jonathan2251 / lbd Goto Github PK

Hi, I encounter a build issue and I don't know how to fix it, maybe it's my environment issue.
Does anybody can help me ? Thanks : )

the CMakeError.log" is :

Checking whether the ASM compiler is GNU using "--version" did not match "(GNU assembler)|(GCC)|(Free Software Foundation)":
clang version 17.0.0
Target: x86_64-unknown-linux-gnu
Thread model: posix
InstalledDir: /media/Workspace/llvm/debug/build/bin
Performing C++ SOURCE FILE Test LLVM_LIBSTDCXX_MIN failed with the following output:
Change Dir: /media/Workspace/llvm/test/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/ninja cmTC_9b05f && [1/2] Building CXX object CMakeFiles/cmTC_9b05f.dir/src.cxx.o
FAILED: CMakeFiles/cmTC_9b05f.dir/src.cxx.o
/media/Workspace/llvm/debug/build/bin/clang++ -DLLVM_LIBSTDCXX_MIN  -std=c++0x -std=c++14 -MD -MT CMakeFiles/cmTC_9b05f.dir/src.cxx.o -MF CMakeFiles/cmTC_9b05f.dir/src.cxx.o.d -o CMakeFiles/cmTC_9b05f.dir/src.cxx.o -c /media/Workspace/llvm/test/build/CMakeFiles/CMakeTmp/src.cxx
/media/Workspace/llvm/test/build/CMakeFiles/CMakeTmp/src.cxx:2:10: fatal error: 'iosfwd' file not found
    2 | #include <iosfwd>
      |          ^~~~~~~~
1 error generated.
ninja: build stopped: subcommand failed.


Source file was:

#include <iosfwd>
#if defined(__GLIBCXX__)
#if !defined(_GLIBCXX_RELEASE) || _GLIBCXX_RELEASE < 7
#error Unsupported libstdc++ version
#endif
#endif
int main() { return 0; }

Hi Jonathan,
I‘m impressed with the amount of effort you have put into writing this tutorial. But to be frank, the amount of spelling and grammatical mistakes in this make it painful to read. Of course, I get where you are coming from given that both of us are non-native English speakers. For me personally, ChatGPT has come through in a lot of different language-related situations. For instance, it may help you with the following prompt:

You are a technical writer for the LLVM infrastructure. The following paragraphs contain numerous linguistical errata.  Please correct the following paragraphs in a scientific and technical tone.

If you want to, I may assist you with correcting your tutorial by opening PRs.

Hello, dear friends!
I am trying to install LLVM and the Cpu0 example code following:
"Appendix A: Getting Started: Installing LLVM and the Cpu0 example code" instructions.
Everything builds ithout a problem, but when the installation is completed, folder Examplecode/lbdex/InputFiles is missing in this location:
~/llvm/test/src/lib/Target/Cpu0/ExampleCode/lbdex/InputFiles.
I downloaded lbdex archive from here: http://jonathan2251.github.io/lbd/lbdex.tar.gz.
I have repeated the installation several times with the same result...
My OS is Debian 4.9.13 Kali

Hope you can help! Thanks in advance!

I'm compiling the Cpu0 code with LLVM 17, and get below error message:

