (!TheExecutionEngine) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + FunctionPassManager OurFPM(TheModule); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + OurFPM.add(new DataLayout(*TheExecutionEngine->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + OurFPM.add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + OurFPM.add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + OurFPM.add(createInstructionCombiningPass()); + // Reassociate expressions. + OurFPM.add(createReassociatePass()); + // Eliminate Common SubExpressions. + OurFPM.add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + OurFPM.add(createCFGSimplificationPass()); + + OurFPM.doInitialization(); + + // Set the global so the code gen can use this. + TheFPM = &OurFPM; + + // Prime the first token. +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + getNextToken(); + + // Run the main "interpreter loop" now. + MainLoop(); + + // Print out all of the generated code. + TheFPM = 0; +#ifndef MINIMAL_STDERR_OUTPUT + TheModule->dump(); +#endif + return 0; +} diff --git a/examples/Kaleidoscope/MCJIT/lazy/toy.cpp b/examples/Kaleidoscope/MCJIT/lazy/toy.cpp new file mode 100644 index 00000000000..0a8d80e25a8 --- /dev/null +++ b/examples/Kaleidoscope/MCJIT/lazy/toy.cpp @@ -0,0 +1,1422 @@ +#define MINIMAL_STDERR_OUTPUT + +#include "llvm/Analysis/Passes.h" +#include "llvm/Analysis/Verifier.h" +#include "llvm/ExecutionEngine/ExecutionEngine.h" +#include "llvm/ExecutionEngine/MCJIT.h" +#include "llvm/ExecutionEngine/SectionMemoryManager.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/PassManager.h" +#include "llvm/Support/TargetSelect.h" +#include "llvm/Transforms/Scalar.h" +#include +#include +#include +#include +using namespace llvm; + +//===----------------------------------------------------------------------===// +// Lexer +//===----------------------------------------------------------------------===// + +// The lexer returns tokens [0-255] if it is an unknown character, otherwise one +// of these for known things. +enum Token { + tok_eof = -1, + + // commands + tok_def = -2, tok_extern = -3, + + // primary + tok_identifier = -4, tok_number = -5, + + // control + tok_if = -6, tok_then = -7, tok_else = -8, + tok_for = -9, tok_in = -10, + + // operators + tok_binary = -11, tok_unary = -12, + + // var definition + tok_var = -13 +}; + +static std::string IdentifierStr; // Filled in if tok_identifier +static double NumVal; // Filled in if tok_number + +/// gettok - Return the next token from standard input. +static int gettok() { + static int LastChar = ' '; + + // Skip any whitespace. + while (isspace(LastChar)) + LastChar = getchar(); + + if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* + IdentifierStr = LastChar; + while (isalnum((LastChar = getchar()))) + IdentifierStr += LastChar; + + if (IdentifierStr == "def") return tok_def; + if (IdentifierStr == "extern") return tok_extern; + if (IdentifierStr == "if") return tok_if; + if (IdentifierStr == "then") return tok_then; + if (IdentifierStr == "else") return tok_else; + if (IdentifierStr == "for") return tok_for; + if (IdentifierStr == "in") return tok_in; + if (IdentifierStr == "binary") return tok_binary; + if (IdentifierStr == "unary") return tok_unary; + if (IdentifierStr == "var") return tok_var; + return tok_identifier; + } + + if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ + std::string NumStr; + do { + NumStr += LastChar; + LastChar = getchar(); + } while (isdigit(LastChar) || LastChar == '.'); + + NumVal = strtod(NumStr.c_str(), 0); + return tok_number; + } + + if (LastChar == '#') { + // Comment until end of line. + do LastChar = getchar(); + while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); + + if (LastChar != EOF) + return gettok(); + } + + // Check for end of file. Don't eat the EOF. + if (LastChar == EOF) + return tok_eof; + + // Otherwise, just return the character as its ascii value. + int ThisChar = LastChar; + LastChar = getchar(); + return ThisChar; +} + +//===----------------------------------------------------------------------===// +// Abstract Syntax Tree (aka Parse Tree) +//===----------------------------------------------------------------------===// + +/// ExprAST - Base class for all expression nodes. +class ExprAST { +public: + virtual ~ExprAST() {} + virtual Value *Codegen() = 0; +}; + +/// NumberExprAST - Expression class for numeric literals like "1.0". +class NumberExprAST : public ExprAST { + double Val; +public: + NumberExprAST(double val) : Val(val) {} + virtual Value *Codegen(); +}; + +/// VariableExprAST - Expression class for referencing a variable, like "a". +class VariableExprAST : public ExprAST { + std::string Name; +public: + VariableExprAST(const std::string &name) : Name(name) {} + const std::string &getName() const { return Name; } + virtual Value *Codegen(); +}; + +/// UnaryExprAST - Expression class for a unary operator. +class UnaryExprAST : public ExprAST { + char Opcode; + ExprAST *Operand; +public: + UnaryExprAST(char opcode, ExprAST *operand) + : Opcode(opcode), Operand(operand) {} + virtual Value *Codegen(); +}; + +/// BinaryExprAST - Expression class for a binary operator. +class BinaryExprAST : public ExprAST { + char Op; + ExprAST *LHS, *RHS; +public: + BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) + : Op(op), LHS(lhs), RHS(rhs) {} + virtual Value *Codegen(); +}; + +/// CallExprAST - Expression class for function calls. +class CallExprAST : public ExprAST { + std::string Callee; + std::vector Args; +public: + CallExprAST(const std::string &callee, std::vector &args) + : Callee(callee), Args(args) {} + virtual Value *Codegen(); +}; + +/// IfExprAST - Expression class for if/then/else. +class IfExprAST : public ExprAST { + ExprAST *Cond, *Then, *Else; +public: + IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) + : Cond(cond), Then(then), Else(_else) {} + virtual Value *Codegen(); +}; + +/// ForExprAST - Expression class for for/in. +class ForExprAST : public ExprAST { + std::string VarName; + ExprAST *Start, *End, *Step, *Body; +public: + ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, + ExprAST *step, ExprAST *body) + : VarName(varname), Start(start), End(end), Step(step), Body(body) {} + virtual Value *Codegen(); +}; + +/// VarExprAST - Expression class for var/in +class VarExprAST : public ExprAST { + std::vector > VarNames; + ExprAST *Body; +public: + VarExprAST(const std::vector > &varnames, + ExprAST *body) + : VarNames(varnames), Body(body) {} + + virtual Value *Codegen(); +}; + +/// PrototypeAST - This class represents the "prototype" for a function, +/// which captures its argument names as well as if it is an operator. +class PrototypeAST { + std::string Name; + std::vector Args; + bool isOperator; + unsigned Precedence; // Precedence if a binary op. +public: + PrototypeAST(const std::string &name, const std::vector &args, + bool isoperator = false, unsigned prec = 0) + : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} + + bool isUnaryOp() const { return isOperator && Args.size() == 1; } + bool isBinaryOp() const { return isOperator && Args.size() == 2; } + + char getOperatorName() const { + assert(isUnaryOp() || isBinaryOp()); + return Name[Name.size()-1]; + } + + unsigned getBinaryPrecedence() const { return Precedence; } + + Function *Codegen(); + + void CreateArgumentAllocas(Function *F); +}; + +/// FunctionAST - This class represents a function definition itself. +class FunctionAST { + PrototypeAST *Proto; + ExprAST *Body; +public: + FunctionAST(PrototypeAST *proto, ExprAST *body) + : Proto(proto), Body(body) {} + + Function *Codegen(); +}; + +//===----------------------------------------------------------------------===// +// Parser +//===----------------------------------------------------------------------===// + +/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current +/// token the parser is looking at. getNextToken reads another token from the +/// lexer and updates CurTok with its results. +static int CurTok; +static int getNextToken() { + return CurTok = gettok(); +} + +/// BinopPrecedence - This holds the precedence for each binary operator that is +/// defined. +static std::map BinopPrecedence; + +/// GetTokPrecedence - Get the precedence of the pending binary operator token. +static int GetTokPrecedence() { + if (!isascii(CurTok)) + return -1; + + // Make sure it's a declared binop. + int TokPrec = BinopPrecedence[CurTok]; + if (TokPrec <= 0) return -1; + return TokPrec; +} + +/// Error* - These are little helper functions for error handling. +ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} +PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } +FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } + +static ExprAST *ParseExpression(); + +/// identifierexpr +/// ::= identifier +/// ::= identifier '(' expression* ')' +static ExprAST *ParseIdentifierExpr() { + std::string IdName = IdentifierStr; + + getNextToken(); // eat identifier. + + if (CurTok != '(') // Simple variable ref. + return new VariableExprAST(IdName); + + // Call. + getNextToken(); // eat ( + std::vector Args; + if (CurTok != ')') { + while (1) { + ExprAST *Arg = ParseExpression(); + if (!Arg) return 0; + Args.push_back(Arg); + + if (CurTok == ')') break; + + if (CurTok != ',') + return Error("Expected ')' or ',' in argument list"); + getNextToken(); + } + } + + // Eat the ')'. + getNextToken(); + + return new CallExprAST(IdName, Args); +} + +/// numberexpr ::= number +static ExprAST *ParseNumberExpr() { + ExprAST *Result = new NumberExprAST(NumVal); + getNextToken(); // consume the number + return Result; +} + +/// parenexpr ::= '(' expression ')' +static ExprAST *ParseParenExpr() { + getNextToken(); // eat (. + ExprAST *V = ParseExpression(); + if (!V) return 0; + + if (CurTok != ')') + return Error("expected ')'"); + getNextToken(); // eat ). + return V; +} + +/// ifexpr ::= 'if' expression 'then' expression 'else' expression +static ExprAST *ParseIfExpr() { + getNextToken(); // eat the if. + + // condition. + ExprAST *Cond = ParseExpression(); + if (!Cond) return 0; + + if (CurTok != tok_then) + return Error("expected then"); + getNextToken(); // eat the then + + ExprAST *Then = ParseExpression(); + if (Then == 0) return 0; + + if (CurTok != tok_else) + return Error("expected else"); + + getNextToken(); + + ExprAST *Else = ParseExpression(); + if (!Else) return 0; + + return new IfExprAST(Cond, Then, Else); +} + +/// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression +static ExprAST *ParseForExpr() { + getNextToken(); // eat the for. + + if (CurTok != tok_identifier) + return Error("expected identifier after for"); + + std::string IdName = IdentifierStr; + getNextToken(); // eat identifier. + + if (CurTok != '=') + return Error("expected '=' after for"); + getNextToken(); // eat '='. + + + ExprAST *Start = ParseExpression(); + if (Start == 0) return 0; + if (CurTok != ',') + return Error("expected ',' after for start value"); + getNextToken(); + + ExprAST *End = ParseExpression(); + if (End == 0) return 0; + + // The step value is optional. + ExprAST *Step = 0; + if (CurTok == ',') { + getNextToken(); + Step = ParseExpression(); + if (Step == 0) return 0; + } + + if (CurTok != tok_in) + return Error("expected 'in' after for"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new ForExprAST(IdName, Start, End, Step, Body); +} + +/// varexpr ::= 'var' identifier ('=' expression)? +// (',' identifier ('=' expression)?)* 'in' expression +static ExprAST *ParseVarExpr() { + getNextToken(); // eat the var. + + std::vector > VarNames; + + // At least one variable name is required. + if (CurTok != tok_identifier) + return Error("expected identifier after var"); + + while (1) { + std::string Name = IdentifierStr; + getNextToken(); // eat identifier. + + // Read the optional initializer. + ExprAST *Init = 0; + if (CurTok == '=') { + getNextToken(); // eat the '='. + + Init = ParseExpression(); + if (Init == 0) return 0; + } + + VarNames.push_back(std::make_pair(Name, Init)); + + // End of var list, exit loop. + if (CurTok != ',') break; + getNextToken(); // eat the ','. + + if (CurTok != tok_identifier) + return Error("expected identifier list after var"); + } + + // At this point, we have to have 'in'. + if (CurTok != tok_in) + return Error("expected 'in' keyword after 'var'"); + getNextToken(); // eat 'in'. + + ExprAST *Body = ParseExpression(); + if (Body == 0) return 0; + + return new VarExprAST(VarNames, Body); +} + +/// primary +/// ::= identifierexpr +/// ::= numberexpr +/// ::= parenexpr +/// ::= ifexpr +/// ::= forexpr +/// ::= varexpr +static ExprAST *ParsePrimary() { + switch (CurTok) { + default: return Error("unknown token when expecting an expression"); + case tok_identifier: return ParseIdentifierExpr(); + case tok_number: return ParseNumberExpr(); + case '(': return ParseParenExpr(); + case tok_if: return ParseIfExpr(); + case tok_for: return ParseForExpr(); + case tok_var: return ParseVarExpr(); + } +} + +/// unary +/// ::= primary +/// ::= '!' unary +static ExprAST *ParseUnary() { + // If the current token is not an operator, it must be a primary expr. + if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') + return ParsePrimary(); + + // If this is a unary operator, read it. + int Opc = CurTok; + getNextToken(); + if (ExprAST *Operand = ParseUnary()) + return new UnaryExprAST(Opc, Operand); + return 0; +} + +/// binoprhs +/// ::= ('+' unary)* +static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { + // If this is a binop, find its precedence. + while (1) { + int TokPrec = GetTokPrecedence(); + + // If this is a binop that binds at least as tightly as the current binop, + // consume it, otherwise we are done. + if (TokPrec < ExprPrec) + return LHS; + + // Okay, we know this is a binop. + int BinOp = CurTok; + getNextToken(); // eat binop + + // Parse the unary expression after the binary operator. + ExprAST *RHS = ParseUnary(); + if (!RHS) return 0; + + // If BinOp binds less tightly with RHS than the operator after RHS, let + // the pending operator take RHS as its LHS. + int NextPrec = GetTokPrecedence(); + if (TokPrec < NextPrec) { + RHS = ParseBinOpRHS(TokPrec+1, RHS); + if (RHS == 0) return 0; + } + + // Merge LHS/RHS. + LHS = new BinaryExprAST(BinOp, LHS, RHS); + } +} + +/// expression +/// ::= unary binoprhs +/// +static ExprAST *ParseExpression() { + ExprAST *LHS = ParseUnary(); + if (!LHS) return 0; + + return ParseBinOpRHS(0, LHS); +} + +/// prototype +/// ::= id '(' id* ')' +/// ::= binary LETTER number? (id, id) +/// ::= unary LETTER (id) +static PrototypeAST *ParsePrototype() { + std::string FnName; + + unsigned Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. + unsigned BinaryPrecedence = 30; + + switch (CurTok) { + default: + return ErrorP("Expected function name in prototype"); + case tok_identifier: + FnName = IdentifierStr; + Kind = 0; + getNextToken(); + break; + case