llvm-beginner

This is a repo to store notes when learning the llvm.

LLVM installation: https://solarianprogrammer.com/2013/01/17/building-clang-libcpp-ubuntu-linux/

Check notes.

llvm-cookbook notes

Learn from llvm-cookbook

Chap4

Reference:

Steps:

Create a new folder at <llvm_src>/lib/Transforms/: FuncBlockCount
Create required files that used to create the library. I just copy the Hello folder and change the name from Hello to FuncBlockCount accordiingly.
Write the pass flowed by llvm-cookbook. There are several additional work to make the pass work.
- Error:
  
  Loop is not declared.
- Solution: add analysis pass header file: #include "llvm/Analysis/LoopInfo.h"
- Error:
  
  opt: .../PassAnalysisSupport.h:235: AnalysisType& llvm::Pass::getAnalysisID(llvm::AnalysisID) const [with AnalysisType = llvm::LoopInfoWrapperPass; llvm::AnalysisID = const void*]: Assertion `ResultPass && "getAnalysis*() called on an analysis that was not " "'required' by pass!"' failed.
- Solution: add getAnalysisUsage function after runOnFunction
```
virtual void getAnalysisUsage(AnalysisUsage& AU) const override {
AU.addRequired<LoopInfoWrapperPass>();
}
```

Modify the CMakelists.txt in this pass folder and the parent folder.

// pass folder
if( NOT LLVM_REQUIRES_RTTI )
if( NOT LLVM_REQUIRES_EH )
    set(LLVM_EXPORTED_SYMBOL_FILE ${CMAKE_CURRENT_SOURCE_DIR}/FuncBlockCount.exports)
endif()
endif()

if(WIN32 OR CYGWIN)
set(LLVM_LINK_COMPONENTS Core Support)
endif()

add_llvm_library( LLVMFuncBlockCount MODULE BUILDTREE_ONLY
FuncBlockCount.cpp

DEPENDS
intrinsics_gen
PLUGIN_TOOL
opt
)
-------------------------------------
// parent folder
add_subdirectory(xxx)

Compile this pass, run make in build/ folder. Or only compile this new pass by using llvm-pass-skeleton.
compile the test source code (I target to arm): clang -O0 -S -emit-llvm target-program.cpp -o target-program3.ll
check our pass: opt -load build/lib/xx.so -xx -disable-output test.ll

chap5

Add pass in the llvm/lib/Transforms/Scalar/

chap6

Life of an LLVM IR instruction:

C code to LLVM IR
→ IR optimization
→ LLVM IR to SelectionDAG (visit each IR instruction to create an SDAGNode)
→ SelectionDAG legalization (make it to support target architecture)
→ Conversion from target-independent DAG to machine DAG (MachineSDNode, machine instructions are described in the target description .td file)
→ Scheduling instructions (covert d DAG into a linear set of instructions)
→ Register allocation (SSA has unlimited registers)
→ Code emission (JIT or llc: generate assembly instructions for a target)

Visualizing LLVM IR CFG using graphviz

$ llc --view-dag-combine-lt-dags test.ll
$ dot -Tpng /tmp/dag._Z8additionv-88c78e.dot > test.png

Set the LLVM to support embedded platform

Choice 1. LLVM embedded toolchain for arm

Install the toolchain

Modify the <repo_root>/sample/src/<proj_name>/Makefile to generate .ll

hello.ll: *.c
  $(BIN_PATH)/clang --config armv6m_soft_nofp_rdimon_baremetal -g -T ../../ldscripts/microbit.ld -O0 -S -emit-llvm hello.c -o hello.ll

Gnereate the .ll file

Add new pass

# Add the folder in repo-1.0/llvm/lib/Transform/Test
$ cp -r Test <root_repo>/repo-1.0/llvm/lib/Transform/Test
# Compile the new pass
$ cd <root_repo>/build-1.0/llvm
$ make -j4

Run new pass

# run the new pass
$ ../../../build-0.1/llvm/bin/opt -load ../../../build-0.1/llvm/lib/LLVMTest.so -test -disable-output hello.ll

Run samples and Debug (needs qemu version >= 4.0.0)

$ make
$ make debug
$ arm-none-eabi-gdb hello.elf
(gdb) target remote :1234

[TODO] Choice 2 . Use flags to specific the target. For example, I install LLVM 9.0.0 and I don't want to reinstall the LLVM.
- Use the clang to compile (learn from the previous repo *cfg files):
```
clang --target=armv8m.main-non-eabi -mfloat-abi=soft -march=armv8m.main+nofp -fno-exceptions -fno-rtti -O0 -S -emit-llvm target-program.cpp -o target-program4.ll
```
- Use the llc to generate the .s
  - run llc -march=arm -mattr=help to check which cpu it supports
  - I select the cortex-m33:
```
llc -mcpu=cortex-m33 target-program4.ll -o target-program4.s
```
- But this is not convenient to generate the executable .elf file since we need to write linker script for the simple tests.

Install arm-gcc and qemu

Install arm gcc toolchain. (Reference)

Download the arm gcc toolchain from link.

sudo tar xjf gcc-arm-none-eabi-your-version.bz2 -C /usr/share/
sudo ln -s /usr/share/gcc-arm-none-eabi-your-version/bin/* /usr/bin/

Install qemu 4.0.0. Reference

Update the python3. Reference.

$ sudo apt update
$ sudo apt install software-properties-common
$ sudo add-apt-repository ppa:deadsnakes/ppa
$ sudo apt update
$ sudo apt install python3.8
---------- Link newly installed python3----
$ which python3.8
/usr/bin/python3.8
$ sudo update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.8 1
$ which python3.5
/usr/bin/python3.5
$ sudo update-alternatives --config python3
[select 2, and enter]

Install Ninja:
```
sudo apt install ninja-build
```
Clear previous installation:
```
sudo apt-get purge "qemu*"
sudo apt-get autoremove
```
- Enable Source Code repositories in Software and Updates (software-properties-gtk) and then:
```
sudo apt-get build-dep qemu
```
- Download qemu and compile (I installed qemu 4.0.0 for microbit is supported since qemu 4.0.0):
```
cd ~/Downloads
wget https://download.qemu.org/qemu-4.0.0.tar.xz
tar -xf qemu-4.0.0.tar.xz
cd qemu-4.0.0/
```
Decide the target we want. I only need the x86 and arm, so specifi the confiration (run ./configure --help to get the suppported tareget list.).
```
./configure --target-list=aarch64-softmmu,aarch64-linux-user,arm-softmmu,arm-linux-user,x86_64-softmmu,x86_64-linux-user 
make
```

Use checkinstall, the qemu package will be generated as a .deb file and can be managed by the package manager. Can be easy to remove with dpkg - r qemu.

sudo apt-get install checkinstall
sudo checkinstall make install
sudo apt-get install ./*.deb

Check the version:

$ qemu-system-arm --version
QEMU emulator version 4.0.0
Copyright (c) 2003-2019 Fabrice Bellard and the QEMU Project developers

Test FreeRTOS with qemu

Refer to FreeRTOS Demo for m3_mps2+

ORTEX_M3_MPS2_QEMU_GCC git:(main) ✗ sudo qemu-system-arm -machine mps2-an385 -monitor null -semihosting \
        --semihosting-config enable=on,target=native \
        -kernel ./build/RTOSDemo.axf \
        -serial stdio -nographic -s -S
blinking
blinking
blinking
blinking

mintancy/llvm-beginner