[llvm] [mlir] add ch1~ch3 note (PR #170624)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Dec 4 00:11:20 PST 2025
https://github.com/hellozmz created https://github.com/llvm/llvm-project/pull/170624
None
>From 0adefd3592874ac8c137d1141e3280e24cb9ff24 Mon Sep 17 00:00:00 2001
From: hellozmz <407190054 at qq.com>
Date: Thu, 4 Dec 2025 08:09:31 +0000
Subject: [PATCH] add ch1~ch3 note
---
build.sh | 17 +++++++++++++++++
mlir/examples/toy/Ch1/include/toy/AST.h | 19 ++++++++++---------
mlir/examples/toy/Ch1/include/toy/Lexer.h | 16 ++++++++--------
mlir/examples/toy/Ch1/toyc.cpp | 6 +++---
mlir/examples/toy/Ch2/compile.sh | 5 +++++
mlir/examples/toy/Ch2/include/toy/Ops.td | 1 +
mlir/examples/toy/Ch2/toyc.cpp | 15 ++++++++-------
mlir/examples/toy/Ch3/toyc.cpp | 2 +-
mlir/test/Examples/Toy/Ch1/test.sh | 7 +++++++
mlir/test/Examples/Toy/Ch2/invalid.mlir | 3 ++-
mlir/test/Examples/Toy/Ch2/test.sh | 9 +++++++++
mlir/test/Examples/Toy/Ch3/test.sh | 20 ++++++++++++++++++++
12 files changed, 91 insertions(+), 29 deletions(-)
create mode 100644 build.sh
create mode 100644 mlir/examples/toy/Ch2/compile.sh
create mode 100644 mlir/test/Examples/Toy/Ch1/test.sh
create mode 100644 mlir/test/Examples/Toy/Ch2/test.sh
create mode 100644 mlir/test/Examples/Toy/Ch3/test.sh
diff --git a/build.sh b/build.sh
new file mode 100644
index 0000000000000..29b576aaee0fc
--- /dev/null
+++ b/build.sh
@@ -0,0 +1,17 @@
+mkdir llvm-project/build
+cd llvm-project/build
+cmake -G Ninja ../llvm \
+ -DLLVM_ENABLE_PROJECTS=mlir \
+ -DLLVM_BUILD_EXAMPLES=ON \
+ -DLLVM_TARGETS_TO_BUILD="Native;" \
+ -DCMAKE_BUILD_TYPE=Release \
+ -DLLVM_ENABLE_ASSERTIONS=ON
+# Using clang and lld speeds up the build, we recommend adding:
+# -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DLLVM_ENABLE_LLD=ON
+# CCache can drastically speed up further rebuilds, try adding:
+# -DLLVM_CCACHE_BUILD=ON
+# Optionally, using ASAN/UBSAN can find bugs early in development, enable with:
+# -DLLVM_USE_SANITIZER="Address;Undefined"
+# Optionally, enabling integration tests as well
+# -DMLIR_INCLUDE_INTEGRATION_TESTS=ON
+cmake --build . --target check-mlir
diff --git a/mlir/examples/toy/Ch1/include/toy/AST.h b/mlir/examples/toy/Ch1/include/toy/AST.h
index d2ba101dea5a4..b91254dd188f3 100644
--- a/mlir/examples/toy/Ch1/include/toy/AST.h
+++ b/mlir/examples/toy/Ch1/include/toy/AST.h
@@ -31,18 +31,19 @@ struct VarType {
std::vector<int64_t> shape;
};
+// 源代码 --> 词法分析(Lexer) --> 语法分析(Parser) --> AST --> 语义分析 --> 中间代码生成
/// Base class for all expression nodes.
