updated on 2022/12/5

• commands
• semantic notes
• semantic checklist
• IR design
• raw IR checklist
• codegen notes
• raw codegen checklist

test.ll(generated by my compiler from test.mx) and builtin.ll are both x86-64 target.

• sh ir-test.sh
• sh ir-draw.sh

The testIR flag in Compiler.java should be set to true.

• python3 ir-auto-test.py

• sh asm-riscv-gen.sh to generate test.s in risv32 target.
• sh codegen-test.sh use test.s, builtin.s and ravel to execute the program.

• to .ll
  • clang -S -emit-llvm test.c
  • clang -S -emit-llvm test.c --target=riscv32
• to .bc
  • clang -c -emit-llvm test.c

• llvm-link test.ll builtin.ll -o linked.bc

• clang test.ll -o test
• clang test.bc -o test

• lli test.ll
• lli test.bc

• llc test.ll -o test.s
• llc test.bc -o test.s

• clang -S test.c
• clang -S test.c --target=riscv32 -march=rv32im i means basic integer instructions, m means multiply and divide, which we will use in code-generation phase.

• llc test.ll -o test.s -march=riscv32 -mattr=+m m tells the compiler not to use __mul but to use mul instruction.