FAILED: lib/Target/Cpu0/MCTargetDesc/CMakeFiles/LLVMCpu0Desc.dir/Cpu0MCTargetDesc.cpp.o 
/usr/sbin/clang++ -DGTEST_HAS_RTTI=0 -D_DEBUG -D_GLIBCXX_ASSERTIONS -D_GNU_SOURCE -D_LIBCPP_ENABLE_ASSERTIONS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -I/root/projects/llvm-project/llvm/build/lib/Target/Cpu0/MCTargetDesc -I/root/projects/llvm-project/llvm/lib/Target/Cpu0/MCTargetDesc -I/root/projects/llvm-project/llvm/lib/Target/Cpu0 -I/root/projects/llvm-project/llvm/build/lib/Target/Cpu0 -I/root/projects/llvm-project/llvm/build/include -I/root/projects/llvm-project/llvm/include -fPIC -fno-semantic-interposition -fvisibility-inlines-hidden -Werror=date-time -Werror=unguarded-availability-new -Wall -Wextra -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wmissing-field-initializers -pedantic -Wno-long-long -Wc++98-compat-extra-semi -Wimplicit-fallthrough -Wcovered-switch-default -Wno-noexcept-type -Wnon-virtual-dtor -Wdelete-non-virtual-dtor -Wsuggest-override -Wno-comment -Wstring-conversion -Wmisleading-indentation -Wctad-maybe-unsupported -fdiagnostics-color -g -fvisibility=hidden  -fno-exceptions -funwind-tables -fno-rtti -std=c++17 -MD -MT lib/Target/Cpu0/MCTargetDesc/CMakeFiles/LLVMCpu0Desc.dir/Cpu0MCTargetDesc.cpp.o -MF lib/Target/Cpu0/MCTargetDesc/CMakeFiles/LLVMCpu0Desc.dir/Cpu0MCTargetDesc.cpp.o.d -o lib/Target/Cpu0/MCTargetDesc/CMakeFiles/LLVMCpu0Desc.dir/Cpu0MCTargetDesc.cpp.o -c /root/projects/llvm-project/llvm/lib/Target/Cpu0/MCTargetDesc/Cpu0MCTargetDesc.cpp
In file included from /root/projects/llvm-project/llvm/lib/Target/Cpu0/MCTargetDesc/Cpu0MCTargetDesc.cpp:34:
/root/projects/llvm-project/llvm/build/lib/Target/Cpu0/Cpu0GenSubtargetInfo.inc:74:57: warning: zero size arrays are an extension [-Wzero-length-array]
extern const llvm::SubtargetSubTypeKV Cpu0SubTypeKV[] = {
                                                        ^
/root/projects/llvm-project/llvm/build/lib/Target/Cpu0/Cpu0GenSubtargetInfo.inc:105:73: error: use of undeclared identifier 'None'
  return new Cpu0GenMCSubtargetInfo(TT, CPU, TuneCPU, FS, std::nullopt, None, 
                                                                        ^
1 warning and 1 error generated.

I've tried to solve it by myself, but as a newbie, it's really hard.

What should I do to fix the "use of undeclared identifier" issue? Seems there is something missing.

Configuration and compilation commands (the working directory is llvm-project/llvm):

$ cmake -S . -B build -G Ninja -DCMAKE_EXPORT_COMPILE_COMMANDS=1 -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Debug
$ cmake --build build --target all

Hello Jonathan,

I am new in LLVM.
Using Ubuntu 20.04, it seems that the build-llvm.sh does not download the llvm-project properly:

bash build-llvm.sh
~/llvm ~/git/lbd/lbdex/install_llvm
Cloning into 'llvm-project'...
fatal: unable to connect to github.com:
github.com[0: 192.30.255.113]: errno=Connection timed out
build-llvm.sh: line 11: cd: llvm-project: No such file or directory
fatal: not a git repository (or any of the parent directories): .git
fatal: not a git repository (or any of the parent directories): .git
cp: cannot stat 'llvm-project/clang': No such file or directory
cp: cannot stat 'llvm-projHello Jonathan,

I am new in LLVM.
Using Ubuntu 20.04, it seems that the build-llvm.sh does not download the llvm-project properly:

bash build-llvm.sh
~/llvm ~/git/lbd/lbdex/install_llvm
Cloning into 'llvm-project'...
fatal: unable to connect to github.com:
github.com[0: 192.30.255.113]: errno=Connection timed out
build-llvm.sh: line 11: cd: llvm-project: No such file or directory
fatal: not a git repository (or any of the parent directories): .git
fatal: not a git repository (or any of the parent directories): .git
cp: cannot stat 'llvm-project/clang': No such file or directory
cp: cannot stat 'llvm-project/llvm': No such file or directory
cp: cannot stat 'llvm-project/compiler-rt': No such file or directory
~/llvm/release/build ~ ~/git/lbd/lbdex/install_llvm
OS = Linux
CMake Error: The source directory "/home//llvm/release/llvm" does not exist.
Specify --help for usage, or press the help button on the CMake GUI.
ninja: error: loading 'build.ninja': No such file or directory
~ ~/git/lbd/lbdex/install_llvm
~/git/lbd/lbdex/install_llvm
Please remember to add ${HOME}/llvm/release/bin to variable ${PATH} to your environment for clang++, clang.ect/llvm': No such file or directory
cp: cannot stat 'llvm-project/compiler-rt': No such file or directory
~/llvm/release/build ~ ~/git/lbd/lbdex/install_llvm
OS = Linux
CMake Error: The source directory "/home//llvm/release/llvm" does not exist.
Specify --help for usage, or press the help button on the CMake GUI.
ninja: error: loading 'build.ninja': No such file or directory
~ ~/git/lbd/lbdex/install_llvm
~/git/lbd/lbdex/install_llvm
Please remember to add ${HOME}/llvm/release/bin to variable ${PATH} to your environment for clang++, clang.