-class ExprAST {
+class ExprAST { // Abstract Syntax Tree，抽象语法树
public:
enum ExprASTKind {
- Expr_VarDecl,
- Expr_Return,
- Expr_Num,
- Expr_Literal,
- Expr_Var,
- Expr_BinOp,
- Expr_Call,
- Expr_Print,
+ Expr_VarDecl, // 定义变量
+ Expr_Return, // 返回操作
+ Expr_Num, // 数值类型
+ Expr_Literal, // 文本类型
+ Expr_Var, // 引用变量
+ Expr_BinOp, // 两元素操作
+ Expr_Call, // 函数调用
+ Expr_Print, // 内置打印功能
};
ExprAST(ExprASTKind kind, Location location)
diff --git a/mlir/examples/toy/Ch1/include/toy/Lexer.h b/mlir/examples/toy/Ch1/include/toy/Lexer.h
index d420a7ebbf3b6..f5bb729a159c7 100644
--- a/mlir/examples/toy/Ch1/include/toy/Lexer.h
+++ b/mlir/examples/toy/Ch1/include/toy/Lexer.h
@@ -140,13 +140,13 @@ class Lexer {
while (isalnum((lastChar = Token(getNextChar()))) || lastChar == '_')
identifierStr += (char)lastChar;
- if (identifierStr == "return")
+ if (identifierStr == "return") // 返回语句
return tok_return;
- if (identifierStr == "def")
+ if (identifierStr == "def") // 函数定义
return tok_def;
- if (identifierStr == "var")
+ if (identifierStr == "var") // 变量定义
return tok_var;
- return tok_identifier;
+ return tok_identifier; // 标识符
}
// Number: [0-9.]+
@@ -158,10 +158,10 @@ class Lexer {
} while (isdigit(lastChar) || lastChar == '.');
numVal = strtod(numStr.c_str(), nullptr);
- return tok_number;
+ return tok_number; // 数字
}
- if (lastChar == '#') {
+ if (lastChar == '#') { // 注释
// Comment until end of line.
do {
lastChar = Token(getNextChar());
@@ -173,12 +173,12 @@ class Lexer {
// Check for end of file. Don't eat the EOF.
if (lastChar == EOF)
- return tok_eof;
+ return tok_eof; // 文件结束
// Otherwise, just return the character as its ascii value.
Token thisChar = Token(lastChar);
lastChar = Token(getNextChar());
- return thisChar;
+ return thisChar; // 其他字符
}
/// The last token read from the input.
diff --git a/mlir/examples/toy/Ch1/toyc.cpp b/mlir/examples/toy/Ch1/toyc.cpp
index fb7b484a92fb9..d9064411934ec 100644
--- a/mlir/examples/toy/Ch1/toyc.cpp
+++ b/mlir/examples/toy/Ch1/toyc.cpp
@@ -31,7 +31,7 @@ static cl::opt<std::string> inputFilename(cl::Positional,
cl::init("-"),
cl::value_desc("filename"));
namespace {
-enum Action { None, DumpAST };
+enum Action { None, DumpAST }; // 两种操作，不操作和输出AST（dump）
} // namespace
static cl::opt<enum Action>
@@ -47,8 +47,8 @@ std::unique_ptr<toy::ModuleAST> parseInputFile(llvm::StringRef filename) {
return nullptr;
}
auto buffer = fileOrErr.get()->getBuffer();
- LexerBuffer lexer(buffer.begin(), buffer.end(), std::string(filename));
- Parser parser(lexer);
+ LexerBuffer lexer(buffer.begin(), buffer.end(), std::string(filename)); // 词法分析
+ Parser parser(lexer); // 语法分析
return parser.parseModule();
}
diff --git a/mlir/examples/toy/Ch2/compile.sh b/mlir/examples/toy/Ch2/compile.sh
new file mode 100644
index 0000000000000..877106d551915
--- /dev/null
+++ b/mlir/examples/toy/Ch2/compile.sh
@@ -0,0 +1,5 @@
+
+build_root=/mnt/data01/zmz/workspace/07ascendnpu/llvm/llvm-project/build
+mlir_src_root=/mnt/data01/zmz/workspace/07ascendnpu/llvm/llvm-project/mlir
+
+${build_root}/bin/mlir-tblgen -gen-op-defs ${mlir_src_root}/examples/toy/Ch2/include/toy/Ops.td -I ${mlir_src_root}/include/
\ No newline at end of file
diff --git a/mlir/examples/toy/Ch2/include/toy/Ops.td b/mlir/examples/toy/Ch2/include/toy/Ops.td
index 91bf83a54df1a..e07930560fe91 100644
--- a/mlir/examples/toy/Ch2/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch2/include/toy/Ops.td
@@ -78,6 +78,7 @@ def ConstantOp : Toy_Op<"constant", [Pure]> {
}]>,
// Build a constant with a given constant floating-point value.