tok_unary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected unary operator"); + FnName = "unary"; + FnName += (char)CurTok; + Kind = 1; + getNextToken(); + break; + case tok_binary: + getNextToken(); + if (!isascii(CurTok)) + return ErrorP("Expected binary operator"); + FnName = "binary"; + FnName += (char)CurTok; + Kind = 2; + getNextToken(); + + // Read the precedence if present. + if (CurTok == tok_number) { + if (NumVal < 1 || NumVal > 100) + return ErrorP("Invalid precedecnce: must be 1..100"); + BinaryPrecedence = (unsigned)NumVal; + getNextToken(); + } + break; + } + + if (CurTok != '(') + return ErrorP("Expected '(' in prototype"); + + std::vector ArgNames; + while (getNextToken() == tok_identifier) + ArgNames.push_back(IdentifierStr); + if (CurTok != ')') + return ErrorP("Expected ')' in prototype"); + + // success. + getNextToken(); // eat ')'. + + // Verify right number of names for operator. + if (Kind && ArgNames.size() != Kind) + return ErrorP("Invalid number of operands for operator"); + + return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); +} + +/// definition ::= 'def' prototype expression +static FunctionAST *ParseDefinition() { + getNextToken(); // eat def. + PrototypeAST *Proto = ParsePrototype(); + if (Proto == 0) return 0; + + if (ExprAST *E = ParseExpression()) + return new FunctionAST(Proto, E); + return 0; +} + +/// toplevelexpr ::= expression +static FunctionAST *ParseTopLevelExpr() { + if (ExprAST *E = ParseExpression()) { + // Make an anonymous proto. + PrototypeAST *Proto = new PrototypeAST("", std::vector()); + return new FunctionAST(Proto, E); + } + return 0; +} + +/// external ::= 'extern' prototype +static PrototypeAST *ParseExtern() { + getNextToken(); // eat extern. + return ParsePrototype(); +} + +//===----------------------------------------------------------------------===// +// Quick and dirty hack +//===----------------------------------------------------------------------===// + +// FIXME: Obviously we can do better than this +std::string GenerateUniqueName(const char *root) +{ + static int i = 0; + char s[16]; + sprintf(s, "%s%d", root, i++); + std::string S = s; + return S; +} + +std::string MakeLegalFunctionName(std::string Name) +{ + std::string NewName; + if (!Name.length()) + return GenerateUniqueName("anon_func_"); + + // Start with what we have + NewName = Name; + + // Look for a numberic first character + if (NewName.find_first_of("0123456789") == 0) { + NewName.insert(0, 1, 'n'); + } + + // Replace illegal characters with their ASCII equivalent + std::string legal_elements = "_abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"; + size_t pos; + while ((pos = NewName.find_first_not_of(legal_elements)) != std::string::npos) { + char old_c = NewName.at(pos); + char new_str[16]; + sprintf(new_str, "%d", (int)old_c); + NewName = NewName.replace(pos, 1, new_str); + } + + return NewName; +} + +//===----------------------------------------------------------------------===// +// MCJIT helper class +//===----------------------------------------------------------------------===// + +class MCJITHelper +{ +public: + MCJITHelper(LLVMContext& C) : Context(C), OpenModule(NULL) {} + ~MCJITHelper(); + + Function *getFunction(const std::string FnName); + Module *getModuleForNewFunction(); + void *getPointerToFunction(Function* F); + void *getPointerToNamedFunction(const std::string &Name); + ExecutionEngine *compileModule(Module *M); + void closeCurrentModule(); + void dump(); + +private: + typedef std::vector ModuleVector; + + LLVMContext &Context; + Module *OpenModule; + ModuleVector Modules; + std::map EngineMap; +}; + +class HelpingMemoryManager : public SectionMemoryManager +{ + HelpingMemoryManager(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + void operator=(const HelpingMemoryManager&) LLVM_DELETED_FUNCTION; + +public: + HelpingMemoryManager(MCJITHelper *Helper) : MasterHelper(Helper) {} + virtual ~HelpingMemoryManager() {} + + /// This method returns the address of the specified function. + /// Our implementation will attempt to find functions in other + /// modules associated with the MCJITHelper to cross link functions + /// from one generated module to another. + /// + /// If \p AbortOnFailure is false and no function with the given name is + /// found, this function returns a null pointer. Otherwise, it prints a + /// message to stderr and aborts. + virtual void *getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure = true); +private: + MCJITHelper *MasterHelper; +}; + +void *HelpingMemoryManager::getPointerToNamedFunction(const std::string &Name, + bool AbortOnFailure) +{ + // Try the standard symbol resolution first, but ask it not to abort. + void *pfn = SectionMemoryManager::getPointerToNamedFunction(Name, false); + if (pfn) + return pfn; + + pfn = MasterHelper->getPointerToNamedFunction(Name); + if (!pfn && AbortOnFailure) + report_fatal_error("Program used external function '" + Name + + "' which could not be resolved!"); + return