Skip to content

Latest commit

 

History

History
26 lines (19 loc) · 791 Bytes

README.md

File metadata and controls

26 lines (19 loc) · 791 Bytes
Raw

Lowering from bitcode to target machine assembly

Now we are going to use the LLVM static compiler llc to lower the code once more.

Run :

$ llc mult.bc -o mult.s

Or using Clang API :

$ clang -S mult.bc -o mult.s -fomit-frame-pointer

We use the -fomit-frame-pointer option because by default Clang does not elimate the frame pointer whereas llc does it by default.

Remarks

Here we did not mention the target architecture, by default the assembly code is generated for the host's architecture. One can toggle different options depending on where the code is going to be run.

  • -march=architecture : target a specific hardware
  • -mcpu=cpu : specify a CPU whitin the architecture
  • -regalloc=basic/greedy/fast/pbqp : type of register allocation