+ // 下面创建的函数，没有函数的实现，只有定义，具体的实现在Dialect.cpp中
OpBuilder<(ins "double":$value)>
];
diff --git a/mlir/examples/toy/Ch2/toyc.cpp b/mlir/examples/toy/Ch2/toyc.cpp
index e33b49b41c5a1..07572d690149b 100644
--- a/mlir/examples/toy/Ch2/toyc.cpp
+++ b/mlir/examples/toy/Ch2/toyc.cpp
@@ -50,7 +50,7 @@ static cl::opt<enum InputType> inputType(
"load the input file as an MLIR file")));
namespace {
-enum Action { None, DumpAST, DumpMLIR };
+enum Action { None, DumpAST, DumpMLIR }; // 增加了dump mlir的选项。现在源代码可以dump成抽象语法树，也可以dump成mlir（这个是重点）。
} // namespace
static cl::opt<enum Action> emitAction(
"emit", cl::desc("Select the kind of output desired"),
@@ -71,14 +71,14 @@ std::unique_ptr<toy::ModuleAST> parseInputFile(llvm::StringRef filename) {
return parser.parseModule();
}
-int dumpMLIR() {
+int dumpMLIR() { // dump mlir逻辑
mlir::MLIRContext context;
// Load our Dialect in this MLIR Context.
- context.getOrLoadDialect<mlir::toy::ToyDialect>();
+ context.getOrLoadDialect<mlir::toy::ToyDialect>(); // 增加ToyDialect方言，完全自定义
// Handle '.toy' input to the compiler.
if (inputType != InputType::MLIR &&
- !llvm::StringRef(inputFilename).ends_with(".mlir")) {
+ !llvm::StringRef(inputFilename).ends_with(".mlir")) { // 处理非mlir文件，也就是toy文件
auto moduleAST = parseInputFile(inputFilename);
if (!moduleAST)
return 6;
@@ -102,13 +102,13 @@ int dumpMLIR() {
llvm::SourceMgr sourceMgr;
sourceMgr.AddNewSourceBuffer(std::move(*fileOrErr), llvm::SMLoc());
mlir::OwningOpRef<mlir::ModuleOp> module =
- mlir::parseSourceFile<mlir::ModuleOp>(sourceMgr, &context);
+ mlir::parseSourceFile<mlir::ModuleOp>(sourceMgr, &context); // 解析mlir文件
if (!module) {
llvm::errs() << "Error can't load file " << inputFilename << "\n";
return 3;
}
- module->dump();
+ module->dump(); // 打印mlir中间表示
return 0;
}
@@ -130,7 +130,8 @@ int main(int argc, char **argv) {
// Register any command line options.
mlir::registerAsmPrinterCLOptions();
mlir::registerMLIRContextCLOptions();
- cl::ParseCommandLineOptions(argc, argv, "toy compiler\n");
+ // example: toyc-ch2 %s -emit=mlir
+ cl::ParseCommandLineOptions(argc, argv, "toy compiler\n"); // 给 inputType 和 emitAction 赋值
switch (emitAction) {
case Action::DumpAST:
diff --git a/mlir/examples/toy/Ch3/toyc.cpp b/mlir/examples/toy/Ch3/toyc.cpp
index f8aa846582267..e31972c8eaa11 100644
--- a/mlir/examples/toy/Ch3/toyc.cpp
+++ b/mlir/examples/toy/Ch3/toyc.cpp
@@ -125,7 +125,7 @@ int dumpMLIR() {
return 4;
// Add a run of the canonicalizer to optimize the mlir module.