pfn; +} + +MCJITHelper::~MCJITHelper() +{ + // Walk the vector of modules. + ModuleVector::iterator it, end; + for (it = Modules.begin(), end = Modules.end(); + it != end; ++it) { + // See if we have an execution engine for this module. + std::map::iterator mapIt = EngineMap.find(*it); + // If we have an EE, the EE owns the module so just delete the EE. + if (mapIt != EngineMap.end()) { + delete mapIt->second; + } else { + // Otherwise, we still own the module. Delete it now. + delete *it; + } + } +} + +Function *MCJITHelper::getFunction(const std::string FnName) { + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(FnName); + if (F) { + if (*it == OpenModule) + return F; + + assert(OpenModule != NULL); + + // This function is in a module that has already been JITed. + // We need to generate a new prototype for external linkage. + Function *PF = OpenModule->getFunction(FnName); + if (PF && !PF->empty()) { + ErrorF("redefinition of function across modules"); + return 0; + } + + // If we don't have a prototype yet, create one. + if (!PF) + PF = Function::Create(F->getFunctionType(), + Function::ExternalLinkage, + FnName, + OpenModule); + return PF; + } + } + return NULL; +} + +Module *MCJITHelper::getModuleForNewFunction() { + // If we have a Module that hasn't been JITed, use that. + if (OpenModule) + return OpenModule; + + // Otherwise create a new Module. + std::string ModName = GenerateUniqueName("mcjit_module_"); + Module *M = new Module(ModName, Context); + Modules.push_back(M); + OpenModule = M; + return M; +} + +void *MCJITHelper::getPointerToFunction(Function* F) { + // Look for this function in an existing module + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + std::string FnName = F->getName(); + for (it = begin; it != end; ++it) { + Function *MF = (*it)->getFunction(FnName); + if (MF == F) { + std::map::iterator eeIt = EngineMap.find(*it); + if (eeIt != EngineMap.end()) { + void *P = eeIt->second->getPointerToFunction(F); + if (P) + return P; + } else { + ExecutionEngine *EE = compileModule(*it); + void *P = EE->getPointerToFunction(F); + if (P) + return P; + } + } + } + return NULL; +} + +void MCJITHelper::closeCurrentModule() { + OpenModule = NULL; +} + +ExecutionEngine *MCJITHelper::compileModule(Module *M) { + if (M == OpenModule) + closeCurrentModule(); + + std::string ErrStr; + ExecutionEngine *NewEngine = EngineBuilder(M) + .setErrorStr(&ErrStr) + .setUseMCJIT(true) + .setMCJITMemoryManager(new HelpingMemoryManager(this)) + .create(); + if (!NewEngine) { + fprintf(stderr, "Could not create ExecutionEngine: %s\n", ErrStr.c_str()); + exit(1); + } + + // Create a function pass manager for this engine + FunctionPassManager *FPM = new FunctionPassManager(M); + + // Set up the optimizer pipeline. Start with registering info about how the + // target lays out data structures. + FPM->add(new DataLayout(*NewEngine->getDataLayout())); + // Provide basic AliasAnalysis support for GVN. + FPM->add(createBasicAliasAnalysisPass()); + // Promote allocas to registers. + FPM->add(createPromoteMemoryToRegisterPass()); + // Do simple "peephole" optimizations and bit-twiddling optzns. + FPM->add(createInstructionCombiningPass()); + // Reassociate expressions. + FPM->add(createReassociatePass()); + // Eliminate Common SubExpressions. + FPM->add(createGVNPass()); + // Simplify the control flow graph (deleting unreachable blocks, etc). + FPM->add(createCFGSimplificationPass()); + FPM->doInitialization(); + + // For each function in the module + Module::iterator it; + Module::iterator end = M->end(); + for (it = M->begin(); it != end; ++it) { + // Run the FPM on this function + FPM->run(*it); + } + + // We don't need this anymore + delete FPM; + + // Store this engine + EngineMap[M] = NewEngine; + NewEngine->finalizeObject(); + + return NewEngine; +} + +void *MCJITHelper::getPointerToNamedFunction(const std::string &Name) +{ + // Look for the functions in our modules, compiling only as necessary + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) { + Function *F = (*it)->getFunction(Name); + if (F && !F->empty()) { + std::map::iterator eeIt = EngineMap.find(*it); + if (eeIt != EngineMap.end()) { + void *P = eeIt->second->getPointerToFunction(F); + if (P) + return P; + } else { + ExecutionEngine *EE = compileModule(*it); + void *P = EE->getPointerToFunction(F); + if (P) + return P; + } + } + } + return NULL; +} + +void MCJITHelper::dump() +{ + ModuleVector::iterator begin = Modules.begin(); + ModuleVector::iterator end = Modules.end(); + ModuleVector::iterator it; + for (it = begin; it != end; ++it) + (*it)->dump(); +} + +//===----------------------------------------------------------------------===// +// Code Generation +//===----------------------------------------------------------------------===// + +static MCJITHelper *TheHelper; +static IRBuilder<> Builder(getGlobalContext()); +static std::map NamedValues; + +Value *ErrorV(const char *Str) { Error(Str); return 0; } + +/// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of +/// the function. This is used for mutable variables etc. +static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction, + const std::string &VarName) { + IRBuilder<> TmpB(&TheFunction->getEntryBlock(), + TheFunction->getEntryBlock().begin()); + return TmpB.CreateAlloca(Type::getDoubleTy(getGlobalContext()), 0, + VarName.c_str()); +} + +Value *NumberExprAST::Codegen() { + return ConstantFP::get(getGlobalContext(), APFloat(Val)); +} + +Value *VariableExprAST::Codegen() { + // Look this variable up in the function. + Value *V = NamedValues[Name]; + char ErrStr[256]; + sprintf(ErrStr, "Unknown variable name %s", Name.c_str()); + if (V == 0) return ErrorV(ErrStr); + + // Load the value. + return Builder.CreateLoad(V, Name.c_str()); +} + +Value *UnaryExprAST::Codegen() { + Value *OperandV = Operand->Codegen(); + if (OperandV == 0) return 0; + + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("unary")+Opcode)); + if (F == 0) + return ErrorV("Unknown unary operator"); + + return Builder.CreateCall(F, OperandV, "unop"); +} + +Value *BinaryExprAST::Codegen() { + // Special case '=' because we don't want to emit the LHS as an expression. + if (Op == '=') { + // Assignment requires the LHS to be an identifier. + VariableExprAST *LHSE = reinterpret_cast(LHS); + if (!LHSE) + return ErrorV("destination of '=' must be a variable"); + // Codegen the RHS. + Value *Val = RHS->Codegen(); + if (Val == 0) return 0; + + // Look up the name. + Value *Variable = NamedValues[LHSE->getName()]; + if (Variable == 0) return ErrorV("Unknown variable name"); + + Builder.CreateStore(Val, Variable); + return Val; + } + + Value *L = LHS->Codegen(); + Value *R = RHS->Codegen(); + if (L == 0 || R == 0) return 0; + + switch (Op) { + case '+': return Builder.CreateFAdd(L, R, "addtmp"); + case '-': return Builder.CreateFSub(L, R, "subtmp"); + case '*': return Builder.CreateFMul(L, R, "multmp"); + case '/': return Builder.CreateFDiv(L, R, "divtmp"); + case '<': + L = Builder.CreateFCmpULT(L, R, "cmptmp"); + // Convert bool 0/1 to double 0.0 or 1.0 + return Builder.CreateUIToFP(L, Type::getDoubleTy(getGlobalContext()), + "booltmp"); + default: break; + } + + // If it wasn't a builtin binary operator, it must be a user defined one. Emit + // a call to it. + Function *F = TheHelper->getFunction(MakeLegalFunctionName(std::string("binary")+Op)); + assert(F && "binary operator not found!"); + + Value *Ops[] = { L, R }; + return Builder.CreateCall(F, Ops, "binop"); +} + +Value *CallExprAST::Codegen() { + // Look up the name in the global module table. + Function *CalleeF = TheHelper->getFunction(Callee); + if (CalleeF == 0) + return ErrorV("Unknown function referenced"); + + // If argument mismatch error. + if (CalleeF->arg_size() != Args.size()) + return ErrorV("Incorrect # arguments passed"); + + std::vector ArgsV; + for (unsigned i = 0, e = Args.size(); i != e; ++i) { + ArgsV.push_back(Args[i]->Codegen()); + if (ArgsV.back() == 0) return 0; + } + + return Builder.CreateCall(CalleeF, ArgsV, "calltmp"); +} + +Value *IfExprAST::Codegen() { + Value *CondV = Cond->Codegen(); + if (CondV == 0) return 0; + + // Convert condition to a bool by comparing equal to 0.0. + CondV = Builder.CreateFCmpONE(CondV, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "ifcond"); + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create blocks for the then and else cases. Insert the 'then' block at the + // end of the function. + BasicBlock *ThenBB = BasicBlock::Create(getGlobalContext(), "then", TheFunction); + BasicBlock *ElseBB = BasicBlock::Create(getGlobalContext(), "else"); + BasicBlock *MergeBB = BasicBlock::Create(getGlobalContext(), "ifcont"); + + Builder.CreateCondBr(CondV, ThenBB, ElseBB); + + // Emit then value. + Builder.SetInsertPoint(ThenBB); + + Value *ThenV = Then->Codegen(); + if (ThenV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Then' can change the current block, update ThenBB for the PHI. + ThenBB = Builder.GetInsertBlock(); + + // Emit else block. + TheFunction->getBasicBlockList().push_back(ElseBB); + Builder.SetInsertPoint(ElseBB); + + Value *ElseV = Else->Codegen(); + if (ElseV == 0) return 0; + + Builder.CreateBr(MergeBB); + // Codegen of 'Else' can change the current block, update ElseBB for the PHI. + ElseBB = Builder.GetInsertBlock(); + + // Emit merge block. + TheFunction->getBasicBlockList().push_back(MergeBB); + Builder.SetInsertPoint(MergeBB); + PHINode *PN = Builder.CreatePHI(Type::getDoubleTy(getGlobalContext()), 2, + "iftmp"); + + PN->addIncoming(ThenV, ThenBB); + PN->addIncoming(ElseV, ElseBB); + return PN; +} + +Value *ForExprAST::Codegen() { + // Output this as: + // var = alloca double + // ... + // start = startexpr + // store start -> var + // goto loop + // loop: + // ... + // bodyexpr + // ... + // loopend: + // step = stepexpr + // endcond = endexpr + // + // curvar = load var + // nextvar = curvar + step + // store nextvar -> var + // br endcond, loop, endloop + // outloop: + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Create an alloca for the variable in the entry block. + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + + // Emit the start code first, without 'variable' in scope. + Value *StartVal = Start->Codegen(); + if (StartVal == 0) return 0; + + // Store the value into the alloca. + Builder.CreateStore(StartVal, Alloca); + + // Make the new basic block for the loop header, inserting after current + // block. + BasicBlock *LoopBB = BasicBlock::Create(getGlobalContext(), "loop", TheFunction); + + // Insert an explicit fall through from the current block to the LoopBB. + Builder.CreateBr(LoopBB); + + // Start insertion in LoopBB. + Builder.SetInsertPoint(LoopBB); + + // Within the loop, the variable is defined equal to the PHI node. If it + // shadows an existing variable, we have to restore it, so save it now. + AllocaInst *OldVal = NamedValues[VarName]; + NamedValues[VarName] = Alloca; + + // Emit the body of the loop. This, like any other expr, can change the + // current BB. Note that we ignore the value computed by the body, but don't + // allow an error. + if (Body->Codegen() == 0) + return 0; + + // Emit the step value. + Value *StepVal; + if (Step) { + StepVal = Step->Codegen(); + if (StepVal == 0) return 0; + } else { + // If not specified, use 1.0. + StepVal = ConstantFP::get(getGlobalContext(), APFloat(1.0)); + } + + // Compute the end condition. + Value *EndCond = End->Codegen(); + if (EndCond == 0) return EndCond; + + // Reload, increment, and restore the alloca. This handles the case where + // the body of the loop mutates the variable. + Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str()); + Value *NextVar = Builder.CreateFAdd(CurVar, StepVal, "nextvar"); + Builder.CreateStore(NextVar, Alloca); + + // Convert condition to a bool by comparing equal to 0.0. + EndCond = Builder.CreateFCmpONE(EndCond, + ConstantFP::get(getGlobalContext(), APFloat(0.0)), + "loopcond"); + + // Create the "after loop" block and insert it. + BasicBlock *AfterBB = BasicBlock::Create(getGlobalContext(), "afterloop", TheFunction); + + // Insert the conditional branch into the end of LoopEndBB. + Builder.CreateCondBr(EndCond, LoopBB, AfterBB); + + // Any new code will be inserted in AfterBB. + Builder.SetInsertPoint(AfterBB); + + // Restore the unshadowed variable. + if (OldVal) + NamedValues[VarName] = OldVal; + else + NamedValues.erase(VarName); + + + // for expr always returns 0.0. + return Constant::getNullValue(Type::getDoubleTy(getGlobalContext())); +} + +Value *VarExprAST::Codegen() { + std::vector OldBindings; + + Function *TheFunction = Builder.GetInsertBlock()->getParent(); + + // Register all variables and emit their initializer. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) { + const std::string &VarName = VarNames[i].first; + ExprAST *Init = VarNames[i].second; + + // Emit the initializer before adding the variable to scope, this prevents + // the initializer from referencing the variable itself, and permits stuff + // like this: + // var a = 1 in + // var a = a in ... # refers to outer 'a'. + Value *InitVal; + if (Init) { + InitVal = Init->Codegen(); + if (InitVal == 0) return 0; + } else { // If not specified, use 0.0. + InitVal = ConstantFP::get(getGlobalContext(), APFloat(0.0)); + } + + AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName); + Builder.CreateStore(InitVal, Alloca); + + // Remember the old variable binding so that we can restore the binding when + // we unrecurse. + OldBindings.push_back(NamedValues[VarName]); + + // Remember this binding. + NamedValues[VarName] = Alloca; + } + + // Codegen the body, now that all vars are in scope. + Value *BodyVal = Body->Codegen(); + if (BodyVal == 0) return 0; + + // Pop all our variables from scope. + for (unsigned i = 0, e = VarNames.size(); i != e; ++i) + NamedValues[VarNames[i].first] = OldBindings[i]; + + // Return the body computation. + return BodyVal; +} + +Function *PrototypeAST::Codegen() { + // Make the function type: double(double,double) etc. + std::vector Doubles(Args.size(), + Type::getDoubleTy(getGlobalContext())); + FunctionType *FT = FunctionType::get(Type::getDoubleTy(getGlobalContext()), + Doubles, false); + + std::string FnName = MakeLegalFunctionName(Name); + + Module* M = TheHelper->getModuleForNewFunction(); + + Function *F = Function::Create(FT, Function::ExternalLinkage, FnName, M); + + // If F conflicted, there was already something named 'FnName'. If it has a + // body, don't allow redefinition or reextern. + if (F->getName() != FnName) { + // Delete the one we just made and get the existing one. + F->eraseFromParent(); + F = M->getFunction(Name); + + // If F already has a body, reject this. + if (!F->empty()) { + ErrorF("redefinition of function"); + return 0; + } + + // If F took a different number of args, reject. + if (F->arg_size() != Args.size()) { + ErrorF("redefinition of function with different # args"); + return 0; + } + } + + // Set names for all arguments. + unsigned