I'm writing a compiler that uses LLVM as a backend, and my compiler generates the following LLVM assembly code
@0 = private 开发者_Go百科constant [25 x i8] c"Hello World to %dntegers\00"
declare void @printf (i8*, i32)
define void @main () {
%1 = getelementptr [25 x i8]* @0, i32 0, i32 0
%2 = 1
tail call void @printf(i8* %1, i32 %2)
ret void
}
But I get the following error:
c.ll:8:8: error: expected instruction opcode
%2 = 1
^
The documentation shows examples like that though.
Do you know how to get this to work? Currently, I'm using instead:
%2 = add i32 0, 1 ; FIXME
See the LLVMdev discussion "How to assign a constant to a register?" As pointed out there by Duncan Sands:
let me say that in general doing this is pointless. Due to SSA form, if %x is set to 5 you can't set it to something else later. Thus everywhere that you use %x you might as well just directly use 5 there instead. A common situation is that you have a register %x, and due to performing optimizations you discover that in fact %x will always have the value 5. You can use RAUW (aka the replaceAllUsesWith method) to replace %x with 5 everywhere.
The thread did generate a couple alternatives to directly using the constant:
- Use
alloca
to get a pointer, store to it, then load the value as needed. - Create a global pointing to the constant, then load that value as needed.
Someone pointed out that the alloca
approach will end up using a register after optimization.
Regardless, using the constant directly as suggested by @SK-logic seems to be the cleanest solution.
精彩评论