- pm.addNestedPass<mlir::toy::FuncOp>(mlir::createCanonicalizerPass());
+ pm.addNestedPass<mlir::toy::FuncOp>(mlir::createCanonicalizerPass()); // getCanonicalizationPatterns
if (mlir::failed(pm.run(*module)))
return 4;
}
diff --git a/mlir/test/Examples/Toy/Ch1/test.sh b/mlir/test/Examples/Toy/Ch1/test.sh
new file mode 100644
index 0000000000000..b1e46cc27d97b
--- /dev/null
+++ b/mlir/test/Examples/Toy/Ch1/test.sh
@@ -0,0 +1,7 @@
+export PATH=/mnt/data01/zmz/workspace/07ascendnpu/llvm/llvm-project/build/bin/:$PATH
+
+toyc-ch1 ./empty.toy -emit=ast 2>&1 | FileCheck ./empty.toy
+
+toyc-ch1 ./ast.toy -emit=ast 2>&1 | FileCheck ./ast.toy
+
+toyc-ch1 ./ast.toy -emit=ast
\ No newline at end of file
diff --git a/mlir/test/Examples/Toy/Ch2/invalid.mlir b/mlir/test/Examples/Toy/Ch2/invalid.mlir
index b3ff353f93041..bddaa2ef41bf3 100644
--- a/mlir/test/Examples/Toy/Ch2/invalid.mlir
+++ b/mlir/test/Examples/Toy/Ch2/invalid.mlir
@@ -5,5 +5,6 @@
// - toy.print should take an argument.
// - There should be a block terminator.
toy.func @main() {
- %0 = "toy.print"() : () -> tensor<2x3xf64>
+ // %0 = "toy.print"() : () -> tensor<2x3xf64>
+ %0 = "toy.print"() // : () -> tensor<2x3xf64>
}
diff --git a/mlir/test/Examples/Toy/Ch2/test.sh b/mlir/test/Examples/Toy/Ch2/test.sh
new file mode 100644
index 0000000000000..b398e9384a084
--- /dev/null
+++ b/mlir/test/Examples/Toy/Ch2/test.sh
@@ -0,0 +1,9 @@
+export PATH=/mnt/data01/zmz/workspace/07ascendnpu/llvm/llvm-project/build/bin/:$PATH
+
+toyc-ch2 ./invalid.mlir -emit=mlir 2>&1
+
+toyc-ch2 ./scalar.toy -emit=mlir 2>&1 # | FileCheck ./scalar.toy
+
+toyc-ch2 ./codegen.toy -emit=mlir -mlir-print-debuginfo 2>&1 # | FileCheck ./codegen.toy
+
+# toyc-ch2 ./codegen.toy -emit=error 2>&1 # | FileCheck ./codegen.toy
diff --git a/mlir/test/Examples/Toy/Ch3/test.sh b/mlir/test/Examples/Toy/Ch3/test.sh
new file mode 100644
index 0000000000000..eacd3566a2a97
--- /dev/null
+++ b/mlir/test/Examples/Toy/Ch3/test.sh
@@ -0,0 +1,20 @@
+export PATH=/mnt/data01/zmz/workspace/07ascendnpu/llvm/llvm-project/build/bin/:$PATH
+
+# toyc-ch3 ./invalid.mlir -emit=mlir 2>&1
+
+toyc-ch3 ./scalar.toy -emit=mlir 2>&1 # | FileCheck ./scalar.toy
+toyc-ch3 ./scalar.toy -emit=mlir 2>&1 | FileCheck ./scalar.toy
+
+# toyc-ch3 ./codegen.toy -emit=mlir -mlir-print-debuginfo 2>&1 # | FileCheck ./codegen.toy
+
+# toyc-ch3 ./codegen.toy -emit=error 2>&1 # | FileCheck ./codegen.toy
+
+toyc-ch3 ./codegen.toy -emit=mlir 2>&1 | FileCheck ./codegen.toy
+
+toyc-ch3 ./transpose_transpose.toy -emit=mlir -opt # 添加优化选项，可以执行SimplifyRedundantTranspose，将两个transpose给消除掉
+
+toyc-ch3 ./transpose_transpose.toy -emit=mlir
+
+toyc-ch3 ./trivial_reshape.toy -emit=mlir
+toyc-ch3 ./trivial_reshape.toy -emit=mlir -opt
+
More information about the llvm-commits
mailing list