Idx = 0; + for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); + ++AI, ++Idx) + AI->setName(Args[Idx]); + + return F; +} + +/// CreateArgumentAllocas - Create an alloca for each argument and register the +/// argument in the symbol table so that references to it will succeed. +void PrototypeAST::CreateArgumentAllocas(Function *F) { + Function::arg_iterator AI = F->arg_begin(); + for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) { + // Create an alloca for this variable. + AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]); + + // Store the initial value into the alloca. + Builder.CreateStore(AI, Alloca); + + // Add arguments to variable symbol table. + NamedValues[Args[Idx]] = Alloca; + } +} + +Function *FunctionAST::Codegen() { + NamedValues.clear(); + + Function *TheFunction = Proto->Codegen(); + if (TheFunction == 0) + return 0; + + // If this is an operator, install it. + if (Proto->isBinaryOp()) + BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); + + // Create a new basic block to start insertion into. + BasicBlock *BB = BasicBlock::Create(getGlobalContext(), "entry", TheFunction); + Builder.SetInsertPoint(BB); + + // Add all arguments to the symbol table and create their allocas. + Proto->CreateArgumentAllocas(TheFunction); + + if (Value *RetVal = Body->Codegen()) { + // Finish off the function. + Builder.CreateRet(RetVal); + + // Validate the generated code, checking for consistency. + verifyFunction(*TheFunction); + + return TheFunction; + } + + // Error reading body, remove function. + TheFunction->eraseFromParent(); + + if (Proto->isBinaryOp()) + BinopPrecedence.erase(Proto->getOperatorName()); + return 0; +} + +//===----------------------------------------------------------------------===// +// Top-Level parsing and JIT Driver +//===----------------------------------------------------------------------===// + +static void HandleDefinition() { + if (FunctionAST *F = ParseDefinition()) { + TheHelper->closeCurrentModule(); + if (Function *LF = F->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read function definition:"); + LF->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleExtern() { + if (PrototypeAST *P = ParseExtern()) { + if (Function *F = P->Codegen()) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "Read extern: "); + F->dump(); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +static void HandleTopLevelExpression() { + // Evaluate a top-level expression into an anonymous function. + if (FunctionAST *F = ParseTopLevelExpr()) { + if (Function *LF = F->Codegen()) { + // JIT the function, returning a function pointer. + void *FPtr = TheHelper->getPointerToFunction(LF); + + // Cast it to the right type (takes no arguments, returns a double) so we + // can call it as a native function. + double (*FP)() = (double (*)())(intptr_t)FPtr; +#ifdef MINIMAL_STDERR_OUTPUT + FP(); +#else + fprintf(stderr, "Evaluated to %f\n", FP()); +#endif + } + } else { + // Skip token for error recovery. + getNextToken(); + } +} + +/// top ::= definition | external | expression | ';' +static void MainLoop() { + while (1) { +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + switch (CurTok) { + case tok_eof: return; + case ';': getNextToken(); break; // ignore top-level semicolons. + case tok_def: HandleDefinition(); break; + case tok_extern: HandleExtern(); break; + default: HandleTopLevelExpression(); break; + } + } +} + +//===----------------------------------------------------------------------===// +// "Library" functions that can be "extern'd" from user code. +//===----------------------------------------------------------------------===// + +/// putchard - putchar that takes a double and returns 0. +extern "C" +double putchard(double X) { + putchar((char)X); + return 0; +} + +/// printd - printf that takes a double prints it as "%f\n", returning 0. +extern "C" +double printd(double X) { + printf("%f", X); + return 0; +} + +extern "C" +double printlf() { + printf("\n"); + return 0; +} + +//===----------------------------------------------------------------------===// +// Main driver code. +//===----------------------------------------------------------------------===// + +int main() { + InitializeNativeTarget(); + InitializeNativeTargetAsmPrinter(); + InitializeNativeTargetAsmParser(); + LLVMContext &Context = getGlobalContext(); + + // Install standard binary operators. + // 1 is lowest precedence. + BinopPrecedence['='] = 2; + BinopPrecedence['<'] = 10; + BinopPrecedence['+'] = 20; + BinopPrecedence['-'] = 20; + BinopPrecedence['/'] = 40; + BinopPrecedence['*'] = 40; // highest. + + // Prime the first token. +#ifndef MINIMAL_STDERR_OUTPUT + fprintf(stderr, "ready> "); +#endif + getNextToken(); + + // Make the helper, which holds all the code. + TheHelper = new MCJITHelper(Context); + + // Run the main "interpreter loop" now. + MainLoop(); + +#ifndef MINIMAL_STDERR_OUTPUT + // Print out all of the generated code. + TheHelper->dump(); +#endif + + return 0; +}