Hey,

I am reading this book and noticed that the Triple.h, which is referenced here (https://jonathan2251.github.io/lbd/llvmstructure.html#cpu0-backend-machine-id-and-relocation-records), is not correct anymore. It was moved to TargetParser.

Old folder: https://github.com/llvm/llvm-project/tree/main/llvm/include/llvm/ADT
New: https://github.com/llvm/llvm-project/blob/main/llvm/include/llvm/TargetParser/Triple.h#L44

Hello, Jonathan. You tell us how to transfer the LLVM IR into the target CPU instructions in this http://jonathan2251.github.io/lbd/llvmstructure.html. I learned a lot. but I am still confused that how to compile the code in chapter2. Could you list the steps to compile the code in chapter2? Thank you very much!

Benji Xu

Hello Jonathan!
I am having problems with vararg support in the Cpu0 backend implementation.

In short, it's this piece of code.
It saves varargs originally passed in registers onto stacks. But it sometimes goes out of the stack frame of the original function. I guess it's because when computing the stack size, varargs are not taken into account.

Below is my example:

For the following C code

#include <stdarg.h>

unsigned sum(unsigned amount, ...)
{
  unsigned i = 0;
  unsigned val = 0;
  unsigned sum = 0;
  va_list vl;
  va_start(vl, amount);
  for (i = 0; i < amount; i++) {
    val = va_arg(vl, unsigned);
    sum += val;
  }
  va_end(vl);
  return sum;
}

llc generates the assembly

......
	.type	sum,@function
	.ent	sum                     # @sum
sum:
	.frame	$fp,24,$lr
	.mask 	0x00001000,-4
	.set	noreorder
	.set	nomacro
# BB#0:
	addiu	$sp, $sp, -24
	st	$fp, 20($sp)            # 4-byte Folded Spill
	move	 $fp, $sp
	st	$5, 28($fp)
	st	$4, 16($fp)
......

The stack frame is only 24 bytes large, but it's trying to do st $5, 28($fp). so it's storing $5 beyond sum's stack frame, which causes segfaults in some cases.

Is it a bug or have I made any mistakes?

My environment:

• Ubuntu 16.04 64bit
• clang 3.8
• llvm 3.9 (the same as in your backend tutorial)

The command I used to generate assembly:

clang -target mips-unknown-linux-gnu -O0 -emit-llvm -c test.c
../llc -march=cpu0el -relocation-model=static -filetype=asm test.bc -o test.s

hi, @Jonathan2251:

When I tried to generate the "Cpu0GenCodeEmitter.inc" with tablegen(LLVM Cpu0GenCodeEmitter.inc -gen-emitter), It failed with the following message.

lib/Target/Cpu0/Cpu0InstrInfo.td:1037:5: error: No operand named addr in record JSUB
def JSUB    : JumpLink<0x3b, "jsub">;
    ^
note: Dumping record for previous error:
much more messages ...

Finally, I found the "FJ" in "Cpu0InstrFormats.td" has defined "bits<24> addr;" field,
however, the "class JumpLink" adopt the "$target" as follows.

 let isCall=1, hasDelaySlot=1 in {
   //@JumpLink {
   class JumpLink<bits<8> op, string instr_asm>:
-    FJ<op, (outs), (ins calltarget:$target, variable_ops),
-       !strconcat(instr_asm, "\t$target"), [(Cpu0JmpLink imm:$target)],
        IIBranch> {
 //#if CH >= CH10_1 2
        let DecoderMethod = "DecodeJumpTarget";

class FJ<bits<8> op, dag outs, dag ins, string asmstr, list<dag> pattern,
         InstrItinClass itin>: Cpu0Inst<outs, ins, asmstr, pattern, itin, FrmJ>
{
  bits<24> addr;

  let Opcode = op;

  let Inst{23-0} = addr;
}

After replacing the "$target" to "$addr", tblgen build successfully.

 let isCall=1, hasDelaySlot=1 in {
   //@JumpLink {
   class JumpLink<bits<8> op, string instr_asm>:
-    FJ<op, (outs), (ins calltarget:$target, variable_ops),
-       !strconcat(instr_asm, "\t$target"), [(Cpu0JmpLink imm:$target)],
+    FJ<op, (outs), (ins calltarget:$addr, variable_ops),
+       !strconcat(instr_asm, "\t$addr"), [(Cpu0JmpLink imm:$addr)],
        IIBranch> {
 //#if CH >= CH10_1 2
        let DecoderMethod = "DecodeJumpTarget";

Have you found this issue, or it's just a regression from the llvm project's evolution?